docs/CHANGELOG.md: cut v1.49.0

vendor: make vendor-update
deployment/docker: update Go builder from v1.15.5 to v1.15.6
2026-06-08 19:33:35 +03:00 · 2020-12-05 13:49:04 +02:00 · 2020-12-05 12:46:54 +02:00 · 2020-12-05 12:37:34 +02:00 · 2020-12-05 12:30:46 +02:00 · 2020-12-05 12:15:52 +02:00
406 changed files with 30959 additions and 4996 deletions
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,115 +0,0 @@
-# tip
-
-
-# [v1.44.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.44.0)
-
-* FEATURE: automatically add missing label filters to binary operands as described at https://utcc.utoronto.ca/~cks/space/blog/sysadmin/PrometheusLabelNonOptimization .
-  This should improve performance for queries with missing label filters in binary operands. For example, the following query should work faster now, because it shouldn't
-  fetch and discard time series for `node_filesystem_files_free` metric without matching labels for the left side of the expression:
-  ```
-     node_filesystem_files{ host="$host", mountpoint="/" } - node_filesystem_files_free
-  ```
-* FEATURE: vmagent: add Docker Swarm service discovery (aka [dockerswarm_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#dockerswarm_sd_config)).
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/656
-* FEATURE: add ability to export data in CSV format. See [these docs](https://victoriametrics.github.io/#how-to-export-csv-data) for details.
-* FEATURE: vmagent: add `-promscrape.suppressDuplicateScrapeTargetErrors` command-line flag for suppressing `duplicate scrape target` errors.
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/651 and https://victoriametrics.github.io/vmagent.html#troubleshooting .
-* FEATURE: vmagent: show original labels before relabeling is applied on `duplicate scrape target` errors. This should simplify debugging for incorrect relabeling.
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/651
-* FEATURE: vmagent: `/targets` page now accepts optional `show_original_labels=1` query arg for displaying original labels for each target before relabeling is applied.
-  This should simplify debugging for target relabeling configs. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/651
-* FEATURE: add `-finalMergeDelay` command-line flag for configuring the delay before final merge for per-month partitions.
-  The final merge is started after no new data is ingested into per-month partition during `-finalMergeDelay`.
-* FEATURE: add `vm_rows_added_to_storage_total` metric, which shows the total number of rows added to storage since app start.
-  The `sum(rate(vm_rows_added_to_storage_total))` can be smaller than `sum(rate(vm_rows_inserted_total))` if certain metrics are dropped
-  due to [relabeling](https://victoriametrics.github.io/#relabeling). The `sum(rate(vm_rows_added_to_storage_total))` can be bigger
-  than `sum(rate(vm_rows_inserted_total))` if [replication](https://victoriametrics.github.io/Cluster-VictoriaMetrics.html#replication-and-data-safety) is enabled.
-* FEATURE: keep metric name after applying [MetricsQL](https://victoriametrics.github.io/MetricsQL.html) functions, which don't change time series meaning.
-  The list of such functions:
-   * `keep_last_value`
-   * `keep_next_value`
-   * `interpolate`
-   * `running_min`
-   * `running_max`
-   * `running_avg`
-   * `range_min`
-   * `range_max`
-   * `range_avg`
-   * `range_first`
-   * `range_last`
-   * `range_quantile`
-   * `smooth_exponential`
-   * `ceil`
-   * `floor`
-   * `round`
-   * `clamp_min`
-   * `clamp_max`
-   * `max_over_time`
-   * `min_over_time`
-   * `avg_over_time`
-   * `quantile_over_time`
-   * `mode_over_time`
-   * `geomean_over_time`
-   * `holt_winters`
-   * `predict_linear`
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/674
-
-* BUGFIX: properly handle stale time series after K8S deployment. Previously such time series could be double-counted.
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/748
-* BUGFIX: return a single time series at max from `absent()` function like Prometheus does.
-* BUGFIX: vmalert: accept days, weeks and years in `for: ` part of config like Prometheus does. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/817
-* BUGFIX: fix `mode_over_time(m[d])` calculations. Previously the function could return incorrect results.
-
-
-# [v1.43.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.43.0)
-
-* FEATURE: reduce CPU usage for repeated queries over sliding time window when no new time series are added to the database.
-  Typical use cases: repeated evaluation of alerting rules in [vmalert](https://victoriametrics.github.io/vmalert.html) or dashboard auto-refresh in Grafana.
-* FEATURE: vmagent: add OpenStack service discovery aka [openstack_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#openstack_sd_config).
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/728 .
-* FEATURE: vmalert: make `-maxIdleConnections` configurable for datasource HTTP client. This option can be used for minimizing connection churn.
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/795 .
-* FEATURE: add `-influx.maxLineSize` command-line flag for configuring the maximum size for a single Influx line during parsing.
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/807
-
-* BUGFIX: properly handle `inf` values during [background merge of LSM parts](https://medium.com/@valyala/how-victoriametrics-makes-instant-snapshots-for-multi-terabyte-time-series-data-e1f3fb0e0282).
-  Previously `Inf` values could result in `NaN` values for adjancent samples in time series. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/805 .
-* BUGFIX: fill gaps on graphs for `range_*` and `running_*` functions. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/806 .
-* BUGFIX: make a copy of label with new name during relabeling with `action: labelmap` in the same way as Prometheus does.
-  Previously the original label name has been replaced. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/812 .
-* BUGFIX: support parsing floating-point timestamp like Graphite Carbon does. Such timestmaps are truncated to seconds.
-
-
-# [v1.42.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.42.0)
-
-* FEATURE: use all the available CPU cores when accepting data via a single TCP connection
-  for [all the supported protocols](https://victoriametrics.github.io/#how-to-import-time-series-data).
-  Previously data ingested via a single TCP connection could use only a single CPU core. This could limit data ingestion performance.
-  The main benefit of this feature is that data can be imported at max speed via a single connection - there is no need to open multiple concurrent
-  connections to VictoriaMetrics or [vmagent](https://victoriametrics.github.io/vmagent.html) in order to achieve the maximum data ingestion speed.
-* FEATURE: cluster: improve performance for data ingestion path from `vminsert` to `vmstorage` nodes. The maximum data ingestion performance
-  for a single connection between `vminsert` and `vmstorage` node scales with the number of available CPU cores on `vmstorage` side.
-  This should help with https://github.com/VictoriaMetrics/VictoriaMetrics/issues/791 .
-* FEATURE: add ability to export / import data in native format via `/api/v1/export/native` and `/api/v1/import/native`.
-  This is the most optimized approach for data migration between VictoriaMetrics instances. Both single-node and cluster instances are supported.
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/787#issuecomment-700632551 .
-* FEATURE: add `reduce_mem_usage` query option to `/api/v1/export` in order to reduce memory usage during data export / import.
-  See [these docs](https://victoriametrics.github.io/#how-to-export-data-in-json-line-format) for details.
-* FEATURE: improve performance for `/api/v1/series` handler when it returns big number of time series.
-* FEATURE: add `vm_merge_need_free_disk_space` metric, which can be used for estimating the number of deferred background data merges due to the lack of free disk space.
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/686 .
-* FEATURE: add OpenBSD support. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/785 .
-
-* BUGFIX: properly apply `-search.maxStalenessInterval` command-line flag value. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/784 .
-* BUGFIX: fix displaying data in Grafana tables. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/720 .
-* BUGFIX: do not adjust the number of detected CPU cores found at `/sys/devices/system/cpu/online`.
-  The adjustement was increasing the resulting GOMAXPROC by 1, which looked confusing to users.
-  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/685#issuecomment-698595309 .
-* BUGFIX: vmagent: do not show `-remoteWrite.url` in initial logs if `-remoteWrite.showURL` isn't set. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/773 .
-* BUGFIX: properly handle case when [/metrics/find](https://victoriametrics.github.io/#graphite-metrics-api-usage) finds both a leaf and a node for the given `query=prefix.*`.
-  In this case only the node must be returned with stripped dot in the end of id as carbonapi does.
-
-
-# Previous releases
-
-See [releases page](https://github.com/VictoriaMetrics/VictoriaMetrics/releases).
--- a/4
+++ b/4
@@ -122,8 +122,8 @@ benchmark-pure:
 	GO111MODULE=on CGO_ENABLED=0 go test -mod=vendor -bench=. ./app/...

 vendor-update:
-	GO111MODULE=on go get -u ./lib/...
-	GO111MODULE=on go get -u ./app/...
+	GO111MODULE=on go get -u -d ./lib/...
+	GO111MODULE=on go get -u -d ./app/...
 	GO111MODULE=on go mod tidy
 	GO111MODULE=on go mod vendor

--- a/README.md
+++ b/README.md
@@ -10,23 +10,27 @@

 ## VictoriaMetrics

-VictoriaMetrics is fast, cost-effective and scalable time-series database.
+VictoriaMetrics is fast, cost-effective and scalable monitoring solution and time series database.

 It is available in [binary releases](https://github.com/VictoriaMetrics/VictoriaMetrics/releases),
-[docker images](https://hub.docker.com/r/victoriametrics/victoria-metrics/) and
-in [source code](https://github.com/VictoriaMetrics/VictoriaMetrics). Just download VictoriaMetrics and see [how to start it](#how-to-start-victoriametrics).
+[docker images](https://hub.docker.com/r/victoriametrics/victoria-metrics/), [Snap package](https://snapcraft.io/victoriametrics)
+and in [source code](https://github.com/VictoriaMetrics/VictoriaMetrics). Just download VictoriaMetrics and see [how to start it](#how-to-start-victoriametrics).
+If you use Ubuntu, then just run `snap install victoriametrics` in order to install and run it.
+Then read [Prometheus setup](#prometheus-setup) and [Grafana setup](#grafana-setup) docs.

 Cluster version is available [here](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/cluster).

 See our [Wiki](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki) for additional documentation.

 [Contact us](mailto:info@victoriametrics.com) if you need paid enterprise support for VictoriaMetrics.
-See [features available for enterprise customers](https://github.com/VictoriaMetrics/VictoriaMetrics/issues?q=is%3Aissue+label%3Aenterprise).
+See [features available for enterprise customers](https://victoriametrics.com/enterprise.html).


 ## Case studies and talks

-* [Adidas](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#adidas)
+Click on a link in order to read the corresponding case study
+
+* [adidas](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#adidas)
 * [CERN](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#cern)
 * [COLOPL](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#colopl)
 * [Zerodha](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#zerodha)
@@ -38,23 +42,29 @@ See [features available for enterprise customers](https://github.com/VictoriaMet
 * [Brandwatch](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#brandwatch)
 * [Adsterra](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#adsterra)
 * [ARNES](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#arnes)
+* [Idealo.de](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#idealode)


 ## Prominent features

 * VictoriaMetrics can be used as long-term storage for Prometheus or for [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md).
  See [these docs](#prometheus-setup) for details.
-* Supports [Prometheus querying API](https://prometheus.io/docs/prometheus/latest/querying/api/), so it can be used as Prometheus drop-in replacement in Grafana.
-  VictoriaMetrics implements [MetricsQL](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/MetricsQL) query language, which is inspired by PromQL.
-* Supports global query view. Multiple Prometheus instances may write data into VictoriaMetrics. Later this data may be used in a single query.
+* VictoriaMetrics supports [Prometheus querying API](https://prometheus.io/docs/prometheus/latest/querying/api/), so it can be used as Prometheus drop-in replacement in Grafana.
+* VictoriaMetrics implements [MetricsQL](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/MetricsQL) query language backwards compatible with PromQL.
+* VictoriaMetrics provides global query view. Multiple Prometheus instances or any other data sources may ingest data into VictoriaMetrics.
+  Later this data may be queried via a single query.
 * High performance and good scalability for both [inserts](https://medium.com/@valyala/high-cardinality-tsdb-benchmarks-victoriametrics-vs-timescaledb-vs-influxdb-13e6ee64dd6b)
  and [selects](https://medium.com/@valyala/when-size-matters-benchmarking-victoriametrics-vs-timescale-and-influxdb-6035811952d4).
  [Outperforms InfluxDB and TimescaleDB by up to 20x](https://medium.com/@valyala/measuring-vertical-scalability-for-time-series-databases-in-google-cloud-92550d78d8ae).
-* [Uses 10x less RAM than InfluxDB](https://medium.com/@valyala/insert-benchmarks-with-inch-influxdb-vs-victoriametrics-e31a41ae2893) when working with millions of unique time series (aka high cardinality).
+* [Uses 10x less RAM than InfluxDB](https://medium.com/@valyala/insert-benchmarks-with-inch-influxdb-vs-victoriametrics-e31a41ae2893)
+  and [up to 7x less RAM than Prometheus, Thanos or Cortex](https://valyala.medium.com/prometheus-vs-victoriametrics-benchmark-on-node-exporter-metrics-4ca29c75590f)
+  when dealing with millions of unique time series (aka high cardinality).
 * Optimized for time series with high churn rate. Think about [prometheus-operator](https://github.com/coreos/prometheus-operator) metrics from frequent deployments in Kubernetes.
 * High data compression, so [up to 70x more data points](https://medium.com/@valyala/when-size-matters-benchmarking-victoriametrics-vs-timescale-and-influxdb-6035811952d4)
-  may be crammed into limited storage comparing to TimescaleDB.
-* Optimized for storage with high-latency IO and low IOPS (HDD and network storage in AWS, Google Cloud, Microsoft Azure, etc). See [graphs from these benchmarks](https://medium.com/@valyala/high-cardinality-tsdb-benchmarks-victoriametrics-vs-timescaledb-vs-influxdb-13e6ee64dd6b).
+  may be crammed into limited storage comparing to TimescaleDB
+  and [up to 7x less storage space is required comparing to Prometheus, Thanos or Cortex](https://valyala.medium.com/prometheus-vs-victoriametrics-benchmark-on-node-exporter-metrics-4ca29c75590f).
+* Optimized for storage with high-latency IO and low IOPS (HDD and network storage in AWS, Google Cloud, Microsoft Azure, etc).
+  See [graphs from these benchmarks](https://medium.com/@valyala/high-cardinality-tsdb-benchmarks-victoriametrics-vs-timescaledb-vs-influxdb-13e6ee64dd6b).
 * A single-node VictoriaMetrics may substitute moderately sized clusters built with competing solutions such as Thanos, M3DB, Cortex, InfluxDB or TimescaleDB.
  See [vertical scalability benchmarks](https://medium.com/@valyala/measuring-vertical-scalability-for-time-series-databases-in-google-cloud-92550d78d8ae),
  [comparing Thanos to VictoriaMetrics cluster](https://medium.com/@valyala/comparing-thanos-to-victoriametrics-cluster-b193bea1683)
@@ -63,7 +73,7 @@ See [features available for enterprise customers](https://github.com/VictoriaMet
 * Easy operation:
  * VictoriaMetrics consists of a single [small executable](https://medium.com/@valyala/stripping-dependency-bloat-in-victoriametrics-docker-image-983fb5912b0d) without external dependencies.
  * All the configuration is done via explicit command-line flags with reasonable defaults.
-  * All the data is stored in a single directory pointed by `-storageDataPath` flag.
+  * All the data is stored in a single directory pointed by `-storageDataPath` command-line flag.
  * Easy and fast backups from [instant snapshots](https://medium.com/@valyala/how-victoriametrics-makes-instant-snapshots-for-multi-terabyte-time-series-data-e1f3fb0e0282)
  to S3 or GCS with [vmbackup](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmbackup/README.md) / [vmrestore](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmrestore/README.md).
  See [this article](https://medium.com/@valyala/speeding-up-backups-for-big-time-series-databases-533c1a927883) for more details.
@@ -86,6 +96,7 @@ See [features available for enterprise customers](https://github.com/VictoriaMet
 * Has open source [cluster version](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/cluster).
 * See also technical [Articles about VictoriaMetrics](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/Articles).

+
 ## Operation

 ### Table of contents
@@ -103,7 +114,9 @@ See [features available for enterprise customers](https://github.com/VictoriaMet
 * [How to send data from OpenTSDB-compatible agents](#how-to-send-data-from-opentsdb-compatible-agents)
 * [Prometheus querying API usage](#prometheus-querying-api-usage)
  * [Prometheus querying API enhancements](#prometheus-querying-api-enhancements)
-* [Graphite Metrics API usage](#graphite-metrics-api-usage)
+* [Graphite API usage](#graphite-api-usage)
+  * [Graphite Metrics API usage](#graphite-metrics-api-usage)
+  * [Graphite Tags API usage](#graphite-tags-api-usage)
 * [How to build from sources](#how-to-build-from-sources)
  * [Development build](#development-build)
  * [Production build](#production-build)
@@ -155,33 +168,36 @@ See [features available for enterprise customers](https://github.com/VictoriaMet
    * [Color Palette](#color-palette)
  * [We kindly ask](#we-kindly-ask)

-### How to start VictoriaMetrics

-Just start VictoriaMetrics [executable](https://github.com/VictoriaMetrics/VictoriaMetrics/releases)
+## How to start VictoriaMetrics
+
+Start VictoriaMetrics [executable](https://github.com/VictoriaMetrics/VictoriaMetrics/releases)
 or [docker image](https://hub.docker.com/r/victoriametrics/victoria-metrics/) with the desired command-line flags.

 The following command-line flags are used the most:

-* `-storageDataPath` - path to data directory. VictoriaMetrics stores all the data in this directory. Default path is `victoria-metrics-data` in current working directory.
-* `-retentionPeriod` - retention period in months for the data. Older data is automatically deleted. Default period is 1 month.
+* `-storageDataPath` - path to data directory. VictoriaMetrics stores all the data in this directory. Default path is `victoria-metrics-data` in the current working directory.
+* `-retentionPeriod` - retention for stored data. Older data is automatically deleted. Default retention is 1 month. See [these docs](#retention) for more details.

-Other flags have good enough default values, so set them only if you really need this.
+Other flags have good enough default values, so set them only if you really need this. Pass `-help` to see all the available flags with description and default values.
+
+See how to [ingest data to VictoriaMetrics](#how-to-import-time-series-data), how to [query VictoriaMetrics](#grafana-setup)
+and how to [handle alerts](#alerting).
 VictoriaMetrics accepts [Prometheus querying API requests](#prometheus-querying-api-usage) on port `8428` by default.

-Pass `-help` to see all the available flags with description and default values.
-
 It is recommended setting up [monitoring](#monitoring) for VictoriaMetrics.

-#### Environment variables
+### Environment variables

-Each flag values can be set thru environment variables by following these rules:
+Each flag value can be set via environment variables according to these rules:

 * The `-envflag.enable` flag must be set
-* Each `.` in flag names must be substituted by `_` (for example `-insert.maxQueueDuration <duration>` will translate to `insert_maxQueueDuration=<duration>`)
-* For repeating flags, an alternative syntax can be used by joining the different values into one using `,` as separator (for example `-storageNode <nodeA> -storageNode <nodeB>` will translate to `storageNode=<nodeA>,<nodeB>`)
+* Each `.` char in flag name must be substituted by `_` (for example `-insert.maxQueueDuration <duration>` will translate to `insert_maxQueueDuration=<duration>`)
+* For repeating flags an alternative syntax can be used by joining the different values into one using `,` char as separator (for example `-storageNode <nodeA> -storageNode <nodeB>` will translate to `storageNode=<nodeA>,<nodeB>`)
 * It is possible setting prefix for environment vars with `-envflag.prefix`. For instance, if `-envflag.prefix=VM_`, then env vars must be prepended with `VM_`

-### Prometheus setup
+
+## Prometheus setup

 Prometheus must be configured with [remote_write](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#remote_write)
 in order to send data to VictoriaMetrics. Add the following lines
@@ -192,15 +208,15 @@ remote_write:
  - url: http://<victoriametrics-addr>:8428/api/v1/write
 ```

-Substitute `<victoriametrics-addr>` with the hostname or IP address of VictoriaMetrics.
-Then apply the new config via the following command:
+Substitute `<victoriametrics-addr>` with hostname or IP address of VictoriaMetrics.
+Then apply new config via the following command:

 ```bash
 kill -HUP `pidof prometheus`
 ```

 Prometheus writes incoming data to local storage and replicates it to remote storage in parallel.
-This means the data remains available in local storage for `--storage.tsdb.retention.time` duration
+This means that data remains available in local storage for `--storage.tsdb.retention.time` duration
 even if remote storage is unavailable.

 If you plan to send data to VictoriaMetrics from multiple Prometheus instances, then add the following lines into `global` section
@@ -213,11 +229,10 @@ global:
 ```

 This instructs Prometheus to add `datacenter=dc-123` label to each time series sent to remote storage.
-The label name may be arbitrary - `datacenter` is just an example. The label value must be unique
+The label name can be arbitrary - `datacenter` is just an example. The label value must be unique
 across Prometheus instances, so those time series may be filtered and grouped by this label.

-For highly loaded Prometheus instances (400k+ samples per second)
-the following tuning may be applied:
+For highly loaded Prometheus instances (400k+ samples per second) the following tuning may be applied:

 ```yaml
 remote_write:
@@ -228,22 +243,20 @@ remote_write:
      max_shards: 30
 ```

-Using remote write increases memory usage for Prometheus up to ~25%
-and depends on the shape of data. If you are experiencing issues with
-too high memory consumption try to lower `max_samples_per_send`
-and `capacity` params (keep in mind that these two params are tightly connected).
+Using remote write increases memory usage for Prometheus up to ~25% and depends on the shape of data. If you are experiencing issues with
+too high memory consumption try to lower `max_samples_per_send` and `capacity` params (keep in mind that these two params are tightly connected).
 Read more about tuning remote write for Prometheus [here](https://prometheus.io/docs/practices/remote_write).

-It is recommended upgrading Prometheus to [v2.12.0](https://github.com/prometheus/prometheus/releases) or newer,
-since the previous versions may have issues with `remote_write`.
+It is recommended upgrading Prometheus to [v2.12.0](https://github.com/prometheus/prometheus/releases) or newer, since previous versions may have issues with `remote_write`.

-Take a look also at [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md),
-which can be used as faster and less resource-hungry alternative to Prometheus in certain cases.
+Take a look also at [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md)
+and [vmalert](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmalert/README.md),
+which can be used as faster and less resource-hungry alternative to Prometheus.


-### Grafana setup
+## Grafana setup

-Create [Prometheus datasource](http://docs.grafana.org/features/datasources/prometheus/) in Grafana with the following Url:
+Create [Prometheus datasource](http://docs.grafana.org/features/datasources/prometheus/) in Grafana with the following url:

 ```url
 http://<victoriametrics-addr>:8428
@@ -251,10 +264,12 @@ http://<victoriametrics-addr>:8428

 Substitute `<victoriametrics-addr>` with the hostname or IP address of VictoriaMetrics.

-Then build graphs with the created datasource using [Prometheus query language](https://prometheus.io/docs/prometheus/latest/querying/basics/).
-VictoriaMetrics supports native PromQL and [extends it with useful features](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/MetricsQL).
+Then build graphs with the created datasource using [PromQL](https://prometheus.io/docs/prometheus/latest/querying/basics/)
+or [MetricsQL](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/MetricsQL). VictoriaMetrics supports [Prometheus querying API](#prometheus-querying-api-usage),
+which is used by Grafana.

-### How to upgrade VictoriaMetrics
+
+## How to upgrade VictoriaMetrics

 It is safe upgrading VictoriaMetrics to new versions unless [release notes](https://github.com/VictoriaMetrics/VictoriaMetrics/releases)
 say otherwise. It is safe skipping multiple versions during the upgrade unless [release notes](https://github.com/VictoriaMetrics/VictoriaMetrics/releases) say otherwise.
@@ -264,25 +279,27 @@ It is also safe downgrading to the previous version unless [release notes](https

 The following steps must be performed during the upgrade / downgrade:

-1) Send `SIGINT` signal to VictoriaMetrics process in order to gracefully stop it.
-2) Wait until the process stops. This can take a few seconds.
-3) Start the upgraded VictoriaMetrics.
+* Send `SIGINT` signal to VictoriaMetrics process in order to gracefully stop it.
+* Wait until the process stops. This can take a few seconds.
+* Start the upgraded VictoriaMetrics.

 Prometheus doesn't drop data during VictoriaMetrics restart.
 See [this article](https://grafana.com/blog/2019/03/25/whats-new-in-prometheus-2.8-wal-based-remote-write/) for details.

-### How to apply new config to VictoriaMetrics

-VictoriaMetrics must be restarted for applying new config:
+## How to apply new config to VictoriaMetrics

-1) Send `SIGINT` signal to VictoriaMetrics process in order to gracefully stop it.
-2) Wait until the process stops. This can take a few seconds.
-3) Start VictoriaMetrics with the new config.
+VictoriaMetrics is configured via command-line flags, so it must be restarted when new command-line flags should be applied:
+
+* Send `SIGINT` signal to VictoriaMetrics process in order to gracefully stop it.
+* Wait until the process stops. This can take a few seconds.
+* Start VictoriaMetrics with the new command-line flags.

 Prometheus doesn't drop data during VictoriaMetrics restart.
 See [this article](https://grafana.com/blog/2019/03/25/whats-new-in-prometheus-2.8-wal-based-remote-write/) for details.

-### How to scrape Prometheus exporters such as [node-exporter](https://github.com/prometheus/node_exporter)
+
+## How to scrape Prometheus exporters such as [node-exporter](https://github.com/prometheus/node_exporter)

 VictoriaMetrics can be used as drop-in replacement for Prometheus for scraping targets configured in `prometheus.yml` config file according to [the specification](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#configuration-file).
 Just set `-promscrape.config` command-line flag to the path to `prometheus.yml` config - and VictoriaMetrics should start scraping the configured targets.
@@ -297,8 +314,10 @@ Currently the following [scrape_config](https://prometheus.io/docs/prometheus/la
 * [dns_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#dns_sd_config)
 * [openstack_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#openstack_sd_config)
 * [dockerswarm_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#dockerswarm_sd_config)
+* [eureka_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#eureka_sd_config)

-In the future other `*_sd_config` types will be supported.
+
+Other `*_sd_config` types will be supported in the future.

 The file pointed by `-promscrape.config` may contain `%{ENV_VAR}` placeholders, which are substituted by the corresponding `ENV_VAR` environment variable values.

@@ -306,9 +325,10 @@ VictoriaMetrics also supports [importing data in Prometheus exposition format](#

 See also [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md), which can be used as drop-in replacement for Prometheus.

-### How to send data from InfluxDB-compatible agents such as [Telegraf](https://www.influxdata.com/time-series-platform/telegraf/)

-Just use `http://<victoriametric-addr>:8428` url instead of InfluxDB url in agents' configs.
+## How to send data from InfluxDB-compatible agents such as [Telegraf](https://www.influxdata.com/time-series-platform/telegraf/)
+
+Use `http://<victoriametric-addr>:8428` url instead of InfluxDB url in agents' configs.
 For instance, put the following lines into `Telegraf` config, so it sends data to VictoriaMetrics instead of InfluxDB:

 ```toml
@@ -316,8 +336,6 @@ For instance, put the following lines into `Telegraf` config, so it sends data t
  urls = ["http://<victoriametrics-addr>:8428"]
 ```

-Do not forget substituting `<victoriametrics-addr>` with the real address where VictoriaMetrics runs.
-
 Another option is to enable TCP and UDP receiver for Influx line protocol via `-influxListenAddr` command-line flag
 and stream plain Influx line protocol data to the configured TCP and/or UDP addresses.

@@ -352,7 +370,7 @@ to local VictoriaMetrics using `curl`:
 curl -d 'measurement,tag1=value1,tag2=value2 field1=123,field2=1.23' -X POST 'http://localhost:8428/write'
 ```

-An arbitrary number of lines delimited by '\n' (aka newline char) may be sent in a single request.
+An arbitrary number of lines delimited by '\n' (aka newline char) can be sent in a single request.
 After that the data may be read via [/api/v1/export](#how-to-export-data-in-json-line-format) endpoint:

 ```bash
@@ -369,16 +387,17 @@ The `/api/v1/export` endpoint should return the following response:
 Note that Influx line protocol expects [timestamps in *nanoseconds* by default](https://docs.influxdata.com/influxdb/v1.7/write_protocols/line_protocol_tutorial/#timestamp),
 while VictoriaMetrics stores them with *milliseconds* precision.

-### How to send data from Graphite-compatible agents such as [StatsD](https://github.com/etsy/statsd)

-1) Enable Graphite receiver in VictoriaMetrics by setting `-graphiteListenAddr` command line flag. For instance,
+## How to send data from Graphite-compatible agents such as [StatsD](https://github.com/etsy/statsd)
+
+Enable Graphite receiver in VictoriaMetrics by setting `-graphiteListenAddr` command line flag. For instance,
 the following command will enable Graphite receiver in VictoriaMetrics on TCP and UDP port `2003`:

 ```bash
 /path/to/victoria-metrics-prod -graphiteListenAddr=:2003
 ```

-2) Use the configured address in Graphite-compatible agents. For instance, set `graphiteHost`
+Use the configured address in Graphite-compatible agents. For instance, set `graphiteHost`
 to the VictoriaMetrics host in `StatsD` configs.

 Example for writing data with Graphite plaintext protocol to local VictoriaMetrics using `nc`:
@@ -388,7 +407,7 @@ echo "foo.bar.baz;tag1=value1;tag2=value2 123 `date +%s`" | nc -N localhost 2003
 ```

 VictoriaMetrics sets the current time if the timestamp is omitted.
-An arbitrary number of lines delimited by `\n` (aka newline char) may be sent in one go.
+An arbitrary number of lines delimited by `\n` (aka newline char) can be sent in one go.
 After that the data may be read via [/api/v1/export](#how-to-export-data-in-json-line-format) endpoint:

 ```bash
@@ -401,30 +420,31 @@ The `/api/v1/export` endpoint should return the following response:
 {"metric":{"__name__":"foo.bar.baz","tag1":"value1","tag2":"value2"},"values":[123],"timestamps":[1560277406000]}
 ```

-### Querying Graphite data
+## Querying Graphite data

 Data sent to VictoriaMetrics via `Graphite plaintext protocol` may be read via the following APIs:

 * [Prometheus querying API](#prometheus-querying-api-usage)
 * Metric names can be explored via [Graphite metrics API](#graphite-metrics-api-usage)
-* [go-graphite/carbonapi](https://github.com/go-graphite/carbonapi/blob/master/cmd/carbonapi/carbonapi.example.prometheus.yaml)
+* Tags can be explored via [Graphite tags API](#graphite-tags-api-usage)
+* [go-graphite/carbonapi](https://github.com/go-graphite/carbonapi/blob/main/cmd/carbonapi/carbonapi.example.victoriametrics.yaml)

-### How to send data from OpenTSDB-compatible agents
+## How to send data from OpenTSDB-compatible agents

 VictoriaMetrics supports [telnet put protocol](http://opentsdb.net/docs/build/html/api_telnet/put.html)
 and [HTTP /api/put requests](http://opentsdb.net/docs/build/html/api_http/put.html) for ingesting OpenTSDB data.
 The same protocol is used for [ingesting data in KairosDB](https://kairosdb.github.io/docs/build/html/PushingData.html).

-#### Sending data via `telnet put` protocol
+### Sending data via `telnet put` protocol

-1) Enable OpenTSDB receiver in VictoriaMetrics by setting `-opentsdbListenAddr` command line flag. For instance,
+Enable OpenTSDB receiver in VictoriaMetrics by setting `-opentsdbListenAddr` command line flag. For instance,
 the following command enables OpenTSDB receiver in VictoriaMetrics on TCP and UDP port `4242`:

 ```bash
 /path/to/victoria-metrics-prod -opentsdbListenAddr=:4242
 ```

-2) Send data to the given address from OpenTSDB-compatible agents.
+Send data to the given address from OpenTSDB-compatible agents.

 Example for writing data with OpenTSDB protocol to local VictoriaMetrics using `nc`:

@@ -432,7 +452,7 @@ Example for writing data with OpenTSDB protocol to local VictoriaMetrics using `
 echo "put foo.bar.baz `date +%s` 123 tag1=value1 tag2=value2" | nc -N localhost 4242
 ```

-An arbitrary number of lines delimited by `\n` (aka newline char) may be sent in one go.
+An arbitrary number of lines delimited by `\n` (aka newline char) can be sent in one go.
 After that the data may be read via [/api/v1/export](#how-to-export-data-in-json-line-format) endpoint:

 ```bash
@@ -445,16 +465,16 @@ The `/api/v1/export` endpoint should return the following response:
 {"metric":{"__name__":"foo.bar.baz","tag1":"value1","tag2":"value2"},"values":[123],"timestamps":[1560277292000]}
 ```

-#### Sending OpenTSDB data via HTTP `/api/put` requests
+### Sending OpenTSDB data via HTTP `/api/put` requests

-1) Enable HTTP server for OpenTSDB `/api/put` requests by setting `-opentsdbHTTPListenAddr` command line flag. For instance,
+Enable HTTP server for OpenTSDB `/api/put` requests by setting `-opentsdbHTTPListenAddr` command line flag. For instance,
 the following command enables OpenTSDB HTTP server on port `4242`:

 ```bash
 /path/to/victoria-metrics-prod -opentsdbHTTPListenAddr=:4242
 ```

-2) Send data to the given address from OpenTSDB-compatible agents.
+Send data to the given address from OpenTSDB-compatible agents.

 Example for writing a single data point:

@@ -483,7 +503,7 @@ The `/api/v1/export` endpoint should return the following response:
 ```


-### Prometheus querying API usage
+## Prometheus querying API usage

 VictoriaMetrics supports the following handlers from [Prometheus querying API](https://prometheus.io/docs/prometheus/latest/querying/api/):

@@ -492,11 +512,14 @@ VictoriaMetrics supports the following handlers from [Prometheus querying API](h
 * [/api/v1/series](https://prometheus.io/docs/prometheus/latest/querying/api/#finding-series-by-label-matchers)
 * [/api/v1/labels](https://prometheus.io/docs/prometheus/latest/querying/api/#getting-label-names)
 * [/api/v1/label/.../values](https://prometheus.io/docs/prometheus/latest/querying/api/#querying-label-values)
-* [/api/v1/status/tsdb](https://prometheus.io/docs/prometheus/latest/querying/api/#tsdb-stats)
+* [/api/v1/status/tsdb](https://prometheus.io/docs/prometheus/latest/querying/api/#tsdb-stats). VictoriaMetrics accepts optional `topN=N` and `date=YYYY-MM-DD`
+  query args for this handler, where `N` is the number of top entries to return in the response and `YYYY-MM-DD` is the date for collecting the stats.
+  By default top 10 entries are returned and the stats is collected for the current day.
+* [/api/v1/targets](https://prometheus.io/docs/prometheus/latest/querying/api/#targets) - see [these docs](#how-to-scrape-prometheus-exporters-such-as-node-exporter) for more details.

 These handlers can be queried from Prometheus-compatible clients such as Grafana or curl.

-#### Prometheus querying API enhancements
+### Prometheus querying API enhancements

 Additionally to unix timestamps and [RFC3339](https://www.ietf.org/rfc/rfc3339.txt) VictoriaMetrics accepts relative times in `time`, `start` and `end` query args.
 For example, the following query would return data for the last 30 minutes: `/api/v1/query_range?start=-30m&query=...`.
@@ -518,6 +541,14 @@ Additionally VictoriaMetrics provides the following handlers:
 * `/api/v1/status/active_queries` - it returns a list of currently running queries.


+## Graphite API usage
+
+VictoriaMetrics supports the following Graphite APIs:
+
+* Metrics API - see [these docs](#graphite-metrics-api-usage).
+* Tags API - see [these docs](#graphite-tags-api-usage).
+
+
 ### Graphite Metrics API usage

 VictoriaMetrics supports the following handlers from [Graphite Metrics API](https://graphite-api.readthedocs.io/en/latest/api.html#the-metrics-api):
@@ -532,42 +563,56 @@ VictoriaMetrics accepts the following additional query args at `/metrics/find` a
    that start with `node_`. By default `delimiter=.`.


-### How to build from sources
+### Graphite Tags API usage
+
+VictoriaMetrics supports the following handlers from [Graphite Tags API](https://graphite.readthedocs.io/en/stable/tags.html):
+
+* [/tags/tagSeries](https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb)
+* [/tags/tagMultiSeries](https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb)
+* [/tags](https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags)
+* [/tags/{tag_name}](https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags)
+* [/tags/findSeries](https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags)
+* [/tags/autoComplete/tags](https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support)
+* [/tags/autoComplete/values](https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support)
+* [/tags/delSeries](https://graphite.readthedocs.io/en/stable/tags.html#removing-series-from-the-tagdb)
+
+
+## How to build from sources

 We recommend using either [binary releases](https://github.com/VictoriaMetrics/VictoriaMetrics/releases) or
 [docker images](https://hub.docker.com/r/victoriametrics/victoria-metrics/) instead of building VictoriaMetrics
 from sources. Building from sources is reasonable when developing additional features specific
 to your needs or when testing bugfixes.

-#### Development build
+### Development build

 1. [Install Go](https://golang.org/doc/install). The minimum supported version is Go 1.13.
 2. Run `make victoria-metrics` from the root folder of the repository.
   It builds `victoria-metrics` binary and puts it into the `bin` folder.

-#### Production build
+### Production build

 1. [Install docker](https://docs.docker.com/install/).
 2. Run `make victoria-metrics-prod` from the root folder of the repository.
   It builds `victoria-metrics-prod` binary and puts it into the `bin` folder.

-#### ARM build
+### ARM build

 ARM build may run on Raspberry Pi or on [energy-efficient ARM servers](https://blog.cloudflare.com/arm-takes-wing/).

-#### Development ARM build
+### Development ARM build

 1. [Install Go](https://golang.org/doc/install). The minimum supported version is Go 1.13.
 2. Run `make victoria-metrics-arm` or `make victoria-metrics-arm64` from the root folder of the repository.
   It builds `victoria-metrics-arm` or `victoria-metrics-arm64` binary respectively and puts it into the `bin` folder.

-#### Production ARM build
+### Production ARM build

 1. [Install docker](https://docs.docker.com/install/).
 2. Run `make victoria-metrics-arm-prod` or `make victoria-metrics-arm64-prod` from the root folder of the repository.
   It builds `victoria-metrics-arm-prod` or `victoria-metrics-arm64-prod` binary respectively and puts it into the `bin` folder.

-#### Pure Go build (CGO_ENABLED=0)
+### Pure Go build (CGO_ENABLED=0)

 `Pure Go` mode builds only Go code without [cgo](https://golang.org/cmd/cgo/) dependencies.
 This is an experimental mode, which may result in a lower compression ratio and slower decompression performance.
@@ -577,7 +622,7 @@ Use it with caution!
 2. Run `make victoria-metrics-pure` from the root folder of the repository.
   It builds `victoria-metrics-pure` binary and puts it into the `bin` folder.

-#### Building docker images
+### Building docker images

 Run `make package-victoria-metrics`. It builds `victoriametrics/victoria-metrics:<PKG_TAG>` docker image locally.
 `<PKG_TAG>` is auto-generated image tag, which depends on source code in the repository.
@@ -591,17 +636,20 @@ For example, the following command builds the image on top of [scratch](https://
 ROOT_IMAGE=scratch make package-victoria-metrics
 ```

-### Start with docker-compose
+## Start with docker-compose

 [Docker-compose](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/deployment/docker/docker-compose.yml)
 helps to spin up VictoriaMetrics, [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md) and Grafana with one command.
 More details may be found [here](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/master/deployment/docker#folder-contains-basic-images-and-tools-for-building-and-running-victoria-metrics-in-docker).

-### Setting up service
+
+## Setting up service

 Read [these instructions](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/43) on how to set up VictoriaMetrics as a service in your OS.
+There is also [snap package for Ubuntu](https://snapcraft.io/victoriametrics).

-### How to work with snapshots
+
+## How to work with snapshots

 VictoriaMetrics can create [instant snapshots](https://medium.com/@valyala/how-victoriametrics-makes-instant-snapshots-for-multi-terabyte-time-series-data-e1f3fb0e0282)
 for all the data stored under `-storageDataPath` directory.
@@ -630,7 +678,7 @@ Steps for restoring from a snapshot:
   to the directory pointed by `-storageDataPath`.
 3. Start VictoriaMetrics.

-### How to delete time series
+## How to delete time series

 Send a request to `http://<victoriametrics-addr>:8428/api/v1/admin/tsdb/delete_series?match[]=<timeseries_selector_for_delete>`,
 where `<timeseries_selector_for_delete>` may contain any [time series selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors)
@@ -662,7 +710,7 @@ It isn't recommended using delete API for the following cases, since it brings n
 It is better using `-retentionPeriod` command-line flag for efficient pruning of old data.


-### Forced merge
+## Forced merge

 VictoriaMetrics performs [data compactions in background](https://medium.com/@valyala/how-victoriametrics-makes-instant-snapshots-for-multi-terabyte-time-series-data-e1f3fb0e0282)
 in order to keep good performance characteristics when accepting new data. These compactions (merges) are performed independently on per-month partitions.
@@ -677,7 +725,7 @@ since VictoriaMetrics automatically performs [optimal merges in background](http
 when new data is ingested into it.


-### How to export time series
+## How to export time series

 VictoriaMetrics provides the following handlers for exporting data:

@@ -687,11 +735,20 @@ VictoriaMetrics provides the following handlers for exporting data:
 * `/api/v1/export/csv` for exporting data in CSV. See [these docs](#how-to-export-csv-data) for details.


-#### How to export data in native format
+### How to export data in native format

 Send a request to `http://<victoriametrics-addr>:8428/api/v1/export/native?match[]=<timeseries_selector_for_export>`,
 where `<timeseries_selector_for_export>` may contain any [time series selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors)
-for metrics to export. Use `{__name__!=""}` selector for fetching all the time series.
+for metrics to export. Use `{__name__=~".*"}` selector for fetching all the time series.
+
+On large databases you may experience problems with limit on unique timeseries (default value is 300000). In this case you need to adjust `-search.maxUniqueTimeseries` parameter:
+
+```bash
+# count unique timeseries in database
+wget -O- -q 'http://your_victoriametrics_instance:8428/api/v1/series/count' | jq '.data[0]'
+
+# relaunch victoriametrics with search.maxUniqueTimeseries more than value from previous command
+```

 Optional `start` and `end` args may be added to the request in order to limit the time frame for the exported data. These args may contain either
 unix timestamp in seconds or [RFC3339](https://www.ietf.org/rfc/rfc3339.txt) values.
@@ -699,7 +756,7 @@ unix timestamp in seconds or [RFC3339](https://www.ietf.org/rfc/rfc3339.txt) val
 The exported data can be imported to VictoriaMetrics via [/api/v1/import/native](#how-to-import-data-in-native-format).


-#### How to export data in JSON line format
+### How to export data in JSON line format

 Consider [exporting data in native format](#how-to-export-data-in-native-format) if big amounts of data must be migrated between VictoriaMetrics instances,
 since exporting in native format usually consumes lower amounts of CPU and memory resources, while the resulting exported data occupies lower amounts of disk space.
@@ -734,7 +791,7 @@ The maximum duration for each request to `/api/v1/export` is limited by `-search
 Exported data can be imported via POST'ing it to [/api/v1/import](#how-to-import-data-in-json-line-format).


-#### How to export CSV data
+### How to export CSV data

 Send a request to `http://<victoriametrics-addr>:8428/api/v1/export/csv?format=<format>&match=<timeseries_selector_for_export>`,
 where:
@@ -758,24 +815,24 @@ unix timestamp in seconds or [RFC3339](https://www.ietf.org/rfc/rfc3339.txt) val
 The exported CSV data can be imported to VictoriaMetrics via [/api/v1/import/csv](#how-to-import-csv-data).


-### How to import time series data
+## How to import time series data

 Time series data can be imported via any supported ingestion protocol:

-* [Prometheus remote_write API](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#remote_write).
+* [Prometheus remote_write API](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#remote_write). See [these docs](#prometheus-setup) for details.
 * Influx line protocol. See [these docs](#how-to-send-data-from-influxdb-compatible-agents-such-as-telegraf) for details.
 * Graphite plaintext protocol. See [these docs](#how-to-send-data-from-graphite-compatible-agents-such-as-statsd) for details.
 * OpenTSDB telnet put protocol. See [these docs](#sending-data-via-telnet-put-protocol) for details.
 * OpenTSDB http `/api/put` protocol. See [these docs](#sending-opentsdb-data-via-http-apiput-requests) for details.
 * `/api/v1/import` for importing data obtained from [/api/v1/export](#how-to-export-data-in-json-line-format).
-  See [these docs](##how-to-import-data-in-json-line-format) for details.
+  See [these docs](#how-to-import-data-in-json-line-format) for details.
 * `/api/v1/import/native` for importing data obtained from [/api/v1/export/native](#how-to-export-data-in-native-format).
  See [these docs](#how-to-import-data-in-native-format) for details.
 * `/api/v1/import/csv` for importing arbitrary CSV data. See [these docs](#how-to-import-csv-data) for details.
 * `/api/v1/import/prometheus` for importing data in Prometheus exposition format. See [these docs](#how-to-import-data-in-prometheus-exposition-format) for details.


-#### How to import data in native format
+### How to import data in native format

 The most efficient protocol for importing data into VictoriaMetrics is `/api/v1/import/native`.
 Example for importing data obtained via [/api/v1/export/native](#how-to-export-data-in-native-format):
@@ -804,7 +861,7 @@ For example, `/api/v1/import/native?extra_label=foo=bar` would add `"foo":"bar"`
 Note that it could be required to flush response cache after importing historical data. See [these docs](#backfilling) for detail.


-#### How to import data in JSON line format
+### How to import data in JSON line format

 Example for importing data obtained via [/api/v1/export](#how-to-export-data-in-json-line-format):

@@ -832,7 +889,7 @@ For example, `/api/v1/import?extra_label=foo=bar` would add `"foo":"bar"` label
 Note that it could be required to flush response cache after importing historical data. See [these docs](#backfilling) for detail.


-#### How to import CSV data
+### How to import CSV data

 Arbitrary CSV data can be imported via `/api/v1/import/csv`. The CSV data is imported according to the provided `format` query arg.
 The `format` query arg must contain comma-separated list of parsing rules for CSV fields. Each rule consists of three parts delimited by a colon:
@@ -885,7 +942,7 @@ For example, `/api/v1/import/csv?extra_label=foo=bar` would add `"foo":"bar"` la
 Note that it could be required to flush response cache after importing historical data. See [these docs](#backfilling) for detail.


-#### How to import data in Prometheus exposition format
+### How to import data in Prometheus exposition format

 VictoriaMetrics accepts data in [Prometheus exposition format](https://github.com/prometheus/docs/blob/master/content/docs/instrumenting/exposition_formats.md#text-based-format)
 via `/api/v1/import/prometheus` path. For example, the following line imports a single line in Prometheus exposition format into VictoriaMetrics:
@@ -920,7 +977,7 @@ VictoriaMetrics also may scrape Prometheus targets - see [these docs](#how-to-sc



-### Relabeling
+## Relabeling

 VictoriaMetrics supports Prometheus-compatible relabeling for all the ingested metrics if `-relabelConfig` command-line flag points
 to a file containing a list of [relabel_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#relabel_config) entries.
@@ -944,7 +1001,7 @@ VictoriaMetrics provides the following extra actions for relabeling rules:
 See also [relabeling in vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md#relabeling).


-### Federation
+## Federation

 VictoriaMetrics exports [Prometheus-compatible federation data](https://prometheus.io/docs/prometheus/latest/federation/)
 at `http://<victoriametrics-addr>:8428/federate?match[]=<timeseries_selector_for_federation>`.
@@ -955,7 +1012,7 @@ on the interval `[now - max_lookback ... now]` is scraped for each time series.
 For instance, `/federate?match[]=up&max_lookback=1h` would return last points on the `[now - 1h ... now]` interval. This may be useful for time series federation
 with scrape intervals exceeding `5m`.

-### Capacity planning
+## Capacity planning

 A rough estimation of the required resources for ingestion path:

@@ -995,10 +1052,11 @@ The required resources for query path:
 * Network usage: depends on the frequency and the type of incoming requests. Typical Grafana dashboards usually
  require negligible network bandwidth.

-### High availability

-1) Install multiple VictoriaMetrics instances in distinct datacenters (availability zones).
-2) Pass addresses of these instances to [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md) via `-remoteWrite.url` command-line flag:
+## High availability
+
+* Install multiple VictoriaMetrics instances in distinct datacenters (availability zones).
+* Pass addresses of these instances to [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md) via `-remoteWrite.url` command-line flag:

 ```bash
 /path/to/vmagent -remoteWrite.url=http://<victoriametrics-addr-1>:8428/api/v1/write -remoteWrite.url=http://<victoriametrics-addr-2>:8428/api/v1/write
@@ -1017,7 +1075,7 @@ remote_write:
      max_samples_per_send: 10000
 ```

-3) Apply the updated config:
+* Apply the updated config:

 ```bash
 kill -HUP `pidof prometheus`
@@ -1025,9 +1083,9 @@ kill -HUP `pidof prometheus`

 It is recommended to use [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md) instead of Prometheus for highly loaded setups.

-4) Now Prometheus should write data into all the configured `remote_write` urls in parallel.
-5) Set up [Promxy](https://github.com/jacksontj/promxy) in front of all the VictoriaMetrics replicas.
-6) Set up Prometheus datasource in Grafana that points to Promxy.
+* Now Prometheus should write data into all the configured `remote_write` urls in parallel.
+* Set up [Promxy](https://github.com/jacksontj/promxy) in front of all the VictoriaMetrics replicas.
+* Set up Prometheus datasource in Grafana that points to Promxy.

 If you have Prometheus HA pairs with replicas `r1` and `r2` in each pair, then configure each `r1`
 to write data to `victoriametrics-addr-1`, while each `r2` should write data to `victoriametrics-addr-2`.
@@ -1036,7 +1094,7 @@ Another option is to write data simultaneously from Prometheus HA pair to a pair
 with the enabled de-duplication. See [this section](#deduplication) for details.


-### Deduplication
+## Deduplication

 VictoriaMetrics de-duplicates data points if `-dedup.minScrapeInterval` command-line flag
 is set to positive duration. For example, `-dedup.minScrapeInterval=60s` would de-duplicate data points
@@ -1046,7 +1104,8 @@ The de-duplication reduces disk space usage if multiple identically configured P
 write data to the same VictoriaMetrics instance. Note that these Prometheus instances must have identical
 `external_labels` section in their configs, so they write data to the same time series.

-### Retention
+
+## Retention

 Retention is configured with `-retentionPeriod` command-line flag. For instance, `-retentionPeriod=3` means
 that the data will be stored for 3 months and then deleted.
@@ -1057,7 +1116,11 @@ For example if `-retentionPeriod` is set to 1, data for January is deleted on Ma
 It is safe to extend `-retentionPeriod` on existing data. If `-retentionPeriod` is set to lower
 value than before then data outside the configured period will be eventually deleted.

-### Multiple retentions
+VictoriaMetrics supports retention smaller than 1 month. For example, `-retentionPeriod=5d` would set data retention for 5 days.
+Older data is eventually deleted during [background merge](https://medium.com/@valyala/how-victoriametrics-makes-instant-snapshots-for-multi-terabyte-time-series-data-e1f3fb0e0282).
+
+
+## Multiple retentions

 Just start multiple VictoriaMetrics instances with distinct values for the following flags:

@@ -1070,7 +1133,7 @@ so it could route requests from particular user to VictoriaMetrics with the desi
 The same scheme could be implemented for multiple tenants in [VictoriaMetrics cluster](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/cluster/README.md).


-### Downsampling
+## Downsampling

 There is no downsampling support at the moment, but:

@@ -1086,11 +1149,13 @@ It is possible to (ab)use [-dedup.minScrapeInterval](#deduplication) for basic d
 For instance, if interval between the ingested data points is 15s, then `-dedup.minScrapeInterval=5m` will leave
 only a single data point out of 20 initial data points per each 5m interval.

-### Multi-tenancy
+
+## Multi-tenancy

 Single-node VictoriaMetrics doesn't support multi-tenancy. Use [cluster version](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/cluster) instead.

-### Scalability and cluster version
+
+## Scalability and cluster version

 Though single-node VictoriaMetrics cannot scale to multiple nodes, it is optimized for resource usage - storage size / bandwidth / IOPS, RAM, CPU.
 This means that a single-node VictoriaMetrics may scale vertically and substitute a moderately sized cluster built with competing solutions
@@ -1100,7 +1165,8 @@ So try single-node VictoriaMetrics at first and then [switch to cluster version]
 horizontally scalable long-term remote storage for really large Prometheus deployments.
 [Contact us](mailto:info@victoriametrics.com) for paid support.

-### Alerting
+
+## Alerting

 It is recommended using [vmalert](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmalert/README.md) for alerting.

@@ -1111,7 +1177,7 @@ Additionally, alerting can be set up with the following tools:
 * With Grafana - see [the corresponding docs](https://grafana.com/docs/alerting/rules/).


-### Security
+## Security

 Do not forget protecting sensitive endpoints in VictoriaMetrics when exposing it to untrusted networks such as the internet.
 Consider setting the following command-line flags:
@@ -1131,7 +1197,7 @@ Prefer authorizing all the incoming requests from untrusted networks with [vmaut
 or similar auth proxy.


-### Tuning
+## Tuning

 * There is no need for VictoriaMetrics tuning since it uses reasonable defaults for command-line flags,
  which are automatically adjusted for the available CPU and RAM resources.
@@ -1147,7 +1213,7 @@ or similar auth proxy.
 mkfs.ext4 ... -O 64bit,huge_file,extent -T huge
 ```

-### Monitoring
+## Monitoring

 VictoriaMetrics exports internal metrics in Prometheus format at `/metrics` page.
 These metrics may be collected by [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md)
@@ -1177,7 +1243,7 @@ The most interesting metrics are:
 VictoriaMetrics also exposes currently running queries with their execution times at `/api/v1/status/active_queries` page.


-### Troubleshooting
+## Troubleshooting

 * It is recommended to use default command-line flag values (i.e. don't set them explicitly) until the need
  of tweaking these flag values arises.
@@ -1187,13 +1253,18 @@ VictoriaMetrics also exposes currently running queries with their execution time

 * It is recommended inspecting logs during troubleshooting, since they may contain useful information.

+* VictoriaMetrics buffers incoming data in memory for up to a few seconds before flushing it to persistent storage.
+  This may lead to the following "issues":
+  * Data becomes available for querying in a few seconds after inserting. It is possible to flush in-memory buffers to persistent storage
+    by requesting `/internal/force_flush` http handler. This handler is mostly needed for testing and debugging purposes.
+  * The last few seconds of inserted data may be lost on unclean shutdown (i.e. OOM, `kill -9` or hardware reset).
+    See [this article for technical details](https://valyala.medium.com/wal-usage-looks-broken-in-modern-time-series-databases-b62a627ab704).
+
 * If VictoriaMetrics works slowly and eats more than a CPU core per 100K ingested data points per second,
  then it is likely you have too many active time series for the current amount of RAM.
  VictoriaMetrics [exposes](#monitoring) `vm_slow_*` metrics, which could be used as an indicator of low amounts of RAM.
  It is recommended increasing the amount of RAM on the node with VictoriaMetrics in order to improve
  ingestion and query performance in this case.
-  Another option is to increase `-memory.allowedPercent` command-line flag value. Be careful with this
-  option, since too big value for `-memory.allowedPercent` may result in high I/O usage.

 * VictoriaMetrics prioritizes data ingestion over data querying. So if it has no enough resources for data ingestion,
  then data querying may slow down significantly.
@@ -1208,9 +1279,9 @@ VictoriaMetrics also exposes currently running queries with their execution time
  which would start background merge if they had more free disk space.

 * If VictoriaMetrics doesn't work because of certain parts are corrupted due to disk errors,
-  then just remove directories with broken parts. This will recover VictoriaMetrics at the cost
-  of data loss stored in the broken parts. In the future, `vmrecover` tool will be created
-  for automatic recovering from such errors.
+  then just remove directories with broken parts. It is safe removing subdirectories under `<-storageDataPath>/data/{big,small}/YYYY_MM` directories
+  when VictoriaMetrics isn't running. This recovers VictoriaMetrics at the cost of data loss stored in the deleted broken parts.
+  In the future, `vmrecover` tool will be created for automatic recovering from such errors.

 * If you see gaps on the graphs, try resetting the cache by sending request to `/internal/resetRollupResultCache`.
  If this removes gaps on the graphs, then it is likely data with timestamps older than `-search.cacheTimestampOffset`
@@ -1232,10 +1303,15 @@ VictoriaMetrics also exposes currently running queries with their execution time
  This prevents from ingesting metrics with too many labels. It is recommended [monitoring](#monitoring) `vm_metrics_with_dropped_labels_total`
  metric in order to determine whether `-maxLabelsPerTimeseries` must be adjusted for your workload.

+* If you store Graphite metrics like `foo.bar.baz` in VictoriaMetrics, then `-search.treatDotsAsIsInRegexps` command-line flag could be useful.
+  By default `.` chars in regexps match any char. If you need matching only dots, then the `\\.` must be used in regexp filters.
+  When `-search.treatDotsAsIsInRegexps` option is enabled, then dots in regexps are automatically escaped in order to match only dots instead of arbitrary chars.
+  This may significantly increase performance when locating time series for the given label filters.
+
 * VictoriaMetrics ignores `NaN` values during data ingestion.


-### Backfilling
+## Backfilling

 VictoriaMetrics accepts historical data in arbitrary order of time via [any supported ingestion method](#how-to-import-time-series-data).
 Make sure that configured `-retentionPeriod` covers timestamps for the backfilled data.
@@ -1251,7 +1327,7 @@ Yet another solution is to increase `-search.cacheTimestampOffset` flag value in
 for data with timestamps close to the current time.


-### Data updates
+## Data updates

 VictoriaMetrics doesn't support updating already existing sample values to new ones. It stores all the ingested data points
 for the same time series with identical timestamps. While is possible substituting old time series with new time series via
@@ -1259,7 +1335,7 @@ for the same time series with identical timestamps. While is possible substituti
 should be used only for one-off updates. It shouldn't be used for frequent updates because of non-zero overhead related to data removal.


-### Replication
+## Replication

 Single-node VictoriaMetrics doesn't support application-level replication. Use cluster version instead.
 See [these docs](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/cluster/README.md#replication-and-data-safety) for details.
@@ -1269,14 +1345,14 @@ Storage-level replication may be offloaded to durable persistent storage such as
 See also [high availability docs](#high-availability) and [backup docs](#backups).


-### Backups
+## Backups

 VictoriaMetrics supports backups via [vmbackup](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmbackup/README.md)
 and [vmrestore](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmrestore/README.md) tools.
-We also provide provide `vmbackuper` tool for paid enterprise subscribers - see [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/466) for details.
+We also provide `vmbackuper` tool for paid enterprise subscribers - see [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/466) for details.


-### Profiling
+## Profiling

 VictoriaMetrics provides handlers for collecting the following [Go profiles](https://blog.golang.org/profiling-go-programs):

@@ -1296,6 +1372,7 @@ The command for collecting CPU profile waits for 30 seconds before returning.

 The collected profiles may be analyzed with [go tool pprof](https://github.com/google/pprof).

+
 ## Integrations

 * [Helm charts for single-node and cluster versions of VictoriaMetrics](https://github.com/VictoriaMetrics/helm-charts).
@@ -1307,6 +1384,8 @@ The collected profiles may be analyzed with [go tool pprof](https://github.com/g
  See [this example](https://github.com/go-graphite/carbonapi/blob/master/cmd/carbonapi/carbonapi.example.prometheus.yaml).
 * [Ansible role for installing single-node VictoriaMetrics](https://github.com/dreamteam-gg/ansible-victoriametrics-role).
 * [Ansible role for installing cluster VictoriaMetrics](https://github.com/Slapper/ansible-victoriametrics-cluster-role).
+* [Snap package for VictoriaMetrics](https://snapcraft.io/victoriametrics).
+

 ## Third-party contributions

@@ -1315,10 +1394,12 @@ The collected profiles may be analyzed with [go tool pprof](https://github.com/g
 * [Prometheus -> VictoriaMetrics exporter #2](https://github.com/AnchorFree/tsdb-remote-write)
 * [Prometheus Oauth proxy](https://gitlab.com/optima_public/prometheus_oauth_proxy) - see [this article](https://medium.com/@richard.holly/powerful-saas-solution-for-detection-metrics-c67b9208d362) for details.

+
 ## Contacts

 Contact us with any questions regarding VictoriaMetrics at [info@victoriametrics.com](mailto:info@victoriametrics.com).

+
 ## Community and contributions

 Feel free asking any questions regarding VictoriaMetrics:
--- a/app/victoria-metrics/Makefile
+++ b/app/victoria-metrics/Makefile
@@ -108,3 +108,10 @@ victoria-metrics-package-deb-rpm-all: \
        victoria-metrics-package-deb-arm64 \
        victoria-metrics-package-rpm \
        victoria-metrics-package-rpm-arm64
+
+### Packaging as snap
+victoria-metrics-package-snap:
+	which snapcraft || snap install snapcraft
+	which multipass || snap install multipass
+	snapcraft
+
--- a/app/victoria-metrics/main.go
+++ b/app/victoria-metrics/main.go
@@ -17,6 +17,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/procutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
 )

@@ -25,6 +26,8 @@ var (
 	minScrapeInterval = flag.Duration("dedup.minScrapeInterval", 0, "Remove superflouos samples from time series if they are located closer to each other than this duration. "+
 		"This may be useful for reducing overhead when multiple identically configured Prometheus instances write data to the same VictoriaMetrics. "+
 		"Deduplication is disabled if the -dedup.minScrapeInterval is 0")
+	dryRun = flag.Bool("dryRun", false, "Whether to check only -promscrape.config and then exit. "+
+		"Unknown config entries are allowed in -promscrape.config by default. This can be changed with -promscrape.config.strictParse")
 )

 func main() {
@@ -34,6 +37,18 @@ func main() {
 	buildinfo.Init()
 	logger.Init()
 	cgroup.UpdateGOMAXPROCSToCPUQuota()
+
+	if promscrape.IsDryRun() {
+		*dryRun = true
+	}
+	if *dryRun {
+		if err := promscrape.CheckConfig(); err != nil {
+			logger.Fatalf("error when checking -promscrape.config: %s", err)
+		}
+		logger.Infof("-promscrape.config is ok; exitting with 0 status code")
+		return
+	}
+
 	logger.Infof("starting VictoriaMetrics at %q...", *httpListenAddr)
 	startTime := time.Now()
 	storage.SetMinScrapeIntervalForDeduplication(*minScrapeInterval)
--- a/app/vmagent/README.md
+++ b/app/vmagent/README.md
@@ -63,6 +63,22 @@ Then send Influx data to `http://vmagent-host:8429`. See [these docs](https://gi
 Pass `-help` to `vmagent` in order to see the full list of supported command-line flags with their descriptions.


+### Configuration update
+
+`vmagent` should be restarted in order to update config options set via command-line args.
+
+`vmagent` supports multiple approaches for reloading configs from updated config files such as `-promscrape.config`, `-remoteWrite.relabelConfig` and `-remoteWrite.urlRelabelConfig`:
+
+* Sending `SUGHUP` signal to `vmagent` process:
+  ```bash
+  kill -SIGHUP `pidof vmagent`
+  ```
+
+* Sending HTTP request to `http://vmagent:8429/-/reload` endpoint.
+
+There is also `-promscrape.configCheckInterval` command-line option, which can be used for automatic reloading configs from updated `-promscrape.config` file.
+
+
 ### Use cases


@@ -153,6 +169,8 @@ The following scrape types in [scrape_config](https://prometheus.io/docs/prometh
  [OpenStack identity API v3](https://docs.openstack.org/api-ref/identity/v3/) is supported only.
 * `dockerswarm_sd_configs` - for scraping Docker Swarm targets.
  See [dockerswarm_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#dockerswarm_sd_config) for details.
+* `eureka_sd_configs` - for scraping targets registered in [Netflix Eureka](https://github.com/Netflix/eureka).
+  See [eureka_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#eureka_sd_config) for details.

 File feature requests at [our issue tracker](https://github.com/VictoriaMetrics/VictoriaMetrics/issues) if you need other service discovery mechanisms to be supported by `vmagent`.

@@ -197,6 +215,7 @@ The relabeling can be defined in the following places:

 Read more about relabeling in the following articles:

+* [How to use Relabeling in Prometheus and VictoriaMetrics](https://valyala.medium.com/how-to-use-relabeling-in-prometheus-and-victoriametrics-8b90fc22c4b2)
 * [Life of a label](https://www.robustperception.io/life-of-a-label)
 * [Discarding targets and timeseries with relabeling](https://www.robustperception.io/relabelling-can-discard-targets-timeseries-and-alerts)
 * [Dropping labels at scrape time](https://www.robustperception.io/dropping-metrics-at-scrape-time-with-prometheus)
@@ -211,9 +230,16 @@ either via `vmagent` itself or via Prometheus, so the exported metrics could be
 Use official [Grafana dashboard](https://grafana.com/grafana/dashboards/12683) for `vmagent` state overview.
 If you have suggestions, improvements or found a bug - feel free to open an issue on github or add review to the dashboard.

-`vmagent` also exports target statuses at `http://vmagent-host:8429/targets` page in plaintext format.
-`/targets` handler accepts optional `show_original_labels=1` query arg, which shows the original labels per each target
-before applying relabeling. This information may be useful for debugging target relabeling.
+`vmagent` also exports target statuses at the following handlers:
+
+* `http://vmagent-host:8429/targets`. This handler returns human-readable plaintext status for every active target.
+This page is convenient to query from command line with `wget`, `curl` or similar tools.
+It accepts optional `show_original_labels=1` query arg, which shows the original labels per each target before applying relabeling.
+This information may be useful for debugging target relabeling.
+* `http://vmagent-host:8429/api/v1/targets`. This handler returns data compatible with [the corresponding page from Prometheus API](https://prometheus.io/docs/prometheus/latest/querying/api/#targets).
+
+* `http://vmagent-host:8429/ready`. This handler returns http 200 status code when `vmagent` finishes initialization for all service_discovery configs.
+It may be useful for performing `vmagent` rolling update without scrape loss.


 ### Troubleshooting
@@ -224,7 +250,32 @@ before applying relabeling. This information may be useful for debugging target
  since `vmagent` establishes at least a single TCP connection per each target.

 * When `vmagent` scrapes many unreliable targets, it can flood error log with scrape errors. These errors can be suppressed
-  by passing `-promscrape.suppressScrapeErrors` command-line flag to `vmagent`. The most recent scrape error per each target can be observed at `http://vmagent-host:8429/targets`.
+  by passing `-promscrape.suppressScrapeErrors` command-line flag to `vmagent`. The most recent scrape error per each target can be observed at `http://vmagent-host:8429/targets`
+  and `http://vmagent-host:8429/api/v1/targets`.
+
+* The `/api/v1/targets` page could be useful for debugging relabeling process for scrape targets.
+  This page contains original labels for targets dropped during relabeling (see "droppedTargets" section in the page output). By default up to `-promscrape.maxDroppedTargets` targets are shown here. If your setup drops more targets during relabeling, then increase `-promscrape.maxDroppedTargets` command-line flag value in order to see all the dropped targets. Note that tracking each dropped target requires up to 10Kb of RAM, so big values for `-promscrape.maxDroppedTargets` may result in increased memory usage if big number of scrape targets are dropped during relabeling.
+
+* If `vmagent` scrapes big number of targets, then `-promscrape.dropOriginalLabels` command-line option may be passed to `vmagent` in order to reduce memory usage.
+  This option drops `"discoveredLabels"` and `"droppedTargets"` lists at `/api/v1/targets` page, which may result in reduced debuggability for improperly configured per-target relabeling.
+
+* If `vmagent` scrapes targets with millions of metrics per each target (for instance, when scraping [federation endpoints](https://prometheus.io/docs/prometheus/latest/federation/)),
+  then it is recommended enabling `stream parsing mode` in order to reduce memory usage during scraping. This mode may be enabled either globally for all the scrape targets
+  by passing `-promscrape.streamParse` command-line flag or on a per-scrape target basis with `stream_parse: true` option. For example:
+
+  ```yml
+  scrape_configs:
+  - job_name: 'big-federate'
+    stream_parse: true
+    static_configs:
+    - targets:
+      - big-prometeus1
+      - big-prometeus2
+    honor_labels: true
+    metrics_path: /federate
+    params:
+      'match[]': ['{__name__!=""}']
+  ```

 * It is recommended to increase `-remoteWrite.queues` if `vmagent_remotewrite_pending_data_bytes` metric exported at `http://vmagent-host:8429/metrics` page constantly grows.

--- a/app/vmagent/common/push_ctx.go
+++ b/app/vmagent/common/push_ctx.go
@@ -5,6 +5,7 @@ import (
 	"sync"

 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
 )

 // PushCtx is a context used for populating WriteRequest.
@@ -28,12 +29,7 @@ func (ctx *PushCtx) Reset() {
 	}
 	ctx.WriteRequest.Timeseries = ctx.WriteRequest.Timeseries[:0]

-	labels := ctx.Labels
-	for i := range labels {
-		label := &labels[i]
-		label.Name = ""
-		label.Value = ""
-	}
+	promrelabel.CleanLabels(ctx.Labels)
 	ctx.Labels = ctx.Labels[:0]

 	ctx.Samples = ctx.Samples[:0]
--- a/app/vmagent/influx/request_handler.go
+++ b/app/vmagent/influx/request_handler.go
@@ -11,6 +11,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/remotewrite"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
 	parser "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/influx"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/writeconcurrencylimiter"
 	"github.com/VictoriaMetrics/metrics"
@@ -135,12 +136,8 @@ type pushCtx struct {
 func (ctx *pushCtx) reset() {
 	ctx.ctx.Reset()

-	commonLabels := ctx.commonLabels
-	for i := range commonLabels {
-		label := &commonLabels[i]
-		label.Name = ""
-		label.Value = ""
-	}
+	promrelabel.CleanLabels(ctx.commonLabels)
+	ctx.commonLabels = ctx.commonLabels[:0]

 	ctx.metricGroupBuf = ctx.metricGroupBuf[:0]
 	ctx.buf = ctx.buf[:0]
--- a/app/vmagent/main.go
+++ b/app/vmagent/main.go
@@ -7,6 +7,7 @@ import (
 	"os"
 	"strconv"
 	"strings"
+	"sync/atomic"
 	"time"

 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/csvimport"
@@ -22,6 +23,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/buildinfo"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/envflag"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	graphiteserver "github.com/VictoriaMetrics/VictoriaMetrics/lib/ingestserver/graphite"
 	influxserver "github.com/VictoriaMetrics/VictoriaMetrics/lib/ingestserver/influx"
@@ -47,7 +49,8 @@ var (
 		"Usually :4242 must be set. Doesn't work if empty")
 	opentsdbHTTPListenAddr = flag.String("opentsdbHTTPListenAddr", "", "TCP address to listen for OpentTSDB HTTP put requests. Usually :4242 must be set. Doesn't work if empty")
 	dryRun                 = flag.Bool("dryRun", false, "Whether to check only config files without running vmagent. The following files are checked: "+
-		"-promscrape.config, -remoteWrite.relabelConfig, -remoteWrite.urlRelabelConfig . See also -promscrape.config.dryRun")
+		"-promscrape.config, -remoteWrite.relabelConfig, -remoteWrite.urlRelabelConfig . "+
+		"Unknown config entries are allowed in -promscrape.config by default. This can be changed with -promscrape.config.strictParse")
 )

 var (
@@ -67,15 +70,19 @@ func main() {
 	logger.Init()
 	cgroup.UpdateGOMAXPROCSToCPUQuota()

-	if *dryRun {
-		if err := flag.Set("promscrape.config.strictParse", "true"); err != nil {
-			logger.Panicf("BUG: cannot set promscrape.config.strictParse=true: %s", err)
+	if promscrape.IsDryRun() {
+		if err := promscrape.CheckConfig(); err != nil {
+			logger.Fatalf("error when checking -promscrape.config: %s", err)
 		}
+		logger.Infof("-promscrape.config is ok; exitting with 0 status code")
+		return
+	}
+	if *dryRun {
 		if err := remotewrite.CheckRelabelConfigs(); err != nil {
 			logger.Fatalf("error when checking relabel configs: %s", err)
 		}
 		if err := promscrape.CheckConfig(); err != nil {
-			logger.Fatalf("error when checking Prometheus config: %s", err)
+			logger.Fatalf("error when checking -promscrape.config: %s", err)
 		}
 		logger.Infof("all the configs are ok; exitting with 0 status code")
 		return
@@ -207,15 +214,31 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 		return true
 	case "/targets":
 		promscrapeTargetsRequests.Inc()
-		w.Header().Set("Content-Type", "text/plain")
+		w.Header().Set("Content-Type", "text/plain; charset=utf-8")
 		showOriginalLabels, _ := strconv.ParseBool(r.FormValue("show_original_labels"))
 		promscrape.WriteHumanReadableTargetsStatus(w, showOriginalLabels)
 		return true
+	case "/api/v1/targets":
+		promscrapeAPIV1TargetsRequests.Inc()
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
+		state := r.FormValue("state")
+		promscrape.WriteAPIV1Targets(w, state)
+		return true
 	case "/-/reload":
 		promscrapeConfigReloadRequests.Inc()
 		procutil.SelfSIGHUP()
 		w.WriteHeader(http.StatusOK)
 		return true
+	case "/ready":
+		if rdy := atomic.LoadInt32(&promscrape.PendingScrapeConfigs); rdy > 0 {
+			errMsg := fmt.Sprintf("waiting for scrapes to init, left: %d", rdy)
+			http.Error(w, errMsg, http.StatusTooEarly)
+		} else {
+			w.Header().Set("Content-Type", "text/plain; charset=utf-8")
+			w.WriteHeader(http.StatusOK)
+			w.Write([]byte("OK"))
+		}
+		return true
 	}
 	return false
 }
@@ -241,7 +264,8 @@ var (

 	influxQueryRequests = metrics.NewCounter(`vmagent_http_requests_total{path="/query", protocol="influx"}`)

-	promscrapeTargetsRequests = metrics.NewCounter(`vmagent_http_requests_total{path="/targets"}`)
+	promscrapeTargetsRequests      = metrics.NewCounter(`vmagent_http_requests_total{path="/targets"}`)
+	promscrapeAPIV1TargetsRequests = metrics.NewCounter(`vmagent_http_requests_total{path="/api/v1/targets"}`)

 	promscrapeConfigReloadRequests = metrics.NewCounter(`vmagent_http_requests_total{path="/-/reload"}`)
 )
@@ -252,8 +276,5 @@ vmagent collects metrics data via popular data ingestion protocols and routes it

 See the docs at https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md .
 `
-
-	f := flag.CommandLine.Output()
-	fmt.Fprintf(f, "%s\n", s)
-	flag.PrintDefaults()
+	flagutil.Usage(s)
 }
--- a/app/vmagent/remotewrite/client.go
+++ b/app/vmagent/remotewrite/client.go
@@ -53,6 +53,7 @@ type client struct {
 	requestDuration *metrics.Histogram
 	requestsOKCount *metrics.Counter
 	errorsCount     *metrics.Counter
+	packetsDropped  *metrics.Counter
 	retriesCount    *metrics.Counter

 	wg     sync.WaitGroup
@@ -114,6 +115,7 @@ func newClient(argIdx int, remoteWriteURL, sanitizedURL string, fq *persistentqu
 	c.requestDuration = metrics.GetOrCreateHistogram(fmt.Sprintf(`vmagent_remotewrite_duration_seconds{url=%q}`, c.sanitizedURL))
 	c.requestsOKCount = metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_requests_total{url=%q, status_code="2XX"}`, c.sanitizedURL))
 	c.errorsCount = metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_errors_total{url=%q}`, c.sanitizedURL))
+	c.packetsDropped = metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_packets_dropped_total{url=%q}`, c.sanitizedURL))
 	c.retriesCount = metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_retries_count_total{url=%q}`, c.sanitizedURL))
 	for i := 0; i < concurrency; i++ {
 		c.wg.Add(1)
@@ -228,10 +230,20 @@ again:
 		c.requestsOKCount.Inc()
 		return
 	}
+	metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_requests_total{url=%q, status_code="%d"}`, c.sanitizedURL, statusCode)).Inc()
+	if statusCode == 409 {
+		// Just drop block on 409 status code like Prometheus does.
+		// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/873
+		body, _ := ioutil.ReadAll(resp.Body)
+		_ = resp.Body.Close()
+		logger.Errorf("unexpected status code received when sending a block with size %d bytes to %q: #%d; dropping the block like Prometheus does; "+
+			"response body=%q", len(block), c.sanitizedURL, statusCode, body)
+		c.packetsDropped.Inc()
+		return
+	}

 	// Unexpected status code returned
 	retriesCount++
-	metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_requests_total{url=%q, status_code="%d"}`, c.sanitizedURL, statusCode)).Inc()
 	retryDuration *= 2
 	if retryDuration > time.Minute {
 		retryDuration = time.Minute
--- a/app/vmagent/remotewrite/pendingseries.go
+++ b/app/vmagent/remotewrite/pendingseries.go
@@ -11,6 +11,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/persistentqueue"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
 	"github.com/VictoriaMetrics/metrics"
 	"github.com/golang/snappy"
 )
@@ -104,11 +105,7 @@ func (wr *writeRequest) reset() {
 	}
 	wr.tss = wr.tss[:0]

-	for i := range wr.labels {
-		label := &wr.labels[i]
-		label.Name = ""
-		label.Value = ""
-	}
+	promrelabel.CleanLabels(wr.labels)
 	wr.labels = wr.labels[:0]

 	wr.samples = wr.samples[:0]
--- a/app/vmagent/remotewrite/relabel.go
+++ b/app/vmagent/remotewrite/relabel.go
@@ -117,12 +117,7 @@ type relabelCtx struct {
 }

 func (rctx *relabelCtx) reset() {
-	labels := rctx.labels
-	for i := range labels {
-		label := &labels[i]
-		label.Name = ""
-		label.Value = ""
-	}
+	promrelabel.CleanLabels(rctx.labels)
 	rctx.labels = rctx.labels[:0]
 }

--- a/app/vmalert/README.md
+++ b/app/vmalert/README.md
@@ -11,6 +11,7 @@ rules against configured address.
 * Prometheus [alerting rules definition format](https://prometheus.io/docs/prometheus/latest/configuration/alerting_rules/#defining-alerting-rules)
 support;
 * Integration with [Alertmanager](https://github.com/prometheus/alertmanager);
+* Keeps the alerts [state on restarts](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/master/app/vmalert#alerts-state-on-restarts);
 * Lightweight without extra dependencies.

 ### Limitations:
@@ -121,14 +122,6 @@ annotations:
  [ <labelname>: <tmpl_string> ]
 ``` 

-`vmalert` has no local storage and alerts state is stored in process memory. Hence, after reloading of `vmalert` process
-alerts state will be lost. To avoid this situation, `vmalert` may be configured via following flags:
-* `-remoteWrite.url` - URL to Victoria Metrics or VMInsert. `vmalert` will persist alerts state into the configured
-address in form of timeseries with name `ALERTS` via remote-write protocol.
-* `-remoteRead.url` - URL to Victoria Metrics or VMSelect. `vmalert` will try to restore alerts state from configured
-address by querying `ALERTS` timeseries.
-
-
 ##### Recording rules

 The syntax for recording rules is following:
@@ -147,6 +140,22 @@ labels:
 For recording rules to work `-remoteWrite.url` must specified.


+#### Alerts state on restarts
+
+`vmalert` has no local storage, so alerts state is stored in the process memory. Hence, after reloading of `vmalert` 
+the process alerts state will be lost. To avoid this situation, `vmalert` should be configured via the following flags:
+* `-remoteWrite.url` - URL to VictoriaMetrics (Single) or VMInsert (Cluster). `vmalert` will persist alerts state 
+into the configured address in the form of time series named `ALERTS` and `ALERTS_FOR_STATE` via remote-write protocol. 
+These are regular time series and may be queried from VM just as any other time series. 
+The state stored to the configured address on every rule evaluation.
+* `-remoteRead.url` - URL to VictoriaMetrics (Single) or VMSelect (Cluster). `vmalert` will try to restore alerts state 
+from configured address by querying time series with name `ALERTS_FOR_STATE`.
+
+Both flags are required for the proper state restoring. Restore process may fail if time series are missing
+in configured `-remoteRead.url`, weren't updated in the last `1h` or received state doesn't match current `vmalert` 
+rules configuration.
+
+
 #### WEB

 `vmalert` runs a web-server (`-httpListenAddr`) for serving metrics and alerts endpoints:
--- a/app/vmalert/alerting.go
+++ b/app/vmalert/alerting.go
@@ -140,7 +140,17 @@ func (ar *AlertingRule) Exec(ctx context.Context, q datasource.Querier, series b
 	updated := make(map[uint64]struct{})
 	// update list of active alerts
 	for _, m := range qMetrics {
+		for k, v := range ar.Labels {
+			// apply extra labels
+			m.SetLabel(k, v)
+		}
+
 		h := hash(m)
+		if _, ok := updated[h]; ok {
+			// duplicate may be caused by extra labels
+			// conflicting with the metric labels
+			return nil, fmt.Errorf("labels %v: %w", m.Labels, errDuplicate)
+		}
 		updated[h] = struct{}{}
 		if a, ok := ar.alerts[h]; ok {
 			if a.Value != m.Value {
@@ -258,25 +268,11 @@ func (ar *AlertingRule) newAlert(m datasource.Metric, start time.Time) (*notifie
 }

 func (ar *AlertingRule) template(a *notifier.Alert) error {
-	// 1. template rule labels with data labels
-	rLabels, err := a.ExecTemplate(ar.Labels)
-	if err != nil {
-		return err
-	}
-
-	// 2. merge data labels and rule labels
-	// metric labels may be overridden by
-	// rule labels
-	for k, v := range rLabels {
-		a.Labels[k] = v
-	}
-
-	// 3. template merged labels
+	var err error
 	a.Labels, err = a.ExecTemplate(a.Labels)
 	if err != nil {
 		return err
 	}
-
 	a.Annotations, err = a.ExecTemplate(ar.Annotations)
 	return err
 }
@@ -403,7 +399,7 @@ func (ar *AlertingRule) Restore(ctx context.Context, q datasource.Querier, lookb
 		labelsFilter += fmt.Sprintf(",%s=%q", k, v)
 	}

-	// Get the last datapoint in range via MetricsQL `last_over_time`.
+	// Get the last data point in range via MetricsQL `last_over_time`.
 	// We don't use plain PromQL since Prometheus doesn't support
 	// remote write protocol which is used for state persistence in vmalert.
 	expr := fmt.Sprintf("last_over_time(%s{alertname=%q%s}[%ds])",
@@ -417,13 +413,9 @@ func (ar *AlertingRule) Restore(ctx context.Context, q datasource.Querier, lookb
 		labels := m.Labels
 		m.Labels = make([]datasource.Label, 0)
 		// drop all extra labels, so hash key will
-		// be identical to timeseries received in Exec
+		// be identical to time series received in Exec
 		for _, l := range labels {
-			if l.Name == alertNameLabel {
-				continue
-			}
-			// drop all overridden labels
-			if _, ok := ar.Labels[l.Name]; ok {
+			if l.Name == alertNameLabel || l.Name == alertGroupNameLabel {
 				continue
 			}
 			m.Labels = append(m.Labels, l)
@@ -436,7 +428,7 @@ func (ar *AlertingRule) Restore(ctx context.Context, q datasource.Querier, lookb
 		a.ID = hash(m)
 		a.State = notifier.StatePending
 		ar.alerts[a.ID] = a
-		logger.Infof("alert %q(%d) restored to state at %v", a.Name, a.ID, a.Start)
+		logger.Infof("alert %q (%d) restored to state at %v", a.Name, a.ID, a.Start)
 	}
 	return nil
 }
--- a/app/vmalert/alerting_test.go
+++ b/app/vmalert/alerting_test.go
@@ -2,6 +2,8 @@ package main

 import (
 	"context"
+	"errors"
+	"strings"
 	"testing"
 	"time"

@@ -218,19 +220,6 @@ func TestAlertingRule_Exec(t *testing.T) {
 				hash(metricWithLabels(t, "name", "foo2")): {State: notifier.StateFiring},
 			},
 		},
-		{
-			newTestAlertingRule("duplicate", 0),
-			[][]datasource.Metric{
-				{
-					// metrics with the same labelset should result in one alert
-					metricWithLabels(t, "name", "foo", "type", "bar"),
-					metricWithLabels(t, "type", "bar", "name", "foo"),
-				},
-			},
-			map[uint64]*notifier.Alert{
-				hash(metricWithLabels(t, "name", "foo", "type", "bar")): {State: notifier.StateFiring},
-			},
-		},
 		{
 			newTestAlertingRule("for-pending", time.Minute),
 			[][]datasource.Metric{
@@ -355,6 +344,7 @@ func TestAlertingRule_Restore(t *testing.T) {
 				metricWithValueAndLabels(t, float64(time.Now().Truncate(time.Hour).Unix()),
 					"__name__", alertForStateMetricName,
 					alertNameLabel, "",
+					alertGroupNameLabel, "groupID",
 					"foo", "bar",
 					"namespace", "baz",
 				),
@@ -375,7 +365,7 @@ func TestAlertingRule_Restore(t *testing.T) {
 					alertNameLabel, "",
 					"foo", "bar",
 					"namespace", "baz",
-					// following pair supposed to be dropped
+					// extra labels set by rule
 					"source", "vm",
 				),
 			},
@@ -383,6 +373,7 @@ func TestAlertingRule_Restore(t *testing.T) {
 				hash(metricWithLabels(t,
 					"foo", "bar",
 					"namespace", "baz",
+					"source", "vm",
 				)): {State: notifier.StatePending,
 					Start: time.Now().Truncate(time.Hour)},
 			},
@@ -441,6 +432,38 @@ func TestAlertingRule_Restore(t *testing.T) {
 	}
 }

+func TestAlertingRule_Exec_Negative(t *testing.T) {
+	fq := &fakeQuerier{}
+	ar := newTestAlertingRule("test", 0)
+	ar.Labels = map[string]string{"job": "test"}
+
+	// successful attempt
+	fq.add(metricWithValueAndLabels(t, 1, "__name__", "foo", "job", "bar"))
+	_, err := ar.Exec(context.TODO(), fq, false)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// label `job` will collide with rule extra label and will make both time series equal
+	fq.add(metricWithValueAndLabels(t, 1, "__name__", "foo", "job", "baz"))
+	_, err = ar.Exec(context.TODO(), fq, false)
+	if !errors.Is(err, errDuplicate) {
+		t.Fatalf("expected to have %s error; got %s", errDuplicate, err)
+	}
+
+	fq.reset()
+
+	expErr := "connection reset by peer"
+	fq.setErr(errors.New(expErr))
+	_, err = ar.Exec(context.TODO(), fq, false)
+	if err == nil {
+		t.Fatalf("expected to get err; got nil")
+	}
+	if !strings.Contains(err.Error(), expErr) {
+		t.Fatalf("expected to get err %q; got %q insterad", expErr, err)
+	}
+}
+
 func newTestRuleWithLabels(name string, labels ...string) *AlertingRule {
 	r := newTestAlertingRule(name, 0)
 	r.Labels = make(map[string]string)
--- a/app/vmalert/config/config.go
+++ b/app/vmalert/config/config.go
@@ -11,6 +11,7 @@ import (
 	"time"

 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/notifier"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/utils"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/envtemplate"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/metricsql"
@@ -94,7 +95,7 @@ type Rule struct {
 	Record      string            `yaml:"record,omitempty"`
 	Alert       string            `yaml:"alert,omitempty"`
 	Expr        string            `yaml:"expr"`
-	For         PromDuration      `yaml:"for,omitempty"`
+	For         PromDuration      `yaml:"for"`
 	Labels      map[string]string `yaml:"labels,omitempty"`
 	Annotations map[string]string `yaml:"annotations,omitempty"`

@@ -114,6 +115,11 @@ func NewPromDuration(d time.Duration) PromDuration {
 	}
 }

+// MarshalYAML implements yaml.Marshaler interface.
+func (pd PromDuration) MarshalYAML() (interface{}, error) {
+	return pd.Duration().String(), nil
+}
+
 // UnmarshalYAML implements yaml.Unmarshaler interface.
 func (pd *PromDuration) UnmarshalYAML(unmarshal func(interface{}) error) error {
 	var s string
@@ -193,25 +199,32 @@ func Parse(pathPatterns []string, validateAnnotations, validateExpressions bool)
 		}
 		fp = append(fp, matches...)
 	}
+	errGroup := new(utils.ErrGroup)
 	var groups []Group
 	for _, file := range fp {
 		uniqueGroups := map[string]struct{}{}
 		gr, err := parseFile(file)
 		if err != nil {
-			return nil, fmt.Errorf("failed to parse file %q: %w", file, err)
+			errGroup.Add(fmt.Errorf("failed to parse file %q: %w", file, err))
+			continue
 		}
 		for _, g := range gr {
 			if err := g.Validate(validateAnnotations, validateExpressions); err != nil {
-				return nil, fmt.Errorf("invalid group %q in file %q: %w", g.Name, file, err)
+				errGroup.Add(fmt.Errorf("invalid group %q in file %q: %w", g.Name, file, err))
+				continue
 			}
 			if _, ok := uniqueGroups[g.Name]; ok {
-				return nil, fmt.Errorf("group name %q duplicate in file %q", g.Name, file)
+				errGroup.Add(fmt.Errorf("group name %q duplicate in file %q", g.Name, file))
+				continue
 			}
 			uniqueGroups[g.Name] = struct{}{}
 			g.File = file
 			groups = append(groups, g)
 		}
 	}
+	if err := errGroup.Err(); err != nil {
+		return nil, err
+	}
 	if len(groups) < 1 {
 		logger.Warnf("no groups found in %s", strings.Join(pathPatterns, ";"))
 	}
--- a/app/vmalert/config/config_test.go
+++ b/app/vmalert/config/config_test.go
@@ -323,34 +323,55 @@ func TestHashRule(t *testing.T) {
 }

 func TestGroupChecksum(t *testing.T) {
-	data := `
+	f := func(t *testing.T, data, newData string) {
+		t.Helper()
+		var g Group
+		if err := yaml.Unmarshal([]byte(data), &g); err != nil {
+			t.Fatalf("failed to unmarshal: %s", err)
+		}
+		if g.Checksum == "" {
+			t.Fatalf("expected to get non-empty checksum")
+		}
+
+		var ng Group
+		if err := yaml.Unmarshal([]byte(newData), &ng); err != nil {
+			t.Fatalf("failed to unmarshal: %s", err)
+		}
+		if g.Checksum == ng.Checksum {
+			t.Fatalf("expected to get different checksums")
+		}
+	}
+	t.Run("Ok", func(t *testing.T) {
+		f(t, `
 name: TestGroup
 rules:
  - alert: ExampleAlertAlwaysFiring
    expr: sum by(job) (up == 1)
  - record: handler:requests:rate5m
    expr: sum(rate(prometheus_http_requests_total[5m])) by (handler)
-`
-	var g Group
-	if err := yaml.Unmarshal([]byte(data), &g); err != nil {
-		t.Fatalf("failed to unmarshal: %s", err)
-	}
-	if g.Checksum == "" {
-		t.Fatalf("expected to get non-empty checksum")
-	}
-	newData := `
+`, `
 name: TestGroup
 rules:
  - record: handler:requests:rate5m
    expr: sum(rate(prometheus_http_requests_total[5m])) by (handler)
  - alert: ExampleAlertAlwaysFiring
    expr: sum by(job) (up == 1)
-`
-	var ng Group
-	if err := yaml.Unmarshal([]byte(newData), &g); err != nil {
-		t.Fatalf("failed to unmarshal: %s", err)
-	}
-	if g.Checksum == ng.Checksum {
-		t.Fatalf("expected to get different checksums")
-	}
+`)
+	})
+
+	t.Run("Ok, `for` must change cs", func(t *testing.T) {
+		f(t, `
+name: TestGroup
+rules:
+  - alert: ExampleAlertWithFor
+    expr: sum by(job) (up == 1)
+    for: 5m
+`, `
+name: TestGroup
+rules:
+  - alert: ExampleAlertWithFor
+    expr: sum by(job) (up == 1)
+`)
+	})
+
 }
--- a/app/vmalert/datasource/datasource.go
+++ b/app/vmalert/datasource/datasource.go
@@ -17,6 +17,23 @@ type Metric struct {
 	Value     float64
 }

+// SetLabel adds or updates existing one label
+// by the given key and label
+func (m *Metric) SetLabel(key, value string) {
+	for i, l := range m.Labels {
+		if l.Name == key {
+			m.Labels[i].Value = value
+			return
+		}
+	}
+	m.AddLabel(key, value)
+}
+
+// AddLabel appends the given label to the label set
+func (m *Metric) AddLabel(key, value string) {
+	m.Labels = append(m.Labels, Label{Name: key, Value: value})
+}
+
 // Label represents metric's label
 type Label struct {
 	Name  string
--- a/app/vmalert/datasource/vm.go
+++ b/app/vmalert/datasource/vm.go
@@ -37,7 +37,7 @@ func (r response) metrics() ([]Metric, error) {
 		}
 		m.Labels = nil
 		for k, v := range r.Data.Result[i].Labels {
-			m.Labels = append(m.Labels, Label{Name: k, Value: v})
+			m.AddLabel(k, v)
 		}
 		m.Timestamp = int64(res.TV[0].(float64))
 		m.Value = f
@@ -82,7 +82,7 @@ func (s *VMStorage) Query(ctx context.Context, query string) ([]Metric, error) {
 	if err != nil {
 		return nil, err
 	}
-	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("Content-Type", "application/json; charset=utf-8")
 	if s.basicAuthPass != "" {
 		req.SetBasicAuth(s.basicAuthUser, s.basicAuthPass)
 	}
--- a/app/vmalert/group_test.go
+++ b/app/vmalert/group_test.go
@@ -172,6 +172,11 @@ func TestGroupStart(t *testing.T) {
 		t.Fatalf("faield to create alert: %s", err)
 	}
 	alert1.State = notifier.StateFiring
+	// add external label
+	alert1.Labels["cluster"] = "east-1"
+	// add rule labels - see config/testdata/rules1-good.rules
+	alert1.Labels["label"] = "bar"
+	alert1.Labels["host"] = inst1
 	alert1.ID = hash(m1)

 	alert2, err := r.newAlert(m2, time.Now())
@@ -179,6 +184,11 @@ func TestGroupStart(t *testing.T) {
 		t.Fatalf("faield to create alert: %s", err)
 	}
 	alert2.State = notifier.StateFiring
+	// add external label
+	alert2.Labels["cluster"] = "east-1"
+	// add rule labels - see config/testdata/rules1-good.rules
+	alert2.Labels["label"] = "bar"
+	alert2.Labels["host"] = inst2
 	alert2.ID = hash(m2)

 	finished := make(chan struct{})
--- a/app/vmalert/main.go
+++ b/app/vmalert/main.go
@@ -10,6 +10,7 @@ import (
 	"strings"
 	"time"

+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/config"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/datasource"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/notifier"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/remoteread"
@@ -47,6 +48,8 @@ eg. 'explore?orgId=1&left=[\"now-1h\",\"now\",\"VictoriaMetrics\",{\"expr\": \"{

 	remoteReadLookBack = flag.Duration("remoteRead.lookback", time.Hour, "Lookback defines how far to look into past for alerts timeseries."+
 		" For example, if lookback=1h then range from now() to now()-1h will be scanned.")
+
+	dryRun = flag.Bool("dryRun", false, "Whether to check only config files without running vmalert. The rules file are validated. The `-rule` flag must be specified.")
 )

 func main() {
@@ -58,6 +61,18 @@ func main() {
 	logger.Init()
 	cgroup.UpdateGOMAXPROCSToCPUQuota()

+	if *dryRun {
+		u, _ := url.Parse("https://victoriametrics.com/")
+		notifier.InitTemplateFunc(u)
+		groups, err := config.Parse(*rulePath, true, true)
+		if err != nil {
+			logger.Fatalf(err.Error())
+		}
+		if len(groups) == 0 {
+			logger.Fatalf("No rules for validation. Please specify path to file(s) with alerting and/or recording rules using `-rule` flag")
+		}
+		return
+	}
 	ctx, cancel := context.WithCancel(context.Background())
 	manager, err := newManager(ctx)
 	if err != nil {
@@ -204,8 +219,5 @@ vmalert processes alerts and recording rules.

 See the docs at https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmalert/README.md .
 `
-
-	f := flag.CommandLine.Output()
-	fmt.Fprintf(f, "%s\n", s)
-	flag.PrintDefaults()
+	flagutil.Usage(s)
 }
--- a/app/vmalert/notifier/alertmanager.go
+++ b/app/vmalert/notifier/alertmanager.go
@@ -28,7 +28,7 @@ func (am *AlertManager) Send(ctx context.Context, alerts []Alert) error {
 	if err != nil {
 		return err
 	}
-	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("Content-Type", "application/json; charset=utf-8")
 	req = req.WithContext(ctx)
 	if am.basicAuthPass != "" {
 		req.SetBasicAuth(am.basicAuthUser, am.basicAuthPass)
--- a/app/vmalert/recording.go
+++ b/app/vmalert/recording.go
@@ -2,7 +2,6 @@ package main

 import (
 	"context"
-	"errors"
 	"fmt"
 	"hash/fnv"
 	"sort"
@@ -79,8 +78,6 @@ func (rr *RecordingRule) Close() {
 	metrics.UnregisterMetric(rr.metrics.errors.name)
 }

-var errDuplicate = errors.New("result contains metrics with the same labelset after applying rule labels")
-
 // Exec executes RecordingRule expression via the given Querier.
 func (rr *RecordingRule) Exec(ctx context.Context, q datasource.Querier, series bool) ([]prompbmarshal.TimeSeries, error) {
 	if !series {
--- a/app/vmalert/rule.go
+++ b/app/vmalert/rule.go
@@ -2,6 +2,7 @@ package main

 import (
 	"context"
+	"errors"

 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/datasource"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
@@ -25,3 +26,5 @@ type Rule interface {
 	// such as metrics unregister
 	Close()
 }
+
+var errDuplicate = errors.New("result contains metrics with the same labelset after applying rule labels")
--- a/app/vmalert/web.go
+++ b/app/vmalert/web.go
@@ -40,7 +40,7 @@ func (rh *requestHandler) handler(w http.ResponseWriter, r *http.Request) bool {
 			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
 			return true
 		}
-		w.Header().Set("Content-Type", "application/json")
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
 		w.Write(data)
 		return true
 	case "/api/v1/alerts":
@@ -49,7 +49,7 @@ func (rh *requestHandler) handler(w http.ResponseWriter, r *http.Request) bool {
 			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
 			return true
 		}
-		w.Header().Set("Content-Type", "application/json")
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
 		w.Write(data)
 		return true
 	case "/-/reload":
@@ -67,7 +67,7 @@ func (rh *requestHandler) handler(w http.ResponseWriter, r *http.Request) bool {
 			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
 			return true
 		}
-		w.Header().Set("Content-Type", "application/json")
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
 		w.Write(data)
 		return true
 	}
--- a/app/vmauth/main.go
+++ b/app/vmauth/main.go
@@ -2,7 +2,6 @@ package main

 import (
 	"flag"
-	"fmt"
 	"net/http"
 	"net/http/httputil"
 	"net/url"
@@ -12,6 +11,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/buildinfo"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/envflag"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/procutil"
@@ -100,8 +100,5 @@ vmauth authenticates and authorizes incoming requests and proxies them to Victor

 See the docs at https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmauth/README.md .
 `
-
-	f := flag.CommandLine.Output()
-	fmt.Fprintf(f, "%s\n", s)
-	flag.PrintDefaults()
+	flagutil.Usage(s)
 }
--- a/app/vmbackup/main.go
+++ b/app/vmbackup/main.go
@@ -10,6 +10,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/backup/actions"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/backup/common"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/backup/fslocal"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/backup/fsnil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/buildinfo"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/envflag"
@@ -42,7 +43,7 @@ func main() {
 	cgroup.UpdateGOMAXPROCSToCPUQuota()

 	if len(*snapshotCreateURL) > 0 {
-		logger.Infof("%s", "Snapshots enabled")
+		logger.Infof("Snapshots enabled")
 		logger.Infof("Snapshot create url %s", *snapshotCreateURL)
 		if len(*snapshotDeleteURL) <= 0 {
 			err := flag.Set("snapshot.deleteURL", strings.Replace(*snapshotCreateURL, "/create", "/delete", 1))
@@ -54,17 +55,17 @@ func main() {

 		name, err := snapshot.Create(*snapshotCreateURL)
 		if err != nil {
-			logger.Fatalf("%s", err)
+			logger.Fatalf("cannot create snapshot: %s", err)
 		}
 		err = flag.Set("snapshotName", name)
 		if err != nil {
-			logger.Fatalf("Failed to set snapshotName flag: %v", err)
+			logger.Fatalf("cannot set snapshotName flag: %v", err)
 		}

 		defer func() {
 			err := snapshot.Delete(*snapshotDeleteURL, name)
 			if err != nil {
-				logger.Fatalf("%s", err)
+				logger.Fatalf("cannot delete snapshot: %s", err)
 			}
 		}()
 	}
@@ -102,10 +103,7 @@ or local filesystem. Backed up data can be restored with vmrestore.

 See the docs at https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmbackup/README.md .
 `
-
-	f := flag.CommandLine.Output()
-	fmt.Fprintf(f, "%s\n", s)
-	flag.PrintDefaults()
+	flagutil.Usage(s)
 }

 func newSrcFS() (*fslocal.FS, error) {
@@ -146,9 +144,9 @@ func newDstFS() (common.RemoteFS, error) {
 	return fs, nil
 }

-func newOriginFS() (common.RemoteFS, error) {
+func newOriginFS() (common.OriginFS, error) {
 	if len(*origin) == 0 {
-		return nil, nil
+		return &fsnil.FS{}, nil
 	}
 	fs, err := actions.NewRemoteFS(*origin)
 	if err != nil {
--- a/app/vmbackup/snapshot/snapshot.go
+++ b/app/vmbackup/snapshot/snapshot.go
@@ -20,26 +20,27 @@ type snapshot struct {
 // Create creates a snapshot and the provided api endpoint and returns
 // the snapshot name
 func Create(createSnapshotURL string) (string, error) {
-	logger.Infof("%s", "Creating snapshot")
+	logger.Infof("Creating snapshot")
 	u, err := url.Parse(createSnapshotURL)
 	if err != nil {
 		return "", err
 	}
-
 	resp, err := http.Get(u.String())
 	if err != nil {
 		return "", err
 	}
-
 	body, err := ioutil.ReadAll(resp.Body)
 	if err != nil {
 		return "", err
 	}
+	if resp.StatusCode != http.StatusOK {
+		return "", fmt.Errorf("unexpected status code returned from %q; expecting %d; got %d; response body: %q", createSnapshotURL, resp.StatusCode, http.StatusOK, body)
+	}

 	snap := snapshot{}
 	err = json.Unmarshal(body, &snap)
 	if err != nil {
-		return "", err
+		return "", fmt.Errorf("cannot parse JSON response from %q: %w; response body: %q", createSnapshotURL, err, body)
 	}

 	if snap.Status == "ok" {
@@ -58,26 +59,26 @@ func Delete(deleteSnapshotURL string, snapshotName string) error {
 	formData := url.Values{
 		"snapshot": {snapshotName},
 	}
-
 	u, err := url.Parse(deleteSnapshotURL)
 	if err != nil {
 		return err
 	}
-
 	resp, err := http.PostForm(u.String(), formData)
 	if err != nil {
 		return err
 	}
-
 	body, err := ioutil.ReadAll(resp.Body)
 	if err != nil {
 		return err
 	}
+	if resp.StatusCode != http.StatusOK {
+		return fmt.Errorf("unexpected status code returned from %q; expecting %d; got %d; response body: %q", deleteSnapshotURL, resp.StatusCode, http.StatusOK, body)
+	}

 	snap := snapshot{}
 	err = json.Unmarshal(body, &snap)
 	if err != nil {
-		return err
+		return fmt.Errorf("cannot parse JSON response from %q: %w; response body: %q", deleteSnapshotURL, err, body)
 	}

 	if snap.Status == "ok" {
--- a/app/vminsert/main.go
+++ b/app/vminsert/main.go
@@ -40,7 +40,7 @@ var (
 		"Telnet put messages and HTTP /api/put messages are simultaneously served on TCP port. "+
 		"Usually :4242 must be set. Doesn't work if empty")
 	opentsdbHTTPListenAddr = flag.String("opentsdbHTTPListenAddr", "", "TCP address to listen for OpentTSDB HTTP put requests. Usually :4242 must be set. Doesn't work if empty")
-	maxLabelsPerTimeseries = flag.Int("maxLabelsPerTimeseries", 30, "The maximum number of labels accepted per time series. Superflouos labels are dropped")
+	maxLabelsPerTimeseries = flag.Int("maxLabelsPerTimeseries", 30, "The maximum number of labels accepted per time series. Superfluous labels are dropped")
 )

 var (
@@ -155,15 +155,31 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		return true
 	case "/targets":
 		promscrapeTargetsRequests.Inc()
-		w.Header().Set("Content-Type", "text/plain")
+		w.Header().Set("Content-Type", "text/plain; charset=utf-8")
 		showOriginalLabels, _ := strconv.ParseBool(r.FormValue("show_original_labels"))
 		promscrape.WriteHumanReadableTargetsStatus(w, showOriginalLabels)
 		return true
+	case "/api/v1/targets":
+		promscrapeAPIV1TargetsRequests.Inc()
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
+		state := r.FormValue("state")
+		promscrape.WriteAPIV1Targets(w, state)
+		return true
 	case "/-/reload":
 		promscrapeConfigReloadRequests.Inc()
 		procutil.SelfSIGHUP()
 		w.WriteHeader(http.StatusNoContent)
 		return true
+	case "/ready":
+		if rdy := atomic.LoadInt32(&promscrape.PendingScrapeConfigs); rdy > 0 {
+			errMsg := fmt.Sprintf("waiting for scrape config to init targets, configs left: %d", rdy)
+			http.Error(w, errMsg, http.StatusTooEarly)
+		} else {
+			w.Header().Set("Content-Type", "text/plain; charset=utf-8")
+			w.WriteHeader(http.StatusOK)
+			w.Write([]byte("OK"))
+		}
+		return true
 	default:
 		// This is not our link
 		return false
@@ -191,7 +207,8 @@ var (

 	influxQueryRequests = metrics.NewCounter(`vm_http_requests_total{path="/query", protocol="influx"}`)

-	promscrapeTargetsRequests = metrics.NewCounter(`vm_http_requests_total{path="/targets"}`)
+	promscrapeTargetsRequests      = metrics.NewCounter(`vm_http_requests_total{path="/targets"}`)
+	promscrapeAPIV1TargetsRequests = metrics.NewCounter(`vm_http_requests_total{path="/api/v1/targets"}`)

 	promscrapeConfigReloadRequests = metrics.NewCounter(`vm_http_requests_total{path="/-/reload"}`)

--- a/app/vminsert/relabel/relabel.go
+++ b/app/vminsert/relabel/relabel.go
@@ -69,12 +69,7 @@ type Ctx struct {

 // Reset resets ctx.
 func (ctx *Ctx) Reset() {
-	labels := ctx.tmpLabels
-	for i := range labels {
-		label := &labels[i]
-		label.Name = ""
-		label.Value = ""
-	}
+	promrelabel.CleanLabels(ctx.tmpLabels)
 	ctx.tmpLabels = ctx.tmpLabels[:0]
 }

--- a/app/vmrestore/README.md
+++ b/app/vmrestore/README.md
@@ -3,7 +3,7 @@
 `vmrestore` restores data from backups created by [vmbackup](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmbackup/README.md).
 VictoriaMetrics `v1.29.0` and newer versions must be used for working with the restored data.

-Restore process can be interrupted at any time. It is automatically resumed from the inerruption point
+Restore process can be interrupted at any time. It is automatically resumed from the interruption point
 when restarting `vmrestore` with the same args.


--- a/app/vmrestore/main.go
+++ b/app/vmrestore/main.go
@@ -62,10 +62,7 @@ vmrestore restores VictoriaMetrics data from backups made by vmbackup.

 See the docs at https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmrestore/README.md .
 `
-
-	f := flag.CommandLine.Output()
-	fmt.Fprintf(f, "%s\n", s)
-	flag.PrintDefaults()
+	flagutil.Usage(s)
 }

 func newDstFS() (*fslocal.FS, error) {
--- a/app/vmselect/graphite/metrics_api.go
+++ b/app/vmselect/graphite/metrics_api.go
@@ -84,10 +84,7 @@ func MetricsFindHandler(startTime time.Time, w http.ResponseWriter, r *http.Requ
 	}
 	paths = deduplicatePaths(paths, delimiter)
 	sortPaths(paths, delimiter)
-	contentType := "application/json"
-	if jsonp != "" {
-		contentType = "text/javascript"
-	}
+	contentType := getContentType(jsonp)
 	w.Header().Set("Content-Type", contentType)
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
@@ -166,10 +163,7 @@ func MetricsExpandHandler(startTime time.Time, w http.ResponseWriter, r *http.Re
 		}
 		m[query] = paths
 	}
-	contentType := "application/json"
-	if jsonp != "" {
-		contentType = "text/javascript"
-	}
+	contentType := getContentType(jsonp)
 	w.Header().Set("Content-Type", contentType)
 	if groupByExpr {
 		for _, paths := range m {
@@ -215,10 +209,7 @@ func MetricsIndexHandler(startTime time.Time, w http.ResponseWriter, r *http.Req
 	if err != nil {
 		return fmt.Errorf(`cannot obtain metric names: %w`, err)
 	}
-	contentType := "application/json"
-	if jsonp != "" {
-		contentType = "text/javascript"
-	}
+	contentType := getContentType(jsonp)
 	w.Header().Set("Content-Type", contentType)
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
@@ -417,3 +408,10 @@ var regexpCache = make(map[regexpCacheKey]*regexpCacheEntry)
 var regexpCacheLock sync.Mutex

 const maxRegexpCacheSize = 10000
+
+func getContentType(jsonp string) string {
+	if jsonp == "" {
+		return "application/json; charset=utf-8"
+	}
+	return "text/javascript; charset=utf-8"
+}
--- a/app/vmselect/graphite/metrics_api_test.go
+++ b/app/vmselect/graphite/metrics_api_test.go
--- a/app/vmselect/graphite/tag_values_response.qtpl
+++ b/app/vmselect/graphite/tag_values_response.qtpl
@@ -0,0 +1,20 @@
+{% stripspace %}
+
+Tags generates response for /tags/<tag_name> handler
+See https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags
+{% func TagValuesResponse(tagName string, tagValues []string) %}
+{
+	"tag":{%q= tagName %},
+	"values":[
+	{% for i, value := range tagValues %}
+		{
+			"count":1,
+			"value":{%q= value %}
+		}
+		{% if i+1 < len(tagValues) %},{% endif %}
+	{% endfor %}
+	]
+}
+{% endfunc %}
+
+{% endstripspace %}
--- a/app/vmselect/graphite/tag_values_response.qtpl.go
+++ b/app/vmselect/graphite/tag_values_response.qtpl.go
@@ -0,0 +1,75 @@
+// Code generated by qtc from "tag_values_response.qtpl". DO NOT EDIT.
+// See https://github.com/valyala/quicktemplate for details.
+
+// Tags generates response for /tags/<tag_name> handlerSee https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags
+
+//line app/vmselect/graphite/tag_values_response.qtpl:5
+package graphite
+
+//line app/vmselect/graphite/tag_values_response.qtpl:5
+import (
+	qtio422016 "io"
+
+	qt422016 "github.com/valyala/quicktemplate"
+)
+
+//line app/vmselect/graphite/tag_values_response.qtpl:5
+var (
+	_ = qtio422016.Copy
+	_ = qt422016.AcquireByteBuffer
+)
+
+//line app/vmselect/graphite/tag_values_response.qtpl:5
+func StreamTagValuesResponse(qw422016 *qt422016.Writer, tagName string, tagValues []string) {
+//line app/vmselect/graphite/tag_values_response.qtpl:5
+	qw422016.N().S(`{"tag":`)
+//line app/vmselect/graphite/tag_values_response.qtpl:7
+	qw422016.N().Q(tagName)
+//line app/vmselect/graphite/tag_values_response.qtpl:7
+	qw422016.N().S(`,"values":[`)
+//line app/vmselect/graphite/tag_values_response.qtpl:9
+	for i, value := range tagValues {
+//line app/vmselect/graphite/tag_values_response.qtpl:9
+		qw422016.N().S(`{"count":1,"value":`)
+//line app/vmselect/graphite/tag_values_response.qtpl:12
+		qw422016.N().Q(value)
+//line app/vmselect/graphite/tag_values_response.qtpl:12
+		qw422016.N().S(`}`)
+//line app/vmselect/graphite/tag_values_response.qtpl:14
+		if i+1 < len(tagValues) {
+//line app/vmselect/graphite/tag_values_response.qtpl:14
+			qw422016.N().S(`,`)
+//line app/vmselect/graphite/tag_values_response.qtpl:14
+		}
+//line app/vmselect/graphite/tag_values_response.qtpl:15
+	}
+//line app/vmselect/graphite/tag_values_response.qtpl:15
+	qw422016.N().S(`]}`)
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+}
+
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+func WriteTagValuesResponse(qq422016 qtio422016.Writer, tagName string, tagValues []string) {
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+	qw422016 := qt422016.AcquireWriter(qq422016)
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+	StreamTagValuesResponse(qw422016, tagName, tagValues)
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+	qt422016.ReleaseWriter(qw422016)
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+}
+
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+func TagValuesResponse(tagName string, tagValues []string) string {
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+	qb422016 := qt422016.AcquireByteBuffer()
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+	WriteTagValuesResponse(qb422016, tagName, tagValues)
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+	qs422016 := string(qb422016.B)
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+	qt422016.ReleaseByteBuffer(qb422016)
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+	return qs422016
+//line app/vmselect/graphite/tag_values_response.qtpl:18
+}
--- a/app/vmselect/graphite/tags_api.go
+++ b/app/vmselect/graphite/tags_api.go
@@ -0,0 +1,504 @@
+package graphite
+
+import (
+	"fmt"
+	"net/http"
+	"regexp"
+	"sort"
+	"strconv"
+	"strings"
+	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/bufferedwriter"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/netstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/searchutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
+	graphiteparser "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/graphite"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/metrics"
+)
+
+// TagsDelSeriesHandler implements /tags/delSeries handler.
+//
+// See https://graphite.readthedocs.io/en/stable/tags.html#removing-series-from-the-tagdb
+func TagsDelSeriesHandler(startTime time.Time, w http.ResponseWriter, r *http.Request) error {
+	if err := r.ParseForm(); err != nil {
+		return fmt.Errorf("cannot parse form values: %w", err)
+	}
+	paths := r.Form["path"]
+	totalDeleted := 0
+	var row graphiteparser.Row
+	var tagsPool []graphiteparser.Tag
+	ct := time.Now().UnixNano() / 1e6
+	for _, path := range paths {
+		var err error
+		tagsPool, err = row.UnmarshalMetricAndTags(path, tagsPool[:0])
+		if err != nil {
+			return fmt.Errorf("cannot parse path=%q: %w", path, err)
+		}
+		tfs := make([]storage.TagFilter, 0, 1+len(row.Tags))
+		tfs = append(tfs, storage.TagFilter{
+			Key:   nil,
+			Value: []byte(row.Metric),
+		})
+		for _, tag := range row.Tags {
+			tfs = append(tfs, storage.TagFilter{
+				Key:   []byte(tag.Key),
+				Value: []byte(tag.Value),
+			})
+		}
+		sq := storage.NewSearchQuery(0, ct, [][]storage.TagFilter{tfs})
+		n, err := netstorage.DeleteSeries(sq)
+		if err != nil {
+			return fmt.Errorf("cannot delete series for %q: %w", sq, err)
+		}
+		totalDeleted += n
+	}
+
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
+	if totalDeleted > 0 {
+		fmt.Fprintf(w, "true")
+	} else {
+		fmt.Fprintf(w, "false")
+	}
+	return nil
+}
+
+// TagsTagSeriesHandler implements /tags/tagSeries handler.
+//
+// See https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb
+func TagsTagSeriesHandler(startTime time.Time, w http.ResponseWriter, r *http.Request) error {
+	return registerMetrics(startTime, w, r, false)
+}
+
+// TagsTagMultiSeriesHandler implements /tags/tagMultiSeries handler.
+//
+// See https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb
+func TagsTagMultiSeriesHandler(startTime time.Time, w http.ResponseWriter, r *http.Request) error {
+	return registerMetrics(startTime, w, r, true)
+}
+
+func registerMetrics(startTime time.Time, w http.ResponseWriter, r *http.Request, isJSONResponse bool) error {
+	if err := r.ParseForm(); err != nil {
+		return fmt.Errorf("cannot parse form values: %w", err)
+	}
+	paths := r.Form["path"]
+	var row graphiteparser.Row
+	var labels []prompb.Label
+	var b []byte
+	var tagsPool []graphiteparser.Tag
+	mrs := make([]storage.MetricRow, len(paths))
+	ct := time.Now().UnixNano() / 1e6
+	canonicalPaths := make([]string, len(paths))
+	for i, path := range paths {
+		var err error
+		tagsPool, err = row.UnmarshalMetricAndTags(path, tagsPool[:0])
+		if err != nil {
+			return fmt.Errorf("cannot parse path=%q: %w", path, err)
+		}
+
+		// Construct canonical path according to https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb
+		sort.Slice(row.Tags, func(i, j int) bool {
+			return row.Tags[i].Key < row.Tags[j].Key
+		})
+		b = append(b[:0], row.Metric...)
+		for _, tag := range row.Tags {
+			b = append(b, ';')
+			b = append(b, tag.Key...)
+			b = append(b, '=')
+			b = append(b, tag.Value...)
+		}
+		canonicalPaths[i] = string(b)
+
+		// Convert parsed metric and tags to labels.
+		labels = append(labels[:0], prompb.Label{
+			Name:  []byte("__name__"),
+			Value: []byte(row.Metric),
+		})
+		for _, tag := range row.Tags {
+			labels = append(labels, prompb.Label{
+				Name:  []byte(tag.Key),
+				Value: []byte(tag.Value),
+			})
+		}
+
+		// Put labels with the current timestamp to MetricRow
+		mr := &mrs[i]
+		mr.MetricNameRaw = storage.MarshalMetricNameRaw(mr.MetricNameRaw[:0], labels)
+		mr.Timestamp = ct
+	}
+	if err := vmstorage.RegisterMetricNames(mrs); err != nil {
+		return fmt.Errorf("cannot register paths: %w", err)
+	}
+
+	// Return response
+	contentType := "text/plain; charset=utf-8"
+	if isJSONResponse {
+		contentType = "application/json; charset=utf-8"
+	}
+	w.Header().Set("Content-Type", contentType)
+	WriteTagsTagMultiSeriesResponse(w, canonicalPaths, isJSONResponse)
+	if isJSONResponse {
+		tagsTagMultiSeriesDuration.UpdateDuration(startTime)
+	} else {
+		tagsTagSeriesDuration.UpdateDuration(startTime)
+	}
+	return nil
+}
+
+var (
+	tagsTagSeriesDuration      = metrics.NewSummary(`vm_request_duration_seconds{path="/tags/tagSeries"}`)
+	tagsTagMultiSeriesDuration = metrics.NewSummary(`vm_request_duration_seconds{path="/tags/tagMultiSeries"}`)
+)
+
+// TagsAutoCompleteValuesHandler implements /tags/autoComplete/values endpoint from Graphite Tags API.
+//
+// See https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support
+func TagsAutoCompleteValuesHandler(startTime time.Time, w http.ResponseWriter, r *http.Request) error {
+	deadline := searchutils.GetDeadlineForQuery(r, startTime)
+	if err := r.ParseForm(); err != nil {
+		return fmt.Errorf("cannot parse form values: %w", err)
+	}
+	limit, err := getInt(r, "limit")
+	if err != nil {
+		return err
+	}
+	if limit <= 0 {
+		// Use limit=100 by default. See https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support
+		limit = 100
+	}
+	tag := r.FormValue("tag")
+	if len(tag) == 0 {
+		return fmt.Errorf("missing `tag` query arg")
+	}
+	valuePrefix := r.FormValue("valuePrefix")
+	exprs := r.Form["expr"]
+	var tagValues []string
+	if len(exprs) == 0 {
+		// Fast path: there are no `expr` filters, so use netstorage.GetGraphiteTagValues.
+		// Escape special chars in tagPrefix as Graphite does.
+		// See https://github.com/graphite-project/graphite-web/blob/3ad279df5cb90b211953e39161df416e54a84948/webapp/graphite/tags/base.py#L228
+		filter := regexp.QuoteMeta(valuePrefix)
+		tagValues, err = netstorage.GetGraphiteTagValues(tag, filter, limit, deadline)
+		if err != nil {
+			return err
+		}
+	} else {
+		// Slow path: use netstorage.SearchMetricNames for applying `expr` filters.
+		sq, err := getSearchQueryForExprs(exprs)
+		if err != nil {
+			return err
+		}
+		mns, err := netstorage.SearchMetricNames(sq, deadline)
+		if err != nil {
+			return fmt.Errorf("cannot fetch metric names for %q: %w", sq, err)
+		}
+		m := make(map[string]struct{})
+		if tag == "name" {
+			tag = "__name__"
+		}
+		for _, mn := range mns {
+			tagValue := mn.GetTagValue(tag)
+			if len(tagValue) == 0 {
+				continue
+			}
+			m[string(tagValue)] = struct{}{}
+		}
+		if len(valuePrefix) > 0 {
+			for tagValue := range m {
+				if !strings.HasPrefix(tagValue, valuePrefix) {
+					delete(m, tagValue)
+				}
+			}
+		}
+		tagValues = make([]string, 0, len(m))
+		for tagValue := range m {
+			tagValues = append(tagValues, tagValue)
+		}
+		sort.Strings(tagValues)
+		if limit > 0 && limit < len(tagValues) {
+			tagValues = tagValues[:limit]
+		}
+	}
+
+	jsonp := r.FormValue("jsonp")
+	contentType := getContentType(jsonp)
+	w.Header().Set("Content-Type", contentType)
+	bw := bufferedwriter.Get(w)
+	defer bufferedwriter.Put(bw)
+	WriteTagsAutoCompleteResponse(bw, tagValues, jsonp)
+	if err := bw.Flush(); err != nil {
+		return err
+	}
+	tagsAutoCompleteValuesDuration.UpdateDuration(startTime)
+	return nil
+}
+
+var tagsAutoCompleteValuesDuration = metrics.NewSummary(`vm_request_duration_seconds{path="/tags/autoComplete/values"}`)
+
+// TagsAutoCompleteTagsHandler implements /tags/autoComplete/tags endpoint from Graphite Tags API.
+//
+// See https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support
+func TagsAutoCompleteTagsHandler(startTime time.Time, w http.ResponseWriter, r *http.Request) error {
+	deadline := searchutils.GetDeadlineForQuery(r, startTime)
+	if err := r.ParseForm(); err != nil {
+		return fmt.Errorf("cannot parse form values: %w", err)
+	}
+	limit, err := getInt(r, "limit")
+	if err != nil {
+		return err
+	}
+	if limit <= 0 {
+		// Use limit=100 by default. See https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support
+		limit = 100
+	}
+	tagPrefix := r.FormValue("tagPrefix")
+	exprs := r.Form["expr"]
+	var labels []string
+	if len(exprs) == 0 {
+		// Fast path: there are no `expr` filters, so use netstorage.GetGraphiteTags.
+
+		// Escape special chars in tagPrefix as Graphite does.
+		// See https://github.com/graphite-project/graphite-web/blob/3ad279df5cb90b211953e39161df416e54a84948/webapp/graphite/tags/base.py#L181
+		filter := regexp.QuoteMeta(tagPrefix)
+		labels, err = netstorage.GetGraphiteTags(filter, limit, deadline)
+		if err != nil {
+			return err
+		}
+	} else {
+		// Slow path: use netstorage.SearchMetricNames for applying `expr` filters.
+		sq, err := getSearchQueryForExprs(exprs)
+		if err != nil {
+			return err
+		}
+		mns, err := netstorage.SearchMetricNames(sq, deadline)
+		if err != nil {
+			return fmt.Errorf("cannot fetch metric names for %q: %w", sq, err)
+		}
+		m := make(map[string]struct{})
+		for _, mn := range mns {
+			m["name"] = struct{}{}
+			for _, tag := range mn.Tags {
+				m[string(tag.Key)] = struct{}{}
+			}
+		}
+		if len(tagPrefix) > 0 {
+			for label := range m {
+				if !strings.HasPrefix(label, tagPrefix) {
+					delete(m, label)
+				}
+			}
+		}
+		labels = make([]string, 0, len(m))
+		for label := range m {
+			labels = append(labels, label)
+		}
+		sort.Strings(labels)
+		if limit > 0 && limit < len(labels) {
+			labels = labels[:limit]
+		}
+	}
+
+	jsonp := r.FormValue("jsonp")
+	contentType := getContentType(jsonp)
+	w.Header().Set("Content-Type", contentType)
+	bw := bufferedwriter.Get(w)
+	defer bufferedwriter.Put(bw)
+	WriteTagsAutoCompleteResponse(bw, labels, jsonp)
+	if err := bw.Flush(); err != nil {
+		return err
+	}
+	tagsAutoCompleteTagsDuration.UpdateDuration(startTime)
+	return nil
+}
+
+var tagsAutoCompleteTagsDuration = metrics.NewSummary(`vm_request_duration_seconds{path="/tags/autoComplete/tags"}`)
+
+// TagsFindSeriesHandler implements /tags/findSeries endpoint from Graphite Tags API.
+//
+// See https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags
+func TagsFindSeriesHandler(startTime time.Time, w http.ResponseWriter, r *http.Request) error {
+	deadline := searchutils.GetDeadlineForQuery(r, startTime)
+	if err := r.ParseForm(); err != nil {
+		return fmt.Errorf("cannot parse form values: %w", err)
+	}
+	limit, err := getInt(r, "limit")
+	if err != nil {
+		return err
+	}
+	exprs := r.Form["expr"]
+	if len(exprs) == 0 {
+		return fmt.Errorf("expecting at least one `expr` query arg")
+	}
+	sq, err := getSearchQueryForExprs(exprs)
+	if err != nil {
+		return err
+	}
+	mns, err := netstorage.SearchMetricNames(sq, deadline)
+	if err != nil {
+		return fmt.Errorf("cannot fetch metric names for %q: %w", sq, err)
+	}
+	paths := getCanonicalPaths(mns)
+	if limit > 0 && limit < len(paths) {
+		paths = paths[:limit]
+	}
+
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
+	bw := bufferedwriter.Get(w)
+	defer bufferedwriter.Put(bw)
+	WriteTagsFindSeriesResponse(bw, paths)
+	if err := bw.Flush(); err != nil {
+		return err
+	}
+	tagsFindSeriesDuration.UpdateDuration(startTime)
+	return nil
+}
+
+func getCanonicalPaths(mns []storage.MetricName) []string {
+	paths := make([]string, 0, len(mns))
+	var b []byte
+	var tags []storage.Tag
+	for _, mn := range mns {
+		b = append(b[:0], mn.MetricGroup...)
+		tags = append(tags[:0], mn.Tags...)
+		sort.Slice(tags, func(i, j int) bool {
+			return string(tags[i].Key) < string(tags[j].Key)
+		})
+		for _, tag := range tags {
+			b = append(b, ';')
+			b = append(b, tag.Key...)
+			b = append(b, '=')
+			b = append(b, tag.Value...)
+		}
+		paths = append(paths, string(b))
+	}
+	sort.Strings(paths)
+	return paths
+}
+
+var tagsFindSeriesDuration = metrics.NewSummary(`vm_request_duration_seconds{path="/tags/findSeries"}`)
+
+// TagValuesHandler implements /tags/<tag_name> endpoint from Graphite Tags API.
+//
+// See https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags
+func TagValuesHandler(startTime time.Time, tagName string, w http.ResponseWriter, r *http.Request) error {
+	deadline := searchutils.GetDeadlineForQuery(r, startTime)
+	if err := r.ParseForm(); err != nil {
+		return fmt.Errorf("cannot parse form values: %w", err)
+	}
+	limit, err := getInt(r, "limit")
+	if err != nil {
+		return err
+	}
+	filter := r.FormValue("filter")
+	tagValues, err := netstorage.GetGraphiteTagValues(tagName, filter, limit, deadline)
+	if err != nil {
+		return err
+	}
+
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
+	bw := bufferedwriter.Get(w)
+	defer bufferedwriter.Put(bw)
+	WriteTagValuesResponse(bw, tagName, tagValues)
+	if err := bw.Flush(); err != nil {
+		return err
+	}
+	tagValuesDuration.UpdateDuration(startTime)
+	return nil
+}
+
+var tagValuesDuration = metrics.NewSummary(`vm_request_duration_seconds{path="/tags/<tag_name>"}`)
+
+// TagsHandler implements /tags endpoint from Graphite Tags API.
+//
+// See https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags
+func TagsHandler(startTime time.Time, w http.ResponseWriter, r *http.Request) error {
+	deadline := searchutils.GetDeadlineForQuery(r, startTime)
+	if err := r.ParseForm(); err != nil {
+		return fmt.Errorf("cannot parse form values: %w", err)
+	}
+	limit, err := getInt(r, "limit")
+	if err != nil {
+		return err
+	}
+	filter := r.FormValue("filter")
+	labels, err := netstorage.GetGraphiteTags(filter, limit, deadline)
+	if err != nil {
+		return err
+	}
+
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
+	bw := bufferedwriter.Get(w)
+	defer bufferedwriter.Put(bw)
+	WriteTagsResponse(bw, labels)
+	if err := bw.Flush(); err != nil {
+		return err
+	}
+	tagsDuration.UpdateDuration(startTime)
+	return nil
+}
+
+var tagsDuration = metrics.NewSummary(`vm_request_duration_seconds{path="/tags"}`)
+
+func getInt(r *http.Request, argName string) (int, error) {
+	argValue := r.FormValue(argName)
+	if len(argValue) == 0 {
+		return 0, nil
+	}
+	n, err := strconv.Atoi(argValue)
+	if err != nil {
+		return 0, fmt.Errorf("cannot parse %q=%q: %w", argName, argValue, err)
+	}
+	return n, nil
+}
+
+func getSearchQueryForExprs(exprs []string) (*storage.SearchQuery, error) {
+	tfs, err := exprsToTagFilters(exprs)
+	if err != nil {
+		return nil, err
+	}
+	ct := time.Now().UnixNano() / 1e6
+	sq := storage.NewSearchQuery(0, ct, [][]storage.TagFilter{tfs})
+	return sq, nil
+}
+
+func exprsToTagFilters(exprs []string) ([]storage.TagFilter, error) {
+	tfs := make([]storage.TagFilter, 0, len(exprs))
+	for _, expr := range exprs {
+		tf, err := parseFilterExpr(expr)
+		if err != nil {
+			return nil, fmt.Errorf("cannot parse `expr` query arg: %w", err)
+		}
+		tfs = append(tfs, *tf)
+	}
+	return tfs, nil
+}
+
+func parseFilterExpr(s string) (*storage.TagFilter, error) {
+	n := strings.Index(s, "=")
+	if n < 0 {
+		return nil, fmt.Errorf("missing tag value in filter expression %q", s)
+	}
+	tagName := s[:n]
+	tagValue := s[n+1:]
+	isNegative := false
+	if strings.HasSuffix(tagName, "!") {
+		isNegative = true
+		tagName = tagName[:len(tagName)-1]
+	}
+	if tagName == "name" {
+		tagName = ""
+	}
+	isRegexp := false
+	if strings.HasPrefix(tagValue, "~") {
+		isRegexp = true
+		tagValue = "^(?:" + tagValue[1:] + ").*"
+	}
+	return &storage.TagFilter{
+		Key:        []byte(tagName),
+		Value:      []byte(tagValue),
+		IsNegative: isNegative,
+		IsRegexp:   isRegexp,
+	}, nil
+}
--- a/app/vmselect/graphite/tags_autocomplete_response.qtpl
+++ b/app/vmselect/graphite/tags_autocomplete_response.qtpl
@@ -0,0 +1,16 @@
+{% stripspace %}
+
+TagsAutoCompleteResponse generates responses for /tags/autoComplete/{tags,values} handlers in Graphite Tags API.
+See https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support
+{% func TagsAutoCompleteResponse(ss []string, jsonp string) %}
+	{% if jsonp != "" %}{%s= jsonp %}({% endif %}
+	[
+		{% for i, s := range ss %}
+			{%q= s %}
+			{% if i+1 < len(ss) %},{% endif %}
+		{% endfor %}
+	]
+	{% if jsonp != "" %}){% endif %}
+{% endfunc %}
+
+{% endstripspace %}
--- a/app/vmselect/graphite/tags_autocomplete_response.qtpl.go
+++ b/app/vmselect/graphite/tags_autocomplete_response.qtpl.go
@@ -0,0 +1,81 @@
+// Code generated by qtc from "tags_autocomplete_response.qtpl". DO NOT EDIT.
+// See https://github.com/valyala/quicktemplate for details.
+
+// TagsAutoCompleteResponse generates responses for /tags/autoComplete/{tags,values} handlers in Graphite Tags API.See https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support
+
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:5
+package graphite
+
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:5
+import (
+	qtio422016 "io"
+
+	qt422016 "github.com/valyala/quicktemplate"
+)
+
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:5
+var (
+	_ = qtio422016.Copy
+	_ = qt422016.AcquireByteBuffer
+)
+
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:5
+func StreamTagsAutoCompleteResponse(qw422016 *qt422016.Writer, ss []string, jsonp string) {
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:6
+	if jsonp != "" {
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:6
+		qw422016.N().S(jsonp)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:6
+		qw422016.N().S(`(`)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:6
+	}
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:6
+	qw422016.N().S(`[`)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:8
+	for i, s := range ss {
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:9
+		qw422016.N().Q(s)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:10
+		if i+1 < len(ss) {
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:10
+			qw422016.N().S(`,`)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:10
+		}
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:11
+	}
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:11
+	qw422016.N().S(`]`)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:13
+	if jsonp != "" {
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:13
+		qw422016.N().S(`)`)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:13
+	}
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+}
+
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+func WriteTagsAutoCompleteResponse(qq422016 qtio422016.Writer, ss []string, jsonp string) {
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+	qw422016 := qt422016.AcquireWriter(qq422016)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+	StreamTagsAutoCompleteResponse(qw422016, ss, jsonp)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+	qt422016.ReleaseWriter(qw422016)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+}
+
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+func TagsAutoCompleteResponse(ss []string, jsonp string) string {
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+	qb422016 := qt422016.AcquireByteBuffer()
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+	WriteTagsAutoCompleteResponse(qb422016, ss, jsonp)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+	qs422016 := string(qb422016.B)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+	qt422016.ReleaseByteBuffer(qb422016)
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+	return qs422016
+//line app/vmselect/graphite/tags_autocomplete_response.qtpl:14
+}
--- a/app/vmselect/graphite/tags_find_series_response.qtpl
+++ b/app/vmselect/graphite/tags_find_series_response.qtpl
@@ -0,0 +1,12 @@
+{% stripspace %}
+
+{% func TagsFindSeriesResponse(paths []string) %}
+[
+	{% for i, path := range paths %}
+		{%q= path %}
+		{% if i+1 < len(paths) %},{% endif %}
+	{% endfor %}
+]
+{% endfunc %}
+
+{% endstripspace %}
--- a/app/vmselect/graphite/tags_find_series_response.qtpl.go
+++ b/app/vmselect/graphite/tags_find_series_response.qtpl.go
@@ -0,0 +1,65 @@
+// Code generated by qtc from "tags_find_series_response.qtpl". DO NOT EDIT.
+// See https://github.com/valyala/quicktemplate for details.
+
+//line app/vmselect/graphite/tags_find_series_response.qtpl:3
+package graphite
+
+//line app/vmselect/graphite/tags_find_series_response.qtpl:3
+import (
+	qtio422016 "io"
+
+	qt422016 "github.com/valyala/quicktemplate"
+)
+
+//line app/vmselect/graphite/tags_find_series_response.qtpl:3
+var (
+	_ = qtio422016.Copy
+	_ = qt422016.AcquireByteBuffer
+)
+
+//line app/vmselect/graphite/tags_find_series_response.qtpl:3
+func StreamTagsFindSeriesResponse(qw422016 *qt422016.Writer, paths []string) {
+//line app/vmselect/graphite/tags_find_series_response.qtpl:3
+	qw422016.N().S(`[`)
+//line app/vmselect/graphite/tags_find_series_response.qtpl:5
+	for i, path := range paths {
+//line app/vmselect/graphite/tags_find_series_response.qtpl:6
+		qw422016.N().Q(path)
+//line app/vmselect/graphite/tags_find_series_response.qtpl:7
+		if i+1 < len(paths) {
+//line app/vmselect/graphite/tags_find_series_response.qtpl:7
+			qw422016.N().S(`,`)
+//line app/vmselect/graphite/tags_find_series_response.qtpl:7
+		}
+//line app/vmselect/graphite/tags_find_series_response.qtpl:8
+	}
+//line app/vmselect/graphite/tags_find_series_response.qtpl:8
+	qw422016.N().S(`]`)
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+}
+
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+func WriteTagsFindSeriesResponse(qq422016 qtio422016.Writer, paths []string) {
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+	qw422016 := qt422016.AcquireWriter(qq422016)
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+	StreamTagsFindSeriesResponse(qw422016, paths)
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+	qt422016.ReleaseWriter(qw422016)
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+}
+
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+func TagsFindSeriesResponse(paths []string) string {
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+	qb422016 := qt422016.AcquireByteBuffer()
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+	WriteTagsFindSeriesResponse(qb422016, paths)
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+	qs422016 := string(qb422016.B)
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+	qt422016.ReleaseByteBuffer(qb422016)
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+	return qs422016
+//line app/vmselect/graphite/tags_find_series_response.qtpl:10
+}
--- a/app/vmselect/graphite/tags_response.qtpl
+++ b/app/vmselect/graphite/tags_response.qtpl
@@ -0,0 +1,16 @@
+{% stripspace %}
+
+Tags generates response for /tags handler
+See https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags
+{% func TagsResponse(tags []string) %}
+[
+	{% for i, tag := range tags %}
+		{
+			"tag":{%q= tag %}
+		}
+		{% if i+1 < len(tags) %},{% endif %}
+	{% endfor %}
+]
+{% endfunc %}
+
+{% endstripspace %}
--- a/app/vmselect/graphite/tags_response.qtpl.go
+++ b/app/vmselect/graphite/tags_response.qtpl.go
@@ -0,0 +1,71 @@
+// Code generated by qtc from "tags_response.qtpl". DO NOT EDIT.
+// See https://github.com/valyala/quicktemplate for details.
+
+// Tags generates response for /tags handlerSee https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags
+
+//line app/vmselect/graphite/tags_response.qtpl:5
+package graphite
+
+//line app/vmselect/graphite/tags_response.qtpl:5
+import (
+	qtio422016 "io"
+
+	qt422016 "github.com/valyala/quicktemplate"
+)
+
+//line app/vmselect/graphite/tags_response.qtpl:5
+var (
+	_ = qtio422016.Copy
+	_ = qt422016.AcquireByteBuffer
+)
+
+//line app/vmselect/graphite/tags_response.qtpl:5
+func StreamTagsResponse(qw422016 *qt422016.Writer, tags []string) {
+//line app/vmselect/graphite/tags_response.qtpl:5
+	qw422016.N().S(`[`)
+//line app/vmselect/graphite/tags_response.qtpl:7
+	for i, tag := range tags {
+//line app/vmselect/graphite/tags_response.qtpl:7
+		qw422016.N().S(`{"tag":`)
+//line app/vmselect/graphite/tags_response.qtpl:9
+		qw422016.N().Q(tag)
+//line app/vmselect/graphite/tags_response.qtpl:9
+		qw422016.N().S(`}`)
+//line app/vmselect/graphite/tags_response.qtpl:11
+		if i+1 < len(tags) {
+//line app/vmselect/graphite/tags_response.qtpl:11
+			qw422016.N().S(`,`)
+//line app/vmselect/graphite/tags_response.qtpl:11
+		}
+//line app/vmselect/graphite/tags_response.qtpl:12
+	}
+//line app/vmselect/graphite/tags_response.qtpl:12
+	qw422016.N().S(`]`)
+//line app/vmselect/graphite/tags_response.qtpl:14
+}
+
+//line app/vmselect/graphite/tags_response.qtpl:14
+func WriteTagsResponse(qq422016 qtio422016.Writer, tags []string) {
+//line app/vmselect/graphite/tags_response.qtpl:14
+	qw422016 := qt422016.AcquireWriter(qq422016)
+//line app/vmselect/graphite/tags_response.qtpl:14
+	StreamTagsResponse(qw422016, tags)
+//line app/vmselect/graphite/tags_response.qtpl:14
+	qt422016.ReleaseWriter(qw422016)
+//line app/vmselect/graphite/tags_response.qtpl:14
+}
+
+//line app/vmselect/graphite/tags_response.qtpl:14
+func TagsResponse(tags []string) string {
+//line app/vmselect/graphite/tags_response.qtpl:14
+	qb422016 := qt422016.AcquireByteBuffer()
+//line app/vmselect/graphite/tags_response.qtpl:14
+	WriteTagsResponse(qb422016, tags)
+//line app/vmselect/graphite/tags_response.qtpl:14
+	qs422016 := string(qb422016.B)
+//line app/vmselect/graphite/tags_response.qtpl:14
+	qt422016.ReleaseByteBuffer(qb422016)
+//line app/vmselect/graphite/tags_response.qtpl:14
+	return qs422016
+//line app/vmselect/graphite/tags_response.qtpl:14
+}
--- a/app/vmselect/graphite/tags_tag_multi_series_response.qtpl
+++ b/app/vmselect/graphite/tags_tag_multi_series_response.qtpl
@@ -0,0 +1,14 @@
+{% stripspace %}
+
+TagsTagMultiSeriesResponse generates response for /tags/tagMultiSeries .
+See https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb
+{% func TagsTagMultiSeriesResponse(canonicalPaths []string, isJSONResponse bool) %}
+	{% if isJSONResponse %}[{% endif %}
+	{% for i, path := range canonicalPaths %}
+		{%q= path %}
+		{% if i+1 < len(canonicalPaths) %},{% endif %}
+	{% endfor %}
+	{% if isJSONResponse %}]{% endif %}
+{% endfunc %}
+
+{% endstripspace %}
--- a/app/vmselect/graphite/tags_tag_multi_series_response.qtpl.go
+++ b/app/vmselect/graphite/tags_tag_multi_series_response.qtpl.go
@@ -0,0 +1,75 @@
+// Code generated by qtc from "tags_tag_multi_series_response.qtpl". DO NOT EDIT.
+// See https://github.com/valyala/quicktemplate for details.
+
+// TagsTagMultiSeriesResponse generates response for /tags/tagMultiSeries .See https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb
+
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:5
+package graphite
+
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:5
+import (
+	qtio422016 "io"
+
+	qt422016 "github.com/valyala/quicktemplate"
+)
+
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:5
+var (
+	_ = qtio422016.Copy
+	_ = qt422016.AcquireByteBuffer
+)
+
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:5
+func StreamTagsTagMultiSeriesResponse(qw422016 *qt422016.Writer, canonicalPaths []string, isJSONResponse bool) {
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:6
+	if isJSONResponse {
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:6
+		qw422016.N().S(`[`)
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:6
+	}
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:7
+	for i, path := range canonicalPaths {
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:8
+		qw422016.N().Q(path)
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:9
+		if i+1 < len(canonicalPaths) {
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:9
+			qw422016.N().S(`,`)
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:9
+		}
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:10
+	}
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:11
+	if isJSONResponse {
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:11
+		qw422016.N().S(`]`)
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:11
+	}
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+}
+
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+func WriteTagsTagMultiSeriesResponse(qq422016 qtio422016.Writer, canonicalPaths []string, isJSONResponse bool) {
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+	qw422016 := qt422016.AcquireWriter(qq422016)
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+	StreamTagsTagMultiSeriesResponse(qw422016, canonicalPaths, isJSONResponse)
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+	qt422016.ReleaseWriter(qw422016)
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+}
+
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+func TagsTagMultiSeriesResponse(canonicalPaths []string, isJSONResponse bool) string {
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+	qb422016 := qt422016.AcquireByteBuffer()
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+	WriteTagsTagMultiSeriesResponse(qb422016, canonicalPaths, isJSONResponse)
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+	qs422016 := string(qb422016.B)
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+	qt422016.ReleaseByteBuffer(qb422016)
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+	return qs422016
+//line app/vmselect/graphite/tags_tag_multi_series_response.qtpl:12
+}
--- a/app/vmselect/main.go
+++ b/app/vmselect/main.go
@@ -10,10 +10,12 @@ import (
 	"time"

 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/graphite"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/netstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/prometheus"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/promql"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/searchutils"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timerpool"
@@ -21,7 +23,7 @@ import (
 )

 var (
-	deleteAuthKey         = flag.String("deleteAuthKey", "", "authKey for metrics' deletion via /api/v1/admin/tsdb/delete_series")
+	deleteAuthKey         = flag.String("deleteAuthKey", "", "authKey for metrics' deletion via /api/v1/admin/tsdb/delete_series and /tags/delSeries")
 	maxConcurrentRequests = flag.Int("search.maxConcurrentRequests", getDefaultMaxConcurrentRequests(), "The maximum number of concurrent search requests. "+
 		"It shouldn't be high, since a single request can saturate all the CPU cores. See also -search.maxQueueDuration")
 	maxQueueDuration = flag.Duration("search.maxQueueDuration", 10*time.Second, "The maximum time the request waits for execution when -search.maxConcurrentRequests "+
@@ -45,6 +47,9 @@ func getDefaultMaxConcurrentRequests() int {

 // Init initializes vmselect
 func Init() {
+	tmpDirPath := *vmstorage.DataPath + "/tmp"
+	fs.RemoveDirContents(tmpDirPath)
+	netstorage.InitTmpBlocksDir(tmpDirPath)
 	promql.InitRollupResultCache(*vmstorage.DataPath + "/cache/rollupResult")

 	concurrencyCh = make(chan struct{}, *maxConcurrentRequests)
@@ -127,6 +132,16 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 			return true
 		}
 	}
+	if strings.HasPrefix(path, "/tags/") && !isGraphiteTagsPath(path) {
+		tagName := r.URL.Path[len("/tags/"):]
+		graphiteTagValuesRequests.Inc()
+		if err := graphite.TagValuesHandler(startTime, tagName, w, r); err != nil {
+			graphiteTagValuesErrors.Inc()
+			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
+			return true
+		}
+		return true
+	}

 	switch path {
 	case "/api/v1/query":
@@ -254,22 +269,85 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 			return true
 		}
 		return true
+	case "/tags/tagSeries":
+		graphiteTagsTagSeriesRequests.Inc()
+		if err := graphite.TagsTagSeriesHandler(startTime, w, r); err != nil {
+			graphiteTagsTagSeriesErrors.Inc()
+			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
+			return true
+		}
+		return true
+	case "/tags/tagMultiSeries":
+		graphiteTagsTagMultiSeriesRequests.Inc()
+		if err := graphite.TagsTagMultiSeriesHandler(startTime, w, r); err != nil {
+			graphiteTagsTagMultiSeriesErrors.Inc()
+			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
+			return true
+		}
+		return true
+	case "/tags":
+		graphiteTagsRequests.Inc()
+		if err := graphite.TagsHandler(startTime, w, r); err != nil {
+			graphiteTagsErrors.Inc()
+			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
+			return true
+		}
+		return true
+	case "/tags/findSeries":
+		graphiteTagsFindSeriesRequests.Inc()
+		if err := graphite.TagsFindSeriesHandler(startTime, w, r); err != nil {
+			graphiteTagsFindSeriesErrors.Inc()
+			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
+			return true
+		}
+		return true
+	case "/tags/autoComplete/tags":
+		graphiteTagsAutoCompleteTagsRequests.Inc()
+		httpserver.EnableCORS(w, r)
+		if err := graphite.TagsAutoCompleteTagsHandler(startTime, w, r); err != nil {
+			graphiteTagsAutoCompleteTagsErrors.Inc()
+			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
+			return true
+		}
+		return true
+	case "/tags/autoComplete/values":
+		graphiteTagsAutoCompleteValuesRequests.Inc()
+		httpserver.EnableCORS(w, r)
+		if err := graphite.TagsAutoCompleteValuesHandler(startTime, w, r); err != nil {
+			graphiteTagsAutoCompleteValuesErrors.Inc()
+			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
+			return true
+		}
+		return true
+	case "/tags/delSeries":
+		graphiteTagsDelSeriesRequests.Inc()
+		authKey := r.FormValue("authKey")
+		if authKey != *deleteAuthKey {
+			httpserver.Errorf(w, r, "invalid authKey %q. It must match the value from -deleteAuthKey command line flag", authKey)
+			return true
+		}
+		if err := graphite.TagsDelSeriesHandler(startTime, w, r); err != nil {
+			graphiteTagsDelSeriesErrors.Inc()
+			httpserver.Errorf(w, r, "error in %q: %s", r.URL.Path, err)
+			return true
+		}
+		return true
 	case "/api/v1/rules":
 		// Return dumb placeholder
 		rulesRequests.Inc()
-		w.Header().Set("Content-Type", "application/json")
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
 		fmt.Fprintf(w, "%s", `{"status":"success","data":{"groups":[]}}`)
 		return true
 	case "/api/v1/alerts":
 		// Return dumb placehloder
 		alertsRequests.Inc()
-		w.Header().Set("Content-Type", "application/json")
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
 		fmt.Fprintf(w, "%s", `{"status":"success","data":{"alerts":[]}}`)
 		return true
 	case "/api/v1/metadata":
 		// Return dumb placeholder
 		metadataRequests.Inc()
-		w.Header().Set("Content-Type", "application/json")
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
 		fmt.Fprintf(w, "%s", `{"status":"success","data":{}}`)
 		return true
 	case "/api/v1/admin/tsdb/delete_series":
@@ -291,10 +369,22 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 	}
 }

+func isGraphiteTagsPath(path string) bool {
+	switch path {
+	// See https://graphite.readthedocs.io/en/stable/tags.html for a list of Graphite Tags API paths.
+	// Do not include `/tags/<tag_name>` here, since this will fool the caller.
+	case "/tags/tagSeries", "/tags/tagMultiSeries", "/tags/findSeries",
+		"/tags/autoComplete/tags", "/tags/autoComplete/values", "/tags/delSeries":
+		return true
+	default:
+		return false
+	}
+}
+
 func sendPrometheusError(w http.ResponseWriter, r *http.Request, err error) {
 	logger.Warnf("error in %q: %s", r.RequestURI, err)

-	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
 	statusCode := http.StatusUnprocessableEntity
 	var esc *httpserver.ErrorWithStatusCode
 	if errors.As(err, &esc) {
@@ -355,6 +445,30 @@ var (
 	graphiteMetricsIndexRequests = metrics.NewCounter(`vm_http_requests_total{path="/metrics/index.json"}`)
 	graphiteMetricsIndexErrors   = metrics.NewCounter(`vm_http_request_errors_total{path="/metrics/index.json"}`)

+	graphiteTagsTagSeriesRequests = metrics.NewCounter(`vm_http_requests_total{path="/tags/tagSeries"}`)
+	graphiteTagsTagSeriesErrors   = metrics.NewCounter(`vm_http_request_errors_total{path="/tags/tagSeries"}`)
+
+	graphiteTagsTagMultiSeriesRequests = metrics.NewCounter(`vm_http_requests_total{path="/tags/tagMultiSeries"}`)
+	graphiteTagsTagMultiSeriesErrors   = metrics.NewCounter(`vm_http_request_errors_total{path="/tags/tagMultiSeries"}`)
+
+	graphiteTagsRequests = metrics.NewCounter(`vm_http_requests_total{path="/tags"}`)
+	graphiteTagsErrors   = metrics.NewCounter(`vm_http_request_errors_total{path="/tags"}`)
+
+	graphiteTagValuesRequests = metrics.NewCounter(`vm_http_requests_total{path="/tags/<tag_name>"}`)
+	graphiteTagValuesErrors   = metrics.NewCounter(`vm_http_request_errors_total{path="/tags/<tag_name>"}`)
+
+	graphiteTagsFindSeriesRequests = metrics.NewCounter(`vm_http_requests_total{path="/tags/findSeries"}`)
+	graphiteTagsFindSeriesErrors   = metrics.NewCounter(`vm_http_request_errors_total{path="/tags/findSeries"}`)
+
+	graphiteTagsAutoCompleteTagsRequests = metrics.NewCounter(`vm_http_requests_total{path="/tags/autoComplete/tags"}`)
+	graphiteTagsAutoCompleteTagsErrors   = metrics.NewCounter(`vm_http_request_errors_total{path="/tags/autoComplete/tags"}`)
+
+	graphiteTagsAutoCompleteValuesRequests = metrics.NewCounter(`vm_http_requests_total{path="/tags/autoComplete/values"}`)
+	graphiteTagsAutoCompleteValuesErrors   = metrics.NewCounter(`vm_http_request_errors_total{path="/tags/autoComplete/values"}`)
+
+	graphiteTagsDelSeriesRequests = metrics.NewCounter(`vm_http_requests_total{path="/tags/delSeries"}`)
+	graphiteTagsDelSeriesErrors   = metrics.NewCounter(`vm_http_request_errors_total{path="/tags/delSeries"}`)
+
 	rulesRequests    = metrics.NewCounter(`vm_http_requests_total{path="/api/v1/rules"}`)
 	alertsRequests   = metrics.NewCounter(`vm_http_requests_total{path="/api/v1/alerts"}`)
 	metadataRequests = metrics.NewCounter(`vm_http_requests_total{path="/api/v1/metadata"}`)
--- a/app/vmselect/netstorage/netstorage.go
+++ b/app/vmselect/netstorage/netstorage.go
@@ -5,6 +5,7 @@ import (
 	"errors"
 	"flag"
 	"fmt"
+	"regexp"
 	"runtime"
 	"sort"
 	"sync"
@@ -57,6 +58,7 @@ type Results struct {

 	packedTimeseries []packedTimeseries
 	sr               *storage.Search
+	tbf              *tmpBlocksFile
 }

 // Len returns the number of results in rss.
@@ -72,6 +74,8 @@ func (rss *Results) Cancel() {
 func (rss *Results) mustClose() {
 	putStorageSearch(rss.sr)
 	rss.sr = nil
+	putTmpBlocksFile(rss.tbf)
+	rss.tbf = nil
 }

 var timeseriesWorkCh = make(chan *timeseriesWork, gomaxprocs*16)
@@ -105,7 +109,7 @@ func timeseriesWorker(workerID uint) {
 			tsw.doneCh <- nil
 			continue
 		}
-		if err := tsw.pts.Unpack(&rs, rss.tr, rss.fetchData); err != nil {
+		if err := tsw.pts.Unpack(&rs, rss.tbf, rss.tr, rss.fetchData); err != nil {
 			tsw.doneCh <- fmt.Errorf("error during time series unpacking: %w", err)
 			continue
 		}
@@ -179,27 +183,29 @@ var gomaxprocs = runtime.GOMAXPROCS(-1)

 type packedTimeseries struct {
 	metricName string
-	brs        []storage.BlockRef
+	brs        []blockRef
 }

 var unpackWorkCh = make(chan *unpackWork, gomaxprocs*128)

 type unpackWorkItem struct {
-	br storage.BlockRef
+	br blockRef
 	tr storage.TimeRange
 }

 type unpackWork struct {
+	tbf    *tmpBlocksFile
 	ws     []unpackWorkItem
 	sbs    []*sortBlock
 	doneCh chan error
 }

 func (upw *unpackWork) reset() {
+	upw.tbf = nil
 	ws := upw.ws
 	for i := range ws {
 		w := &ws[i]
-		w.br = storage.BlockRef{}
+		w.br = blockRef{}
 		w.tr = storage.TimeRange{}
 	}
 	upw.ws = upw.ws[:0]
@@ -216,7 +222,7 @@ func (upw *unpackWork) reset() {
 func (upw *unpackWork) unpack(tmpBlock *storage.Block) {
 	for _, w := range upw.ws {
 		sb := getSortBlock()
-		if err := sb.unpackFrom(tmpBlock, w.br, w.tr); err != nil {
+		if err := sb.unpackFrom(tmpBlock, upw.tbf, w.br, w.tr); err != nil {
 			putSortBlock(sb)
 			upw.doneCh <- fmt.Errorf("cannot unpack block: %w", err)
 			return
@@ -262,7 +268,7 @@ func unpackWorker() {
 var unpackBatchSize = 8 * runtime.GOMAXPROCS(-1)

 // Unpack unpacks pts to dst.
-func (pts *packedTimeseries) Unpack(dst *Result, tr storage.TimeRange, fetchData bool) error {
+func (pts *packedTimeseries) Unpack(dst *Result, tbf *tmpBlocksFile, tr storage.TimeRange, fetchData bool) error {
 	dst.reset()
 	if err := dst.MetricName.Unmarshal(bytesutil.ToUnsafeBytes(pts.metricName)); err != nil {
 		return fmt.Errorf("cannot unmarshal metricName %q: %w", pts.metricName, err)
@@ -276,11 +282,13 @@ func (pts *packedTimeseries) Unpack(dst *Result, tr storage.TimeRange, fetchData
 	brsLen := len(pts.brs)
 	upws := make([]*unpackWork, 0, 1+brsLen/unpackBatchSize)
 	upw := getUnpackWork()
+	upw.tbf = tbf
 	for _, br := range pts.brs {
 		if len(upw.ws) >= unpackBatchSize {
 			unpackWorkCh <- upw
 			upws = append(upws, upw)
 			upw = getUnpackWork()
+			upw.tbf = tbf
 		}
 		upw.ws = append(upw.ws, unpackWorkItem{
 			br: br,
@@ -397,9 +405,10 @@ func (sb *sortBlock) reset() {
 	sb.NextIdx = 0
 }

-func (sb *sortBlock) unpackFrom(tmpBlock *storage.Block, br storage.BlockRef, tr storage.TimeRange) error {
+func (sb *sortBlock) unpackFrom(tmpBlock *storage.Block, tbf *tmpBlocksFile, br blockRef, tr storage.TimeRange) error {
 	tmpBlock.Reset()
-	br.MustReadBlock(tmpBlock, true)
+	brReal := tbf.MustReadBlockRefAt(br.partRef, br.addr)
+	brReal.MustReadBlock(tmpBlock, true)
 	if err := tmpBlock.UnmarshalData(); err != nil {
 		return fmt.Errorf("cannot unmarshal block: %w", err)
 	}
@@ -445,6 +454,55 @@ func DeleteSeries(sq *storage.SearchQuery) (int, error) {
 	return vmstorage.DeleteMetrics(tfss)
 }

+// GetLabelsOnTimeRange returns labels for the given tr until the given deadline.
+func GetLabelsOnTimeRange(tr storage.TimeRange, deadline searchutils.Deadline) ([]string, error) {
+	if deadline.Exceeded() {
+		return nil, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
+	}
+	labels, err := vmstorage.SearchTagKeysOnTimeRange(tr, *maxTagKeysPerSearch, deadline.Deadline())
+	if err != nil {
+		return nil, fmt.Errorf("error during labels search on time range: %w", err)
+	}
+	// Substitute "" with "__name__"
+	for i := range labels {
+		if labels[i] == "" {
+			labels[i] = "__name__"
+		}
+	}
+	// Sort labels like Prometheus does
+	sort.Strings(labels)
+	return labels, nil
+}
+
+// GetGraphiteTags returns Graphite tags until the given deadline.
+func GetGraphiteTags(filter string, limit int, deadline searchutils.Deadline) ([]string, error) {
+	if deadline.Exceeded() {
+		return nil, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
+	}
+	labels, err := GetLabels(deadline)
+	if err != nil {
+		return nil, err
+	}
+	// Substitute "__name__" with "name" for Graphite compatibility
+	for i := range labels {
+		if labels[i] == "__name__" {
+			labels[i] = "name"
+			sort.Strings(labels)
+			break
+		}
+	}
+	if len(filter) > 0 {
+		labels, err = applyGraphiteRegexpFilter(filter, labels)
+		if err != nil {
+			return nil, err
+		}
+	}
+	if limit > 0 && limit < len(labels) {
+		labels = labels[:limit]
+	}
+	return labels, nil
+}
+
 // GetLabels returns labels until the given deadline.
 func GetLabels(deadline searchutils.Deadline) ([]string, error) {
 	if deadline.Exceeded() {
@@ -454,20 +512,60 @@ func GetLabels(deadline searchutils.Deadline) ([]string, error) {
 	if err != nil {
 		return nil, fmt.Errorf("error during labels search: %w", err)
 	}
-
 	// Substitute "" with "__name__"
 	for i := range labels {
 		if labels[i] == "" {
 			labels[i] = "__name__"
 		}
 	}
-
 	// Sort labels like Prometheus does
 	sort.Strings(labels)
-
 	return labels, nil
 }

+// GetLabelValuesOnTimeRange returns label values for the given labelName on the given tr
+// until the given deadline.
+func GetLabelValuesOnTimeRange(labelName string, tr storage.TimeRange, deadline searchutils.Deadline) ([]string, error) {
+	if deadline.Exceeded() {
+		return nil, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
+	}
+	if labelName == "__name__" {
+		labelName = ""
+	}
+	// Search for tag values
+	labelValues, err := vmstorage.SearchTagValuesOnTimeRange([]byte(labelName), tr, *maxTagValuesPerSearch, deadline.Deadline())
+	if err != nil {
+		return nil, fmt.Errorf("error during label values search on time range for labelName=%q: %w", labelName, err)
+	}
+	// Sort labelValues like Prometheus does
+	sort.Strings(labelValues)
+	return labelValues, nil
+}
+
+// GetGraphiteTagValues returns tag values for the given tagName until the given deadline.
+func GetGraphiteTagValues(tagName, filter string, limit int, deadline searchutils.Deadline) ([]string, error) {
+	if deadline.Exceeded() {
+		return nil, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
+	}
+	if tagName == "name" {
+		tagName = ""
+	}
+	tagValues, err := GetLabelValues(tagName, deadline)
+	if err != nil {
+		return nil, err
+	}
+	if len(filter) > 0 {
+		tagValues, err = applyGraphiteRegexpFilter(filter, tagValues)
+		if err != nil {
+			return nil, err
+		}
+	}
+	if limit > 0 && limit < len(tagValues) {
+		tagValues = tagValues[:limit]
+	}
+	return tagValues, nil
+}
+
 // GetLabelValues returns label values for the given labelName
 // until the given deadline.
 func GetLabelValues(labelName string, deadline searchutils.Deadline) ([]string, error) {
@@ -477,16 +575,13 @@ func GetLabelValues(labelName string, deadline searchutils.Deadline) ([]string,
 	if labelName == "__name__" {
 		labelName = ""
 	}
-
 	// Search for tag values
 	labelValues, err := vmstorage.SearchTagValues([]byte(labelName), *maxTagValuesPerSearch, deadline.Deadline())
 	if err != nil {
 		return nil, fmt.Errorf("error during label values search for labelName=%q: %w", labelName, err)
 	}
-
 	// Sort labelValues like Prometheus does
 	sort.Strings(labelValues)
-
 	return labelValues, nil
 }

@@ -683,6 +778,32 @@ var exportWorkPool = &sync.Pool{
 	},
 }

+// SearchMetricNames returns all the metric names matching sq until the given deadline.
+func SearchMetricNames(sq *storage.SearchQuery, deadline searchutils.Deadline) ([]storage.MetricName, error) {
+	if deadline.Exceeded() {
+		return nil, fmt.Errorf("timeout exceeded before starting to search metric names: %s", deadline.String())
+	}
+
+	// Setup search.
+	tfss, err := setupTfss(sq.TagFilterss)
+	if err != nil {
+		return nil, err
+	}
+	tr := storage.TimeRange{
+		MinTimestamp: sq.MinTimestamp,
+		MaxTimestamp: sq.MaxTimestamp,
+	}
+	if err := vmstorage.CheckTimeRange(tr); err != nil {
+		return nil, err
+	}
+
+	mns, err := vmstorage.SearchMetricNames(tfss, tr, *maxMetricsPerSearch, deadline.Deadline())
+	if err != nil {
+		return nil, fmt.Errorf("cannot find metric names: %w", err)
+	}
+	return mns, nil
+}
+
 // ProcessSearchQuery performs sq until the given deadline.
 //
 // Results.RunParallel or Results.Cancel must be called on the returned Results.
@@ -709,19 +830,31 @@ func ProcessSearchQuery(sq *storage.SearchQuery, fetchData bool, deadline search

 	sr := getStorageSearch()
 	maxSeriesCount := sr.Init(vmstorage.Storage, tfss, tr, *maxMetricsPerSearch, deadline.Deadline())
-
-	m := make(map[string][]storage.BlockRef, maxSeriesCount)
+	m := make(map[string][]blockRef, maxSeriesCount)
 	orderedMetricNames := make([]string, 0, maxSeriesCount)
 	blocksRead := 0
+	tbf := getTmpBlocksFile()
+	var buf []byte
 	for sr.NextMetricBlock() {
 		blocksRead++
 		if deadline.Exceeded() {
+			putTmpBlocksFile(tbf)
 			putStorageSearch(sr)
 			return nil, fmt.Errorf("timeout exceeded while fetching data block #%d from storage: %s", blocksRead, deadline.String())
 		}
+		buf = sr.MetricBlockRef.BlockRef.Marshal(buf[:0])
+		addr, err := tbf.WriteBlockRefData(buf)
+		if err != nil {
+			putTmpBlocksFile(tbf)
+			putStorageSearch(sr)
+			return nil, fmt.Errorf("cannot write %d bytes to temporary file: %w", len(buf), err)
+		}
 		metricName := sr.MetricBlockRef.MetricName
 		brs := m[string(metricName)]
-		brs = append(brs, *sr.MetricBlockRef.BlockRef)
+		brs = append(brs, blockRef{
+			partRef: sr.MetricBlockRef.BlockRef.PartRef(),
+			addr:    addr,
+		})
 		if len(brs) > 1 {
 			// An optimization: do not allocate a string for already existing metricName key in m
 			m[string(metricName)] = brs
@@ -733,12 +866,18 @@ func ProcessSearchQuery(sq *storage.SearchQuery, fetchData bool, deadline search
 		}
 	}
 	if err := sr.Error(); err != nil {
+		putTmpBlocksFile(tbf)
 		putStorageSearch(sr)
 		if errors.Is(err, storage.ErrDeadlineExceeded) {
 			return nil, fmt.Errorf("timeout exceeded during the query: %s", deadline.String())
 		}
 		return nil, fmt.Errorf("search error after reading %d data blocks: %w", blocksRead, err)
 	}
+	if err := tbf.Finalize(); err != nil {
+		putTmpBlocksFile(tbf)
+		putStorageSearch(sr)
+		return nil, fmt.Errorf("cannot finalize temporary file: %w", err)
+	}

 	var rss Results
 	rss.tr = tr
@@ -753,9 +892,15 @@ func ProcessSearchQuery(sq *storage.SearchQuery, fetchData bool, deadline search
 	}
 	rss.packedTimeseries = pts
 	rss.sr = sr
+	rss.tbf = tbf
 	return &rss, nil
 }

+type blockRef struct {
+	partRef storage.PartRef
+	addr    tmpBlockAddr
+}
+
 func setupTfss(tagFilterss [][]storage.TagFilter) ([]*storage.TagFilters, error) {
 	tfss := make([]*storage.TagFilters, 0, len(tagFilterss))
 	for _, tagFilters := range tagFilterss {
@@ -771,3 +916,20 @@ func setupTfss(tagFilterss [][]storage.TagFilter) ([]*storage.TagFilters, error)
 	}
 	return tfss, nil
 }
+
+func applyGraphiteRegexpFilter(filter string, ss []string) ([]string, error) {
+	// Anchor filter regexp to the beginning of the string as Graphite does.
+	// See https://github.com/graphite-project/graphite-web/blob/3ad279df5cb90b211953e39161df416e54a84948/webapp/graphite/tags/localdatabase.py#L157
+	filter = "^(?:" + filter + ")"
+	re, err := regexp.Compile(filter)
+	if err != nil {
+		return nil, fmt.Errorf("cannot parse regexp filter=%q: %w", filter, err)
+	}
+	dst := ss[:0]
+	for _, s := range ss {
+		if re.MatchString(s) {
+			dst = append(dst, s)
+		}
+	}
+	return dst, nil
+}
--- a/app/vmselect/netstorage/tmp_blocks_file.go
+++ b/app/vmselect/netstorage/tmp_blocks_file.go
@@ -0,0 +1,185 @@
+package netstorage
+
+import (
+	"fmt"
+	"io/ioutil"
+	"os"
+	"sync"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/memory"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/metrics"
+)
+
+// InitTmpBlocksDir initializes directory to store temporary search results.
+//
+// It stores data in system-defined temporary directory if tmpDirPath is empty.
+func InitTmpBlocksDir(tmpDirPath string) {
+	if len(tmpDirPath) == 0 {
+		tmpDirPath = os.TempDir()
+	}
+	tmpBlocksDir = tmpDirPath + "/searchResults"
+	fs.MustRemoveAll(tmpBlocksDir)
+	if err := fs.MkdirAllIfNotExist(tmpBlocksDir); err != nil {
+		logger.Panicf("FATAL: cannot create %q: %s", tmpBlocksDir, err)
+	}
+}
+
+var tmpBlocksDir string
+
+func maxInmemoryTmpBlocksFile() int {
+	mem := memory.Allowed()
+	maxLen := mem / 1024
+	if maxLen < 64*1024 {
+		return 64 * 1024
+	}
+	if maxLen > 4*1024*1024 {
+		return 4 * 1024 * 1024
+	}
+	return maxLen
+}
+
+var _ = metrics.NewGauge(`vm_tmp_blocks_max_inmemory_file_size_bytes`, func() float64 {
+	return float64(maxInmemoryTmpBlocksFile())
+})
+
+type tmpBlocksFile struct {
+	buf []byte
+
+	f *os.File
+	r *fs.ReaderAt
+
+	offset uint64
+}
+
+func getTmpBlocksFile() *tmpBlocksFile {
+	v := tmpBlocksFilePool.Get()
+	if v == nil {
+		return &tmpBlocksFile{
+			buf: make([]byte, 0, maxInmemoryTmpBlocksFile()),
+		}
+	}
+	return v.(*tmpBlocksFile)
+}
+
+func putTmpBlocksFile(tbf *tmpBlocksFile) {
+	tbf.MustClose()
+	tbf.buf = tbf.buf[:0]
+	tbf.f = nil
+	tbf.r = nil
+	tbf.offset = 0
+	tmpBlocksFilePool.Put(tbf)
+}
+
+var tmpBlocksFilePool sync.Pool
+
+type tmpBlockAddr struct {
+	offset uint64
+	size   int
+}
+
+func (addr tmpBlockAddr) String() string {
+	return fmt.Sprintf("offset %d, size %d", addr.offset, addr.size)
+}
+
+var (
+	tmpBlocksFilesCreated = metrics.NewCounter(`vm_tmp_blocks_files_created_total`)
+	_                     = metrics.NewGauge(`vm_tmp_blocks_files_directory_free_bytes`, func() float64 {
+		return float64(fs.MustGetFreeSpace(tmpBlocksDir))
+	})
+)
+
+// WriteBlockRefData writes br to tbf.
+//
+// It returns errors since the operation may fail on space shortage
+// and this must be handled.
+func (tbf *tmpBlocksFile) WriteBlockRefData(b []byte) (tmpBlockAddr, error) {
+	var addr tmpBlockAddr
+	addr.offset = tbf.offset
+	addr.size = len(b)
+	tbf.offset += uint64(addr.size)
+	if len(tbf.buf)+len(b) <= cap(tbf.buf) {
+		// Fast path - the data fits tbf.buf
+		tbf.buf = append(tbf.buf, b...)
+		return addr, nil
+	}
+
+	// Slow path: flush the data from tbf.buf to file.
+	if tbf.f == nil {
+		f, err := ioutil.TempFile(tmpBlocksDir, "")
+		if err != nil {
+			return addr, err
+		}
+		tbf.f = f
+		tmpBlocksFilesCreated.Inc()
+	}
+	_, err := tbf.f.Write(tbf.buf)
+	tbf.buf = append(tbf.buf[:0], b...)
+	if err != nil {
+		return addr, fmt.Errorf("cannot write block to %q: %w", tbf.f.Name(), err)
+	}
+	return addr, nil
+}
+
+func (tbf *tmpBlocksFile) Finalize() error {
+	if tbf.f == nil {
+		return nil
+	}
+	fname := tbf.f.Name()
+	if _, err := tbf.f.Write(tbf.buf); err != nil {
+		return fmt.Errorf("cannot write the remaining %d bytes to %q: %w", len(tbf.buf), fname, err)
+	}
+	tbf.buf = tbf.buf[:0]
+	r := fs.MustOpenReaderAt(fname)
+	// Hint the OS that the file is read almost sequentiallly.
+	// This should reduce the number of disk seeks, which is important
+	// for HDDs.
+	r.MustFadviseSequentialRead(true)
+	tbf.r = r
+	return nil
+}
+
+func (tbf *tmpBlocksFile) MustReadBlockRefAt(partRef storage.PartRef, addr tmpBlockAddr) storage.BlockRef {
+	var buf []byte
+	if tbf.f == nil {
+		buf = tbf.buf[addr.offset : addr.offset+uint64(addr.size)]
+	} else {
+		bb := tmpBufPool.Get()
+		defer tmpBufPool.Put(bb)
+		bb.B = bytesutil.Resize(bb.B, addr.size)
+		tbf.r.MustReadAt(bb.B, int64(addr.offset))
+		buf = bb.B
+	}
+	var br storage.BlockRef
+	if err := br.Init(partRef, buf); err != nil {
+		logger.Panicf("FATAL: cannot initialize BlockRef: %s", err)
+	}
+	return br
+}
+
+var tmpBufPool bytesutil.ByteBufferPool
+
+func (tbf *tmpBlocksFile) MustClose() {
+	if tbf.f == nil {
+		return
+	}
+	if tbf.r != nil {
+		// tbf.r could be nil if Finalize wasn't called.
+		tbf.r.MustClose()
+	}
+	fname := tbf.f.Name()
+
+	// Remove the file at first, then close it.
+	// This way the OS shouldn't try to flush file contents to storage
+	// on close.
+	if err := os.Remove(fname); err != nil {
+		logger.Panicf("FATAL: cannot remove %q: %s", fname, err)
+	}
+	if err := tbf.f.Close(); err != nil {
+		logger.Panicf("FATAL: cannot close %q: %s", fname, err)
+	}
+	tbf.f = nil
+}
--- a/app/vmselect/prometheus/prometheus.go
+++ b/app/vmselect/prometheus/prometheus.go
@@ -78,17 +78,13 @@ func FederateHandler(startTime time.Time, w http.ResponseWriter, r *http.Request
 	if err != nil {
 		return err
 	}
-	sq := &storage.SearchQuery{
-		MinTimestamp: start,
-		MaxTimestamp: end,
-		TagFilterss:  tagFilterss,
-	}
+	sq := storage.NewSearchQuery(start, end, tagFilterss)
 	rss, err := netstorage.ProcessSearchQuery(sq, true, deadline)
 	if err != nil {
 		return fmt.Errorf("cannot fetch data for %q: %w", sq, err)
 	}

-	w.Header().Set("Content-Type", "text/plain")
+	w.Header().Set("Content-Type", "text/plain; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
 	err = rss.RunParallel(func(rs *netstorage.Result, workerID uint) error {
@@ -146,12 +142,8 @@ func ExportCSVHandler(startTime time.Time, w http.ResponseWriter, r *http.Reques
 	if err != nil {
 		return err
 	}
-	sq := &storage.SearchQuery{
-		MinTimestamp: start,
-		MaxTimestamp: end,
-		TagFilterss:  tagFilterss,
-	}
-	w.Header().Set("Content-Type", "text/csv")
+	sq := storage.NewSearchQuery(start, end, tagFilterss)
+	w.Header().Set("Content-Type", "text/csv; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)

@@ -227,11 +219,7 @@ func ExportNativeHandler(startTime time.Time, w http.ResponseWriter, r *http.Req
 	if err != nil {
 		return err
 	}
-	sq := &storage.SearchQuery{
-		MinTimestamp: start,
-		MaxTimestamp: end,
-		TagFilterss:  tagFilterss,
-	}
+	sq := storage.NewSearchQuery(start, end, tagFilterss)
 	w.Header().Set("Content-Type", "VictoriaMetrics/native")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
@@ -331,9 +319,9 @@ func exportHandler(w http.ResponseWriter, matches []string, start, end int64, fo
 		WriteExportJSONLine(bb, xb)
 		resultsCh <- bb
 	}
-	contentType := "application/stream+json"
+	contentType := "application/stream+json; charset=utf-8"
 	if format == "prometheus" {
-		contentType = "text/plain"
+		contentType = "text/plain; charset=utf-8"
 		writeLineFunc = func(xb *exportBlock, resultsCh chan<- *quicktemplate.ByteBuffer) {
 			bb := quicktemplate.AcquireByteBuffer()
 			WriteExportPrometheusLine(bb, xb)
@@ -381,11 +369,7 @@ func exportHandler(w http.ResponseWriter, matches []string, start, end int64, fo
 	if err != nil {
 		return err
 	}
-	sq := &storage.SearchQuery{
-		MinTimestamp: start,
-		MaxTimestamp: end,
-		TagFilterss:  tagFilterss,
-	}
+	sq := storage.NewSearchQuery(start, end, tagFilterss)
 	w.Header().Set("Content-Type", contentType)
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
@@ -486,9 +470,7 @@ func DeleteHandler(startTime time.Time, r *http.Request) error {
 	if err != nil {
 		return err
 	}
-	sq := &storage.SearchQuery{
-		TagFilterss: tagFilterss,
-	}
+	sq := storage.NewSearchQuery(0, 0, tagFilterss)
 	deletedCount, err := netstorage.DeleteSeries(sq)
 	if err != nil {
 		return fmt.Errorf("cannot delete time series matching %q: %w", matches, err)
@@ -511,11 +493,31 @@ func LabelValuesHandler(startTime time.Time, labelName string, w http.ResponseWr
 		return fmt.Errorf("cannot parse form values: %w", err)
 	}
 	var labelValues []string
-	if len(r.Form["match[]"]) == 0 && len(r.Form["start"]) == 0 && len(r.Form["end"]) == 0 {
-		var err error
-		labelValues, err = netstorage.GetLabelValues(labelName, deadline)
-		if err != nil {
-			return fmt.Errorf(`cannot obtain label values for %q: %w`, labelName, err)
+	if len(r.Form["match[]"]) == 0 {
+		if len(r.Form["start"]) == 0 && len(r.Form["end"]) == 0 {
+			var err error
+			labelValues, err = netstorage.GetLabelValues(labelName, deadline)
+			if err != nil {
+				return fmt.Errorf(`cannot obtain label values for %q: %w`, labelName, err)
+			}
+		} else {
+			ct := startTime.UnixNano() / 1e6
+			end, err := searchutils.GetTime(r, "end", ct)
+			if err != nil {
+				return err
+			}
+			start, err := searchutils.GetTime(r, "start", end-defaultStep)
+			if err != nil {
+				return err
+			}
+			tr := storage.TimeRange{
+				MinTimestamp: start,
+				MaxTimestamp: end,
+			}
+			labelValues, err = netstorage.GetLabelValuesOnTimeRange(labelName, tr, deadline)
+			if err != nil {
+				return fmt.Errorf(`cannot obtain label values on time range for %q: %w`, labelName, err)
+			}
 		}
 	} else {
 		// Extended functionality that allows filtering by label filters and time range
@@ -541,7 +543,7 @@ func LabelValuesHandler(startTime time.Time, labelName string, w http.ResponseWr
 		}
 	}

-	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
 	WriteLabelValuesResponse(bw, labelValues)
@@ -576,32 +578,41 @@ func labelValuesWithMatches(labelName string, matches []string, start, end int64
 	if start >= end {
 		end = start + defaultStep
 	}
-	sq := &storage.SearchQuery{
-		MinTimestamp: start,
-		MaxTimestamp: end,
-		TagFilterss:  tagFilterss,
-	}
-	rss, err := netstorage.ProcessSearchQuery(sq, false, deadline)
-	if err != nil {
-		return nil, fmt.Errorf("cannot fetch data for %q: %w", sq, err)
-	}
-
+	sq := storage.NewSearchQuery(start, end, tagFilterss)
 	m := make(map[string]struct{})
-	var mLock sync.Mutex
-	err = rss.RunParallel(func(rs *netstorage.Result, workerID uint) error {
-		labelValue := rs.MetricName.GetTagValue(labelName)
-		if len(labelValue) == 0 {
-			return nil
+	if end-start > 24*3600*1000 {
+		// It is cheaper to call SearchMetricNames on time ranges exceeding a day.
+		mns, err := netstorage.SearchMetricNames(sq, deadline)
+		if err != nil {
+			return nil, fmt.Errorf("cannot fetch time series for %q: %w", sq, err)
+		}
+		for _, mn := range mns {
+			labelValue := mn.GetTagValue(labelName)
+			if len(labelValue) == 0 {
+				continue
+			}
+			m[string(labelValue)] = struct{}{}
+		}
+	} else {
+		rss, err := netstorage.ProcessSearchQuery(sq, false, deadline)
+		if err != nil {
+			return nil, fmt.Errorf("cannot fetch data for %q: %w", sq, err)
+		}
+		var mLock sync.Mutex
+		err = rss.RunParallel(func(rs *netstorage.Result, workerID uint) error {
+			labelValue := rs.MetricName.GetTagValue(labelName)
+			if len(labelValue) == 0 {
+				return nil
+			}
+			mLock.Lock()
+			m[string(labelValue)] = struct{}{}
+			mLock.Unlock()
+			return nil
+		})
+		if err != nil {
+			return nil, fmt.Errorf("error when data fetching: %w", err)
 		}
-		mLock.Lock()
-		m[string(labelValue)] = struct{}{}
-		mLock.Unlock()
-		return nil
-	})
-	if err != nil {
-		return nil, fmt.Errorf("error when data fetching: %w", err)
 	}
-
 	labelValues := make([]string, 0, len(m))
 	for labelValue := range m {
 		labelValues = append(labelValues, labelValue)
@@ -619,7 +630,7 @@ func LabelsCountHandler(startTime time.Time, w http.ResponseWriter, r *http.Requ
 	if err != nil {
 		return fmt.Errorf(`cannot obtain label entries: %w`, err)
 	}
-	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
 	WriteLabelsCountResponse(bw, labelEntries)
@@ -670,7 +681,7 @@ func TSDBStatusHandler(startTime time.Time, w http.ResponseWriter, r *http.Reque
 	if err != nil {
 		return fmt.Errorf(`cannot obtain tsdb status for date=%d, topN=%d: %w`, date, topN, err)
 	}
-	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
 	WriteTSDBStatusResponse(bw, status)
@@ -692,11 +703,31 @@ func LabelsHandler(startTime time.Time, w http.ResponseWriter, r *http.Request)
 		return fmt.Errorf("cannot parse form values: %w", err)
 	}
 	var labels []string
-	if len(r.Form["match[]"]) == 0 && len(r.Form["start"]) == 0 && len(r.Form["end"]) == 0 {
-		var err error
-		labels, err = netstorage.GetLabels(deadline)
-		if err != nil {
-			return fmt.Errorf("cannot obtain labels: %w", err)
+	if len(r.Form["match[]"]) == 0 {
+		if len(r.Form["start"]) == 0 && len(r.Form["end"]) == 0 {
+			var err error
+			labels, err = netstorage.GetLabels(deadline)
+			if err != nil {
+				return fmt.Errorf("cannot obtain labels: %w", err)
+			}
+		} else {
+			ct := startTime.UnixNano() / 1e6
+			end, err := searchutils.GetTime(r, "end", ct)
+			if err != nil {
+				return err
+			}
+			start, err := searchutils.GetTime(r, "start", end-defaultStep)
+			if err != nil {
+				return err
+			}
+			tr := storage.TimeRange{
+				MinTimestamp: start,
+				MaxTimestamp: end,
+			}
+			labels, err = netstorage.GetLabelsOnTimeRange(tr, deadline)
+			if err != nil {
+				return fmt.Errorf("cannot obtain labels on time range: %w", err)
+			}
 		}
 	} else {
 		// Extended functionality that allows filtering by label filters and time range
@@ -720,7 +751,7 @@ func LabelsHandler(startTime time.Time, w http.ResponseWriter, r *http.Request)
 		}
 	}

-	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
 	WriteLabelsResponse(bw, labels)
@@ -742,33 +773,41 @@ func labelsWithMatches(matches []string, start, end int64, deadline searchutils.
 	if start >= end {
 		end = start + defaultStep
 	}
-	sq := &storage.SearchQuery{
-		MinTimestamp: start,
-		MaxTimestamp: end,
-		TagFilterss:  tagFilterss,
-	}
-	rss, err := netstorage.ProcessSearchQuery(sq, false, deadline)
-	if err != nil {
-		return nil, fmt.Errorf("cannot fetch data for %q: %w", sq, err)
-	}
-
+	sq := storage.NewSearchQuery(start, end, tagFilterss)
 	m := make(map[string]struct{})
-	var mLock sync.Mutex
-	err = rss.RunParallel(func(rs *netstorage.Result, workerID uint) error {
-		mLock.Lock()
-		tags := rs.MetricName.Tags
-		for i := range tags {
-			t := &tags[i]
-			m[string(t.Key)] = struct{}{}
+	if end-start > 24*3600*1000 {
+		// It is cheaper to call SearchMetricNames on time ranges exceeding a day.
+		mns, err := netstorage.SearchMetricNames(sq, deadline)
+		if err != nil {
+			return nil, fmt.Errorf("cannot fetch time series for %q: %w", sq, err)
+		}
+		for _, mn := range mns {
+			for _, tag := range mn.Tags {
+				m[string(tag.Key)] = struct{}{}
+			}
+		}
+		if len(mns) > 0 {
+			m["__name__"] = struct{}{}
+		}
+	} else {
+		rss, err := netstorage.ProcessSearchQuery(sq, false, deadline)
+		if err != nil {
+			return nil, fmt.Errorf("cannot fetch data for %q: %w", sq, err)
+		}
+		var mLock sync.Mutex
+		err = rss.RunParallel(func(rs *netstorage.Result, workerID uint) error {
+			mLock.Lock()
+			for _, tag := range rs.MetricName.Tags {
+				m[string(tag.Key)] = struct{}{}
+			}
+			m["__name__"] = struct{}{}
+			mLock.Unlock()
+			return nil
+		})
+		if err != nil {
+			return nil, fmt.Errorf("error when data fetching: %w", err)
 		}
-		m["__name__"] = struct{}{}
-		mLock.Unlock()
-		return nil
-	})
-	if err != nil {
-		return nil, fmt.Errorf("error when data fetching: %w", err)
 	}
-
 	labels := make([]string, 0, len(m))
 	for label := range m {
 		labels = append(labels, label)
@@ -786,7 +825,7 @@ func SeriesCountHandler(startTime time.Time, w http.ResponseWriter, r *http.Requ
 	if err != nil {
 		return fmt.Errorf("cannot obtain series count: %w", err)
 	}
-	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
 	WriteSeriesCountResponse(bw, n)
@@ -833,17 +872,39 @@ func SeriesHandler(startTime time.Time, w http.ResponseWriter, r *http.Request)
 	if start >= end {
 		end = start + defaultStep
 	}
-	sq := &storage.SearchQuery{
-		MinTimestamp: start,
-		MaxTimestamp: end,
-		TagFilterss:  tagFilterss,
+	sq := storage.NewSearchQuery(start, end, tagFilterss)
+	if end-start > 24*3600*1000 {
+		// It is cheaper to call SearchMetricNames on time ranges exceeding a day.
+		mns, err := netstorage.SearchMetricNames(sq, deadline)
+		if err != nil {
+			return fmt.Errorf("cannot fetch time series for %q: %w", sq, err)
+		}
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
+		bw := bufferedwriter.Get(w)
+		defer bufferedwriter.Put(bw)
+		resultsCh := make(chan *quicktemplate.ByteBuffer)
+		go func() {
+			for i := range mns {
+				bb := quicktemplate.AcquireByteBuffer()
+				writemetricNameObject(bb, &mns[i])
+				resultsCh <- bb
+			}
+			close(resultsCh)
+		}()
+		// WriteSeriesResponse must consume all the data from resultsCh.
+		WriteSeriesResponse(bw, resultsCh)
+		if err := bw.Flush(); err != nil {
+			return err
+		}
+		seriesDuration.UpdateDuration(startTime)
+		return nil
 	}
 	rss, err := netstorage.ProcessSearchQuery(sq, false, deadline)
 	if err != nil {
 		return fmt.Errorf("cannot fetch data for %q: %w", sq, err)
 	}

-	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
 	resultsCh := make(chan *quicktemplate.ByteBuffer)
@@ -980,7 +1041,7 @@ func QueryHandler(startTime time.Time, w http.ResponseWriter, r *http.Request) e
 		}
 	}

-	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
 	WriteQueryResponse(bw, result)
@@ -1079,7 +1140,7 @@ func queryRangeHandler(startTime time.Time, w http.ResponseWriter, query string,
 	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/153
 	result = removeEmptyValuesAndTimeseries(result)

-	w.Header().Set("Content-Type", "application/json")
+	w.Header().Set("Content-Type", "application/json; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
 	WriteQueryRangeResponse(bw, result)
--- a/app/vmselect/promql/aggr.go
+++ b/app/vmselect/promql/aggr.go
@@ -69,7 +69,9 @@ func newAggrFunc(afe func(tss []*timeseries) []*timeseries) aggrFunc {
 		if err != nil {
 			return nil, err
 		}
-		return aggrFuncExt(afe, tss, &afa.ae.Modifier, afa.ae.Limit, false)
+		return aggrFuncExt(func(tss []*timeseries, modififer *metricsql.ModifierExpr) []*timeseries {
+			return afe(tss)
+		}, tss, &afa.ae.Modifier, afa.ae.Limit, false)
 	}
 }

@@ -98,7 +100,8 @@ func removeGroupTags(metricName *storage.MetricName, modifier *metricsql.Modifie
 	}
 }

-func aggrFuncExt(afe func(tss []*timeseries) []*timeseries, argOrig []*timeseries, modifier *metricsql.ModifierExpr, maxSeries int, keepOriginal bool) ([]*timeseries, error) {
+func aggrFuncExt(afe func(tss []*timeseries, modifier *metricsql.ModifierExpr) []*timeseries, argOrig []*timeseries,
+	modifier *metricsql.ModifierExpr, maxSeries int, keepOriginal bool) ([]*timeseries, error) {
 	arg := copyTimeseriesMetricNames(argOrig, keepOriginal)

 	// Perform grouping.
@@ -124,7 +127,7 @@ func aggrFuncExt(afe func(tss []*timeseries) []*timeseries, argOrig []*timeserie
 	dstTssCount := 0
 	rvs := make([]*timeseries, 0, len(m))
 	for _, tss := range m {
-		rv := afe(tss)
+		rv := afe(tss, modifier)
 		rvs = append(rvs, rv...)
 		srcTssCount += len(tss)
 		dstTssCount += len(rv)
@@ -141,7 +144,7 @@ func aggrFuncAny(afa *aggrFuncArg) ([]*timeseries, error) {
 	if err != nil {
 		return nil, err
 	}
-	afe := func(tss []*timeseries) []*timeseries {
+	afe := func(tss []*timeseries, modifier *metricsql.ModifierExpr) []*timeseries {
 		return tss[:1]
 	}
 	limit := afa.ae.Limit
@@ -178,10 +181,11 @@ func aggrFuncSum(tss []*timeseries) []*timeseries {
 		sum := float64(0)
 		count := 0
 		for _, ts := range tss {
-			if math.IsNaN(ts.Values[i]) {
+			v := ts.Values[i]
+			if math.IsNaN(v) {
 				continue
 			}
-			sum += ts.Values[i]
+			sum += v
 			count++
 		}
 		if count == 0 {
@@ -449,7 +453,7 @@ func aggrFuncZScore(afa *aggrFuncArg) ([]*timeseries, error) {
 	if err != nil {
 		return nil, err
 	}
-	afe := func(tss []*timeseries) []*timeseries {
+	afe := func(tss []*timeseries, modifier *metricsql.ModifierExpr) []*timeseries {
 		for i := range tss[0].Values {
 			// Calculate avg and stddev for tss points at position i.
 			// See `Rapid calculation methods` at https://en.wikipedia.org/wiki/Standard_deviation
@@ -550,7 +554,7 @@ func aggrFuncCountValues(afa *aggrFuncArg) ([]*timeseries, error) {
 		// Do nothing
 	}

-	afe := func(tss []*timeseries) []*timeseries {
+	afe := func(tss []*timeseries, modififer *metricsql.ModifierExpr) []*timeseries {
 		m := make(map[float64]bool)
 		for _, ts := range tss {
 			for _, v := range ts.Values {
@@ -602,7 +606,7 @@ func newAggrFuncTopK(isReverse bool) aggrFunc {
 		if err != nil {
 			return nil, err
 		}
-		afe := func(tss []*timeseries) []*timeseries {
+		afe := func(tss []*timeseries, modififer *metricsql.ModifierExpr) []*timeseries {
 			for n := range tss[0].Values {
 				sort.Slice(tss, func(i, j int) bool {
 					a := tss[i].Values[n]
@@ -623,21 +627,32 @@ func newAggrFuncTopK(isReverse bool) aggrFunc {
 func newAggrFuncRangeTopK(f func(values []float64) float64, isReverse bool) aggrFunc {
 	return func(afa *aggrFuncArg) ([]*timeseries, error) {
 		args := afa.args
-		if err := expectTransformArgsNum(args, 2); err != nil {
-			return nil, err
+		if len(args) < 2 {
+			return nil, fmt.Errorf(`unexpected number of args; got %d; want at least %d`, len(args), 2)
+		}
+		if len(args) > 3 {
+			return nil, fmt.Errorf(`unexpected number of args; got %d; want no more than %d`, len(args), 3)
 		}
 		ks, err := getScalar(args[0], 0)
 		if err != nil {
 			return nil, err
 		}
-		afe := func(tss []*timeseries) []*timeseries {
-			return getRangeTopKTimeseries(tss, ks, f, isReverse)
+		remainingSumTagName := ""
+		if len(args) == 3 {
+			remainingSumTagName, err = getString(args[2], 2)
+			if err != nil {
+				return nil, err
+			}
+		}
+		afe := func(tss []*timeseries, modifier *metricsql.ModifierExpr) []*timeseries {
+			return getRangeTopKTimeseries(tss, modifier, ks, remainingSumTagName, f, isReverse)
 		}
 		return aggrFuncExt(afe, args[1], &afa.ae.Modifier, afa.ae.Limit, true)
 	}
 }

-func getRangeTopKTimeseries(tss []*timeseries, ks []float64, f func(values []float64) float64, isReverse bool) []*timeseries {
+func getRangeTopKTimeseries(tss []*timeseries, modifier *metricsql.ModifierExpr, ks []float64, remainingSumTagName string,
+	f func(values []float64) float64, isReverse bool) []*timeseries {
 	type tsWithValue struct {
 		ts    *timeseries
 		value float64
@@ -661,35 +676,74 @@ func getRangeTopKTimeseries(tss []*timeseries, ks []float64, f func(values []flo
 	for i := range maxs {
 		tss[i] = maxs[i].ts
 	}
+	remainingSumTS := getRemainingSumTimeseries(tss, modifier, ks, remainingSumTagName)
 	for i, k := range ks {
 		fillNaNsAtIdx(i, k, tss)
 	}
+	if remainingSumTS != nil {
+		tss = append(tss, remainingSumTS)
+	}
 	return removeNaNs(tss)
 }

+func getRemainingSumTimeseries(tss []*timeseries, modifier *metricsql.ModifierExpr, ks []float64, remainingSumTagName string) *timeseries {
+	if len(remainingSumTagName) == 0 || len(tss) == 0 {
+		return nil
+	}
+	var dst timeseries
+	dst.CopyFromShallowTimestamps(tss[0])
+	removeGroupTags(&dst.MetricName, modifier)
+	dst.MetricName.RemoveTag(remainingSumTagName)
+	dst.MetricName.AddTag(remainingSumTagName, remainingSumTagName)
+	for i, k := range ks {
+		kn := getIntK(k, len(tss))
+		var sum float64
+		count := 0
+		for _, ts := range tss[:len(tss)-kn] {
+			v := ts.Values[i]
+			if math.IsNaN(v) {
+				continue
+			}
+			sum += v
+			count++
+		}
+		if count == 0 {
+			sum = nan
+		}
+		dst.Values[i] = sum
+	}
+	return &dst
+}
+
 func fillNaNsAtIdx(idx int, k float64, tss []*timeseries) {
-	if math.IsNaN(k) {
-		k = 0
-	}
-	kn := int(k)
-	if kn < 0 {
-		kn = 0
-	}
-	if kn > len(tss) {
-		kn = len(tss)
-	}
+	kn := getIntK(k, len(tss))
 	for _, ts := range tss[:len(tss)-kn] {
 		ts.Values[idx] = nan
 	}
 }

-func minValue(values []float64) float64 {
-	if len(values) == 0 {
-		return nan
+func getIntK(k float64, kMax int) int {
+	if math.IsNaN(k) {
+		return 0
 	}
-	min := values[0]
-	for _, v := range values[1:] {
-		if v < min {
+	kn := int(k)
+	if kn < 0 {
+		return 0
+	}
+	if kn > kMax {
+		return kMax
+	}
+	return kn
+}
+
+func minValue(values []float64) float64 {
+	min := nan
+	for len(values) > 0 && math.IsNaN(min) {
+		min = values[0]
+		values = values[1:]
+	}
+	for _, v := range values {
+		if !math.IsNaN(v) && v < min {
 			min = v
 		}
 	}
@@ -697,12 +751,13 @@ func minValue(values []float64) float64 {
 }

 func maxValue(values []float64) float64 {
-	if len(values) == 0 {
-		return nan
+	max := nan
+	for len(values) > 0 && math.IsNaN(max) {
+		max = values[0]
+		values = values[1:]
 	}
-	max := values[0]
-	for _, v := range values[1:] {
-		if v > max {
+	for _, v := range values {
+		if !math.IsNaN(v) && v > max {
 			max = v
 		}
 	}
@@ -746,7 +801,7 @@ func aggrFuncOutliersK(afa *aggrFuncArg) ([]*timeseries, error) {
 	if err != nil {
 		return nil, err
 	}
-	afe := func(tss []*timeseries) []*timeseries {
+	afe := func(tss []*timeseries, modifier *metricsql.ModifierExpr) []*timeseries {
 		// Calculate medians for each point across tss.
 		medians := make([]float64, len(ks))
 		h := histogram.GetFast()
@@ -771,7 +826,7 @@ func aggrFuncOutliersK(afa *aggrFuncArg) ([]*timeseries, error) {
 			}
 			return sum2
 		}
-		return getRangeTopKTimeseries(tss, ks, f, false)
+		return getRangeTopKTimeseries(tss, &afa.ae.Modifier, ks, "", f, false)
 	}
 	return aggrFuncExt(afe, args[1], &afa.ae.Modifier, afa.ae.Limit, true)
 }
@@ -792,7 +847,7 @@ func aggrFuncLimitK(afa *aggrFuncArg) ([]*timeseries, error) {
 			maxK = k
 		}
 	}
-	afe := func(tss []*timeseries) []*timeseries {
+	afe := func(tss []*timeseries, modifier *metricsql.ModifierExpr) []*timeseries {
 		if len(tss) > maxK {
 			tss = tss[:maxK]
 		}
@@ -833,8 +888,8 @@ func aggrFuncMedian(afa *aggrFuncArg) ([]*timeseries, error) {
 	return aggrFuncExt(afe, tss, &afa.ae.Modifier, afa.ae.Limit, false)
 }

-func newAggrQuantileFunc(phis []float64) func(tss []*timeseries) []*timeseries {
-	return func(tss []*timeseries) []*timeseries {
+func newAggrQuantileFunc(phis []float64) func(tss []*timeseries, modifier *metricsql.ModifierExpr) []*timeseries {
+	return func(tss []*timeseries, modifier *metricsql.ModifierExpr) []*timeseries {
 		dst := tss[0]
 		h := histogram.GetFast()
 		defer histogram.PutFast(h)
--- a/app/vmselect/promql/binary_op.go
+++ b/app/vmselect/promql/binary_op.go
@@ -62,6 +62,9 @@ func newBinaryOpCmpFunc(cf func(left, right float64) bool) binaryOpFunc {
 		if cf(left, right) {
 			return 1
 		}
+		if math.IsNaN(left) {
+			return nan
+		}
 		return 0
 	}
 	return newBinaryOpFunc(cfe)
--- a/app/vmselect/promql/eval.go
+++ b/app/vmselect/promql/eval.go
@@ -653,11 +653,7 @@ func evalRollupFuncWithMetricExpr(ec *EvalConfig, name string, rf rollupFunc,
 	} else {
 		minTimestamp -= ec.Step
 	}
-	sq := &storage.SearchQuery{
-		MinTimestamp: minTimestamp,
-		MaxTimestamp: ec.End,
-		TagFilterss:  [][]storage.TagFilter{tfs},
-	}
+	sq := storage.NewSearchQuery(minTimestamp, ec.End, [][]storage.TagFilter{tfs})
 	rss, err := netstorage.ProcessSearchQuery(sq, true, ec.Deadline)
 	if err != nil {
 		return nil, err
--- a/app/vmselect/promql/exec.go
+++ b/app/vmselect/promql/exec.go
@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"math"
 	"sort"
+	"strings"
 	"sync"
 	"sync/atomic"
 	"time"
@@ -15,7 +16,13 @@ import (
 	"github.com/VictoriaMetrics/metricsql"
 )

-var logSlowQueryDuration = flag.Duration("search.logSlowQueryDuration", 5*time.Second, "Log queries with execution time exceeding this value. Zero disables slow query logging")
+var (
+	logSlowQueryDuration   = flag.Duration("search.logSlowQueryDuration", 5*time.Second, "Log queries with execution time exceeding this value. Zero disables slow query logging")
+	treatDotsAsIsInRegexps = flag.Bool("search.treatDotsAsIsInRegexps", false, "Whether to treat dots as is in regexp label filters used in queries. "+
+		`For example, foo{bar=~"a.b.c"} will be automatically converted to foo{bar=~"a\\.b\\.c"}, i.e. all the dots in regexp filters will be automatically escaped `+
+		`in order to match only dot char instead of matching any char. Dots in ".+", ".*" and ".{n}" regexps aren't escaped. `+
+		`Such escaping can be useful when querying Graphite data`)
+)

 var slowQueries = metrics.NewCounter(`vm_slow_queries_total`)

@@ -26,8 +33,8 @@ func Exec(ec *EvalConfig, q string, isFirstPointOnly bool) ([]netstorage.Result,
 		defer func() {
 			d := time.Since(startTime)
 			if d >= *logSlowQueryDuration {
-				logger.Warnf("slow query according to -search.logSlowQueryDuration=%s: duration=%.3f seconds, start=%d, end=%d, step=%d, query=%q",
-					*logSlowQueryDuration, d.Seconds(), ec.Start/1000, ec.End/1000, ec.Step/1000, q)
+				logger.Warnf("slow query according to -search.logSlowQueryDuration=%s: remoteAddr=%s, duration=%.3f seconds, start=%d, end=%d, step=%d, query=%q",
+					*logSlowQueryDuration, ec.QuotedRemoteAddr, d.Seconds(), ec.Start/1000, ec.End/1000, ec.Step/1000, q)
 				slowQueries.Inc()
 			}
 		}()
@@ -142,11 +149,11 @@ func adjustCmpOps(e metricsql.Expr) metricsql.Expr {
 		if !metricsql.IsBinaryOpCmp(be.Op) {
 			return
 		}
-		if _, ok := be.Left.(*metricsql.NumberExpr); !ok {
+		if isNumberExpr(be.Right) || !isScalarExpr(be.Left) {
 			return
 		}
 		// Convert 'num cmpOp query' expression to `query reverseCmpOp num` expression
-		// like Prometheus does. For isntance, `0.5 < foo` must be converted to `foo > 0.5`
+		// like Prometheus does. For instance, `0.5 < foo` must be converted to `foo > 0.5`
 		// in order to return valid values for `foo` that are bigger than 0.5.
 		be.Right, be.Left = be.Left, be.Right
 		be.Op = getReverseCmpOp(be.Op)
@@ -154,6 +161,22 @@ func adjustCmpOps(e metricsql.Expr) metricsql.Expr {
 	return e
 }

+func isNumberExpr(e metricsql.Expr) bool {
+	_, ok := e.(*metricsql.NumberExpr)
+	return ok
+}
+
+func isScalarExpr(e metricsql.Expr) bool {
+	if isNumberExpr(e) {
+		return true
+	}
+	if fe, ok := e.(*metricsql.FuncExpr); ok {
+		// time() returns scalar in PromQL - see https://prometheus.io/docs/prometheus/latest/querying/functions/#time
+		return strings.ToLower(fe.Name) == "time"
+	}
+	return false
+}
+
 func getReverseCmpOp(op string) string {
 	switch op {
 	case ">":
@@ -177,6 +200,9 @@ func parsePromQLWithCache(q string) (metricsql.Expr, error) {
 		if err == nil {
 			e = metricsql.Optimize(e)
 			e = adjustCmpOps(e)
+			if *treatDotsAsIsInRegexps {
+				e = escapeDotsInRegexpLabelFilters(e)
+			}
 		}
 		pcv = &parseCacheValue{
 			e:   e,
@@ -190,6 +216,41 @@ func parsePromQLWithCache(q string) (metricsql.Expr, error) {
 	return pcv.e, nil
 }

+func escapeDotsInRegexpLabelFilters(e metricsql.Expr) metricsql.Expr {
+	metricsql.VisitAll(e, func(expr metricsql.Expr) {
+		me, ok := expr.(*metricsql.MetricExpr)
+		if !ok {
+			return
+		}
+		for i := range me.LabelFilters {
+			f := &me.LabelFilters[i]
+			if f.IsRegexp {
+				f.Value = escapeDots(f.Value)
+			}
+		}
+	})
+	return e
+}
+
+func escapeDots(s string) string {
+	dotsCount := strings.Count(s, ".")
+	if dotsCount <= 0 {
+		return s
+	}
+	result := make([]byte, 0, len(s)+2*dotsCount)
+	for i := 0; i < len(s); i++ {
+		if s[i] == '.' && (i == 0 || s[i-1] != '\\') && (i+1 == len(s) || i+1 < len(s) && s[i+1] != '*' && s[i+1] != '+' && s[i+1] != '{') {
+			// Escape a dot if the following conditions are met:
+			// - if it isn't escaped already, i.e. if there is no `\` char before the dot.
+			// - if there is no regexp modifiers such as '+', '*' or '{' after the dot.
+			result = append(result, '\\', '.')
+		} else {
+			result = append(result, s[i])
+		}
+	}
+	return string(result)
+}
+
 var parseCacheV = func() *parseCache {
 	pc := &parseCache{
 		m: make(map[string]*parseCacheValue),
--- a/app/vmselect/promql/exec_test.go
+++ b/app/vmselect/promql/exec_test.go
@@ -7,8 +7,46 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/netstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/searchutils"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/metricsql"
 )

+func TestEscapeDots(t *testing.T) {
+	f := func(s, resultExpected string) {
+		t.Helper()
+		result := escapeDots(s)
+		if result != resultExpected {
+			t.Fatalf("unexpected result for escapeDots(%q); got\n%s\nwant\n%s", s, result, resultExpected)
+		}
+	}
+	f("", "")
+	f("a", "a")
+	f("foobar", "foobar")
+	f(".", `\.`)
+	f(".*", `.*`)
+	f(".+", `.+`)
+	f("..", `\.\.`)
+	f("foo.b.{2}ar..+baz.*", `foo\.b.{2}ar\..+baz.*`)
+}
+
+func TestEscapeDotsInRegexpLabelFilters(t *testing.T) {
+	f := func(s, resultExpected string) {
+		t.Helper()
+		e, err := metricsql.Parse(s)
+		if err != nil {
+			t.Fatalf("unexpected error in metricsql.Parse(%q): %s", s, err)
+		}
+		e = escapeDotsInRegexpLabelFilters(e)
+		result := e.AppendString(nil)
+		if string(result) != resultExpected {
+			t.Fatalf("unexpected result for escapeDotsInRegexpLabelFilters(%q);\ngot\n%s\nwant\n%s", s, result, resultExpected)
+		}
+	}
+	f("2", "2")
+	f(`foo.bar + 123`, `foo.bar + 123`)
+	f(`foo{bar=~"baz.xx.yyy"}`, `foo{bar=~"baz\\.xx\\.yyy"}`)
+	f(`foo(a.b{c="d.e",x=~"a.b.+[.a]",y!~"aaa.bb|cc.dd"}) + x.y(1,sum({x=~"aa.bb"}))`, `foo(a.b{c="d.e", x=~"a\\.b.+[\\.a]", y!~"aaa\\.bb|cc\\.dd"}) + x.y(1, sum({x=~"aa\\.bb"}))`)
+}
+
 func TestExecSuccess(t *testing.T) {
 	start := int64(1000e3)
 	end := int64(2000e3)
@@ -488,6 +526,17 @@ func TestExecSuccess(t *testing.T) {
 		resultExpected := []netstorage.Result{r}
 		f(q, resultExpected)
 	})
+	t.Run(`minute(series_with_NaNs)`, func(t *testing.T) {
+		t.Parallel()
+		q := `minute(time() <= 1200 or time() > 1600)`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{16, 20, nan, nan, 30, 33},
+			Timestamps: timestampsExpected,
+		}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
 	t.Run("rate({})", func(t *testing.T) {
 		t.Parallel()
 		q := `rate({})`
@@ -1041,6 +1090,62 @@ func TestExecSuccess(t *testing.T) {
 		resultExpected := []netstorage.Result{r1, r2, r3, r4, r5}
 		f(q, resultExpected)
 	})
+	t.Run(`label_uppercase`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_uppercase(
+			label_set(time(), "foo", "bAr", "XXx", "yyy", "zzz", "abc"),
+			"foo", "XXx", "aaa"
+		)`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{1000, 1200, 1400, 1600, 1800, 2000},
+			Timestamps: timestampsExpected,
+		}
+		r.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("XXx"),
+				Value: []byte("YYY"),
+			},
+			{
+				Key:   []byte("foo"),
+				Value: []byte("BAR"),
+			},
+			{
+				Key:   []byte("zzz"),
+				Value: []byte("abc"),
+			},
+		}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`label_lowercase`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_lowercase(
+			label_set(time(), "foo", "bAr", "XXx", "yyy", "zzz", "aBc"),
+			"foo", "XXx", "aaa"
+		)`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{1000, 1200, 1400, 1600, 1800, 2000},
+			Timestamps: timestampsExpected,
+		}
+		r.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("XXx"),
+				Value: []byte("yyy"),
+			},
+			{
+				Key:   []byte("foo"),
+				Value: []byte("bar"),
+			},
+			{
+				Key:   []byte("zzz"),
+				Value: []byte("aBc"),
+			},
+		}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
 	t.Run(`label_copy(new_tag)`, func(t *testing.T) {
 		t.Parallel()
 		q := `label_copy(
@@ -1734,10 +1839,32 @@ func TestExecSuccess(t *testing.T) {
 	})
 	t.Run(`time() > scalar`, func(t *testing.T) {
 		t.Parallel()
-		q := `time() > 123`
+		q := `time() > 1234`
 		r := netstorage.Result{
 			MetricName: metricNameExpected,
-			Values:     []float64{1000, 1200, 1400, 1600, 1800, 2000},
+			Values:     []float64{nan, nan, 1400, 1600, 1800, 2000},
+			Timestamps: timestampsExpected,
+		}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`time() >bool scalar`, func(t *testing.T) {
+		t.Parallel()
+		q := `time() >bool 1234`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{0, 0, 1, 1, 1, 1},
+			Timestamps: timestampsExpected,
+		}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`nan >bool scalar1`, func(t *testing.T) {
+		t.Parallel()
+		q := `(time() > 1234) >bool 1450`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{nan, nan, 0, 1, 1, 1},
 			Timestamps: timestampsExpected,
 		}
 		resultExpected := []netstorage.Result{r}
@@ -4008,6 +4135,28 @@ func TestExecSuccess(t *testing.T) {
 		resultExpected := []netstorage.Result{r}
 		f(q, resultExpected)
 	})
+	t.Run(`count_eq_over_time`, func(t *testing.T) {
+		t.Parallel()
+		q := `count_eq_over_time(round(5*rand(0))[200s:10s], 1)`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{2, 4, 5, 2, 6, 6},
+			Timestamps: timestampsExpected,
+		}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`count_ne_over_time`, func(t *testing.T) {
+		t.Parallel()
+		q := `count_ne_over_time(round(5*rand(0))[200s:10s], 1)`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{18, 16, 15, 18, 14, 14},
+			Timestamps: timestampsExpected,
+		}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
 	t.Run(`increases_over_time`, func(t *testing.T) {
 		t.Parallel()
 		q := `increases_over_time(rand(0)[200s:10s])`
@@ -4193,7 +4342,7 @@ func TestExecSuccess(t *testing.T) {
 	})
 	t.Run(`topk_max(1)`, func(t *testing.T) {
 		t.Parallel()
-		q := `sort(topk_max(1, label_set(10, "foo", "bar") or label_set(time()/150, "baz", "sss")))`
+		q := `topk_max(1, label_set(10, "foo", "bar") or label_set(time()/150, "baz", "sss"))`
 		r1 := netstorage.Result{
 			MetricName: metricNameExpected,
 			Values:     []float64{nan, nan, nan, 10.666666666666666, 12, 13.333333333333334},
@@ -4206,6 +4355,84 @@ func TestExecSuccess(t *testing.T) {
 		resultExpected := []netstorage.Result{r1}
 		f(q, resultExpected)
 	})
+	t.Run(`topk_max(1, remaining_sum)`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_desc(topk_max(1, label_set(10, "foo", "bar") or label_set(time()/150, "baz", "sss"), "remaining_sum"))`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{nan, nan, nan, 10.666666666666666, 12, 13.333333333333334},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("baz"),
+			Value: []byte("sss"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{10, 10, 10, 10, 10, 10},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("remaining_sum"),
+				Value: []byte("remaining_sum"),
+			},
+		}
+		resultExpected := []netstorage.Result{r1, r2}
+		f(q, resultExpected)
+	})
+	t.Run(`topk_max(2, remaining_sum)`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_desc(topk_max(2, label_set(10, "foo", "bar") or label_set(time()/150, "baz", "sss"), "remaining_sum"))`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{nan, nan, nan, 10.666666666666666, 12, 13.333333333333334},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("baz"),
+			Value: []byte("sss"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{10, 10, 10, 10, 10, 10},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("foo"),
+				Value: []byte("bar"),
+			},
+		}
+		resultExpected := []netstorage.Result{r1, r2}
+		f(q, resultExpected)
+	})
+	t.Run(`topk_max(3, remaining_sum)`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_desc(topk_max(3, label_set(10, "foo", "bar") or label_set(time()/150, "baz", "sss"), "remaining_sum"))`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{nan, nan, nan, 10.666666666666666, 12, 13.333333333333334},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("baz"),
+			Value: []byte("sss"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{10, 10, 10, 10, 10, 10},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("foo"),
+				Value: []byte("bar"),
+			},
+		}
+		resultExpected := []netstorage.Result{r1, r2}
+		f(q, resultExpected)
+	})
 	t.Run(`bottomk_max(1)`, func(t *testing.T) {
 		t.Parallel()
 		q := `sort(bottomk_max(1, label_set(10, "foo", "bar") or label_set(time()/150, "baz", "sss")))`
@@ -6032,6 +6259,8 @@ func TestExecError(t *testing.T) {
 	f(`share_gt_over_time()`)
 	f(`count_le_over_time()`)
 	f(`count_gt_over_time()`)
+	f(`count_eq_over_time()`)
+	f(`count_ne_over_time()`)

 	// Invalid argument type
 	f(`median_over_time({}, 2)`)
@@ -6068,6 +6297,8 @@ func TestExecError(t *testing.T) {
 	f(`label_transform(1, "foo", "invalid(regexp", "baz`)
 	f(`label_match(1, 2, 3)`)
 	f(`label_mismatch(1, 2, 3)`)
+	f(`label_uppercase()`)
+	f(`label_lowercase()`)
 	f(`alias(1, 2)`)
 	f(`aggr_over_time(1, 2)`)
 	f(`aggr_over_time(("foo", "bar"), 3)`)
--- a/app/vmselect/promql/rollup.go
+++ b/app/vmselect/promql/rollup.go
@@ -15,7 +15,7 @@ import (
 	"github.com/valyala/histogram"
 )

-var minStalenessInterval = flag.Duration("search.minStalenessInterval", 0, "The mimimum interval for staleness calculations. "+
+var minStalenessInterval = flag.Duration("search.minStalenessInterval", 0, "The minimum interval for staleness calculations. "+
 	"This flag could be useful for removing gaps on graphs generated from time series with irregular intervals between samples. "+
 	"See also '-search.maxStalenessInterval'")

@@ -64,6 +64,8 @@ var rollupFuncs = map[string]newRollupFunc{
 	"share_gt_over_time":    newRollupShareGT,
 	"count_le_over_time":    newRollupCountLE,
 	"count_gt_over_time":    newRollupCountGT,
+	"count_eq_over_time":    newRollupCountEQ,
+	"count_ne_over_time":    newRollupCountNE,
 	"histogram_over_time":   newRollupFuncOneArg(rollupHistogram),
 	"rollup":                newRollupFuncOneArg(rollupFake),
 	"rollup_rate":           newRollupFuncOneArg(rollupFake), // + rollupFuncsRemoveCounterResets
@@ -258,15 +260,16 @@ func getRollupConfigs(name string, rf rollupFunc, expr metricsql.Expr, start, en
 	}
 	newRollupConfig := func(rf rollupFunc, tagValue string) *rollupConfig {
 		return &rollupConfig{
-			TagValue:        tagValue,
-			Func:            rf,
-			Start:           start,
-			End:             end,
-			Step:            step,
-			Window:          window,
-			MayAdjustWindow: !rollupFuncsCannotAdjustWindow[name],
-			LookbackDelta:   lookbackDelta,
-			Timestamps:      sharedTimestamps,
+			TagValue:          tagValue,
+			Func:              rf,
+			Start:             start,
+			End:               end,
+			Step:              step,
+			Window:            window,
+			MayAdjustWindow:   !rollupFuncsCannotAdjustWindow[name],
+			CanDropLastSample: name == "default_rollup",
+			LookbackDelta:     lookbackDelta,
+			Timestamps:        sharedTimestamps,
 		}
 	}
 	appendRollupConfigs := func(dst []*rollupConfig) []*rollupConfig {
@@ -325,14 +328,32 @@ func getRollupFunc(funcName string) newRollupFunc {
 }

 type rollupFuncArg struct {
-	prevValue     float64
-	prevTimestamp int64
-	values        []float64
-	timestamps    []int64
+	// The value preceeding values if it fits staleness interval.
+	prevValue float64

+	// The timestamp for prevValue.
+	prevTimestamp int64
+
+	// Values that fit window ending at currTimestamp.
+	values []float64
+
+	// Timestamps for values.
+	timestamps []int64
+
+	// Real value preceeding values without restrictions on staleness interval.
+	realPrevValue float64
+
+	// Real value which goes after values.
+	realNextValue float64
+
+	// Current timestamp for rollup evaluation.
 	currTimestamp int64
-	idx           int
-	window        int64
+
+	// Index for the currently evaluated point relative to time range for query evaluation.
+	idx int
+
+	// Time window for rollup calculations.
+	window int64

 	tsm *timeseriesMap
 }
@@ -370,6 +391,11 @@ type rollupConfig struct {
 	// when using window smaller than 2 x scrape_interval.
 	MayAdjustWindow bool

+	// Whether the last sample can be dropped during rollup calculations.
+	// The last sample can be dropped for `default_rollup()` function only.
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/748 .
+	CanDropLastSample bool
+
 	Timestamps []int64

 	// LoookbackDelta is the analog to `-query.lookback-delta` from Prometheus world.
@@ -501,6 +527,9 @@ func (rc *rollupConfig) doInternal(dstValues []float64, tsm *timeseriesMap, valu
 	ni := 0
 	nj := 0
 	stalenessInterval := int64(float64(scrapeInterval) * 0.9)
+	// Do not drop trailing data points for queries, which return 2 or 1 point (aka instant queries).
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/845
+	canDropLastSample := rc.CanDropLastSample && len(rc.Timestamps) > 2
 	for _, tEnd := range rc.Timestamps {
 		tStart := tEnd - window
 		ni = seekFirstTimestampIdxAfter(timestamps[i:], tStart, ni)
@@ -519,15 +548,26 @@ func (rc *rollupConfig) doInternal(dstValues []float64, tsm *timeseriesMap, valu
 		}
 		rfa.values = values[i:j]
 		rfa.timestamps = timestamps[i:j]
-		if j == len(timestamps) && i < j && tEnd-timestamps[j-1] > stalenessInterval {
-			// Do not take into account the last data point in time series if the distance between this data point
-			// and tEnd exceeds stalenessInterval.
+		if canDropLastSample && j == len(timestamps) && j > 0 && (tEnd-timestamps[j-1] > stalenessInterval || i == j && len(timestamps) == 1) {
+			// Drop trailing data points in the following cases:
+			// - if the distance between the last raw sample and tEnd exceeds stalenessInterval
+			// - if time series contains only a single raw sample
 			// This should prevent from double counting when a label changes in time series (for instance,
 			// during new deployment in K8S). See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/748
 			rfa.prevValue = nan
 			rfa.values = nil
 			rfa.timestamps = nil
 		}
+		if i > 0 {
+			rfa.realPrevValue = values[i-1]
+		} else {
+			rfa.realPrevValue = nan
+		}
+		if j < len(values) {
+			rfa.realNextValue = values[j]
+		} else {
+			rfa.realNextValue = nan
+		}
 		rfa.currTimestamp = tEnd
 		value := rc.Func(rfa)
 		rfa.idx++
@@ -857,6 +897,26 @@ func countFilterGT(values []float64, gt float64) int {
 	return n
 }

+func countFilterEQ(values []float64, eq float64) int {
+	n := 0
+	for _, v := range values {
+		if v == eq {
+			n++
+		}
+	}
+	return n
+}
+
+func countFilterNE(values []float64, ne float64) int {
+	n := 0
+	for _, v := range values {
+		if v != ne {
+			n++
+		}
+	}
+	return n
+}
+
 func newRollupShareFilter(args []interface{}, countFilter func(values []float64, limit float64) int) (rollupFunc, error) {
 	rf, err := newRollupCountFilter(args, countFilter)
 	if err != nil {
@@ -876,6 +936,14 @@ func newRollupCountGT(args []interface{}) (rollupFunc, error) {
 	return newRollupCountFilter(args, countFilterGT)
 }

+func newRollupCountEQ(args []interface{}) (rollupFunc, error) {
+	return newRollupCountFilter(args, countFilterEQ)
+}
+
+func newRollupCountNE(args []interface{}) (rollupFunc, error) {
+	return newRollupCountFilter(args, countFilterNE)
+}
+
 func newRollupCountFilter(args []interface{}, countFilter func(values []float64, limit float64) int) (rollupFunc, error) {
 	if err := expectRollupArgsNum(args, 2); err != nil {
 		return nil, err
@@ -1104,10 +1172,10 @@ func rollupSum(rfa *rollupFuncArg) float64 {
 	// before calling rollup funcs.
 	values := rfa.values
 	if len(values) == 0 {
-		if math.IsNaN(rfa.prevValue) {
-			return nan
-		}
-		return 0
+		// Do not take into account rfa.prevValue, since it may lead
+		// to inconsistent results comparing to Prometheus on broken time series
+		// with irregular data points.
+		return nan
 	}
 	var sum float64
 	for _, v := range values {
@@ -1234,6 +1302,12 @@ func rollupDelta(rfa *rollupFuncArg) float64 {
 		if len(values) == 0 {
 			return nan
 		}
+		if !math.IsNaN(rfa.realPrevValue) {
+			// Assume that the value didn't change during the current gap.
+			// This should fix high delta() and increase() values at the end of gaps.
+			// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/894
+			return values[len(values)-1] - rfa.realPrevValue
+		}
 		// Assume that the previous non-existing value was 0
 		// only if the first value doesn't exceed too much the delta with the next value.
 		//
@@ -1246,6 +1320,8 @@ func rollupDelta(rfa *rollupFuncArg) float64 {
 		d := float64(10)
 		if len(values) > 1 {
 			d = values[1] - values[0]
+		} else if !math.IsNaN(rfa.realNextValue) {
+			d = rfa.realNextValue - values[0]
 		}
 		if math.Abs(values[0]) < 10*(math.Abs(d)+1) {
 			prevValue = 0
--- a/app/vmselect/promql/rollup_test.go
+++ b/app/vmselect/promql/rollup_test.go
@@ -285,6 +285,44 @@ func TestRollupCountGTOverTime(t *testing.T) {
 	f(1000, 0)
 }

+func TestRollupCountEQOverTime(t *testing.T) {
+	f := func(eq, vExpected float64) {
+		t.Helper()
+		eqs := []*timeseries{{
+			Values:     []float64{eq},
+			Timestamps: []int64{123},
+		}}
+		var me metricsql.MetricExpr
+		args := []interface{}{&metricsql.RollupExpr{Expr: &me}, eqs}
+		testRollupFunc(t, "count_eq_over_time", args, &me, vExpected)
+	}
+
+	f(-123, 0)
+	f(0, 0)
+	f(34, 4)
+	f(123, 1)
+	f(12, 1)
+}
+
+func TestRollupCountNEOverTime(t *testing.T) {
+	f := func(ne, vExpected float64) {
+		t.Helper()
+		nes := []*timeseries{{
+			Values:     []float64{ne},
+			Timestamps: []int64{123},
+		}}
+		var me metricsql.MetricExpr
+		args := []interface{}{&metricsql.RollupExpr{Expr: &me}, nes}
+		testRollupFunc(t, "count_ne_over_time", args, &me, vExpected)
+	}
+
+	f(-123, 12)
+	f(0, 12)
+	f(34, 8)
+	f(123, 11)
+	f(12, 11)
+}
+
 func TestRollupQuantileOverTime(t *testing.T) {
 	f := func(phi, vExpected float64) {
 		t.Helper()
@@ -1103,11 +1141,13 @@ func testRowsEqual(t *testing.T, values []float64, timestamps []int64, valuesExp
 }

 func TestRollupDelta(t *testing.T) {
-	f := func(prevValue float64, values []float64, resultExpected float64) {
+	f := func(prevValue, realPrevValue, realNextValue float64, values []float64, resultExpected float64) {
 		t.Helper()
 		rfa := &rollupFuncArg{
-			prevValue: prevValue,
-			values:    values,
+			prevValue:     prevValue,
+			values:        values,
+			realPrevValue: realPrevValue,
+			realNextValue: realNextValue,
 		}
 		result := rollupDelta(rfa)
 		if math.IsNaN(result) {
@@ -1120,22 +1160,36 @@ func TestRollupDelta(t *testing.T) {
 			t.Fatalf("unexpected result; got %v; want %v", result, resultExpected)
 		}
 	}
-	f(nan, nil, nan)
+	f(nan, nan, nan, nil, nan)

 	// Small initial value
-	f(nan, []float64{1}, 1)
-	f(nan, []float64{10}, 10)
-	f(nan, []float64{100}, 100)
-	f(nan, []float64{1, 2, 3}, 3)
-	f(1, []float64{1, 2, 3}, 2)
-	f(nan, []float64{5, 6, 8}, 8)
-	f(2, []float64{5, 6, 8}, 6)
+	f(nan, nan, nan, []float64{1}, 1)
+	f(nan, nan, nan, []float64{10}, 10)
+	f(nan, nan, nan, []float64{100}, 100)
+	f(nan, nan, nan, []float64{1, 2, 3}, 3)
+	f(1, nan, nan, []float64{1, 2, 3}, 2)
+	f(nan, nan, nan, []float64{5, 6, 8}, 8)
+	f(2, nan, nan, []float64{5, 6, 8}, 6)

 	// Too big initial value must be skipped.
-	f(nan, []float64{1000}, 0)
-	f(nan, []float64{1000, 1001, 1002}, 2)
+	f(nan, nan, nan, []float64{1000}, 0)
+	f(nan, nan, nan, []float64{1000, 1001, 1002}, 2)
+
+	// Non-nan realPrevValue
+	f(nan, 900, nan, []float64{1000}, 100)
+	f(nan, 1000, nan, []float64{1000}, 0)
+	f(nan, 1100, nan, []float64{1000}, -100)
+	f(nan, 900, nan, []float64{1000, 1001, 1002}, 102)
+
+	// Small delta between realNextValue and values
+	f(nan, nan, 990, []float64{1000}, 0)
+	f(nan, nan, 1005, []float64{1000}, 0)
+
+	// Big delta between relaNextValue and values
+	f(nan, nan, 800, []float64{1000}, 1000)
+	f(nan, nan, 1300, []float64{1000}, 1000)

 	// Empty values
-	f(1, nil, 0)
-	f(100, nil, 0)
+	f(1, nan, nan, nil, 0)
+	f(100, nan, nan, nil, 0)
 }
--- a/app/vmselect/promql/transform.go
+++ b/app/vmselect/promql/transform.go
@@ -73,6 +73,8 @@ var transformFuncs = map[string]transformFunc{
 	// New funcs
 	"label_set":          transformLabelSet,
 	"label_map":          transformLabelMap,
+	"label_uppercase":    transformLabelUppercase,
+	"label_lowercase":    transformLabelLowercase,
 	"label_del":          transformLabelDel,
 	"label_keep":         transformLabelKeep,
 	"label_copy":         transformLabelCopy,
@@ -265,6 +267,9 @@ func newTransformFuncDateTime(f func(t time.Time) int) transformFunc {
 		}
 		tf := func(values []float64) {
 			for i, v := range values {
+				if math.IsNaN(v) {
+					continue
+				}
 				t := time.Unix(int64(v), 0).UTC()
 				values[i] = float64(f(t))
 			}
@@ -1193,6 +1198,42 @@ func transformLabelSet(tfa *transformFuncArg) ([]*timeseries, error) {
 	return rvs, nil
 }

+func transformLabelUppercase(tfa *transformFuncArg) ([]*timeseries, error) {
+	return transformLabelValueFunc(tfa, strings.ToUpper)
+}
+
+func transformLabelLowercase(tfa *transformFuncArg) ([]*timeseries, error) {
+	return transformLabelValueFunc(tfa, strings.ToLower)
+}
+
+func transformLabelValueFunc(tfa *transformFuncArg, f func(string) string) ([]*timeseries, error) {
+	args := tfa.args
+	if len(args) < 2 {
+		return nil, fmt.Errorf(`not enough args; got %d; want at least %d`, len(args), 2)
+	}
+	labels := make([]string, 0, len(args)-1)
+	for i := 1; i < len(args); i++ {
+		label, err := getString(args[i], i)
+		if err != nil {
+			return nil, err
+		}
+		labels = append(labels, label)
+	}
+
+	rvs := args[0]
+	for _, ts := range rvs {
+		mn := &ts.MetricName
+		for _, label := range labels {
+			dstValue := getDstValue(mn, label)
+			*dstValue = append((*dstValue)[:0], f(string(*dstValue))...)
+			if len(*dstValue) == 0 {
+				mn.RemoveTag(label)
+			}
+		}
+	}
+	return rvs, nil
+}
+
 func transformLabelMap(tfa *transformFuncArg) ([]*timeseries, error) {
 	args := tfa.args
 	if len(args) < 2 {
--- a/app/vmselect/searchutils/searchutils.go
+++ b/app/vmselect/searchutils/searchutils.go
@@ -184,5 +184,7 @@ func (d *Deadline) Deadline() uint64 {

 // String returns human-readable string representation for d.
 func (d *Deadline) String() string {
-	return fmt.Sprintf("%.3f seconds; the timeout can be adjusted with `%s` command-line flag", d.timeout.Seconds(), d.flagHint)
+	startTime := time.Unix(int64(d.deadline), 0).Add(-d.timeout)
+	elapsed := time.Since(startTime)
+	return fmt.Sprintf("%.3f seconds (elapsed %.3f seconds); the timeout can be adjusted with `%s` command-line flag", d.timeout.Seconds(), elapsed.Seconds(), d.flagHint)
 }
--- a/app/vmstorage/main.go
+++ b/app/vmstorage/main.go
@@ -10,6 +10,7 @@ import (

 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
@@ -19,9 +20,10 @@ import (
 )

 var (
-	retentionPeriod   = flag.Int("retentionPeriod", 1, "Retention period in months")
+	retentionPeriod   = flagutil.NewDuration("retentionPeriod", 1, "Data with timestamps outside the retentionPeriod is automatically deleted")
 	snapshotAuthKey   = flag.String("snapshotAuthKey", "", "authKey, which must be passed in query string to /snapshot* pages")
 	forceMergeAuthKey = flag.String("forceMergeAuthKey", "", "authKey, which must be passed in query string to /internal/force_merge pages")
+	forceFlushAuthKey = flag.String("forceFlushAuthKey", "", "authKey, which must be passed in query string to /internal/force_flush pages")

 	precisionBits = flag.Int("precisionBits", 64, "The number of precision bits to store per each value. Lower precision bits improves data compression at the cost of precision loss")

@@ -44,12 +46,12 @@ func CheckTimeRange(tr storage.TimeRange) error {
 	if !*denyQueriesOutsideRetention {
 		return nil
 	}
-	minAllowedTimestamp := (int64(fasttime.UnixTimestamp()) - int64(*retentionPeriod)*3600*24*30) * 1000
+	minAllowedTimestamp := int64(fasttime.UnixTimestamp()*1000) - retentionPeriod.Msecs
 	if tr.MinTimestamp > minAllowedTimestamp {
 		return nil
 	}
 	return &httpserver.ErrorWithStatusCode{
-		Err:        fmt.Errorf("the given time range %s is outside the allowed retention of %d months according to -denyQueriesOutsideRetention", &tr, *retentionPeriod),
+		Err:        fmt.Errorf("the given time range %s is outside the allowed -retentionPeriod=%s according to -denyQueriesOutsideRetention", &tr, retentionPeriod),
 		StatusCode: http.StatusServiceUnavailable,
 	}
 }
@@ -72,12 +74,12 @@ func InitWithoutMetrics() {
 	storage.SetBigMergeWorkersCount(*bigMergeConcurrency)
 	storage.SetSmallMergeWorkersCount(*smallMergeConcurrency)

-	logger.Infof("opening storage at %q with retention period %d months", *DataPath, *retentionPeriod)
+	logger.Infof("opening storage at %q with -retentionPeriod=%s", *DataPath, retentionPeriod)
 	startTime := time.Now()
 	WG = syncwg.WaitGroup{}
-	strg, err := storage.OpenStorage(*DataPath, *retentionPeriod)
+	strg, err := storage.OpenStorage(*DataPath, retentionPeriod.Msecs)
 	if err != nil {
-		logger.Fatalf("cannot open a storage at %s with retention period %d months: %s", *DataPath, *retentionPeriod, err)
+		logger.Fatalf("cannot open a storage at %s with -retentionPeriod=%s: %s", *DataPath, retentionPeriod, err)
 	}
 	Storage = strg

@@ -111,6 +113,14 @@ func AddRows(mrs []storage.MetricRow) error {
 	return err
 }

+// RegisterMetricNames registers all the metrics from mrs in the storage.
+func RegisterMetricNames(mrs []storage.MetricRow) error {
+	WG.Add(1)
+	err := Storage.RegisterMetricNames(mrs)
+	WG.Done()
+	return err
+}
+
 // DeleteMetrics deletes metrics matching tfss.
 //
 // Returns the number of deleted metrics.
@@ -121,6 +131,22 @@ func DeleteMetrics(tfss []*storage.TagFilters) (int, error) {
 	return n, err
 }

+// SearchMetricNames returns metric names for the given tfss on the given tr.
+func SearchMetricNames(tfss []*storage.TagFilters, tr storage.TimeRange, maxMetrics int, deadline uint64) ([]storage.MetricName, error) {
+	WG.Add(1)
+	mns, err := Storage.SearchMetricNames(tfss, tr, maxMetrics, deadline)
+	WG.Done()
+	return mns, err
+}
+
+// SearchTagKeysOnTimeRange searches for tag keys on tr.
+func SearchTagKeysOnTimeRange(tr storage.TimeRange, maxTagKeys int, deadline uint64) ([]string, error) {
+	WG.Add(1)
+	keys, err := Storage.SearchTagKeysOnTimeRange(tr, maxTagKeys, deadline)
+	WG.Done()
+	return keys, err
+}
+
 // SearchTagKeys searches for tag keys
 func SearchTagKeys(maxTagKeys int, deadline uint64) ([]string, error) {
 	WG.Add(1)
@@ -129,6 +155,14 @@ func SearchTagKeys(maxTagKeys int, deadline uint64) ([]string, error) {
 	return keys, err
 }

+// SearchTagValuesOnTimeRange searches for tag values for the given tagKey on tr.
+func SearchTagValuesOnTimeRange(tagKey []byte, tr storage.TimeRange, maxTagValues int, deadline uint64) ([]string, error) {
+	WG.Add(1)
+	values, err := Storage.SearchTagValuesOnTimeRange(tagKey, tr, maxTagValues, deadline)
+	WG.Done()
+	return values, err
+}
+
 // SearchTagValues searches for tag values for the given tagKey
 func SearchTagValues(tagKey []byte, maxTagValues int, deadline uint64) ([]string, error) {
 	WG.Add(1)
@@ -205,6 +239,16 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		}()
 		return true
 	}
+	if path == "/internal/force_flush" {
+		authKey := r.FormValue("authKey")
+		if authKey != *forceFlushAuthKey {
+			httpserver.Errorf(w, r, "invalid authKey %q. It must match the value from -forceFlushAuthKey command line flag", authKey)
+			return true
+		}
+		logger.Infof("flushing storage to make pending data available for reading")
+		Storage.DebugFlush()
+		return true
+	}
 	prometheusCompatibleResponse := false
 	if path == "/api/v1/admin/tsdb/snapshot" {
 		// Handle Prometheus API - https://prometheus.io/docs/prometheus/latest/querying/api/#snapshot .
@@ -223,7 +267,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {

 	switch path {
 	case "/create":
-		w.Header().Set("Content-Type", "application/json")
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
 		snapshotPath, err := Storage.CreateSnapshot()
 		if err != nil {
 			err = fmt.Errorf("cannot create snapshot: %w", err)
@@ -237,7 +281,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		}
 		return true
 	case "/list":
-		w.Header().Set("Content-Type", "application/json")
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
 		snapshots, err := Storage.ListSnapshots()
 		if err != nil {
 			err = fmt.Errorf("cannot list snapshots: %w", err)
@@ -254,7 +298,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		fmt.Fprintf(w, `]}`)
 		return true
 	case "/delete":
-		w.Header().Set("Content-Type", "application/json")
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
 		snapshotName := r.FormValue("snapshot")
 		if err := Storage.DeleteSnapshot(snapshotName); err != nil {
 			err = fmt.Errorf("cannot delete snapshot %q: %w", snapshotName, err)
@@ -264,7 +308,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		fmt.Fprintf(w, `{"status":"ok"}`)
 		return true
 	case "/delete_all":
-		w.Header().Set("Content-Type", "application/json")
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
 		snapshots, err := Storage.ListSnapshots()
 		if err != nil {
 			err = fmt.Errorf("cannot list snapshots: %w", err)
--- a/dashboards/victoriametrics.json
+++ b/dashboards/victoriametrics.json
@@ -56,7 +56,7 @@
  "gnetId": 10229,
  "graphTooltip": 0,
  "id": null,
-  "iteration": 1599034965731,
+  "iteration": 1603307754894,
  "links": [
    {
      "icon": "doc",
@@ -925,7 +925,7 @@
      "dashLength": 10,
      "dashes": false,
      "datasource": "$ds",
-      "description": "Shows how many ongoing insertions are taking place.\n* `max` - equal to number of CPU * 2\n* `current` - current number of goroutines busy with inserting rows into storage\n\nWhen `current` hits `max` constantly, it means storage is overloaded and require more CPU.",
+      "description": "Shows how many ongoing insertions (not API /write calls) on disk are taking place, where:\n* `max` - equal to number of CPUs;\n* `current` - current number of goroutines busy with inserting rows into underlying storage.\n\nEvery successful API /write call results into flush on disk. However, these two actions are separated and controlled via different concurrency limiters. The `max` on this panel can't be changed and always equal to number of CPUs. \n\nWhen `current` hits `max` constantly, it means storage is overloaded and requires more CPU.\n\n",
      "fieldConfig": {
        "defaults": {
          "custom": {},
@@ -979,6 +979,7 @@
        {
          "expr": "sum(vm_concurrent_addrows_capacity{job=\"$job\", instance=\"$instance\"})",
          "format": "time_series",
+          "interval": "",
          "intervalFactor": 1,
          "legendFormat": "max",
          "refId": "A"
@@ -995,7 +996,7 @@
      "timeFrom": null,
      "timeRegions": [],
      "timeShift": null,
-      "title": "Concurrent inserts ($instance)",
+      "title": "Concurrent flushes on disk ($instance)",
      "tooltip": {
        "shared": true,
        "sort": 2,
@@ -1164,7 +1165,7 @@
            "h": 8,
            "w": 12,
            "x": 0,
-            "y": 36
+            "y": 3
          },
          "hiddenSeries": false,
          "id": 10,
@@ -1250,7 +1251,7 @@
          "dashLength": 10,
          "dashes": false,
          "datasource": "$ds",
-          "description": "How many datapoints are in RAM queue waiting to be written into storage. The number of pending data points should be in the range from 0 to `2*<ingestion_rate>`, since VictoriaMetrics pushes pending data to persistent storage every second.",
+          "description": "Shows the time needed to reach the 100% of disk capacity based on the following params:\n* free disk space;\n* row ingestion rate;\n* dedup rate;\n* compression.\n\nUse this panel for capacity planning in order to estimate the time remaining for running out of the disk space.\n\n",
          "fieldConfig": {
            "defaults": {
              "custom": {},
@@ -1264,63 +1265,53 @@
            "h": 8,
            "w": 12,
            "x": 12,
-            "y": 36
+            "y": 3
          },
          "hiddenSeries": false,
-          "id": 34,
+          "id": 73,
          "legend": {
-            "avg": false,
-            "current": false,
+            "alignAsTable": true,
+            "avg": true,
+            "current": true,
+            "hideZero": true,
            "max": false,
            "min": false,
            "show": false,
            "total": false,
-            "values": false
+            "values": true
          },
          "lines": true,
          "linewidth": 1,
          "links": [],
-          "nullPointMode": "null",
+          "nullPointMode": "null as zero",
          "percentage": false,
          "pluginVersion": "7.1.1",
          "pointradius": 2,
          "points": false,
          "renderer": "flot",
-          "seriesOverrides": [
-            {
-              "alias": "pending index entries",
-              "yaxis": 2
-            }
-          ],
+          "seriesOverrides": [],
          "spaceLength": 10,
          "stack": false,
          "steppedLine": false,
          "targets": [
            {
-              "expr": "vm_pending_rows{job=\"$job\", instance=~\"$instance\", type=\"storage\"}",
+              "expr": "vm_free_disk_space_bytes{job=\"$job\", instance=\"$instance\"} / ignoring(path) ((rate(vm_rows_added_to_storage_total{job=\"$job\", instance=\"$instance\"}[1d]) - ignoring(type) rate(vm_deduplicated_samples_total{job=\"$job\", instance=\"$instance\", type=\"merge\"}[1d])) * (sum(vm_data_size_bytes{job=\"$job\", instance=\"$instance\", type!=\"indexdb\"}) / sum(vm_rows{job=\"$job\", instance=\"$instance\", type!=\"indexdb\"})))",
              "format": "time_series",
              "hide": false,
+              "interval": "",
              "intervalFactor": 1,
-              "legendFormat": "pending datapoints",
+              "legendFormat": "",
              "refId": "A"
-            },
-            {
-              "expr": "vm_pending_rows{job=\"$job\", instance=~\"$instance\", type=\"indexdb\"}",
-              "format": "time_series",
-              "hide": false,
-              "intervalFactor": 1,
-              "legendFormat": "pending index entries",
-              "refId": "B"
            }
          ],
          "thresholds": [],
          "timeFrom": null,
          "timeRegions": [],
          "timeShift": null,
-          "title": "Pending datapoints ($instance)",
+          "title": "Storage full ETA ($instance)",
          "tooltip": {
            "shared": true,
-            "sort": 0,
+            "sort": 2,
            "value_type": "individual"
          },
          "type": "graph",
@@ -1333,7 +1324,8 @@
          },
          "yaxes": [
            {
-              "format": "short",
+              "decimals": null,
+              "format": "s",
              "label": null,
              "logBase": 1,
              "max": null,
@@ -1341,8 +1333,7 @@
              "show": true
            },
            {
-              "decimals": 3,
-              "format": "none",
+              "format": "short",
              "label": null,
              "logBase": 1,
              "max": null,
@@ -1375,7 +1366,7 @@
            "h": 8,
            "w": 12,
            "x": 0,
-            "y": 44
+            "y": 11
          },
          "hiddenSeries": false,
          "id": 30,
@@ -1472,7 +1463,7 @@
          "dashLength": 10,
          "dashes": false,
          "datasource": "$ds",
-          "description": "Data parts of LSM tree.\nHigh number of parts could be an evidence of slow merge performance - check the resource utilization.\n* `indexdb` - inverted index\n* `storage/small` - recently added parts of data ingested into storage(hot data)\n* `storage/big` -  small parts gradually merged into big parts (cold data)",
+          "description": "How many datapoints are in RAM queue waiting to be written into storage. The number of pending data points should be in the range from 0 to `2*<ingestion_rate>`, since VictoriaMetrics pushes pending data to persistent storage every second.",
          "fieldConfig": {
            "defaults": {
              "custom": {},
@@ -1486,16 +1477,16 @@
            "h": 8,
            "w": 12,
            "x": 12,
-            "y": 44
+            "y": 11
          },
          "hiddenSeries": false,
-          "id": 36,
+          "id": 34,
          "legend": {
            "avg": false,
            "current": false,
            "max": false,
            "min": false,
-            "show": true,
+            "show": false,
            "total": false,
            "values": false
          },
@@ -1508,27 +1499,41 @@
          "pointradius": 2,
          "points": false,
          "renderer": "flot",
-          "seriesOverrides": [],
+          "seriesOverrides": [
+            {
+              "alias": "pending index entries",
+              "yaxis": 2
+            }
+          ],
          "spaceLength": 10,
          "stack": false,
          "steppedLine": false,
          "targets": [
            {
-              "expr": "sum(vm_parts{job=\"$job\", instance=\"$instance\"}) by (type)",
+              "expr": "vm_pending_rows{job=\"$job\", instance=~\"$instance\", type=\"storage\"}",
              "format": "time_series",
+              "hide": false,
              "intervalFactor": 1,
-              "legendFormat": "{{type}}",
+              "legendFormat": "pending datapoints",
              "refId": "A"
+            },
+            {
+              "expr": "vm_pending_rows{job=\"$job\", instance=~\"$instance\", type=\"indexdb\"}",
+              "format": "time_series",
+              "hide": false,
+              "intervalFactor": 1,
+              "legendFormat": "pending index entries",
+              "refId": "B"
            }
          ],
          "thresholds": [],
          "timeFrom": null,
          "timeRegions": [],
          "timeShift": null,
-          "title": "LSM parts ($instance)",
+          "title": "Pending datapoints ($instance)",
          "tooltip": {
            "shared": true,
-            "sort": 2,
+            "sort": 0,
            "value_type": "individual"
          },
          "type": "graph",
@@ -1549,7 +1554,8 @@
              "show": true
            },
            {
-              "format": "short",
+              "decimals": 3,
+              "format": "none",
              "label": null,
              "logBase": 1,
              "max": null,
@@ -1582,7 +1588,7 @@
            "h": 8,
            "w": 12,
            "x": 0,
-            "y": 52
+            "y": 19
          },
          "hiddenSeries": false,
          "id": 53,
@@ -1669,6 +1675,196 @@
            "alignLevel": null
          }
        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$ds",
+          "description": "Data parts of LSM tree.\nHigh number of parts could be an evidence of slow merge performance - check the resource utilization.\n* `indexdb` - inverted index\n* `storage/small` - recently added parts of data ingested into storage(hot data)\n* `storage/big` -  small parts gradually merged into big parts (cold data)",
+          "fieldConfig": {
+            "defaults": {
+              "custom": {},
+              "links": []
+            },
+            "overrides": []
+          },
+          "fill": 1,
+          "fillGradient": 0,
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 19
+          },
+          "hiddenSeries": false,
+          "id": 36,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pluginVersion": "7.1.1",
+          "pointradius": 2,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "sum(vm_parts{job=\"$job\", instance=\"$instance\"}) by (type)",
+              "format": "time_series",
+              "intervalFactor": 1,
+              "legendFormat": "{{type}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeRegions": [],
+          "timeShift": null,
+          "title": "LSM parts ($instance)",
+          "tooltip": {
+            "shared": true,
+            "sort": 2,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$ds",
+          "description": "The number of on-going merges in storage nodes.  It is expected to have high numbers for `storage/small` metric.",
+          "fieldConfig": {
+            "defaults": {
+              "custom": {},
+              "links": []
+            },
+            "overrides": []
+          },
+          "fill": 1,
+          "fillGradient": 0,
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 0,
+            "y": 27
+          },
+          "hiddenSeries": false,
+          "id": 62,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "nullPointMode": "null",
+          "percentage": false,
+          "pluginVersion": "7.1.1",
+          "pointradius": 2,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "sum(vm_active_merges{job=\"$job\", instance=\"$instance\"}) by(type)",
+              "legendFormat": "{{type}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeRegions": [],
+          "timeShift": null,
+          "title": "Active merges ($instance)",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "decimals": 0,
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
        {
          "aliasColors": {},
          "bars": false,
@@ -1689,7 +1885,7 @@
            "h": 8,
            "w": 12,
            "x": 12,
-            "y": 52
+            "y": 27
          },
          "hiddenSeries": false,
          "id": 55,
@@ -1764,194 +1960,6 @@
            "alignLevel": null
          }
        },
-        {
-          "aliasColors": {},
-          "bars": false,
-          "dashLength": 10,
-          "dashes": false,
-          "datasource": "$ds",
-          "description": "The number of on-going merges in storage nodes.  It is expected to have high numbers for `storage/small` metric.",
-          "fieldConfig": {
-            "defaults": {
-              "custom": {},
-              "links": []
-            },
-            "overrides": []
-          },
-          "fill": 1,
-          "fillGradient": 0,
-          "gridPos": {
-            "h": 8,
-            "w": 12,
-            "x": 0,
-            "y": 60
-          },
-          "hiddenSeries": false,
-          "id": 62,
-          "legend": {
-            "avg": false,
-            "current": false,
-            "max": false,
-            "min": false,
-            "show": true,
-            "total": false,
-            "values": false
-          },
-          "lines": true,
-          "linewidth": 1,
-          "nullPointMode": "null",
-          "percentage": false,
-          "pluginVersion": "7.1.1",
-          "pointradius": 2,
-          "points": false,
-          "renderer": "flot",
-          "seriesOverrides": [],
-          "spaceLength": 10,
-          "stack": false,
-          "steppedLine": false,
-          "targets": [
-            {
-              "expr": "sum(vm_active_merges{job=\"$job\", instance=\"$instance\"}) by(type)",
-              "legendFormat": "{{type}}",
-              "refId": "A"
-            }
-          ],
-          "thresholds": [],
-          "timeFrom": null,
-          "timeRegions": [],
-          "timeShift": null,
-          "title": "Active merges ($instance)",
-          "tooltip": {
-            "shared": true,
-            "sort": 0,
-            "value_type": "individual"
-          },
-          "type": "graph",
-          "xaxis": {
-            "buckets": null,
-            "mode": "time",
-            "name": null,
-            "show": true,
-            "values": []
-          },
-          "yaxes": [
-            {
-              "decimals": 0,
-              "format": "short",
-              "label": null,
-              "logBase": 1,
-              "max": null,
-              "min": "0",
-              "show": true
-            },
-            {
-              "format": "short",
-              "label": null,
-              "logBase": 1,
-              "max": null,
-              "min": "0",
-              "show": true
-            }
-          ],
-          "yaxis": {
-            "align": false,
-            "alignLevel": null
-          }
-        },
-        {
-          "aliasColors": {},
-          "bars": false,
-          "dashLength": 10,
-          "dashes": false,
-          "datasource": "$ds",
-          "description": "The number of rows merged per second by storage nodes.",
-          "fieldConfig": {
-            "defaults": {
-              "custom": {},
-              "links": []
-            },
-            "overrides": []
-          },
-          "fill": 1,
-          "fillGradient": 0,
-          "gridPos": {
-            "h": 8,
-            "w": 12,
-            "x": 12,
-            "y": 60
-          },
-          "hiddenSeries": false,
-          "id": 64,
-          "legend": {
-            "avg": false,
-            "current": false,
-            "max": false,
-            "min": false,
-            "show": true,
-            "total": false,
-            "values": false
-          },
-          "lines": true,
-          "linewidth": 1,
-          "nullPointMode": "null",
-          "percentage": false,
-          "pluginVersion": "7.1.1",
-          "pointradius": 2,
-          "points": false,
-          "renderer": "flot",
-          "seriesOverrides": [],
-          "spaceLength": 10,
-          "stack": false,
-          "steppedLine": false,
-          "targets": [
-            {
-              "expr": "sum(rate(vm_rows_merged_total{job=\"$job\", instance=\"$instance\"}[5m])) by(type)",
-              "legendFormat": "{{type}}",
-              "refId": "A"
-            }
-          ],
-          "thresholds": [],
-          "timeFrom": null,
-          "timeRegions": [],
-          "timeShift": null,
-          "title": "Merge speed ($instance)",
-          "tooltip": {
-            "shared": true,
-            "sort": 0,
-            "value_type": "individual"
-          },
-          "type": "graph",
-          "xaxis": {
-            "buckets": null,
-            "mode": "time",
-            "name": null,
-            "show": true,
-            "values": []
-          },
-          "yaxes": [
-            {
-              "decimals": 0,
-              "format": "short",
-              "label": null,
-              "logBase": 1,
-              "max": null,
-              "min": "0",
-              "show": true
-            },
-            {
-              "format": "short",
-              "label": null,
-              "logBase": 1,
-              "max": null,
-              "min": "0",
-              "show": true
-            }
-          ],
-          "yaxis": {
-            "align": false,
-            "alignLevel": null
-          }
-        },
        {
          "aliasColors": {},
          "bars": false,
@@ -1972,7 +1980,7 @@
            "h": 8,
            "w": 12,
            "x": 0,
-            "y": 68
+            "y": 35
          },
          "hiddenSeries": false,
          "id": 58,
@@ -2050,6 +2058,100 @@
            "alignLevel": null
          }
        },
+        {
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "$ds",
+          "description": "The number of rows merged per second by storage nodes.",
+          "fieldConfig": {
+            "defaults": {
+              "custom": {},
+              "links": []
+            },
+            "overrides": []
+          },
+          "fill": 1,
+          "fillGradient": 0,
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 35
+          },
+          "hiddenSeries": false,
+          "id": 64,
+          "legend": {
+            "avg": false,
+            "current": false,
+            "max": false,
+            "min": false,
+            "show": true,
+            "total": false,
+            "values": false
+          },
+          "lines": true,
+          "linewidth": 1,
+          "nullPointMode": "null",
+          "percentage": false,
+          "pluginVersion": "7.1.1",
+          "pointradius": 2,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "sum(rate(vm_rows_merged_total{job=\"$job\", instance=\"$instance\"}[5m])) by(type)",
+              "legendFormat": "{{type}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeRegions": [],
+          "timeShift": null,
+          "title": "Merge speed ($instance)",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "decimals": 0,
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": "0",
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
+        },
        {
          "aliasColors": {},
          "bars": false,
@@ -2070,7 +2172,7 @@
            "h": 8,
            "w": 12,
            "x": 12,
-            "y": 68
+            "y": 43
          },
          "hiddenSeries": false,
          "id": 67,
--- a/deployment/docker/Makefile
+++ b/deployment/docker/Makefile
@@ -2,9 +2,9 @@

 DOCKER_NAMESPACE := victoriametrics

-ROOT_IMAGE ?= alpine:3.12
-CERTS_IMAGE := alpine:3.12
-GO_BUILDER_IMAGE := golang:1.15.2
+ROOT_IMAGE ?= alpine:3.12.1
+CERTS_IMAGE := alpine:3.12.1
+GO_BUILDER_IMAGE := golang:1.15.6
 BUILDER_IMAGE := local/builder:2.0.0-$(shell echo $(GO_BUILDER_IMAGE) | tr : _)
 BASE_IMAGE := local/base:1.1.1-$(shell echo $(ROOT_IMAGE) | tr : _)-$(shell echo $(CERTS_IMAGE) | tr : _)

--- a/deployment/docker/alerts.yml
+++ b/deployment/docker/alerts.yml
@@ -0,0 +1,23 @@
+groups:
+  - name: groupGorSingleAlert
+    rules:
+      - alert: VMRows
+        for: 10s
+        expr: vm_rows > 0
+        labels:
+          label: bar
+          host: "{{ $labels.instance }}"
+        annotations:
+          summary: "{{ $value|humanize }}"
+          description: "{{$labels}}"
+  - name: TestGroup
+    rules:
+      - alert: Conns
+        expr: sum(vm_tcplistener_conns) by(instance) > 1
+        for: 5s
+        annotations:
+          summary: "Too high connection number for {{$labels.instance}}"
+          description: "It is {{ $value }} connections for {{$labels.instance}}"
+      - alert: ExampleAlertAlwaysFiring
+        expr: sum by(job)
+          (up == 1)
--- a/deployment/docker/docker-compose.yml
+++ b/deployment/docker/docker-compose.yml
@@ -21,7 +21,10 @@ services:
    image: victoriametrics/victoria-metrics
    ports:
      - 8428:8428
+      - 8089:8089
+      - 8089:8089/udp
      - 2003:2003
+      - 2003:2003/udp
      - 4242:4242
    volumes:
      - vmdata:/storage
@@ -30,6 +33,7 @@ services:
      - '--graphiteListenAddr=:2003'
      - '--opentsdbListenAddr=:4242'
      - '--httpListenAddr=:8428'
+      - '--influxListenAddr=:8089'
    networks:
      - vm_net
    restart: always
@@ -48,6 +52,36 @@ services:
    networks:
      - vm_net
    restart: always
+  vmalert:
+    container_name: vmalert
+    image: victoriametrics/vmalert
+    depends_on:
+      - "victoriametrics"
+      - "alertmanager"
+    ports:
+      - 8880:8880
+    volumes:
+      - ./alerts.yml:/etc/alerts/alerts.yml
+    command:
+      - '--datasource.url=http://victoriametrics:8428/'
+      - '--remoteRead.url=http://victoriametrics:8428/'
+      - '--remoteWrite.url=http://victoriametrics:8428/'
+      - '--notifier.url=http://alertmanager:9093/'
+      - '--rule=/etc/alerts/*.yml'
+      # display source of alerts in grafana
+      - '-external.url=http://127.0.0.1:3000' #grafana outside container
+      - '--external.alert.source=explore?orgId=1&left=["now-1h","now","VictoriaMetrics",{"expr":"{{$$expr|quotesEscape|pathEscape}}"},{"mode":"Metrics"},{"ui":[true,true,true,"none"]}]' ## when copypaste the line be aware of '$$' for escaping in '$expr'
+    networks:
+      - vm_net
+    restart: always
+  alertmanager:
+    container_name: alertmanager
+    image:  prom/alertmanager
+    ports:
+      - 9093:9093
+    networks:
+      - vm_net
+    restart: always
 volumes:
  vmagentdata: {}
  vmdata: {}
--- a/docs/Articles.md
+++ b/docs/Articles.md
@@ -1,5 +1,27 @@
 # Articles

+## Third-party articles and slides about VictoriaMetrics
+
+* [Foiled by the Firewall: A Tale of Transition From Prometheus to VictoriaMetrics](https://www.percona.com/blog/2020/12/01/foiled-by-the-firewall-a-tale-of-transition-from-prometheus-to-victoriametrics/)
+* [Better Prometheus rate() function with VictoriaMetrics](https://www.percona.com/blog/2020/02/28/better-prometheus-rate-function-with-victoriametrics/)
+* [Making peace with Prometheus rate()](https://blog.doit-intl.com/making-peace-with-prometheus-rate-43a3ea75c4cf)
+* [Infrastructure monitoring with Prometheus at Zerodha](https://zerodha.tech/blog/infra-monitoring-at-zerodha/)
+* [Sismology: Iguana Solutions’ Monitoring System](https://medium.com/@IG1.com/sismology-iguana-solutions-monitoring-system-f46e4170447f)
+* [Prometheus High Availability and Fault Tolerance strategy, long term storage with VictoriaMetrics](https://medium.com/miro-engineering/prometheus-high-availability-and-fault-tolerance-strategy-long-term-storage-with-victoriametrics-82f6f3f0409e)
+* [How we improved our Kubernetes monitoring at Smarkets, and how you could too](https://smarketshq.com/monitoring-kubernetes-clusters-41a4b24c19e3)
+* [Monitoring K8S with VictoriaMetrics](https://docs.google.com/presentation/d/1g7yUyVEaAp4tPuRy-MZbPXKqJ1z78_5VKuV841aQfsg/edit)
+* [CMS monitoring R&D: Real-time monitoring and alerts](https://indico.cern.ch/event/877333/contributions/3696707/attachments/1972189/3281133/CMS_mon_RD_for_opInt.pdf)
+* [The CMS monitoring infrastructure and applications](https://arxiv.org/pdf/2007.03630.pdf)
+* [Disk usage: VictoriaMetrics vs Prometheus](https://stas.starikevich.com/posts/disk-usage-for-vm-versus-prometheus/)
+* [Benchmarking time series workloads on Apache Kudu using TSBS](https://blog.cloudera.com/benchmarking-time-series-workloads-on-apache-kudu-using-tsbs/)
+* [What are Open Source Time Series Databases?](https://www.iunera.com/kraken/fabric/time-series-database/)
+* [Evaluating performance and correctness](https://www.robustperception.io/evaluating-performance-and-correctness)
+* [Running VictoriaMetrics on Raspberry PI](https://stas.starikevich.com/posts/raspberry-pi-4-prometheus/)
+* [Calculating the Error of Quantile Estimation with Histograms](https://linuxczar.net/blog/2020/08/13/histogram-error/)
+* [Monitoring private clouds with VictoriaMetrics at LeroyMerlin](https://www.youtube.com/watch?v=74swsWqf0Uc)
+* [Monitoring Kubernetes with VictoriaMetrics+Prometheus](https://speakerdeck.com/bo0km4n/victoriametrics-plus-prometheusdegou-zhu-surufu-shu-kubernetesfalsejian-shi-ji-pan)
+* [High-performance Graphite storage solution on top of VictoriaMetrics](https://golangexample.com/a-high-performance-graphite-storage-solution/)
+

 ## Our articles

@@ -27,23 +49,7 @@
 * [How to migrate data from Prometheus to VictoriaMetrics](https://medium.com/@romanhavronenko/victoriametrics-how-to-migrate-data-from-prometheus-d44a6728f043)
 * [Filtering and modifying time series during import to VictoriaMetrics](https://medium.com/@romanhavronenko/victoriametrics-how-to-migrate-data-from-prometheus-filtering-and-modifying-time-series-6d40cea4bf21)
 * [Anomaly Detection in VictoriaMetrics](https://medium.com/@VictoriaMetrics/anomaly-detection-in-victoriametrics-9528538786a7)
-
-
-## Third-party articles and slides
-
-* [Better Prometheus rate() function with VictoriaMetrics](https://www.percona.com/blog/2020/02/28/better-prometheus-rate-function-with-victoriametrics/)
-* [Making peace with Prometheus rate()](https://blog.doit-intl.com/making-peace-with-prometheus-rate-43a3ea75c4cf)
-* [Infrastructure monitoring with Prometheus at Zerodha](https://zerodha.tech/blog/infra-monitoring-at-zerodha/)
-* [Sismology: Iguana Solutions’ Monitoring System](https://medium.com/@IG1.com/sismology-iguana-solutions-monitoring-system-f46e4170447f)
-* [Prometheus High Availability and Fault Tolerance strategy, long term storage with VictoriaMetrics](https://medium.com/miro-engineering/prometheus-high-availability-and-fault-tolerance-strategy-long-term-storage-with-victoriametrics-82f6f3f0409e)
-* [Monitoring K8S with VictoriaMetrics](https://docs.google.com/presentation/d/1g7yUyVEaAp4tPuRy-MZbPXKqJ1z78_5VKuV841aQfsg/edit)
-* [CMS monitoring R&D: Real-time monitoring and alerts](https://indico.cern.ch/event/877333/contributions/3696707/attachments/1972189/3281133/CMS_mon_RD_for_opInt.pdf)
-* [The CMS monitoring infrastructure and applications](https://arxiv.org/pdf/2007.03630.pdf)
-* [Disk usage: VictoriaMetrics vs Prometheus](https://stas.starikevich.com/posts/disk-usage-for-vm-versus-prometheus/)
-* [Benchmarking time series workloads on Apache Kudu using TSBS](https://blog.cloudera.com/benchmarking-time-series-workloads-on-apache-kudu-using-tsbs/)
-* [What are Open Source Time Series Databases?](https://www.iunera.com/kraken/fabric/time-series-database/)
-* [Evaluating performance and correctness](https://www.robustperception.io/evaluating-performance-and-correctness)
-* [Running VictoriaMetrics on Raspberry PI](https://stas.starikevich.com/posts/raspberry-pi-4-prometheus/)
-* [Calculating the Error of Quantile Estimation with Histograms](https://linuxczar.net/blog/2020/08/13/histogram-error/)
-* [Monitoring private clouds with VictoriaMetrics at LeroyMerlin](https://www.youtube.com/watch?v=74swsWqf0Uc)
-* [Monitoring Kubernetes with VictoriaMetrics+Prometheus](https://speakerdeck.com/bo0km4n/victoriametrics-plus-prometheusdegou-zhu-surufu-shu-kubernetesfalsejian-shi-ji-pan)
+* [How to use relabeling in Prometheus and VictoriaMetrics](https://valyala.medium.com/how-to-use-relabeling-in-prometheus-and-victoriametrics-8b90fc22c4b2)
+* [First look at performance comparison between InfluxDB IOx and VictoriaMetrics](https://medium.com/@VictoriaMetrics/first-look-at-perfomance-comparassion-between-influxdb-iox-and-victoriametrics-e590f847935b)
+* [Prometheus vs VictoriaMetrics benchmark on node-exporter metrics](https://valyala.medium.com/prometheus-vs-victoriametrics-benchmark-on-node-exporter-metrics-4ca29c75590f)
+* [How to monitor Go applications with VictoriaMetrics](https://victoriametrics.medium.com/how-to-monitor-go-applications-with-victoriametrics-c04703110870)
--- a/docs/CHANGELOG.md
+++ b/docs/CHANGELOG.md
@@ -0,0 +1,245 @@
+# CHANGELOG
+
+# tip
+
+
+# [v1.49.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.49.0)
+
+* FEATURE: optimize Consul service discovery speed when discovering big number of services. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/574
+* FEATURE: add `label_uppercase(q, label1, ... labelN)` and `label_lowercase(q, label1, ... labelN)` function to [MetricsQL](https://victoriametrics.github.io/MetricsQL.html)
+  for uppercasing and lowercasing values for the given labels. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/936
+* FEATURE: add `count_eq_over_time(m[d], N)` and `count_ne_over_time(m[d], N)` for counting the number of samples for `m` over `d` that (equal / not equal) to `N`.
+* FEATURE: do not print usage info for all the command-line flags when incorrect command-line flag is passed. Previously it could be hard reading the error message
+  about incorrect command-line flag because of too big usage info for all the flags.
+* FEATURE: upgrade Go builder from v1.15.5 to v1.15.6 . This fixes [issues found in Go since v1.15.5](https://github.com/golang/go/issues?q=milestone%3AGo1.15.6+label%3ACherryPickApproved).
+
+* BUGFIX: properly parse timestamps in OpenMetrics format - they are exposed as floating-point number in seconds instead of integer milliseconds
+  unlike in Prometheus exposition format. See [the docs](https://github.com/OpenObservability/OpenMetrics/blob/master/specification/OpenMetrics.md#timestamps).
+* BUGFIX: return `nan` for `a >bool b` query when `a` equals to `nan` like Prometheus does. Previously `0` was returned in this case. This applies to any comparison operation
+  with `bool` modifier. See [these docs](https://prometheus.io/docs/prometheus/latest/querying/operators/#comparison-binary-operators) for details.
+* BUGFIX: properly parse hex numbers in MetricsQL. Previously hex numbers with non-decimal digits such as `0x3b` couldn't be parsed.
+* BUGFIX: handle `time() cmp_op metric` like Prometheus does - i.e. return `metric` value if `cmp_op` comparison is true. Previously `time()` value was returned.
+* BUGFIX: return `nan` for `minute(m)` query when `m` equals to `nan` like Prometheus does. This applies to all the time-related functions such as `day_of_month`, `day_of_week`,
+  `days_in_month`, `hour`, `month` and `year`.
+
+
+# [v1.48.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.48.0)
+
+* FEATURE: added [Snap package for single-node VictoriaMetrics](https://snapcraft.io/victoriametrics). This simplifies installation under Ubuntu to a single command:
+  ```bash
+  snap install victoriametrics
+  ```
+* FEATURE: vmselect: add `-replicationFactor` command-line flag for reducing query duration when replication is enabled and a part of vmstorage nodes
+  are temporarily slow and/or temporarily unavailable. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/711
+* FEATURE: vminsert: export `vm_rpc_vmstorage_is_reachable` metric, which can be used for monitoring reachability of vmstorage nodes from vminsert nodes.
+* FEATURE: vmagent: add [Netflix Eureka](https://github.com/Netflix/eureka) service discovery (aka [eureka_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#eureka_sd_config)). See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/851
+* FEATURE: add `filters` option to `dockerswarm_sd_config` like Prometheus did in v2.23.0 - see https://github.com/prometheus/prometheus/pull/8074
+* FEATURE: expose `__meta_ec2_ipv6_addresses` label for `ec2_sd_config` like Prometheus will do in the next release.
+* FEATURE: add `-loggerWarnsPerSecondLimit` command-line flag for rate limiting of WARN messages in logs. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/905
+* FEATURE: apply `loggerErrorsPerSecondLimit` and `-loggerWarnsPerSecondLimit` rate limit per caller. I.e. log messages are suppressed if the same caller logs the same message
+  at the rate exceeding the given limit. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/905#issuecomment-729395855
+* FEATURE: add remoteAddr to slow query log in order to simplify identifying the client that sends slow queries to VictoriaMetrics.
+  Slow query logging is controlled with `-search.logSlowQueryDuration` command-line flag.
+* FEATURE: add `/tags/delSeries` handler from Graphite Tags API. See https://victoriametrics.github.io/#graphite-tags-api-usage
+* FEATURE: log metric name plus all its labels when the metric timestamp is out of the configured retention. This should simplify detecting the source of metrics with unexpected timestamps.
+* FEATURE: add `-dryRun` command-line flag to single-node VictoriaMetrics in order to check config file pointed by `-promscrape.config`.
+
+* BUGFIX: properly parse Prometheus metrics with [exemplars](https://github.com/OpenObservability/OpenMetrics/blob/master/OpenMetrics.md#exemplars-1) such as `foo 123 # {bar="baz"} 1`.
+* BUGFIX: properly parse "infinity" values in [OpenMetrics format](https://github.com/OpenObservability/OpenMetrics/blob/master/OpenMetrics.md#abnf).
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/924
+
+
+# [v1.47.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.47.0)
+
+* FEATURE: vmselect: return the original error from `vmstorage` node in query response if `-search.denyPartialResponse` is set.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/891
+* FEATURE: vmselect: add `"isPartial":{true|false}` field in JSON output for `/api/v1/*` functions
+  from [Prometheus querying API](https://prometheus.io/docs/prometheus/latest/querying/api/). `"isPartial":true` is set if the response contains partial data
+  because of a part of `vmstorage` nodes were unavailable during query processing.
+* FEATURE: improve performance for `/api/v1/series`, `/api/v1/labels` and `/api/v1/label/<labelName>/values` on time ranges exceeding one day.
+* FEATURE: vmagent: reduce memory usage when service discovery detects big number of scrape targets and the set of discovered targets changes over time.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/825
+* FEATURE: vmagent: add `-promscrape.dropOriginalLabels` command-line option, which can be used for reducing memory usage when scraping big number of targets.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/825#issuecomment-724308361 for details.
+* FEATURE: vmalert: explicitly set extra labels to alert entities. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/870
+* FEATURE: add `-search.treatDotsAsIsInRegexps` command-line flag, which can be used for automatic escaping of dots in regexp label filters used in queries.
+  For example, if `-search.treatDotsAsIsInRegexps` is set, then the query `foo{bar=~"aaa.bb.cc|dd.eee"}` is automatically converted to `foo{bar=~"aaa\\.bb\\.cc|dd\\.eee"}`.
+  This may be useful for querying Graphite data.
+* FEATURE: consistently return text-based HTTP responses such as `plain/text` and `application/json` with `charset=utf-8`.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/pull/897
+* FEATURE: update Go builder from v1.15.4 to v1.15.5. This should fix [these issues in Go](https://github.com/golang/go/issues?q=milestone%3AGo1.15.5+label%3ACherryPickApproved).
+* FEATURE: added `/internal/force_flush` http handler for flushing recently ingested data from in-memory buffers to persistent storage.
+  See [troubleshooting docs](https://victoriametrics.github.io/#troubleshooting) for more details.
+* FEATURE: added [Graphite Tags API](https://graphite.readthedocs.io/en/stable/tags.html) support.
+  See [these docs](https://victoriametrics.github.io/#graphite-tags-api-usage) for details.
+
+* BUGFIX: do not return data points in the end of the selected time range for time series ending in the middle of the selected time range.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/887 and https://github.com/VictoriaMetrics/VictoriaMetrics/issues/845
+* BUGFIX: remove spikes at the end of time series gaps for `increase()` or `delta()` functions. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/894
+* BUGFIX: vminsert: properly return HTTP 503 status code when all the vmstorage nodes are unavailable. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/896
+
+
+# [v1.46.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.46.0)
+
+* FEATURE: optimize requests to `/api/v1/labels` and `/api/v1/label/<name>/values` when `start` and `end` args are set.
+* FEATURE: reduce memory usage when query touches big number of time series.
+* FEATURE: vmagent: reduce memory usage when `kubernetes_sd_config` discovers big number of scrape targets (e.g. hundreds of thousands) and the majority of these targets (99%)
+  are dropped during relabeling. Previously labels for all the dropped targets were displayed at `/api/v1/targets` page. Now only up to `-promscrape.maxDroppedTargets` such
+  targets are displayed. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/878 for details.
+* FEATURE: vmagent: reduce memory usage when scraping big number of targets with big number of temporary labels starting with `__`.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/825
+* FEATURE: vmagent: add `/ready` HTTP endpoint, which returns 200 OK status code when all the service discovery has been initialized.
+  This may be useful during rolling upgrades. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/875
+
+* BUGFIX: vmagent: eliminate data race when `-promscrape.streamParse` command-line is set. Previously this mode could result in scraped metrics with garbage labels.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/825#issuecomment-723198247 for details.
+* BUGFIX: properly calculate `topk_*` and `bottomk_*` functions from [MetricsQL](https://victoriametrics.github.io/MetricsQL.html) for time series with gaps.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/pull/883
+
+
+# [v1.45.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.45.0)
+
+* FEATURE: allow setting `-retentionPeriod` smaller than one month. I.e. `-retentionPeriod=3d`, `-retentionPeriod=2w`, etc. is supported now.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/173
+* FEATURE: optimize more cases according to https://utcc.utoronto.ca/~cks/space/blog/sysadmin/PrometheusLabelNonOptimization . Now the following cases are optimized too:
+  * `rollup_func(foo{filters}[d]) op bar` -> `rollup_func(foo{filters}[d]) op bar{filters}`
+  * `transform_func(foo{filters}) op bar` -> `transform_func(foo{filters}) op bar{filters}`
+  * `num_or_scalar op foo{filters} op bar` -> `num_or_scalar op foo{filters} op bar{filters}`
+* FEATURE: improve time series search for queries with multiple label filters. I.e. `foo{label1="value", label2=~"regexp"}`.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/781
+* FEATURE: vmagent: add `stream parse` mode. This mode allows reducing memory usage when individual scrape targets expose tens of millions of metrics.
+  For example, during scraping Prometheus in [federation](https://prometheus.io/docs/prometheus/latest/federation/) mode.
+  See `-promscrape.streamParse` command-line option and `stream_parse: true` config option for `scrape_config` section in `-promscrape.config`.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/825 and [troubleshooting docs for vmagent](https://victoriametrics.github.io/vmagent.html#troubleshooting).
+* FEATURE: vmalert: add `-dryRun` command-line option for validating the provided config files without the need to start `vmalert` service.
+* FEATURE: accept optional third argument of string type at `topk_*` and `bottomk_*` functions. This is label name for additional time series to return with the sum of time series outside top/bottom K. See [MetricsQL docs](https://victoriametrics.github.io/MetricsQL.html) for more details.
+* FEATURE: vmagent: expose `/api/v1/targets` page according to [the corresponding Prometheus API](https://prometheus.io/docs/prometheus/latest/querying/api/#targets).
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/643
+
+* BUGFIX: vmagent: properly handle OpenStack endpoint ending with `v3.0` such as `https://ostack.example.com:5000/v3.0`
+  in the same way as Prometheus does. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/728#issuecomment-709914803
+* BUGFIX: drop trailing data points for time series with a single raw sample. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/748
+* BUGFIX: do not drop trailing data points for instant queries to `/api/v1/query`. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/845
+* BUGFIX: vmbackup: fix panic when `-origin` isn't specified. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/856
+* BUGFIX: vmalert: skip automatically added labels on alerts restore. Label `alertgroup` was introduced in [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/611)
+  and automatically added to generated time series. By mistake, this new label wasn't correctly purged on restore event and affected alert's ID uniqueness.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/870
+* BUGFIX: vmagent: fix panic at scrape error body formating. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/864
+* BUGFIX: vmagent: add leading missing slash to metrics path like Prometheus does. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/835
+* BUGFIX: vmagent: drop packet if remote storage returns 4xx status code. This make the behaviour consistent with Prometheus.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/873
+* BUGFIX: vmagent: properly handle 301 redirects. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/869
+
+
+# [v1.44.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.44.0)
+
+* FEATURE: automatically add missing label filters to binary operands as described at https://utcc.utoronto.ca/~cks/space/blog/sysadmin/PrometheusLabelNonOptimization .
+  This should improve performance for queries with missing label filters in binary operands. For example, the following query should work faster now, because it shouldn't
+  fetch and discard time series for `node_filesystem_files_free` metric without matching labels for the left side of the expression:
+  ```
+     node_filesystem_files{ host="$host", mountpoint="/" } - node_filesystem_files_free
+  ```
+* FEATURE: vmagent: add Docker Swarm service discovery (aka [dockerswarm_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#dockerswarm_sd_config)).
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/656
+* FEATURE: add ability to export data in CSV format. See [these docs](https://victoriametrics.github.io/#how-to-export-csv-data) for details.
+* FEATURE: vmagent: add `-promscrape.suppressDuplicateScrapeTargetErrors` command-line flag for suppressing `duplicate scrape target` errors.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/651 and https://victoriametrics.github.io/vmagent.html#troubleshooting .
+* FEATURE: vmagent: show original labels before relabeling is applied on `duplicate scrape target` errors. This should simplify debugging for incorrect relabeling.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/651
+* FEATURE: vmagent: `/targets` page now accepts optional `show_original_labels=1` query arg for displaying original labels for each target before relabeling is applied.
+  This should simplify debugging for target relabeling configs. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/651
+* FEATURE: add `-finalMergeDelay` command-line flag for configuring the delay before final merge for per-month partitions.
+  The final merge is started after no new data is ingested into per-month partition during `-finalMergeDelay`.
+* FEATURE: add `vm_rows_added_to_storage_total` metric, which shows the total number of rows added to storage since app start.
+  The `sum(rate(vm_rows_added_to_storage_total))` can be smaller than `sum(rate(vm_rows_inserted_total))` if certain metrics are dropped
+  due to [relabeling](https://victoriametrics.github.io/#relabeling). The `sum(rate(vm_rows_added_to_storage_total))` can be bigger
+  than `sum(rate(vm_rows_inserted_total))` if [replication](https://victoriametrics.github.io/Cluster-VictoriaMetrics.html#replication-and-data-safety) is enabled.
+* FEATURE: keep metric name after applying [MetricsQL](https://victoriametrics.github.io/MetricsQL.html) functions, which don't change time series meaning.
+  The list of such functions:
+   * `keep_last_value`
+   * `keep_next_value`
+   * `interpolate`
+   * `running_min`
+   * `running_max`
+   * `running_avg`
+   * `range_min`
+   * `range_max`
+   * `range_avg`
+   * `range_first`
+   * `range_last`
+   * `range_quantile`
+   * `smooth_exponential`
+   * `ceil`
+   * `floor`
+   * `round`
+   * `clamp_min`
+   * `clamp_max`
+   * `max_over_time`
+   * `min_over_time`
+   * `avg_over_time`
+   * `quantile_over_time`
+   * `mode_over_time`
+   * `geomean_over_time`
+   * `holt_winters`
+   * `predict_linear`
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/674
+
+* BUGFIX: properly handle stale time series after K8S deployment. Previously such time series could be double-counted.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/748
+* BUGFIX: return a single time series at max from `absent()` function like Prometheus does.
+* BUGFIX: vmalert: accept days, weeks and years in `for: ` part of config like Prometheus does. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/817
+* BUGFIX: fix `mode_over_time(m[d])` calculations. Previously the function could return incorrect results.
+
+
+# [v1.43.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.43.0)
+
+* FEATURE: reduce CPU usage for repeated queries over sliding time window when no new time series are added to the database.
+  Typical use cases: repeated evaluation of alerting rules in [vmalert](https://victoriametrics.github.io/vmalert.html) or dashboard auto-refresh in Grafana.
+* FEATURE: vmagent: add OpenStack service discovery aka [openstack_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#openstack_sd_config).
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/728 .
+* FEATURE: vmalert: make `-maxIdleConnections` configurable for datasource HTTP client. This option can be used for minimizing connection churn.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/795 .
+* FEATURE: add `-influx.maxLineSize` command-line flag for configuring the maximum size for a single Influx line during parsing.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/807
+
+* BUGFIX: properly handle `inf` values during [background merge of LSM parts](https://medium.com/@valyala/how-victoriametrics-makes-instant-snapshots-for-multi-terabyte-time-series-data-e1f3fb0e0282).
+  Previously `Inf` values could result in `NaN` values for adjancent samples in time series. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/805 .
+* BUGFIX: fill gaps on graphs for `range_*` and `running_*` functions. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/806 .
+* BUGFIX: make a copy of label with new name during relabeling with `action: labelmap` in the same way as Prometheus does.
+  Previously the original label name has been replaced. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/812 .
+* BUGFIX: support parsing floating-point timestamp like Graphite Carbon does. Such timestmaps are truncated to seconds.
+
+
+# [v1.42.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.42.0)
+
+* FEATURE: use all the available CPU cores when accepting data via a single TCP connection
+  for [all the supported protocols](https://victoriametrics.github.io/#how-to-import-time-series-data).
+  Previously data ingested via a single TCP connection could use only a single CPU core. This could limit data ingestion performance.
+  The main benefit of this feature is that data can be imported at max speed via a single connection - there is no need to open multiple concurrent
+  connections to VictoriaMetrics or [vmagent](https://victoriametrics.github.io/vmagent.html) in order to achieve the maximum data ingestion speed.
+* FEATURE: cluster: improve performance for data ingestion path from `vminsert` to `vmstorage` nodes. The maximum data ingestion performance
+  for a single connection between `vminsert` and `vmstorage` node scales with the number of available CPU cores on `vmstorage` side.
+  This should help with https://github.com/VictoriaMetrics/VictoriaMetrics/issues/791 .
+* FEATURE: add ability to export / import data in native format via `/api/v1/export/native` and `/api/v1/import/native`.
+  This is the most optimized approach for data migration between VictoriaMetrics instances. Both single-node and cluster instances are supported.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/787#issuecomment-700632551 .
+* FEATURE: add `reduce_mem_usage` query option to `/api/v1/export` in order to reduce memory usage during data export / import.
+  See [these docs](https://victoriametrics.github.io/#how-to-export-data-in-json-line-format) for details.
+* FEATURE: improve performance for `/api/v1/series` handler when it returns big number of time series.
+* FEATURE: add `vm_merge_need_free_disk_space` metric, which can be used for estimating the number of deferred background data merges due to the lack of free disk space.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/686 .
+* FEATURE: add OpenBSD support. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/785 .
+
+* BUGFIX: properly apply `-search.maxStalenessInterval` command-line flag value. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/784 .
+* BUGFIX: fix displaying data in Grafana tables. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/720 .
+* BUGFIX: do not adjust the number of detected CPU cores found at `/sys/devices/system/cpu/online`.
+  The adjustement was increasing the resulting GOMAXPROC by 1, which looked confusing to users.
+  See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/685#issuecomment-698595309 .
+* BUGFIX: vmagent: do not show `-remoteWrite.url` in initial logs if `-remoteWrite.showURL` isn't set. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/773 .
+* BUGFIX: properly handle case when [/metrics/find](https://victoriametrics.github.io/#graphite-metrics-api-usage) finds both a leaf and a node for the given `query=prefix.*`.
+  In this case only the node must be returned with stripped dot in the end of id as carbonapi does.
+
+
+# Previous releases
+
+See [releases page](https://github.com/VictoriaMetrics/VictoriaMetrics/releases).
--- a/docs/CaseStudies.md
+++ b/docs/CaseStudies.md
@@ -3,10 +3,24 @@
 Below are approved public case studies and talks from VictoriaMetrics users. Join our [community Slack channel](http://slack.victoriametrics.com/)
 and feel free asking for references, reviews and additional case studies from real VictoriaMetrics users there.

-See also [articles about VictoriaMetrics](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/Articles).
+See also [articles about VictoriaMetrics from our users](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/Articles#third-party-articles-and-slides).


-## Adidas
+* [adidas](#adidas)
+* [CERN](#cern)
+* [COLOPL](#colopl)
+* [Zerodha](#zerodha)
+* [Wix.com](#wixcom)
+* [Wedos.com](#wedoscom)
+* [Synthesio](#synthesio)
+* [Dreamteam](#dreamteam)
+* [Brandwatch](#brandwatch)
+* [Adsterra](#adsterra)
+* [ARNES](#arnes)
+* [Idealo.de](#idealode)
+
+
+## adidas

 See [slides](https://promcon.io/2019-munich/slides/remote-write-storage-wars.pdf) and [video](https://youtu.be/OsH6gPdxR4s)
 from [Remote Write Storage Wars](https://promcon.io/2019-munich/talks/remote-write-storage-wars/) talk at [PromCon 2019](https://promcon.io/2019-munich/).
@@ -74,12 +88,12 @@ See [Monitoring K8S with VictoriaMetrics](https://docs.google.com/presentation/d
 Numbers:

 * The number of active time series per VictoriaMetrics instance is 40M.
-* The total number of time series per VictoriaMetrics instance is 400M+.
+* The total number of time series per VictoriaMetrics instance is 5000M+.
 * Ingestion rate per VictoriaMetrics instance is 1M data points per second.
-* The total number of datapoints per VictoriaMetrics instance is 8 trillions.
-* The average time series churn rate is ~3M per day.
+* The total number of datapoints per VictoriaMetrics instance is 8.5 trillions.
+* The average time series churn rate is ~80M per day.
 * The average query rate is ~100 per second (mostly alert queries).
-* Query duration: median is ~70ms, 99th percentile is ~1.5sec.
+* Query duration: median is ~20ms, 99th percentile is ~1.5sec.
 * Retention: 3 months.

 > Alternatives that we’ve played with before choosing VictoriaMetrics are: federated Prometheus, Cortex, IronDB and Thanos.
@@ -304,3 +318,21 @@ Grafana has a LB infront, so if one DC has problems, we can still view all metri

 We are still in the process of migration, but we are really happy with the whole stack. It has proven as an essential piece
 for insight into our services during COVID-19 and has enabled us to provide better service and spot problems faster.
+
+
+## Idealo.de
+
+[idealo.de](https://www.idealo.de/) is the leading price comparison website in Germany. We use Prometheus for metrics on our container platform.
+When we introduced Prometheus at idealo we started with m3db as a longterm storage. In our setup m3db was quite unstable and consumed a lot of resources.
+
+VictoriaMetrics runs very stable for us and uses only a fraction of the resources. Although we also increased our retention time from 1 month to 13 months.
+
+Numbers:
+
+- The number of active time series per VictoriaMetrics instance is 21M.
+- Total ingestion rate 120k metrics per second.
+- The total number of datapoints 3.1 trillion.
+- The average time series churn rate is ~9M per day.
+- The average query rate is ~20 per second. Response time for 99th quantile is 120ms.
+- Retention: 13 months.
+- Size of all datapoints: 3.5 TB
--- a/docs/Cluster-VictoriaMetrics.md
+++ b/docs/Cluster-VictoriaMetrics.md
@@ -181,7 +181,7 @@ or [an alternative dashboard for VictoriaMetrics cluster](https://grafana.com/gr
     - `prometheus/api/v1/import/csv` - for importing arbitrary CSV data. See [these docs](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/README.md#how-to-import-csv-data) for details.
     - `prometheus/api/v1/import/prometheus` - for importing data in Prometheus exposition format. See [these docs](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/README.md#how-to-import-data-in-prometheus-exposition-format) for details.

-* URLs for [Prmetheus querying API](https://prometheus.io/docs/prometheus/latest/querying/api/): `http://<vmselect>:8481/select/<accountID>/prometheus/<suffix>`, where:
+* URLs for [Prometheus querying API](https://prometheus.io/docs/prometheus/latest/querying/api/): `http://<vmselect>:8481/select/<accountID>/prometheus/<suffix>`, where:
  - `<accountID>` is an arbitrary number identifying data namespace for the query (aka tenant)
  - `<suffix>` may have the following values:
    - `api/v1/query` - performs [PromQL instant query](https://prometheus.io/docs/prometheus/latest/querying/api/#instant-queries).
@@ -194,6 +194,8 @@ or [an alternative dashboard for VictoriaMetrics cluster](https://grafana.com/gr
    - `api/v1/export/native` - exports raw data in native binary format. It may be imported into another VictoriaMetrics via `api/v1/import/native` (see above).
    - `api/v1/export/csv` - exports data in CSV. It may be imported into another VictoriaMetrics via `api/v1/import/csv` (see above).
    - `api/v1/status/tsdb` - for time series stats. See [these docs](https://prometheus.io/docs/prometheus/latest/querying/api/#tsdb-stats) for details.
+      VictoriaMetrics accepts optional `topN=N` and `date=YYYY-MM-DD` query args for this handler, where `N` is the number of top entries to return in the response
+      and `YYYY-MM-DD` is the date for collecting the stats. By default the stats is collected for the current day.
    - `api/v1/status/active_queries` - for currently executed active queries. Note that every `vmselect` maintains an independent list of active queries,
      which is returned in the response.

@@ -203,6 +205,14 @@ or [an alternative dashboard for VictoriaMetrics cluster](https://grafana.com/gr
      - `metrics/find` - searches Graphite metrics. See [these docs](https://graphite-api.readthedocs.io/en/latest/api.html#metrics-find).
      - `metrics/expand` - expands Graphite metrics. See [these docs](https://graphite-api.readthedocs.io/en/latest/api.html#metrics-expand).
      - `metrics/index.json` - returns all the metric names. See [these docs](https://graphite-api.readthedocs.io/en/latest/api.html#metrics-index-json).
+      - `tags/tagSeries` - registers time series. See [these docs](https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb).
+      - `tags/tagMultiSeries` - register multiple time series. See [these docs](https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb).
+      - `tags` - returns tag names. See [these docs](https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags).
+      - `tags/<tag_name>` - returns tag values for the given `<tag_name>`. See [these docs](https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags).
+      - `tags/findSeries` - returns series matching the given `expr`. See [these docs](https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags).
+      - `tags/autoComplete/tags` - returns tags matching the given `tagPrefix` and/or `expr`. See [these docs](https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support).
+      - `tags/autoComplete/values` - returns tag values matching the given `valuePrefix` and/or `expr`. See [these docs](https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support).
+      - `tags/delSeries` - deletes series matching the given `path`. See [these docs](https://graphite.readthedocs.io/en/stable/tags.html#removing-series-from-the-tagdb).

 * URL for time series deletion: `http://<vmselect>:8481/delete/<accountID>/prometheus/api/v1/admin/tsdb/delete_series?match[]=<timeseries_selector_for_delete>`.
  Note that the `delete_series` handler should be used only in exceptional cases such as deletion of accidentally ingested incorrect time series. It shouldn't
@@ -278,7 +288,7 @@ Each instance type - `vminsert`, `vmselect` and `vmstorage` - can run on the mos
 #### vmstorage

 * The recommended total number of vCPU cores for all the `vmstorage` instances can be calculated from the ingestion rate: `vCPUs = ingestion_rate / 150K`.
-* The recommended total amount of RAM for all the `vmstorage` instances can be calculated from the number of active time series: `RAM = active_time_series * 1KB`.
+* The recommended total amount of RAM for all the `vmstorage` instances can be calculated from the number of active time series: `RAM = 2 * active_time_series * 1KB`.
  Time series is active if it received at least a single data point during the last hour or if it has been queried during the last hour.
  The required RAM per each `vmstorage` should be multiplied by `-replicationFactor` if [replication](#replication-and-data-safety) is enabled.
  Additional RAM can be required for query processing.
@@ -328,8 +338,9 @@ In order to enable application-level replication, `-replicationFactor=N` command
 This guarantees that all the data remains available for querying if up to `N-1` `vmstorage` nodes are unavailable.
 For example, when `-replicationFactor=3` is passed to `vminsert`, then it replicates all the ingested data to 3 distinct `vmstorage` nodes.

-When the replication is enabled, `-dedup.minScrapeInterval=1ms` command-line flag must be passed to `vmselect`
-in order to de-duplicate replicated data during queries. It is OK if `-dedup.minScrapeInterval` exceeds 1ms
+When the replication is enabled, `-replicationFactor=N` and `-dedup.minScrapeInterval=1ms` command-line flag must be passed to `vmselect` nodes.
+The `-replicationFactor=N` improves query performance when a part of vmstorage nodes respond slowly and/or temporarily unavailable.
+The `-dedup.minScrapeInterval=1ms` de-duplicates replicated data during queries. It is OK if `-dedup.minScrapeInterval` exceeds 1ms
 when [deduplication](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/README.md#deduplication) is used additionally to replication.

 Note that [replication doesn't save from disaster](https://medium.com/@valyala/speeding-up-backups-for-big-time-series-databases-533c1a927883),
--- a/docs/FAQ.md
+++ b/docs/FAQ.md
@@ -90,7 +90,7 @@ The main differences between Cortex and VictoriaMetrics:
 - Cortex may lose up to 12 hours of recent data on Ingestor failure - see [the corresponding docs](https://github.com/cortexproject/cortex/blob/fe56f1420099aa1bf1ce09316c186e05bddee879/docs/architecture.md#ingesters-failure-and-data-loss).
  VictoriaMetrics may lose only a few seconds of recent data, which isn't synced to persistent storage yet.
  See [this article for details](https://medium.com/@valyala/wal-usage-looks-broken-in-modern-time-series-databases-b62a627ab704).
- Cortex is usually slower and requires more CPU and RAM than VictoriaMetrics. See [this talk from Adidas at PromCon 2019](https://promcon.io/2019-munich/talks/remote-write-storage-wars/) and [other case studies](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies).
+- Cortex is usually slower and requires more CPU and RAM than VictoriaMetrics. See [this talk from adidas at PromCon 2019](https://promcon.io/2019-munich/talks/remote-write-storage-wars/) and [other case studies](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies).
 - VictoriaMetrics accepts data in multiple popular data ingestion protocols additionally to Prometheus remote_write protocol - InfluxDB, OpenTSDB, Graphite, CSV, JSON, native binary.
  See [these docs](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/README.md#how-to-import-time-series-data) for details.

@@ -111,7 +111,7 @@ The main differences between Cortex and VictoriaMetrics:
  which is much easier to setup and operate than Thanos components.
 - Thanos may be harder to setup and operate comparing to VictoriaMetrics, since it has more moving parts, which can be connected with less reliable networks.
  See [this article for details](https://medium.com/faun/comparing-thanos-to-victoriametrics-cluster-b193bea1683).
- Thanos is usually slower and requires more CPU and RAM than VictoriaMetrics. See [this talk from Adidas at PromCon 2019](https://promcon.io/2019-munich/talks/remote-write-storage-wars/).
+- Thanos is usually slower and requires more CPU and RAM than VictoriaMetrics. See [this talk from adidas at PromCon 2019](https://promcon.io/2019-munich/talks/remote-write-storage-wars/).
 - VictoriaMetrics accepts data in multiple popular data ingestion protocols additionally to Prometheus remote_write protocol - InfluxDB, OpenTSDB, Graphite, CSV, JSON, native binary.
  See [these docs](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/README.md#how-to-import-time-series-data) for details.

@@ -142,6 +142,7 @@ The architecture is [optimized for storing and querying large amounts of time se

 Yes:

+* [Prometheus vs VictoriaMetrics benchmark on node-exporter metrics](https://valyala.medium.com/prometheus-vs-victoriametrics-benchmark-on-node-exporter-metrics-4ca29c75590f)
 * [Benchmarking time series workloads on Apache Kudu using TSBS](https://blog.cloudera.com/benchmarking-time-series-workloads-on-apache-kudu-using-tsbs/)
 * [Billy: how VictoriaMetrics deals with more than 500 billion rows](https://medium.com/@valyala/billy-how-victoriametrics-deals-with-more-than-500-billion-rows-e82ff8f725da)
 * [Measuring vertical scalability for time series databases: VictoriaMetrics vs InfluxDB vs TimescaleDB](https://medium.com/@valyala/measuring-vertical-scalability-for-time-series-databases-in-google-cloud-92550d78d8ae).
--- a/docs/MetricsQL.md
+++ b/docs/MetricsQL.md
@@ -28,7 +28,7 @@ Feel free [filing a feature request](https://github.com/VictoriaMetrics/Victoria

 This functionality can be tried at [an editable Grafana dashboard](http://play-grafana.victoriametrics.com:3000/d/4ome8yJmz/node-exporter-on-victoriametrics-demo).

- [`WITH` templates](https://play.victoriametrics.com/promql/expand-with-exprs). This feature simplifies writing and managing complex queries. Go to [`WITH` templates playground](https://victoriametrics.com/promql/expand-with-exprs) and try it.
+- [`WITH` templates](https://play.victoriametrics.com/promql/expand-with-exprs). This feature simplifies writing and managing complex queries. Go to [`WITH` templates playground](https://play.victoriametrics.com/promql/expand-with-exprs) and try it.
 - Range duration in functions such as [rate](https://prometheus.io/docs/prometheus/latest/querying/functions/#rate()) may be omitted. VictoriaMetrics automatically selects range duration depending on the current step used for building the graph. For instance, the following query is valid in VictoriaMetrics: `rate(node_network_receive_bytes_total)`.
 - All the aggregate functions support optional `limit N` suffix in order to limit the number of output series. For example, `sum(x) by (y) limit 10` limits
  the number of output time series after the aggregation to 10. All the other time series are dropped.
@@ -57,6 +57,8 @@ This functionality can be tried at [an editable Grafana dashboard](http://play-g
  - `alias(q, name)` for setting metric name across all the time series `q`.
  - `label_set(q, label1, value1, ... labelN, valueN)` for setting the given values for the given labels on `q`.
  - `label_map(q, label, srcValue1, dstValue1, ... srcValueN, dstValueN)` for mapping `label` values from `src*` to `dst*`.
+  - `label_uppercase(q, label1, ... labelN)` for uppercasing values for the given labels.
+  - `label_lowercase(q, label2, ... labelN)` for lowercasing value for the given labels.
  - `label_del(q, label1, ... labelN)` for deleting the given labels from `q`.
  - `label_keep(q, label1, ... labelN)` for deleting all the labels except the given labels from `q`.
  - `label_copy(q, src_label1, dst_label1, ... src_labelN, dst_labelN)` for copying label values from `src_*` to `dst_*`.
@@ -68,15 +70,15 @@ This functionality can be tried at [an editable Grafana dashboard](http://play-g
 - `step()` function for returning the step in seconds used in the query.
 - `start()` and `end()` functions for returning the start and end timestamps of the `[start ... end]` range used in the query.
 - `integrate(m[d])` for returning integral over the given duration `d` for the given metric `m`.
- `ideriv(m)` - for calculating `instant` derivative for `m`.
+- `ideriv(m[d])` - for calculating `instant` derivative for the metric `m` over the duration `d`.
 - `deriv_fast(m[d])` - for calculating `fast` derivative for `m` based on the first and the last points from duration `d`.
 - `running_` functions - `running_sum`, `running_min`, `running_max`, `running_avg` - for calculating [running values](https://en.wikipedia.org/wiki/Running_total) on the selected time range.
- `range_` functions - `range_sum`, `range_min`, `range_max`, `range_avg`, `range_first`, `range_last`, `range_median`, `range_quantile` - for calculating global value over the selected time range.
+- `range_` functions - `range_sum`, `range_min`, `range_max`, `range_avg`, `range_first`, `range_last`, `range_median`, `range_quantile` - for calculating global value over the selected time range. Note that global value is based on calculated datapoints for the inner query. The calculated datapoints can differ from raw datapoints stored in the database. See [these docs](https://prometheus.io/docs/prometheus/latest/querying/basics/#staleness) for details.
 - `smooth_exponential(q, sf)` - smooths `q` using [exponential moving average](https://en.wikipedia.org/wiki/Moving_average#Exponential_moving_average) with the given smooth factor `sf`.
 - `remove_resets(q)` - removes counter resets from `q`.
- `lag(q[d])` - returns lag between the current timestamp and the timestamp from the previous data point in `q` over `d`.
- `lifetime(q[d])` - returns lifetime of `q` over `d` in seconds. It is expected that `d` exceeds the lifetime of `q`.
- `scrape_interval(q[d])` - returns the average interval in seconds between data points of `q` over `d` aka `scrape interval`.
+- `lag(m[d])` - returns lag between the current timestamp and the timestamp from the previous data point in `m` over `d`.
+- `lifetime(m[d])` - returns lifetime of `q` over `d` in seconds. It is expected that `d` exceeds the lifetime of `m`.
+- `scrape_interval(m[d])` - returns the average interval in seconds between data points of `m` over `d` aka `scrape interval`.
 - Trigonometric functions - `sin(q)`, `cos(q)`, `asin(q)`, `acos(q)` and `pi()`.
 - `range_over_time(m[d])` - returns value range for `m` over `d` time window, i.e. `max_over_time(m[d])-min_over_time(m[d])`.
 - `median_over_time(m[d])` - calculates median values for `m` over `d` time window. Shorthand to `quantile_over_time(0.5, m[d])`.
@@ -101,8 +103,8 @@ This functionality can be tried at [an editable Grafana dashboard](http://play-g
 - `histogram_over_time(m[d])` - calculates [VictoriaMetrics histogram](https://godoc.org/github.com/VictoriaMetrics/metrics#Histogram) for `m` over `d`.
  For example, the following query calculates median temperature by country over the last 24 hours:
  `histogram_quantile(0.5, sum(histogram_over_time(temperature[24h])) by (vmbucket, country))`.
- `histogram_share(le, buckets)` - returns share (in the range 0..1) for `buckets`. Useful for calculating SLI and SLO.
-  For instance, the following query returns the share of requests which are performed under 1.5 seconds: `histogram_share(1.5, sum(request_duration_seconds_bucket) by (le))`.
+- `histogram_share(le, buckets)` - returns share (in the range 0..1) for `buckets` that fall below `le`. Useful for calculating SLI and SLO.
+  For instance, the following query returns the share of requests which are performed under 1.5 seconds during the last 5 minutes: `histogram_share(1.5, sum(rate(request_duration_seconds_bucket[5m])) by (le))`.
 - `topk_*` and `bottomk_*` aggregate functions, which return up to K time series. Note that the standard `topk` function may return more than K time series -
   see [this article](https://www.robustperception.io/graph-top-n-time-series-in-grafana) for details.
   - `topk_min(k, q)` - returns top K time series with the max minimums on the given time range
@@ -112,13 +114,17 @@ This functionality can be tried at [an editable Grafana dashboard](http://play-g
   - `bottomk_min(k, q)` - returns bottom K time series with the min minimums on the given time range
   - `bottomk_max(k, q)` - returns bottom K time series with the min maximums on the given time range
   - `bottomk_avg(k, q)` - returns bottom K time series with the min averages on the given time range
-   - `bottomk_median(k, q)` - returns bottom K time series with the min medians on the given time range
+   - `bottomk_median(k, q)` - returns bottom K time series with the min medians on the given time range.
+
+  All the `topk_*` and `bottomk_*` functions accept optional third argument - label name for the sum of the remaining time series outside top K or bottom K time series. For example, `topk_max(3, process_resident_memory_bytes, "remaining_sum")` would return up to 3 time series with the maximum value for `process_resident_memory_bytes` plus fourth time series with the sum of the remaining time series if any. The fourth time series will contain `remaining_sum="remaining_sum"` additional label.
 - `share_le_over_time(m[d], le)` - returns share (in the range 0..1) of values in `m` over `d`, which are smaller or equal to `le`. Useful for calculating SLI and SLO.
  Example: `share_le_over_time(memory_usage_bytes[24h], 100*1024*1024)` returns the share of time series values for the last 24 hours when memory usage was below or equal to 100MB.
 - `share_gt_over_time(m[d], gt)` - returns share (in the range 0..1) of values in `m` over `d`, which are bigger than `gt`. Useful for calculating SLI and SLO.
  Example: `share_gt_over_time(up[24h], 0)` - returns service availability for the last 24 hours.
 - `count_le_over_time(m[d], le)` - returns the number of raw samples for `m` over `d`, which don't exceed `le`.
 - `count_gt_over_time(m[d], gt)` - returns the number of raw samples for `m` over `d`, which are bigger than `gt`.
+- `count_eq_over_time(m[d], N)` - returns the number of raw samples for `m` over `d` with values equal to `N`.
+- `count_ne_over_time(m[d], N)` - returns the number of raw samples for `m` over `d` with values not equal to `N`.
 - `tmin_over_time(m[d])` - returns timestamp for the minimum value for `m` over `d` time range.
 - `tmax_over_time(m[d])` - returns timestamp for the maximum value for `m` over `d` time range.
 - `aggr_over_time(("aggr_func1", "aggr_func2", ...), m[d])` - simultaneously calculates all the listed `aggr_func*` for `m` over `d` time range.
--- a/docs/Quick-Start.md
+++ b/docs/Quick-Start.md
@@ -1,14 +1,16 @@
 # Quick Start

-1. Download the latest VictoriaMetrics release from [releases page](https://github.com/VictoriaMetrics/VictoriaMetrics/releases),
+1. If you run Ubuntu, then just run `snap install victoriametrics` command in order to install and start VictoriaMetrics, then read [these docs](https://snapcraft.io/victoriametrics).
+   Otherwise download the latest VictoriaMetrics release from [releases page](https://github.com/VictoriaMetrics/VictoriaMetrics/releases),
   from [Docker hub](https://hub.docker.com/r/victoriametrics/victoria-metrics/)
   or [build it from sources](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/Single-server-VictoriaMetrics#how-to-build-from-sources).

-2. Run the binary or Docker image with the desired command-line flags. Pass `-help` in order to see description for all the available flags
+2. This step isn't needed if you run VictoriaMetrics via `snap install victoriametrics` as described above.
+   Otherwise run the binary or Docker image with the desired command-line flags. Pass `-help` in order to see description for all the available flags
   and their default values. Default flag values should fit the majoirty of cases. The minimum required flags to configure are:

   * `-storageDataPath` - path to directory where VictoriaMetrics stores all the data.
-   * `-retentionPeriod` - data retention in months.
+   * `-retentionPeriod` - data retention.

   For instance:

--- a/docs/Release-Guide.md
+++ b/docs/Release-Guide.md
@@ -2,7 +2,7 @@ Release process guidance

 ## Release version and Docker images

-0. Document all the changes for new release in [CHANGELOG.md](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/CHANGELOG.md).
+0. Document all the changes for new release in [CHANGELOG.md](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/docs/CHANGELOG.md).
 1. Create release tag with `git tag v1.xx.y`.
 2. Run `make release` for creating `*.tar.gz` release archive with the corresponding `_checksums.txt` inside `bin` directory.
 3. Run `make publish` for creating and publishing Docker images.
--- a/docs/Single-server-VictoriaMetrics.md
+++ b/docs/Single-server-VictoriaMetrics.md
@@ -10,23 +10,27 @@

 ## VictoriaMetrics

-VictoriaMetrics is fast, cost-effective and scalable time-series database.
+VictoriaMetrics is fast, cost-effective and scalable monitoring solution and time series database.

 It is available in [binary releases](https://github.com/VictoriaMetrics/VictoriaMetrics/releases),
-[docker images](https://hub.docker.com/r/victoriametrics/victoria-metrics/) and
-in [source code](https://github.com/VictoriaMetrics/VictoriaMetrics). Just download VictoriaMetrics and see [how to start it](#how-to-start-victoriametrics).
+[docker images](https://hub.docker.com/r/victoriametrics/victoria-metrics/), [Snap package](https://snapcraft.io/victoriametrics)
+and in [source code](https://github.com/VictoriaMetrics/VictoriaMetrics). Just download VictoriaMetrics and see [how to start it](#how-to-start-victoriametrics).
+If you use Ubuntu, then just run `snap install victoriametrics` in order to install and run it.
+Then read [Prometheus setup](#prometheus-setup) and [Grafana setup](#grafana-setup) docs.

 Cluster version is available [here](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/cluster).

 See our [Wiki](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki) for additional documentation.

 [Contact us](mailto:info@victoriametrics.com) if you need paid enterprise support for VictoriaMetrics.
-See [features available for enterprise customers](https://github.com/VictoriaMetrics/VictoriaMetrics/issues?q=is%3Aissue+label%3Aenterprise).
+See [features available for enterprise customers](https://victoriametrics.com/enterprise.html).


 ## Case studies and talks

-* [Adidas](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#adidas)
+Click on a link in order to read the corresponding case study
+
+* [adidas](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#adidas)
 * [CERN](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#cern)
 * [COLOPL](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#colopl)
 * [Zerodha](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#zerodha)
@@ -38,23 +42,29 @@ See [features available for enterprise customers](https://github.com/VictoriaMet
 * [Brandwatch](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#brandwatch)
 * [Adsterra](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#adsterra)
 * [ARNES](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#arnes)
+* [Idealo.de](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/CaseStudies#idealode)


 ## Prominent features

 * VictoriaMetrics can be used as long-term storage for Prometheus or for [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md).
  See [these docs](#prometheus-setup) for details.
-* Supports [Prometheus querying API](https://prometheus.io/docs/prometheus/latest/querying/api/), so it can be used as Prometheus drop-in replacement in Grafana.
-  VictoriaMetrics implements [MetricsQL](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/MetricsQL) query language, which is inspired by PromQL.
-* Supports global query view. Multiple Prometheus instances may write data into VictoriaMetrics. Later this data may be used in a single query.
+* VictoriaMetrics supports [Prometheus querying API](https://prometheus.io/docs/prometheus/latest/querying/api/), so it can be used as Prometheus drop-in replacement in Grafana.
+* VictoriaMetrics implements [MetricsQL](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/MetricsQL) query language backwards compatible with PromQL.
+* VictoriaMetrics provides global query view. Multiple Prometheus instances or any other data sources may ingest data into VictoriaMetrics.
+  Later this data may be queried via a single query.
 * High performance and good scalability for both [inserts](https://medium.com/@valyala/high-cardinality-tsdb-benchmarks-victoriametrics-vs-timescaledb-vs-influxdb-13e6ee64dd6b)
  and [selects](https://medium.com/@valyala/when-size-matters-benchmarking-victoriametrics-vs-timescale-and-influxdb-6035811952d4).
  [Outperforms InfluxDB and TimescaleDB by up to 20x](https://medium.com/@valyala/measuring-vertical-scalability-for-time-series-databases-in-google-cloud-92550d78d8ae).
-* [Uses 10x less RAM than InfluxDB](https://medium.com/@valyala/insert-benchmarks-with-inch-influxdb-vs-victoriametrics-e31a41ae2893) when working with millions of unique time series (aka high cardinality).
+* [Uses 10x less RAM than InfluxDB](https://medium.com/@valyala/insert-benchmarks-with-inch-influxdb-vs-victoriametrics-e31a41ae2893)
+  and [up to 7x less RAM than Prometheus, Thanos or Cortex](https://valyala.medium.com/prometheus-vs-victoriametrics-benchmark-on-node-exporter-metrics-4ca29c75590f)
+  when dealing with millions of unique time series (aka high cardinality).
 * Optimized for time series with high churn rate. Think about [prometheus-operator](https://github.com/coreos/prometheus-operator) metrics from frequent deployments in Kubernetes.
 * High data compression, so [up to 70x more data points](https://medium.com/@valyala/when-size-matters-benchmarking-victoriametrics-vs-timescale-and-influxdb-6035811952d4)
-  may be crammed into limited storage comparing to TimescaleDB.
-* Optimized for storage with high-latency IO and low IOPS (HDD and network storage in AWS, Google Cloud, Microsoft Azure, etc). See [graphs from these benchmarks](https://medium.com/@valyala/high-cardinality-tsdb-benchmarks-victoriametrics-vs-timescaledb-vs-influxdb-13e6ee64dd6b).
+  may be crammed into limited storage comparing to TimescaleDB
+  and [up to 7x less storage space is required comparing to Prometheus, Thanos or Cortex](https://valyala.medium.com/prometheus-vs-victoriametrics-benchmark-on-node-exporter-metrics-4ca29c75590f).
+* Optimized for storage with high-latency IO and low IOPS (HDD and network storage in AWS, Google Cloud, Microsoft Azure, etc).
+  See [graphs from these benchmarks](https://medium.com/@valyala/high-cardinality-tsdb-benchmarks-victoriametrics-vs-timescaledb-vs-influxdb-13e6ee64dd6b).
 * A single-node VictoriaMetrics may substitute moderately sized clusters built with competing solutions such as Thanos, M3DB, Cortex, InfluxDB or TimescaleDB.
  See [vertical scalability benchmarks](https://medium.com/@valyala/measuring-vertical-scalability-for-time-series-databases-in-google-cloud-92550d78d8ae),
  [comparing Thanos to VictoriaMetrics cluster](https://medium.com/@valyala/comparing-thanos-to-victoriametrics-cluster-b193bea1683)
@@ -63,7 +73,7 @@ See [features available for enterprise customers](https://github.com/VictoriaMet
 * Easy operation:
  * VictoriaMetrics consists of a single [small executable](https://medium.com/@valyala/stripping-dependency-bloat-in-victoriametrics-docker-image-983fb5912b0d) without external dependencies.
  * All the configuration is done via explicit command-line flags with reasonable defaults.
-  * All the data is stored in a single directory pointed by `-storageDataPath` flag.
+  * All the data is stored in a single directory pointed by `-storageDataPath` command-line flag.
  * Easy and fast backups from [instant snapshots](https://medium.com/@valyala/how-victoriametrics-makes-instant-snapshots-for-multi-terabyte-time-series-data-e1f3fb0e0282)
  to S3 or GCS with [vmbackup](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmbackup/README.md) / [vmrestore](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmrestore/README.md).
  See [this article](https://medium.com/@valyala/speeding-up-backups-for-big-time-series-databases-533c1a927883) for more details.
@@ -86,6 +96,7 @@ See [features available for enterprise customers](https://github.com/VictoriaMet
 * Has open source [cluster version](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/cluster).
 * See also technical [Articles about VictoriaMetrics](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/Articles).

+
 ## Operation

 ### Table of contents
@@ -103,7 +114,9 @@ See [features available for enterprise customers](https://github.com/VictoriaMet
 * [How to send data from OpenTSDB-compatible agents](#how-to-send-data-from-opentsdb-compatible-agents)
 * [Prometheus querying API usage](#prometheus-querying-api-usage)
  * [Prometheus querying API enhancements](#prometheus-querying-api-enhancements)
-* [Graphite Metrics API usage](#graphite-metrics-api-usage)
+* [Graphite API usage](#graphite-api-usage)
+  * [Graphite Metrics API usage](#graphite-metrics-api-usage)
+  * [Graphite Tags API usage](#graphite-tags-api-usage)
 * [How to build from sources](#how-to-build-from-sources)
  * [Development build](#development-build)
  * [Production build](#production-build)
@@ -155,33 +168,36 @@ See [features available for enterprise customers](https://github.com/VictoriaMet
    * [Color Palette](#color-palette)
  * [We kindly ask](#we-kindly-ask)

-### How to start VictoriaMetrics

-Just start VictoriaMetrics [executable](https://github.com/VictoriaMetrics/VictoriaMetrics/releases)
+## How to start VictoriaMetrics
+
+Start VictoriaMetrics [executable](https://github.com/VictoriaMetrics/VictoriaMetrics/releases)
 or [docker image](https://hub.docker.com/r/victoriametrics/victoria-metrics/) with the desired command-line flags.

 The following command-line flags are used the most:

-* `-storageDataPath` - path to data directory. VictoriaMetrics stores all the data in this directory. Default path is `victoria-metrics-data` in current working directory.
-* `-retentionPeriod` - retention period in months for the data. Older data is automatically deleted. Default period is 1 month.
+* `-storageDataPath` - path to data directory. VictoriaMetrics stores all the data in this directory. Default path is `victoria-metrics-data` in the current working directory.
+* `-retentionPeriod` - retention for stored data. Older data is automatically deleted. Default retention is 1 month. See [these docs](#retention) for more details.

-Other flags have good enough default values, so set them only if you really need this.
+Other flags have good enough default values, so set them only if you really need this. Pass `-help` to see all the available flags with description and default values.
+
+See how to [ingest data to VictoriaMetrics](#how-to-import-time-series-data), how to [query VictoriaMetrics](#grafana-setup)
+and how to [handle alerts](#alerting).
 VictoriaMetrics accepts [Prometheus querying API requests](#prometheus-querying-api-usage) on port `8428` by default.

-Pass `-help` to see all the available flags with description and default values.
-
 It is recommended setting up [monitoring](#monitoring) for VictoriaMetrics.

-#### Environment variables
+### Environment variables

-Each flag values can be set thru environment variables by following these rules:
+Each flag value can be set via environment variables according to these rules:

 * The `-envflag.enable` flag must be set
-* Each `.` in flag names must be substituted by `_` (for example `-insert.maxQueueDuration <duration>` will translate to `insert_maxQueueDuration=<duration>`)
-* For repeating flags, an alternative syntax can be used by joining the different values into one using `,` as separator (for example `-storageNode <nodeA> -storageNode <nodeB>` will translate to `storageNode=<nodeA>,<nodeB>`)
+* Each `.` char in flag name must be substituted by `_` (for example `-insert.maxQueueDuration <duration>` will translate to `insert_maxQueueDuration=<duration>`)
+* For repeating flags an alternative syntax can be used by joining the different values into one using `,` char as separator (for example `-storageNode <nodeA> -storageNode <nodeB>` will translate to `storageNode=<nodeA>,<nodeB>`)
 * It is possible setting prefix for environment vars with `-envflag.prefix`. For instance, if `-envflag.prefix=VM_`, then env vars must be prepended with `VM_`

-### Prometheus setup
+
+## Prometheus setup

 Prometheus must be configured with [remote_write](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#remote_write)
 in order to send data to VictoriaMetrics. Add the following lines
@@ -192,15 +208,15 @@ remote_write:
  - url: http://<victoriametrics-addr>:8428/api/v1/write
 ```

-Substitute `<victoriametrics-addr>` with the hostname or IP address of VictoriaMetrics.
-Then apply the new config via the following command:
+Substitute `<victoriametrics-addr>` with hostname or IP address of VictoriaMetrics.
+Then apply new config via the following command:

 ```bash
 kill -HUP `pidof prometheus`
 ```

 Prometheus writes incoming data to local storage and replicates it to remote storage in parallel.
-This means the data remains available in local storage for `--storage.tsdb.retention.time` duration
+This means that data remains available in local storage for `--storage.tsdb.retention.time` duration
 even if remote storage is unavailable.

 If you plan to send data to VictoriaMetrics from multiple Prometheus instances, then add the following lines into `global` section
@@ -213,11 +229,10 @@ global:
 ```

 This instructs Prometheus to add `datacenter=dc-123` label to each time series sent to remote storage.
-The label name may be arbitrary - `datacenter` is just an example. The label value must be unique
+The label name can be arbitrary - `datacenter` is just an example. The label value must be unique
 across Prometheus instances, so those time series may be filtered and grouped by this label.

-For highly loaded Prometheus instances (400k+ samples per second)
-the following tuning may be applied:
+For highly loaded Prometheus instances (400k+ samples per second) the following tuning may be applied:

 ```yaml
 remote_write:
@@ -228,22 +243,20 @@ remote_write:
      max_shards: 30
 ```

-Using remote write increases memory usage for Prometheus up to ~25%
-and depends on the shape of data. If you are experiencing issues with
-too high memory consumption try to lower `max_samples_per_send`
-and `capacity` params (keep in mind that these two params are tightly connected).
+Using remote write increases memory usage for Prometheus up to ~25% and depends on the shape of data. If you are experiencing issues with
+too high memory consumption try to lower `max_samples_per_send` and `capacity` params (keep in mind that these two params are tightly connected).
 Read more about tuning remote write for Prometheus [here](https://prometheus.io/docs/practices/remote_write).

-It is recommended upgrading Prometheus to [v2.12.0](https://github.com/prometheus/prometheus/releases) or newer,
-since the previous versions may have issues with `remote_write`.
+It is recommended upgrading Prometheus to [v2.12.0](https://github.com/prometheus/prometheus/releases) or newer, since previous versions may have issues with `remote_write`.

-Take a look also at [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md),
-which can be used as faster and less resource-hungry alternative to Prometheus in certain cases.
+Take a look also at [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md)
+and [vmalert](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmalert/README.md),
+which can be used as faster and less resource-hungry alternative to Prometheus.


-### Grafana setup
+## Grafana setup

-Create [Prometheus datasource](http://docs.grafana.org/features/datasources/prometheus/) in Grafana with the following Url:
+Create [Prometheus datasource](http://docs.grafana.org/features/datasources/prometheus/) in Grafana with the following url:

 ```url
 http://<victoriametrics-addr>:8428
@@ -251,10 +264,12 @@ http://<victoriametrics-addr>:8428

 Substitute `<victoriametrics-addr>` with the hostname or IP address of VictoriaMetrics.

-Then build graphs with the created datasource using [Prometheus query language](https://prometheus.io/docs/prometheus/latest/querying/basics/).
-VictoriaMetrics supports native PromQL and [extends it with useful features](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/MetricsQL).
+Then build graphs with the created datasource using [PromQL](https://prometheus.io/docs/prometheus/latest/querying/basics/)
+or [MetricsQL](https://github.com/VictoriaMetrics/VictoriaMetrics/wiki/MetricsQL). VictoriaMetrics supports [Prometheus querying API](#prometheus-querying-api-usage),
+which is used by Grafana.

-### How to upgrade VictoriaMetrics
+
+## How to upgrade VictoriaMetrics

 It is safe upgrading VictoriaMetrics to new versions unless [release notes](https://github.com/VictoriaMetrics/VictoriaMetrics/releases)
 say otherwise. It is safe skipping multiple versions during the upgrade unless [release notes](https://github.com/VictoriaMetrics/VictoriaMetrics/releases) say otherwise.
@@ -264,25 +279,27 @@ It is also safe downgrading to the previous version unless [release notes](https

 The following steps must be performed during the upgrade / downgrade:

-1) Send `SIGINT` signal to VictoriaMetrics process in order to gracefully stop it.
-2) Wait until the process stops. This can take a few seconds.
-3) Start the upgraded VictoriaMetrics.
+* Send `SIGINT` signal to VictoriaMetrics process in order to gracefully stop it.
+* Wait until the process stops. This can take a few seconds.
+* Start the upgraded VictoriaMetrics.

 Prometheus doesn't drop data during VictoriaMetrics restart.
 See [this article](https://grafana.com/blog/2019/03/25/whats-new-in-prometheus-2.8-wal-based-remote-write/) for details.

-### How to apply new config to VictoriaMetrics

-VictoriaMetrics must be restarted for applying new config:
+## How to apply new config to VictoriaMetrics

-1) Send `SIGINT` signal to VictoriaMetrics process in order to gracefully stop it.
-2) Wait until the process stops. This can take a few seconds.
-3) Start VictoriaMetrics with the new config.
+VictoriaMetrics is configured via command-line flags, so it must be restarted when new command-line flags should be applied:
+
+* Send `SIGINT` signal to VictoriaMetrics process in order to gracefully stop it.
+* Wait until the process stops. This can take a few seconds.
+* Start VictoriaMetrics with the new command-line flags.

 Prometheus doesn't drop data during VictoriaMetrics restart.
 See [this article](https://grafana.com/blog/2019/03/25/whats-new-in-prometheus-2.8-wal-based-remote-write/) for details.

-### How to scrape Prometheus exporters such as [node-exporter](https://github.com/prometheus/node_exporter)
+
+## How to scrape Prometheus exporters such as [node-exporter](https://github.com/prometheus/node_exporter)

 VictoriaMetrics can be used as drop-in replacement for Prometheus for scraping targets configured in `prometheus.yml` config file according to [the specification](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#configuration-file).
 Just set `-promscrape.config` command-line flag to the path to `prometheus.yml` config - and VictoriaMetrics should start scraping the configured targets.
@@ -297,8 +314,10 @@ Currently the following [scrape_config](https://prometheus.io/docs/prometheus/la
 * [dns_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#dns_sd_config)
 * [openstack_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#openstack_sd_config)
 * [dockerswarm_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#dockerswarm_sd_config)
+* [eureka_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#eureka_sd_config)

-In the future other `*_sd_config` types will be supported.
+
+Other `*_sd_config` types will be supported in the future.

 The file pointed by `-promscrape.config` may contain `%{ENV_VAR}` placeholders, which are substituted by the corresponding `ENV_VAR` environment variable values.

@@ -306,9 +325,10 @@ VictoriaMetrics also supports [importing data in Prometheus exposition format](#

 See also [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md), which can be used as drop-in replacement for Prometheus.

-### How to send data from InfluxDB-compatible agents such as [Telegraf](https://www.influxdata.com/time-series-platform/telegraf/)

-Just use `http://<victoriametric-addr>:8428` url instead of InfluxDB url in agents' configs.
+## How to send data from InfluxDB-compatible agents such as [Telegraf](https://www.influxdata.com/time-series-platform/telegraf/)
+
+Use `http://<victoriametric-addr>:8428` url instead of InfluxDB url in agents' configs.
 For instance, put the following lines into `Telegraf` config, so it sends data to VictoriaMetrics instead of InfluxDB:

 ```toml
@@ -316,8 +336,6 @@ For instance, put the following lines into `Telegraf` config, so it sends data t
  urls = ["http://<victoriametrics-addr>:8428"]
 ```

-Do not forget substituting `<victoriametrics-addr>` with the real address where VictoriaMetrics runs.
-
 Another option is to enable TCP and UDP receiver for Influx line protocol via `-influxListenAddr` command-line flag
 and stream plain Influx line protocol data to the configured TCP and/or UDP addresses.

@@ -352,7 +370,7 @@ to local VictoriaMetrics using `curl`:
 curl -d 'measurement,tag1=value1,tag2=value2 field1=123,field2=1.23' -X POST 'http://localhost:8428/write'
 ```

-An arbitrary number of lines delimited by '\n' (aka newline char) may be sent in a single request.
+An arbitrary number of lines delimited by '\n' (aka newline char) can be sent in a single request.
 After that the data may be read via [/api/v1/export](#how-to-export-data-in-json-line-format) endpoint:

 ```bash
@@ -369,16 +387,17 @@ The `/api/v1/export` endpoint should return the following response:
 Note that Influx line protocol expects [timestamps in *nanoseconds* by default](https://docs.influxdata.com/influxdb/v1.7/write_protocols/line_protocol_tutorial/#timestamp),
 while VictoriaMetrics stores them with *milliseconds* precision.

-### How to send data from Graphite-compatible agents such as [StatsD](https://github.com/etsy/statsd)

-1) Enable Graphite receiver in VictoriaMetrics by setting `-graphiteListenAddr` command line flag. For instance,
+## How to send data from Graphite-compatible agents such as [StatsD](https://github.com/etsy/statsd)
+
+Enable Graphite receiver in VictoriaMetrics by setting `-graphiteListenAddr` command line flag. For instance,
 the following command will enable Graphite receiver in VictoriaMetrics on TCP and UDP port `2003`:

 ```bash
 /path/to/victoria-metrics-prod -graphiteListenAddr=:2003
 ```

-2) Use the configured address in Graphite-compatible agents. For instance, set `graphiteHost`
+Use the configured address in Graphite-compatible agents. For instance, set `graphiteHost`
 to the VictoriaMetrics host in `StatsD` configs.

 Example for writing data with Graphite plaintext protocol to local VictoriaMetrics using `nc`:
@@ -388,7 +407,7 @@ echo "foo.bar.baz;tag1=value1;tag2=value2 123 `date +%s`" | nc -N localhost 2003
 ```

 VictoriaMetrics sets the current time if the timestamp is omitted.
-An arbitrary number of lines delimited by `\n` (aka newline char) may be sent in one go.
+An arbitrary number of lines delimited by `\n` (aka newline char) can be sent in one go.
 After that the data may be read via [/api/v1/export](#how-to-export-data-in-json-line-format) endpoint:

 ```bash
@@ -401,30 +420,31 @@ The `/api/v1/export` endpoint should return the following response:
 {"metric":{"__name__":"foo.bar.baz","tag1":"value1","tag2":"value2"},"values":[123],"timestamps":[1560277406000]}
 ```

-### Querying Graphite data
+## Querying Graphite data

 Data sent to VictoriaMetrics via `Graphite plaintext protocol` may be read via the following APIs:

 * [Prometheus querying API](#prometheus-querying-api-usage)
 * Metric names can be explored via [Graphite metrics API](#graphite-metrics-api-usage)
-* [go-graphite/carbonapi](https://github.com/go-graphite/carbonapi/blob/master/cmd/carbonapi/carbonapi.example.prometheus.yaml)
+* Tags can be explored via [Graphite tags API](#graphite-tags-api-usage)
+* [go-graphite/carbonapi](https://github.com/go-graphite/carbonapi/blob/main/cmd/carbonapi/carbonapi.example.victoriametrics.yaml)

-### How to send data from OpenTSDB-compatible agents
+## How to send data from OpenTSDB-compatible agents

 VictoriaMetrics supports [telnet put protocol](http://opentsdb.net/docs/build/html/api_telnet/put.html)
 and [HTTP /api/put requests](http://opentsdb.net/docs/build/html/api_http/put.html) for ingesting OpenTSDB data.
 The same protocol is used for [ingesting data in KairosDB](https://kairosdb.github.io/docs/build/html/PushingData.html).

-#### Sending data via `telnet put` protocol
+### Sending data via `telnet put` protocol

-1) Enable OpenTSDB receiver in VictoriaMetrics by setting `-opentsdbListenAddr` command line flag. For instance,
+Enable OpenTSDB receiver in VictoriaMetrics by setting `-opentsdbListenAddr` command line flag. For instance,
 the following command enables OpenTSDB receiver in VictoriaMetrics on TCP and UDP port `4242`:

 ```bash
 /path/to/victoria-metrics-prod -opentsdbListenAddr=:4242
 ```

-2) Send data to the given address from OpenTSDB-compatible agents.
+Send data to the given address from OpenTSDB-compatible agents.

 Example for writing data with OpenTSDB protocol to local VictoriaMetrics using `nc`:

@@ -432,7 +452,7 @@ Example for writing data with OpenTSDB protocol to local VictoriaMetrics using `
 echo "put foo.bar.baz `date +%s` 123 tag1=value1 tag2=value2" | nc -N localhost 4242
 ```

-An arbitrary number of lines delimited by `\n` (aka newline char) may be sent in one go.
+An arbitrary number of lines delimited by `\n` (aka newline char) can be sent in one go.
 After that the data may be read via [/api/v1/export](#how-to-export-data-in-json-line-format) endpoint:

 ```bash
@@ -445,16 +465,16 @@ The `/api/v1/export` endpoint should return the following response:
 {"metric":{"__name__":"foo.bar.baz","tag1":"value1","tag2":"value2"},"values":[123],"timestamps":[1560277292000]}
 ```

-#### Sending OpenTSDB data via HTTP `/api/put` requests
+### Sending OpenTSDB data via HTTP `/api/put` requests

-1) Enable HTTP server for OpenTSDB `/api/put` requests by setting `-opentsdbHTTPListenAddr` command line flag. For instance,
+Enable HTTP server for OpenTSDB `/api/put` requests by setting `-opentsdbHTTPListenAddr` command line flag. For instance,
 the following command enables OpenTSDB HTTP server on port `4242`:

 ```bash
 /path/to/victoria-metrics-prod -opentsdbHTTPListenAddr=:4242
 ```

-2) Send data to the given address from OpenTSDB-compatible agents.
+Send data to the given address from OpenTSDB-compatible agents.

 Example for writing a single data point:

@@ -483,7 +503,7 @@ The `/api/v1/export` endpoint should return the following response:
 ```


-### Prometheus querying API usage
+## Prometheus querying API usage

 VictoriaMetrics supports the following handlers from [Prometheus querying API](https://prometheus.io/docs/prometheus/latest/querying/api/):

@@ -492,11 +512,14 @@ VictoriaMetrics supports the following handlers from [Prometheus querying API](h
 * [/api/v1/series](https://prometheus.io/docs/prometheus/latest/querying/api/#finding-series-by-label-matchers)
 * [/api/v1/labels](https://prometheus.io/docs/prometheus/latest/querying/api/#getting-label-names)
 * [/api/v1/label/.../values](https://prometheus.io/docs/prometheus/latest/querying/api/#querying-label-values)
-* [/api/v1/status/tsdb](https://prometheus.io/docs/prometheus/latest/querying/api/#tsdb-stats)
+* [/api/v1/status/tsdb](https://prometheus.io/docs/prometheus/latest/querying/api/#tsdb-stats). VictoriaMetrics accepts optional `topN=N` and `date=YYYY-MM-DD`
+  query args for this handler, where `N` is the number of top entries to return in the response and `YYYY-MM-DD` is the date for collecting the stats.
+  By default top 10 entries are returned and the stats is collected for the current day.
+* [/api/v1/targets](https://prometheus.io/docs/prometheus/latest/querying/api/#targets) - see [these docs](#how-to-scrape-prometheus-exporters-such-as-node-exporter) for more details.

 These handlers can be queried from Prometheus-compatible clients such as Grafana or curl.

-#### Prometheus querying API enhancements
+### Prometheus querying API enhancements

 Additionally to unix timestamps and [RFC3339](https://www.ietf.org/rfc/rfc3339.txt) VictoriaMetrics accepts relative times in `time`, `start` and `end` query args.
 For example, the following query would return data for the last 30 minutes: `/api/v1/query_range?start=-30m&query=...`.
@@ -518,6 +541,14 @@ Additionally VictoriaMetrics provides the following handlers:
 * `/api/v1/status/active_queries` - it returns a list of currently running queries.


+## Graphite API usage
+
+VictoriaMetrics supports the following Graphite APIs:
+
+* Metrics API - see [these docs](#graphite-metrics-api-usage).
+* Tags API - see [these docs](#graphite-tags-api-usage).
+
+
 ### Graphite Metrics API usage

 VictoriaMetrics supports the following handlers from [Graphite Metrics API](https://graphite-api.readthedocs.io/en/latest/api.html#the-metrics-api):
@@ -532,42 +563,56 @@ VictoriaMetrics accepts the following additional query args at `/metrics/find` a
    that start with `node_`. By default `delimiter=.`.


-### How to build from sources
+### Graphite Tags API usage
+
+VictoriaMetrics supports the following handlers from [Graphite Tags API](https://graphite.readthedocs.io/en/stable/tags.html):
+
+* [/tags/tagSeries](https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb)
+* [/tags/tagMultiSeries](https://graphite.readthedocs.io/en/stable/tags.html#adding-series-to-the-tagdb)
+* [/tags](https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags)
+* [/tags/{tag_name}](https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags)
+* [/tags/findSeries](https://graphite.readthedocs.io/en/stable/tags.html#exploring-tags)
+* [/tags/autoComplete/tags](https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support)
+* [/tags/autoComplete/values](https://graphite.readthedocs.io/en/stable/tags.html#auto-complete-support)
+* [/tags/delSeries](https://graphite.readthedocs.io/en/stable/tags.html#removing-series-from-the-tagdb)
+
+
+## How to build from sources

 We recommend using either [binary releases](https://github.com/VictoriaMetrics/VictoriaMetrics/releases) or
 [docker images](https://hub.docker.com/r/victoriametrics/victoria-metrics/) instead of building VictoriaMetrics
 from sources. Building from sources is reasonable when developing additional features specific
 to your needs or when testing bugfixes.

-#### Development build
+### Development build

 1. [Install Go](https://golang.org/doc/install). The minimum supported version is Go 1.13.
 2. Run `make victoria-metrics` from the root folder of the repository.
   It builds `victoria-metrics` binary and puts it into the `bin` folder.

-#### Production build
+### Production build

 1. [Install docker](https://docs.docker.com/install/).
 2. Run `make victoria-metrics-prod` from the root folder of the repository.
   It builds `victoria-metrics-prod` binary and puts it into the `bin` folder.

-#### ARM build
+### ARM build

 ARM build may run on Raspberry Pi or on [energy-efficient ARM servers](https://blog.cloudflare.com/arm-takes-wing/).

-#### Development ARM build
+### Development ARM build

 1. [Install Go](https://golang.org/doc/install). The minimum supported version is Go 1.13.
 2. Run `make victoria-metrics-arm` or `make victoria-metrics-arm64` from the root folder of the repository.
   It builds `victoria-metrics-arm` or `victoria-metrics-arm64` binary respectively and puts it into the `bin` folder.

-#### Production ARM build
+### Production ARM build

 1. [Install docker](https://docs.docker.com/install/).
 2. Run `make victoria-metrics-arm-prod` or `make victoria-metrics-arm64-prod` from the root folder of the repository.
   It builds `victoria-metrics-arm-prod` or `victoria-metrics-arm64-prod` binary respectively and puts it into the `bin` folder.

-#### Pure Go build (CGO_ENABLED=0)
+### Pure Go build (CGO_ENABLED=0)

 `Pure Go` mode builds only Go code without [cgo](https://golang.org/cmd/cgo/) dependencies.
 This is an experimental mode, which may result in a lower compression ratio and slower decompression performance.
@@ -577,7 +622,7 @@ Use it with caution!
 2. Run `make victoria-metrics-pure` from the root folder of the repository.
   It builds `victoria-metrics-pure` binary and puts it into the `bin` folder.

-#### Building docker images
+### Building docker images

 Run `make package-victoria-metrics`. It builds `victoriametrics/victoria-metrics:<PKG_TAG>` docker image locally.
 `<PKG_TAG>` is auto-generated image tag, which depends on source code in the repository.
@@ -591,17 +636,20 @@ For example, the following command builds the image on top of [scratch](https://
 ROOT_IMAGE=scratch make package-victoria-metrics
 ```

-### Start with docker-compose
+## Start with docker-compose

 [Docker-compose](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/deployment/docker/docker-compose.yml)
 helps to spin up VictoriaMetrics, [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md) and Grafana with one command.
 More details may be found [here](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/master/deployment/docker#folder-contains-basic-images-and-tools-for-building-and-running-victoria-metrics-in-docker).

-### Setting up service
+
+## Setting up service

 Read [these instructions](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/43) on how to set up VictoriaMetrics as a service in your OS.
+There is also [snap package for Ubuntu](https://snapcraft.io/victoriametrics).

-### How to work with snapshots
+
+## How to work with snapshots

 VictoriaMetrics can create [instant snapshots](https://medium.com/@valyala/how-victoriametrics-makes-instant-snapshots-for-multi-terabyte-time-series-data-e1f3fb0e0282)
 for all the data stored under `-storageDataPath` directory.
@@ -630,7 +678,7 @@ Steps for restoring from a snapshot:
   to the directory pointed by `-storageDataPath`.
 3. Start VictoriaMetrics.

-### How to delete time series
+## How to delete time series

 Send a request to `http://<victoriametrics-addr>:8428/api/v1/admin/tsdb/delete_series?match[]=<timeseries_selector_for_delete>`,
 where `<timeseries_selector_for_delete>` may contain any [time series selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors)
@@ -662,7 +710,7 @@ It isn't recommended using delete API for the following cases, since it brings n
 It is better using `-retentionPeriod` command-line flag for efficient pruning of old data.


-### Forced merge
+## Forced merge

 VictoriaMetrics performs [data compactions in background](https://medium.com/@valyala/how-victoriametrics-makes-instant-snapshots-for-multi-terabyte-time-series-data-e1f3fb0e0282)
 in order to keep good performance characteristics when accepting new data. These compactions (merges) are performed independently on per-month partitions.
@@ -677,7 +725,7 @@ since VictoriaMetrics automatically performs [optimal merges in background](http
 when new data is ingested into it.


-### How to export time series
+## How to export time series

 VictoriaMetrics provides the following handlers for exporting data:

@@ -687,11 +735,20 @@ VictoriaMetrics provides the following handlers for exporting data:
 * `/api/v1/export/csv` for exporting data in CSV. See [these docs](#how-to-export-csv-data) for details.


-#### How to export data in native format
+### How to export data in native format

 Send a request to `http://<victoriametrics-addr>:8428/api/v1/export/native?match[]=<timeseries_selector_for_export>`,
 where `<timeseries_selector_for_export>` may contain any [time series selector](https://prometheus.io/docs/prometheus/latest/querying/basics/#time-series-selectors)
-for metrics to export. Use `{__name__!=""}` selector for fetching all the time series.
+for metrics to export. Use `{__name__=~".*"}` selector for fetching all the time series.
+
+On large databases you may experience problems with limit on unique timeseries (default value is 300000). In this case you need to adjust `-search.maxUniqueTimeseries` parameter:
+
+```bash
+# count unique timeseries in database
+wget -O- -q 'http://your_victoriametrics_instance:8428/api/v1/series/count' | jq '.data[0]'
+
+# relaunch victoriametrics with search.maxUniqueTimeseries more than value from previous command
+```

 Optional `start` and `end` args may be added to the request in order to limit the time frame for the exported data. These args may contain either
 unix timestamp in seconds or [RFC3339](https://www.ietf.org/rfc/rfc3339.txt) values.
@@ -699,7 +756,7 @@ unix timestamp in seconds or [RFC3339](https://www.ietf.org/rfc/rfc3339.txt) val
 The exported data can be imported to VictoriaMetrics via [/api/v1/import/native](#how-to-import-data-in-native-format).


-#### How to export data in JSON line format
+### How to export data in JSON line format

 Consider [exporting data in native format](#how-to-export-data-in-native-format) if big amounts of data must be migrated between VictoriaMetrics instances,
 since exporting in native format usually consumes lower amounts of CPU and memory resources, while the resulting exported data occupies lower amounts of disk space.
@@ -734,7 +791,7 @@ The maximum duration for each request to `/api/v1/export` is limited by `-search
 Exported data can be imported via POST'ing it to [/api/v1/import](#how-to-import-data-in-json-line-format).


-#### How to export CSV data
+### How to export CSV data

 Send a request to `http://<victoriametrics-addr>:8428/api/v1/export/csv?format=<format>&match=<timeseries_selector_for_export>`,
 where:
@@ -758,24 +815,24 @@ unix timestamp in seconds or [RFC3339](https://www.ietf.org/rfc/rfc3339.txt) val
 The exported CSV data can be imported to VictoriaMetrics via [/api/v1/import/csv](#how-to-import-csv-data).


-### How to import time series data
+## How to import time series data

 Time series data can be imported via any supported ingestion protocol:

-* [Prometheus remote_write API](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#remote_write).
+* [Prometheus remote_write API](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#remote_write). See [these docs](#prometheus-setup) for details.
 * Influx line protocol. See [these docs](#how-to-send-data-from-influxdb-compatible-agents-such-as-telegraf) for details.
 * Graphite plaintext protocol. See [these docs](#how-to-send-data-from-graphite-compatible-agents-such-as-statsd) for details.
 * OpenTSDB telnet put protocol. See [these docs](#sending-data-via-telnet-put-protocol) for details.
 * OpenTSDB http `/api/put` protocol. See [these docs](#sending-opentsdb-data-via-http-apiput-requests) for details.
 * `/api/v1/import` for importing data obtained from [/api/v1/export](#how-to-export-data-in-json-line-format).
-  See [these docs](##how-to-import-data-in-json-line-format) for details.
+  See [these docs](#how-to-import-data-in-json-line-format) for details.
 * `/api/v1/import/native` for importing data obtained from [/api/v1/export/native](#how-to-export-data-in-native-format).
  See [these docs](#how-to-import-data-in-native-format) for details.
 * `/api/v1/import/csv` for importing arbitrary CSV data. See [these docs](#how-to-import-csv-data) for details.
 * `/api/v1/import/prometheus` for importing data in Prometheus exposition format. See [these docs](#how-to-import-data-in-prometheus-exposition-format) for details.


-#### How to import data in native format
+### How to import data in native format

 The most efficient protocol for importing data into VictoriaMetrics is `/api/v1/import/native`.
 Example for importing data obtained via [/api/v1/export/native](#how-to-export-data-in-native-format):
@@ -804,7 +861,7 @@ For example, `/api/v1/import/native?extra_label=foo=bar` would add `"foo":"bar"`
 Note that it could be required to flush response cache after importing historical data. See [these docs](#backfilling) for detail.


-#### How to import data in JSON line format
+### How to import data in JSON line format

 Example for importing data obtained via [/api/v1/export](#how-to-export-data-in-json-line-format):

@@ -832,7 +889,7 @@ For example, `/api/v1/import?extra_label=foo=bar` would add `"foo":"bar"` label
 Note that it could be required to flush response cache after importing historical data. See [these docs](#backfilling) for detail.


-#### How to import CSV data
+### How to import CSV data

 Arbitrary CSV data can be imported via `/api/v1/import/csv`. The CSV data is imported according to the provided `format` query arg.
 The `format` query arg must contain comma-separated list of parsing rules for CSV fields. Each rule consists of three parts delimited by a colon:
@@ -885,7 +942,7 @@ For example, `/api/v1/import/csv?extra_label=foo=bar` would add `"foo":"bar"` la
 Note that it could be required to flush response cache after importing historical data. See [these docs](#backfilling) for detail.


-#### How to import data in Prometheus exposition format
+### How to import data in Prometheus exposition format

 VictoriaMetrics accepts data in [Prometheus exposition format](https://github.com/prometheus/docs/blob/master/content/docs/instrumenting/exposition_formats.md#text-based-format)
 via `/api/v1/import/prometheus` path. For example, the following line imports a single line in Prometheus exposition format into VictoriaMetrics:
@@ -920,7 +977,7 @@ VictoriaMetrics also may scrape Prometheus targets - see [these docs](#how-to-sc



-### Relabeling
+## Relabeling

 VictoriaMetrics supports Prometheus-compatible relabeling for all the ingested metrics if `-relabelConfig` command-line flag points
 to a file containing a list of [relabel_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#relabel_config) entries.
@@ -944,7 +1001,7 @@ VictoriaMetrics provides the following extra actions for relabeling rules:
 See also [relabeling in vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md#relabeling).


-### Federation
+## Federation

 VictoriaMetrics exports [Prometheus-compatible federation data](https://prometheus.io/docs/prometheus/latest/federation/)
 at `http://<victoriametrics-addr>:8428/federate?match[]=<timeseries_selector_for_federation>`.
@@ -955,7 +1012,7 @@ on the interval `[now - max_lookback ... now]` is scraped for each time series.
 For instance, `/federate?match[]=up&max_lookback=1h` would return last points on the `[now - 1h ... now]` interval. This may be useful for time series federation
 with scrape intervals exceeding `5m`.

-### Capacity planning
+## Capacity planning

 A rough estimation of the required resources for ingestion path:

@@ -995,10 +1052,11 @@ The required resources for query path:
 * Network usage: depends on the frequency and the type of incoming requests. Typical Grafana dashboards usually
  require negligible network bandwidth.

-### High availability

-1) Install multiple VictoriaMetrics instances in distinct datacenters (availability zones).
-2) Pass addresses of these instances to [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md) via `-remoteWrite.url` command-line flag:
+## High availability
+
+* Install multiple VictoriaMetrics instances in distinct datacenters (availability zones).
+* Pass addresses of these instances to [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md) via `-remoteWrite.url` command-line flag:

 ```bash
 /path/to/vmagent -remoteWrite.url=http://<victoriametrics-addr-1>:8428/api/v1/write -remoteWrite.url=http://<victoriametrics-addr-2>:8428/api/v1/write
@@ -1017,7 +1075,7 @@ remote_write:
      max_samples_per_send: 10000
 ```

-3) Apply the updated config:
+* Apply the updated config:

 ```bash
 kill -HUP `pidof prometheus`
@@ -1025,9 +1083,9 @@ kill -HUP `pidof prometheus`

 It is recommended to use [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md) instead of Prometheus for highly loaded setups.

-4) Now Prometheus should write data into all the configured `remote_write` urls in parallel.
-5) Set up [Promxy](https://github.com/jacksontj/promxy) in front of all the VictoriaMetrics replicas.
-6) Set up Prometheus datasource in Grafana that points to Promxy.
+* Now Prometheus should write data into all the configured `remote_write` urls in parallel.
+* Set up [Promxy](https://github.com/jacksontj/promxy) in front of all the VictoriaMetrics replicas.
+* Set up Prometheus datasource in Grafana that points to Promxy.

 If you have Prometheus HA pairs with replicas `r1` and `r2` in each pair, then configure each `r1`
 to write data to `victoriametrics-addr-1`, while each `r2` should write data to `victoriametrics-addr-2`.
@@ -1036,7 +1094,7 @@ Another option is to write data simultaneously from Prometheus HA pair to a pair
 with the enabled de-duplication. See [this section](#deduplication) for details.


-### Deduplication
+## Deduplication

 VictoriaMetrics de-duplicates data points if `-dedup.minScrapeInterval` command-line flag
 is set to positive duration. For example, `-dedup.minScrapeInterval=60s` would de-duplicate data points
@@ -1046,7 +1104,8 @@ The de-duplication reduces disk space usage if multiple identically configured P
 write data to the same VictoriaMetrics instance. Note that these Prometheus instances must have identical
 `external_labels` section in their configs, so they write data to the same time series.

-### Retention
+
+## Retention

 Retention is configured with `-retentionPeriod` command-line flag. For instance, `-retentionPeriod=3` means
 that the data will be stored for 3 months and then deleted.
@@ -1057,7 +1116,11 @@ For example if `-retentionPeriod` is set to 1, data for January is deleted on Ma
 It is safe to extend `-retentionPeriod` on existing data. If `-retentionPeriod` is set to lower
 value than before then data outside the configured period will be eventually deleted.

-### Multiple retentions
+VictoriaMetrics supports retention smaller than 1 month. For example, `-retentionPeriod=5d` would set data retention for 5 days.
+Older data is eventually deleted during [background merge](https://medium.com/@valyala/how-victoriametrics-makes-instant-snapshots-for-multi-terabyte-time-series-data-e1f3fb0e0282).
+
+
+## Multiple retentions

 Just start multiple VictoriaMetrics instances with distinct values for the following flags:

@@ -1070,7 +1133,7 @@ so it could route requests from particular user to VictoriaMetrics with the desi
 The same scheme could be implemented for multiple tenants in [VictoriaMetrics cluster](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/cluster/README.md).


-### Downsampling
+## Downsampling

 There is no downsampling support at the moment, but:

@@ -1086,11 +1149,13 @@ It is possible to (ab)use [-dedup.minScrapeInterval](#deduplication) for basic d
 For instance, if interval between the ingested data points is 15s, then `-dedup.minScrapeInterval=5m` will leave
 only a single data point out of 20 initial data points per each 5m interval.

-### Multi-tenancy
+
+## Multi-tenancy

 Single-node VictoriaMetrics doesn't support multi-tenancy. Use [cluster version](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/cluster) instead.

-### Scalability and cluster version
+
+## Scalability and cluster version

 Though single-node VictoriaMetrics cannot scale to multiple nodes, it is optimized for resource usage - storage size / bandwidth / IOPS, RAM, CPU.
 This means that a single-node VictoriaMetrics may scale vertically and substitute a moderately sized cluster built with competing solutions
@@ -1100,7 +1165,8 @@ So try single-node VictoriaMetrics at first and then [switch to cluster version]
 horizontally scalable long-term remote storage for really large Prometheus deployments.
 [Contact us](mailto:info@victoriametrics.com) for paid support.

-### Alerting
+
+## Alerting

 It is recommended using [vmalert](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmalert/README.md) for alerting.

@@ -1111,7 +1177,7 @@ Additionally, alerting can be set up with the following tools:
 * With Grafana - see [the corresponding docs](https://grafana.com/docs/alerting/rules/).


-### Security
+## Security

 Do not forget protecting sensitive endpoints in VictoriaMetrics when exposing it to untrusted networks such as the internet.
 Consider setting the following command-line flags:
@@ -1131,7 +1197,7 @@ Prefer authorizing all the incoming requests from untrusted networks with [vmaut
 or similar auth proxy.


-### Tuning
+## Tuning

 * There is no need for VictoriaMetrics tuning since it uses reasonable defaults for command-line flags,
  which are automatically adjusted for the available CPU and RAM resources.
@@ -1147,7 +1213,7 @@ or similar auth proxy.
 mkfs.ext4 ... -O 64bit,huge_file,extent -T huge
 ```

-### Monitoring
+## Monitoring

 VictoriaMetrics exports internal metrics in Prometheus format at `/metrics` page.
 These metrics may be collected by [vmagent](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmagent/README.md)
@@ -1177,7 +1243,7 @@ The most interesting metrics are:
 VictoriaMetrics also exposes currently running queries with their execution times at `/api/v1/status/active_queries` page.


-### Troubleshooting
+## Troubleshooting

 * It is recommended to use default command-line flag values (i.e. don't set them explicitly) until the need
  of tweaking these flag values arises.
@@ -1187,13 +1253,18 @@ VictoriaMetrics also exposes currently running queries with their execution time

 * It is recommended inspecting logs during troubleshooting, since they may contain useful information.

+* VictoriaMetrics buffers incoming data in memory for up to a few seconds before flushing it to persistent storage.
+  This may lead to the following "issues":
+  * Data becomes available for querying in a few seconds after inserting. It is possible to flush in-memory buffers to persistent storage
+    by requesting `/internal/force_flush` http handler. This handler is mostly needed for testing and debugging purposes.
+  * The last few seconds of inserted data may be lost on unclean shutdown (i.e. OOM, `kill -9` or hardware reset).
+    See [this article for technical details](https://valyala.medium.com/wal-usage-looks-broken-in-modern-time-series-databases-b62a627ab704).
+
 * If VictoriaMetrics works slowly and eats more than a CPU core per 100K ingested data points per second,
  then it is likely you have too many active time series for the current amount of RAM.
  VictoriaMetrics [exposes](#monitoring) `vm_slow_*` metrics, which could be used as an indicator of low amounts of RAM.
  It is recommended increasing the amount of RAM on the node with VictoriaMetrics in order to improve
  ingestion and query performance in this case.
-  Another option is to increase `-memory.allowedPercent` command-line flag value. Be careful with this
-  option, since too big value for `-memory.allowedPercent` may result in high I/O usage.

 * VictoriaMetrics prioritizes data ingestion over data querying. So if it has no enough resources for data ingestion,
  then data querying may slow down significantly.
@@ -1208,9 +1279,9 @@ VictoriaMetrics also exposes currently running queries with their execution time
  which would start background merge if they had more free disk space.

 * If VictoriaMetrics doesn't work because of certain parts are corrupted due to disk errors,
-  then just remove directories with broken parts. This will recover VictoriaMetrics at the cost
-  of data loss stored in the broken parts. In the future, `vmrecover` tool will be created
-  for automatic recovering from such errors.
+  then just remove directories with broken parts. It is safe removing subdirectories under `<-storageDataPath>/data/{big,small}/YYYY_MM` directories
+  when VictoriaMetrics isn't running. This recovers VictoriaMetrics at the cost of data loss stored in the deleted broken parts.
+  In the future, `vmrecover` tool will be created for automatic recovering from such errors.

 * If you see gaps on the graphs, try resetting the cache by sending request to `/internal/resetRollupResultCache`.
  If this removes gaps on the graphs, then it is likely data with timestamps older than `-search.cacheTimestampOffset`
@@ -1232,10 +1303,15 @@ VictoriaMetrics also exposes currently running queries with their execution time
  This prevents from ingesting metrics with too many labels. It is recommended [monitoring](#monitoring) `vm_metrics_with_dropped_labels_total`
  metric in order to determine whether `-maxLabelsPerTimeseries` must be adjusted for your workload.

+* If you store Graphite metrics like `foo.bar.baz` in VictoriaMetrics, then `-search.treatDotsAsIsInRegexps` command-line flag could be useful.
+  By default `.` chars in regexps match any char. If you need matching only dots, then the `\\.` must be used in regexp filters.
+  When `-search.treatDotsAsIsInRegexps` option is enabled, then dots in regexps are automatically escaped in order to match only dots instead of arbitrary chars.
+  This may significantly increase performance when locating time series for the given label filters.
+
 * VictoriaMetrics ignores `NaN` values during data ingestion.


-### Backfilling
+## Backfilling

 VictoriaMetrics accepts historical data in arbitrary order of time via [any supported ingestion method](#how-to-import-time-series-data).
 Make sure that configured `-retentionPeriod` covers timestamps for the backfilled data.
@@ -1251,7 +1327,7 @@ Yet another solution is to increase `-search.cacheTimestampOffset` flag value in
 for data with timestamps close to the current time.


-### Data updates
+## Data updates

 VictoriaMetrics doesn't support updating already existing sample values to new ones. It stores all the ingested data points
 for the same time series with identical timestamps. While is possible substituting old time series with new time series via
@@ -1259,7 +1335,7 @@ for the same time series with identical timestamps. While is possible substituti
 should be used only for one-off updates. It shouldn't be used for frequent updates because of non-zero overhead related to data removal.


-### Replication
+## Replication

 Single-node VictoriaMetrics doesn't support application-level replication. Use cluster version instead.
 See [these docs](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/cluster/README.md#replication-and-data-safety) for details.
@@ -1269,14 +1345,14 @@ Storage-level replication may be offloaded to durable persistent storage such as
 See also [high availability docs](#high-availability) and [backup docs](#backups).


-### Backups
+## Backups

 VictoriaMetrics supports backups via [vmbackup](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmbackup/README.md)
 and [vmrestore](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmrestore/README.md) tools.
-We also provide provide `vmbackuper` tool for paid enterprise subscribers - see [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/466) for details.
+We also provide `vmbackuper` tool for paid enterprise subscribers - see [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/466) for details.


-### Profiling
+## Profiling

 VictoriaMetrics provides handlers for collecting the following [Go profiles](https://blog.golang.org/profiling-go-programs):

@@ -1296,6 +1372,7 @@ The command for collecting CPU profile waits for 30 seconds before returning.

 The collected profiles may be analyzed with [go tool pprof](https://github.com/google/pprof).

+
 ## Integrations

 * [Helm charts for single-node and cluster versions of VictoriaMetrics](https://github.com/VictoriaMetrics/helm-charts).
@@ -1307,6 +1384,8 @@ The collected profiles may be analyzed with [go tool pprof](https://github.com/g
  See [this example](https://github.com/go-graphite/carbonapi/blob/master/cmd/carbonapi/carbonapi.example.prometheus.yaml).
 * [Ansible role for installing single-node VictoriaMetrics](https://github.com/dreamteam-gg/ansible-victoriametrics-role).
 * [Ansible role for installing cluster VictoriaMetrics](https://github.com/Slapper/ansible-victoriametrics-cluster-role).
+* [Snap package for VictoriaMetrics](https://snapcraft.io/victoriametrics).
+

 ## Third-party contributions

@@ -1315,10 +1394,12 @@ The collected profiles may be analyzed with [go tool pprof](https://github.com/g
 * [Prometheus -> VictoriaMetrics exporter #2](https://github.com/AnchorFree/tsdb-remote-write)
 * [Prometheus Oauth proxy](https://gitlab.com/optima_public/prometheus_oauth_proxy) - see [this article](https://medium.com/@richard.holly/powerful-saas-solution-for-detection-metrics-c67b9208d362) for details.

+
 ## Contacts

 Contact us with any questions regarding VictoriaMetrics at [info@victoriametrics.com](mailto:info@victoriametrics.com).

+
 ## Community and contributions

 Feel free asking any questions regarding VictoriaMetrics:
--- a/docs/vmagent.md
+++ b/docs/vmagent.md
@@ -63,6 +63,22 @@ Then send Influx data to `http://vmagent-host:8429`. See [these docs](https://gi
 Pass `-help` to `vmagent` in order to see the full list of supported command-line flags with their descriptions.


+### Configuration update
+
+`vmagent` should be restarted in order to update config options set via command-line args.
+
+`vmagent` supports multiple approaches for reloading configs from updated config files such as `-promscrape.config`, `-remoteWrite.relabelConfig` and `-remoteWrite.urlRelabelConfig`:
+
+* Sending `SUGHUP` signal to `vmagent` process:
+  ```bash
+  kill -SIGHUP `pidof vmagent`
+  ```
+
+* Sending HTTP request to `http://vmagent:8429/-/reload` endpoint.
+
+There is also `-promscrape.configCheckInterval` command-line option, which can be used for automatic reloading configs from updated `-promscrape.config` file.
+
+
 ### Use cases


@@ -153,6 +169,8 @@ The following scrape types in [scrape_config](https://prometheus.io/docs/prometh
  [OpenStack identity API v3](https://docs.openstack.org/api-ref/identity/v3/) is supported only.
 * `dockerswarm_sd_configs` - for scraping Docker Swarm targets.
  See [dockerswarm_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#dockerswarm_sd_config) for details.
+* `eureka_sd_configs` - for scraping targets registered in [Netflix Eureka](https://github.com/Netflix/eureka).
+  See [eureka_sd_config](https://prometheus.io/docs/prometheus/latest/configuration/configuration/#eureka_sd_config) for details.

 File feature requests at [our issue tracker](https://github.com/VictoriaMetrics/VictoriaMetrics/issues) if you need other service discovery mechanisms to be supported by `vmagent`.

@@ -197,6 +215,7 @@ The relabeling can be defined in the following places:

 Read more about relabeling in the following articles:

+* [How to use Relabeling in Prometheus and VictoriaMetrics](https://valyala.medium.com/how-to-use-relabeling-in-prometheus-and-victoriametrics-8b90fc22c4b2)
 * [Life of a label](https://www.robustperception.io/life-of-a-label)
 * [Discarding targets and timeseries with relabeling](https://www.robustperception.io/relabelling-can-discard-targets-timeseries-and-alerts)
 * [Dropping labels at scrape time](https://www.robustperception.io/dropping-metrics-at-scrape-time-with-prometheus)
@@ -211,9 +230,16 @@ either via `vmagent` itself or via Prometheus, so the exported metrics could be
 Use official [Grafana dashboard](https://grafana.com/grafana/dashboards/12683) for `vmagent` state overview.
 If you have suggestions, improvements or found a bug - feel free to open an issue on github or add review to the dashboard.

-`vmagent` also exports target statuses at `http://vmagent-host:8429/targets` page in plaintext format.
-`/targets` handler accepts optional `show_original_labels=1` query arg, which shows the original labels per each target
-before applying relabeling. This information may be useful for debugging target relabeling.
+`vmagent` also exports target statuses at the following handlers:
+
+* `http://vmagent-host:8429/targets`. This handler returns human-readable plaintext status for every active target.
+This page is convenient to query from command line with `wget`, `curl` or similar tools.
+It accepts optional `show_original_labels=1` query arg, which shows the original labels per each target before applying relabeling.
+This information may be useful for debugging target relabeling.
+* `http://vmagent-host:8429/api/v1/targets`. This handler returns data compatible with [the corresponding page from Prometheus API](https://prometheus.io/docs/prometheus/latest/querying/api/#targets).
+
+* `http://vmagent-host:8429/ready`. This handler returns http 200 status code when `vmagent` finishes initialization for all service_discovery configs.
+It may be useful for performing `vmagent` rolling update without scrape loss.


 ### Troubleshooting
@@ -224,7 +250,32 @@ before applying relabeling. This information may be useful for debugging target
  since `vmagent` establishes at least a single TCP connection per each target.

 * When `vmagent` scrapes many unreliable targets, it can flood error log with scrape errors. These errors can be suppressed
-  by passing `-promscrape.suppressScrapeErrors` command-line flag to `vmagent`. The most recent scrape error per each target can be observed at `http://vmagent-host:8429/targets`.
+  by passing `-promscrape.suppressScrapeErrors` command-line flag to `vmagent`. The most recent scrape error per each target can be observed at `http://vmagent-host:8429/targets`
+  and `http://vmagent-host:8429/api/v1/targets`.
+
+* The `/api/v1/targets` page could be useful for debugging relabeling process for scrape targets.
+  This page contains original labels for targets dropped during relabeling (see "droppedTargets" section in the page output). By default up to `-promscrape.maxDroppedTargets` targets are shown here. If your setup drops more targets during relabeling, then increase `-promscrape.maxDroppedTargets` command-line flag value in order to see all the dropped targets. Note that tracking each dropped target requires up to 10Kb of RAM, so big values for `-promscrape.maxDroppedTargets` may result in increased memory usage if big number of scrape targets are dropped during relabeling.
+
+* If `vmagent` scrapes big number of targets, then `-promscrape.dropOriginalLabels` command-line option may be passed to `vmagent` in order to reduce memory usage.
+  This option drops `"discoveredLabels"` and `"droppedTargets"` lists at `/api/v1/targets` page, which may result in reduced debuggability for improperly configured per-target relabeling.
+
+* If `vmagent` scrapes targets with millions of metrics per each target (for instance, when scraping [federation endpoints](https://prometheus.io/docs/prometheus/latest/federation/)),
+  then it is recommended enabling `stream parsing mode` in order to reduce memory usage during scraping. This mode may be enabled either globally for all the scrape targets
+  by passing `-promscrape.streamParse` command-line flag or on a per-scrape target basis with `stream_parse: true` option. For example:
+
+  ```yml
+  scrape_configs:
+  - job_name: 'big-federate'
+    stream_parse: true
+    static_configs:
+    - targets:
+      - big-prometeus1
+      - big-prometeus2
+    honor_labels: true
+    metrics_path: /federate
+    params:
+      'match[]': ['{__name__!=""}']
+  ```

 * It is recommended to increase `-remoteWrite.queues` if `vmagent_remotewrite_pending_data_bytes` metric exported at `http://vmagent-host:8429/metrics` page constantly grows.

--- a/docs/vmalert.md
+++ b/docs/vmalert.md
@@ -11,6 +11,7 @@ rules against configured address.
 * Prometheus [alerting rules definition format](https://prometheus.io/docs/prometheus/latest/configuration/alerting_rules/#defining-alerting-rules)
 support;
 * Integration with [Alertmanager](https://github.com/prometheus/alertmanager);
+* Keeps the alerts [state on restarts](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/master/app/vmalert#alerts-state-on-restarts);
 * Lightweight without extra dependencies.

 ### Limitations:
@@ -121,14 +122,6 @@ annotations:
  [ <labelname>: <tmpl_string> ]
 ``` 

-`vmalert` has no local storage and alerts state is stored in process memory. Hence, after reloading of `vmalert` process
-alerts state will be lost. To avoid this situation, `vmalert` may be configured via following flags:
-* `-remoteWrite.url` - URL to Victoria Metrics or VMInsert. `vmalert` will persist alerts state into the configured
-address in form of timeseries with name `ALERTS` via remote-write protocol.
-* `-remoteRead.url` - URL to Victoria Metrics or VMSelect. `vmalert` will try to restore alerts state from configured
-address by querying `ALERTS` timeseries.
-
-
 ##### Recording rules

 The syntax for recording rules is following:
@@ -147,6 +140,22 @@ labels:
 For recording rules to work `-remoteWrite.url` must specified.


+#### Alerts state on restarts
+
+`vmalert` has no local storage, so alerts state is stored in the process memory. Hence, after reloading of `vmalert` 
+the process alerts state will be lost. To avoid this situation, `vmalert` should be configured via the following flags:
+* `-remoteWrite.url` - URL to VictoriaMetrics (Single) or VMInsert (Cluster). `vmalert` will persist alerts state 
+into the configured address in the form of time series named `ALERTS` and `ALERTS_FOR_STATE` via remote-write protocol. 
+These are regular time series and may be queried from VM just as any other time series. 
+The state stored to the configured address on every rule evaluation.
+* `-remoteRead.url` - URL to VictoriaMetrics (Single) or VMSelect (Cluster). `vmalert` will try to restore alerts state 
+from configured address by querying time series with name `ALERTS_FOR_STATE`.
+
+Both flags are required for the proper state restoring. Restore process may fail if time series are missing
+in configured `-remoteRead.url`, weren't updated in the last `1h` or received state doesn't match current `vmalert` 
+rules configuration.
+
+
 #### WEB

 `vmalert` runs a web-server (`-httpListenAddr`) for serving metrics and alerts endpoints:
--- a/docs/vmrestore.md
+++ b/docs/vmrestore.md
@@ -3,7 +3,7 @@
 `vmrestore` restores data from backups created by [vmbackup](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmbackup/README.md).
 VictoriaMetrics `v1.29.0` and newer versions must be used for working with the restored data.

-Restore process can be interrupted at any time. It is automatically resumed from the inerruption point
+Restore process can be interrupted at any time. It is automatically resumed from the interruption point
 when restarting `vmrestore` with the same args.


--- a/go.mod
+++ b/go.mod
@@ -1,34 +1,33 @@
 module github.com/VictoriaMetrics/VictoriaMetrics

 require (
-	cloud.google.com/go v0.68.0 // indirect
+	cloud.google.com/go v0.73.0 // indirect
 	cloud.google.com/go/storage v1.12.0
 	github.com/VictoriaMetrics/fastcache v1.5.7

 	// Do not use the original github.com/valyala/fasthttp because of issues
 	// like https://github.com/valyala/fasthttp/commit/996610f021ff45fdc98c2ce7884d5fa4e7f9199b
-	github.com/VictoriaMetrics/fasthttp v1.0.7
+	github.com/VictoriaMetrics/fasthttp v1.0.9
 	github.com/VictoriaMetrics/metrics v1.12.3
-	github.com/VictoriaMetrics/metricsql v0.7.1
-	github.com/aws/aws-sdk-go v1.35.5
+	github.com/VictoriaMetrics/metricsql v0.9.0
+	github.com/aws/aws-sdk-go v1.36.2
 	github.com/cespare/xxhash/v2 v2.1.1
 	github.com/golang/snappy v0.0.2
-	github.com/klauspost/compress v1.11.1
-	github.com/stretchr/testify v1.5.1 // indirect
-	github.com/valyala/fastjson v1.6.1
+	github.com/klauspost/compress v1.11.3
+	github.com/valyala/fastjson v1.6.3
 	github.com/valyala/fastrand v1.0.0
 	github.com/valyala/fasttemplate v1.2.1
 	github.com/valyala/gozstd v1.8.3
 	github.com/valyala/histogram v1.1.2
 	github.com/valyala/quicktemplate v1.6.3
-	go.opencensus.io v0.22.5 // indirect
-	golang.org/x/net v0.0.0-20201006153459-a7d1128ccaa0 // indirect
-	golang.org/x/oauth2 v0.0.0-20200902213428-5d25da1a8d43
-	golang.org/x/sys v0.0.0-20201008064518-c1f3e3309c71
-	golang.org/x/tools v0.0.0-20201008025239-9df69603baec // indirect
-	google.golang.org/api v0.32.0
-	google.golang.org/genproto v0.0.0-20201007142714-5c0e72c5e71e // indirect
-	gopkg.in/yaml.v2 v2.3.0
+	golang.org/x/net v0.0.0-20201202161906-c7110b5ffcbb // indirect
+	golang.org/x/oauth2 v0.0.0-20201203001011-0b49973bad19
+	golang.org/x/sys v0.0.0-20201204225414-ed752295db88
+	golang.org/x/tools v0.0.0-20201204222352-654352759326 // indirect
+	google.golang.org/api v0.36.0
+	google.golang.org/genproto v0.0.0-20201204160425-06b3db808446 // indirect
+	google.golang.org/grpc v1.34.0 // indirect
+	gopkg.in/yaml.v2 v2.4.0
 )

 go 1.13
--- a/go.sum
+++ b/go.sum
@@ -9,65 +9,50 @@ cloud.google.com/go v0.50.0/go.mod h1:r9sluTvynVuxRIOHXQEHMFffphuXHOMZMycpNR5e6T
 cloud.google.com/go v0.52.0/go.mod h1:pXajvRH/6o3+F9jDHZWQ5PbGhn+o8w9qiu/CffaVdO4=
 cloud.google.com/go v0.53.0/go.mod h1:fp/UouUEsRkN6ryDKNW/Upv/JBKnv6WDthjR6+vze6M=
 cloud.google.com/go v0.54.0/go.mod h1:1rq2OEkV3YMf6n/9ZvGWI3GWw0VoqH/1x2nd8Is/bPc=
-cloud.google.com/go v0.56.0 h1:WRz29PgAsVEyPSDHyk+0fpEkwEFyfhHn+JbksT6gIL4=
 cloud.google.com/go v0.56.0/go.mod h1:jr7tqZxxKOVYizybht9+26Z/gUq7tiRzu+ACVAMbKVk=
-cloud.google.com/go v0.57.0 h1:EpMNVUorLiZIELdMZbCYX/ByTFCdoYopYAGxaGVz9ms=
 cloud.google.com/go v0.57.0/go.mod h1:oXiQ6Rzq3RAkkY7N6t3TcE6jE+CIBBbA36lwQ1JyzZs=
-cloud.google.com/go v0.62.0 h1:RmDygqvj27Zf3fCQjQRtLyC7KwFcHkeJitcO0OoGOcA=
 cloud.google.com/go v0.62.0/go.mod h1:jmCYTdRCQuc1PHIIJ/maLInMho30T/Y0M4hTdTShOYc=
-cloud.google.com/go v0.65.0 h1:Dg9iHVQfrhq82rUNu9ZxUDrJLaxFUe/HlCVaLyRruq8=
 cloud.google.com/go v0.65.0/go.mod h1:O5N8zS7uWy9vkA9vayVHs65eM1ubvY4h553ofrNHObY=
-cloud.google.com/go v0.66.0 h1:DZeAkuQGQqnm9Xv36SbMJEU8aFBz4wL04UpMWPWwjzg=
 cloud.google.com/go v0.66.0/go.mod h1:dgqGAjKCDxyhGTtC9dAREQGUJpkceNm1yt590Qno0Ko=
-cloud.google.com/go v0.68.0 h1:AnVkaPGAuWaIY/8a75HlNzZNrHDee6YL4rWkwS+CeyE=
-cloud.google.com/go v0.68.0/go.mod h1:91NO4SCDjUfe1zeC0f4/dpckkUNpuNEyqm4X2KLrzNQ=
+cloud.google.com/go v0.72.0/go.mod h1:M+5Vjvlc2wnp6tjzE102Dw08nGShTscUx2nZMufOKPI=
+cloud.google.com/go v0.73.0 h1:sGvc4e0Cmm4+DKQR76a9VwNukpacQK8TOl5pDl0Pcn0=
+cloud.google.com/go v0.73.0/go.mod h1:BkDh9dFvGjCitVw03TNjKbBxXNKULXXIq6orU6HrJ4Q=
 cloud.google.com/go/bigquery v1.0.1/go.mod h1:i/xbL2UlR5RvWAURpBYZTtm/cXjCha9lbfbpx4poX+o=
 cloud.google.com/go/bigquery v1.3.0/go.mod h1:PjpwJnslEMmckchkHFfq+HTD2DmtT67aNFKH1/VBDHE=
-cloud.google.com/go/bigquery v1.4.0 h1:xE3CPsOgttP4ACBePh79zTKALtXwn/Edhcr16R5hMWU=
 cloud.google.com/go/bigquery v1.4.0/go.mod h1:S8dzgnTigyfTmLBfrtrhyYhwRxG72rYxvftPBK2Dvzc=
 cloud.google.com/go/bigquery v1.5.0/go.mod h1:snEHRnqQbz117VIFhE8bmtwIDY80NLUZUMb4Nv6dBIg=
-cloud.google.com/go/bigquery v1.7.0 h1:a/O/bK/vWrYGOTFtH8di4rBxMZnmkjy+Y5LxpDwo+dA=
 cloud.google.com/go/bigquery v1.7.0/go.mod h1://okPTzCYNXSlb24MZs83e2Do+h+VXtc4gLoIoXIAPc=
-cloud.google.com/go/bigquery v1.8.0 h1:PQcPefKFdaIzjQFbiyOgAqyx8q5djaE7x9Sqe712DPA=
 cloud.google.com/go/bigquery v1.8.0/go.mod h1:J5hqkt3O0uAFnINi6JXValWIb1v0goeZM77hZzJN/fQ=
 cloud.google.com/go/datastore v1.0.0/go.mod h1:LXYbyblFSglQ5pkeyhO+Qmw7ukd3C+pD7TKLgZqpHYE=
-cloud.google.com/go/datastore v1.1.0 h1:/May9ojXjRkPBNVrq+oWLqmWCkr4OU5uRY29bu0mRyQ=
 cloud.google.com/go/datastore v1.1.0/go.mod h1:umbIZjpQpHh4hmRpGhH4tLFup+FVzqBi1b3c64qFpCk=
 cloud.google.com/go/pubsub v1.0.1/go.mod h1:R0Gpsv3s54REJCy4fxDixWD93lHJMoZTyQ2kNxGRt3I=
 cloud.google.com/go/pubsub v1.1.0/go.mod h1:EwwdRX2sKPjnvnqCa270oGRyludottCI76h+R3AArQw=
-cloud.google.com/go/pubsub v1.2.0 h1:Lpy6hKgdcl7a3WGSfJIFmxmcdjSpP6OmBEfcOv1Y680=
 cloud.google.com/go/pubsub v1.2.0/go.mod h1:jhfEVHT8odbXTkndysNHCcx0awwzvfOlguIAii9o8iA=
-cloud.google.com/go/pubsub v1.3.1 h1:ukjixP1wl0LpnZ6LWtZJ0mX5tBmjp1f8Sqer8Z2OMUU=
 cloud.google.com/go/pubsub v1.3.1/go.mod h1:i+ucay31+CNRpDW4Lu78I4xXG+O1r/MAHgjpRVR+TSU=
 cloud.google.com/go/storage v1.0.0/go.mod h1:IhtSnM/ZTZV8YYJWCY8RULGVqBDmpoyjwiyrjsg+URw=
 cloud.google.com/go/storage v1.5.0/go.mod h1:tpKbwo567HUNpVclU5sGELwQWBDZ8gh0ZeosJ0Rtdos=
-cloud.google.com/go/storage v1.6.0 h1:UDpwYIwla4jHGzZJaEJYx1tOejbgSoNqsAfHAUYe2r8=
 cloud.google.com/go/storage v1.6.0/go.mod h1:N7U0C8pVQ/+NIKOBQyamJIeKQKkZ+mxpohlUTyfDhBk=
-cloud.google.com/go/storage v1.8.0 h1:86K1Gel7BQ9/WmNWn7dTKMvTLFzwtBe5FNqYbi9X35g=
 cloud.google.com/go/storage v1.8.0/go.mod h1:Wv1Oy7z6Yz3DshWRJFhqM/UCfaWIRTdp0RXyy7KQOVs=
-cloud.google.com/go/storage v1.10.0 h1:STgFzyU5/8miMl0//zKh2aQeTyeaUH3WN9bSUiJ09bA=
 cloud.google.com/go/storage v1.10.0/go.mod h1:FLPqc6j+Ki4BU591ie1oL6qBQGu2Bl/tZ9ullr3+Kg0=
 cloud.google.com/go/storage v1.12.0 h1:4y3gHptW1EHVtcPAVE0eBBlFuGqEejTTG3KdIE0lUX4=
 cloud.google.com/go/storage v1.12.0/go.mod h1:fFLk2dp2oAhDz8QFKwqrjdJvxSp/W2g7nillojlL5Ho=
 dmitri.shuralyov.com/gpu/mtl v0.0.0-20190408044501-666a987793e9/go.mod h1:H6x//7gZCb22OMCxBHrMx7a5I7Hp++hsVxbQ4BYO7hU=
-github.com/BurntSushi/toml v0.3.1 h1:WXkYYl6Yr3qBf1K79EBnL4mak0OimBfB0XUf9Vl28OQ=
 github.com/BurntSushi/toml v0.3.1/go.mod h1:xHWCNGjB5oqiDr8zfno3MHue2Ht5sIBksp03qcyfWMU=
 github.com/BurntSushi/xgb v0.0.0-20160522181843-27f122750802/go.mod h1:IVnqGOEym/WlBOVXweHU+Q+/VP0lqqI8lqeDx9IjBqo=
 github.com/VictoriaMetrics/fastcache v1.5.7 h1:4y6y0G8PRzszQUYIQHHssv/jgPHAb5qQuuDNdCbyAgw=
 github.com/VictoriaMetrics/fastcache v1.5.7/go.mod h1:ptDBkNMQI4RtmVo8VS/XwRY6RoTu1dAWCbrk+6WsEM8=
-github.com/VictoriaMetrics/fasthttp v1.0.7 h1:9RntF8jE9z+ooyXy0tb3Pl76jan4DLaCQoCQFp/cIUE=
-github.com/VictoriaMetrics/fasthttp v1.0.7/go.mod h1:eaGv8oDDOWE8JQa6GgBgpCwjJhRxavmprtf2oWw8b3o=
-github.com/VictoriaMetrics/metrics v1.12.2 h1:SG8iAmqavDNuh7GIdHPoGHUhDL23KeKfvSZSozucNeA=
+github.com/VictoriaMetrics/fasthttp v1.0.9 h1:Fja1tfcNMNoUD7RJDYpjGx2CsSfXkUbISKY4kNafdN4=
+github.com/VictoriaMetrics/fasthttp v1.0.9/go.mod h1:3SeUL4zwB/p/a9aEeRc6gdlbrtNHXBJR6N376EgiSHU=
 github.com/VictoriaMetrics/metrics v1.12.2/go.mod h1:Z1tSfPfngDn12bTfZSCqArT3OPY3u88J12hSoOhuiRE=
 github.com/VictoriaMetrics/metrics v1.12.3 h1:Fe6JHC6MSEKa+BtLhPN8WIvS+HKPzMc2evEpNeCGy7I=
 github.com/VictoriaMetrics/metrics v1.12.3/go.mod h1:Z1tSfPfngDn12bTfZSCqArT3OPY3u88J12hSoOhuiRE=
-github.com/VictoriaMetrics/metricsql v0.7.1 h1:2V7EbbfKkU2pDzs+D/S0IKYvNSQzDYBlbS8afXD7ntE=
-github.com/VictoriaMetrics/metricsql v0.7.1/go.mod h1:ylO7YITho/Iw6P71oEaGyHbO94bGoGtzWfLGqFhMIg8=
+github.com/VictoriaMetrics/metricsql v0.9.0 h1:mO4YmVRVHQmipTHcSMlCJ7Rctsol7vlu1l2ifh+ibqI=
+github.com/VictoriaMetrics/metricsql v0.9.0/go.mod h1:ylO7YITho/Iw6P71oEaGyHbO94bGoGtzWfLGqFhMIg8=
 github.com/allegro/bigcache v1.2.1-0.20190218064605-e24eb225f156 h1:eMwmnE/GDgah4HI848JfFxHt+iPb26b4zyfspmqY0/8=
 github.com/allegro/bigcache v1.2.1-0.20190218064605-e24eb225f156/go.mod h1:Cb/ax3seSYIx7SuZdm2G2xzfwmv3TPSk2ucNfQESPXM=
 github.com/andybalholm/brotli v1.0.0/go.mod h1:loMXtMfwqflxFJPmdbJO0a3KNoPuLBgiu3qAvBg8x/Y=
-github.com/aws/aws-sdk-go v1.35.5 h1:doSEOxC0UkirPcle20Rc+1kAhJ4Ip+GSEeZ3nKl7Qlk=
-github.com/aws/aws-sdk-go v1.35.5/go.mod h1:tlPOdRjfxPBpNIwqDj61rmsnA85v9jc0Ps9+muhnW+k=
+github.com/aws/aws-sdk-go v1.36.2 h1:UAeFPct+jHqWM+tgiqDrC9/sfbWj6wkcvpsJ+zdcsvA=
+github.com/aws/aws-sdk-go v1.36.2/go.mod h1:hcU610XS61/+aQV88ixoOzUoG7v3b31pl2zKMmprdro=
 github.com/census-instrumentation/opencensus-proto v0.2.1/go.mod h1:f6KPmirojxKA12rnyqOA5BBL4O983OfeGPqjHWSTneU=
 github.com/cespare/xxhash/v2 v2.1.1 h1:6MnRN8NT7+YBpUIWxHtefFZOKTAPgGjpQSxqLNn0+qY=
 github.com/cespare/xxhash/v2 v2.1.1/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
@@ -76,12 +61,14 @@ github.com/chzyer/readline v0.0.0-20180603132655-2972be24d48e/go.mod h1:nSuG5e5P
 github.com/chzyer/test v0.0.0-20180213035817-a1ea475d72b1/go.mod h1:Q3SI9o4m/ZMnBNeIyt5eFwwo7qiLfzFZmjNmxjkiQlU=
 github.com/client9/misspell v0.3.4/go.mod h1:qj6jICC3Q7zFZvVWo7KLAzC3yx5G7kyvSDkc90ppPyw=
 github.com/cncf/udpa/go v0.0.0-20191209042840-269d4d468f6f/go.mod h1:M8M6+tZqaGXZJjfX53e64911xZQV5JYwmTeXPW+k8Sc=
+github.com/cncf/udpa/go v0.0.0-20200629203442-efcf912fb354/go.mod h1:WmhPx2Nbnhtbo57+VJT5O0JRkEi1Wbu0z5j0R8u5Hbk=
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/envoyproxy/go-control-plane v0.9.0/go.mod h1:YTl/9mNaCwkRvm6d1a2C3ymFceY/DCBVvsKhRF0iEA4=
 github.com/envoyproxy/go-control-plane v0.9.1-0.20191026205805-5f8ba28d4473/go.mod h1:YTl/9mNaCwkRvm6d1a2C3ymFceY/DCBVvsKhRF0iEA4=
 github.com/envoyproxy/go-control-plane v0.9.4/go.mod h1:6rpuAdCZL397s3pYoYcLgu1mIlRU8Am5FuJP05cCM98=
+github.com/envoyproxy/go-control-plane v0.9.7/go.mod h1:cwu0lG7PUMfa9snN8LXBig5ynNVH9qI8YYLbd1fK2po=
 github.com/envoyproxy/protoc-gen-validate v0.1.0/go.mod h1:iSmxcyjqTsJpI2R4NaDN7+kN2VEUnK/pcBlmesArF7c=
 github.com/go-gl/glfw v0.0.0-20190409004039-e6da0acd62b1/go.mod h1:vR7hzQXu2zJy9AVAgeJqvqgH9Q5CA+iKCZ2gyEVpxRU=
 github.com/go-gl/glfw/v3.3/glfw v0.0.0-20191125211704-12ad95a8df72/go.mod h1:tQ2UAYgL5IevRw8kRxooKSPJfGvJ9fJQFa0TUsXzTg8=
@@ -110,9 +97,9 @@ github.com/golang/protobuf v1.4.0-rc.2/go.mod h1:LlEzMj4AhA7rCAGe4KMBDvJI+AwstrU
 github.com/golang/protobuf v1.4.0-rc.4.0.20200313231945-b860323f09d0/go.mod h1:WU3c8KckQ9AFe+yFwt9sWVRKCVIyN9cPHBJSNnbL67w=
 github.com/golang/protobuf v1.4.0/go.mod h1:jodUvKwWbYaEsadDk5Fwe5c77LiNKVO9IDvqG2KuDX0=
 github.com/golang/protobuf v1.4.1/go.mod h1:U8fpvMrcmy5pZrNK1lt4xCsGvpyWQ/VVv6QDs8UjoX8=
-github.com/golang/protobuf v1.4.2 h1:+Z5KGCizgyZCbGh1KZqA0fcLLkwbsjIzS4aV2v7wJX0=
 github.com/golang/protobuf v1.4.2/go.mod h1:oDoupMAO8OvCJWAcko0GGGIgR6R6ocIYbsSw735rRwI=
-github.com/golang/snappy v0.0.1 h1:Qgr9rKW7uDUkrbSmQeiDsGa8SjGyCOGtuasMWwvp2P4=
+github.com/golang/protobuf v1.4.3 h1:JjCZWpVbqXDqFVmTfYWEVTMIYrL/NPdPSCHPJ0T/raM=
+github.com/golang/protobuf v1.4.3/go.mod h1:oDoupMAO8OvCJWAcko0GGGIgR6R6ocIYbsSw735rRwI=
 github.com/golang/snappy v0.0.1/go.mod h1:/XxbfmMg8lxefKM7IXC3fBNl/7bRcc72aCRzEWrmP2Q=
 github.com/golang/snappy v0.0.2 h1:aeE13tS0IiQgFjYdoL8qN3K1N2bXXtI6Vi51/y7BpMw=
 github.com/golang/snappy v0.0.2/go.mod h1:/XxbfmMg8lxefKM7IXC3fBNl/7bRcc72aCRzEWrmP2Q=
@@ -124,14 +111,15 @@ github.com/google/go-cmp v0.3.1/go.mod h1:8QqcDgzrUqlUb/G2PQTWiueGozuR1884gddMyw
 github.com/google/go-cmp v0.4.0/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
 github.com/google/go-cmp v0.4.1/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
 github.com/google/go-cmp v0.5.0/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
-github.com/google/go-cmp v0.5.1 h1:JFrFEBb2xKufg6XkJsJr+WbKb4FQlURi5RUcBveYu9k=
 github.com/google/go-cmp v0.5.1/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
-github.com/google/go-cmp v0.5.2 h1:X2ev0eStA3AbceY54o37/0PQ/UWqKEiiO2dKL5OPaFM=
 github.com/google/go-cmp v0.5.2/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
+github.com/google/go-cmp v0.5.4 h1:L8R9j+yAqZuZjsqh/z+F1NCffTKKLShY6zXTItVIZ8M=
+github.com/google/go-cmp v0.5.4/go.mod h1:v8dTdLbMG2kIc/vJvl+f65V22dbkXbowE6jgT/gNBxE=
 github.com/google/martian v2.1.0+incompatible h1:/CP5g8u/VJHijgedC/Legn3BAbAaWPgecwXBIDzw5no=
 github.com/google/martian v2.1.0+incompatible/go.mod h1:9I4somxYTbIHy5NJKHRl3wXiIaQGbYVAs8BPL6v8lEs=
-github.com/google/martian/v3 v3.0.0 h1:pMen7vLs8nvgEYhywH3KDWJIJTeEr2ULsVWHWYHQyBs=
 github.com/google/martian/v3 v3.0.0/go.mod h1:y5Zk1BBys9G+gd6Jrk0W3cC1+ELVxBWuIGO+w/tUAp0=
+github.com/google/martian/v3 v3.1.0 h1:wCKgOCHuUEVfsaQLpPSJb7VdYCdTVZQAuOdYm1yc/60=
+github.com/google/martian/v3 v3.1.0/go.mod h1:y5Zk1BBys9G+gd6Jrk0W3cC1+ELVxBWuIGO+w/tUAp0=
 github.com/google/pprof v0.0.0-20181206194817-3ea8567a2e57/go.mod h1:zfwlbNMJ+OItoe0UupaVj+oy1omPYYDuagoSzA8v9mc=
 github.com/google/pprof v0.0.0-20190515194954-54271f7e092f/go.mod h1:zfwlbNMJ+OItoe0UupaVj+oy1omPYYDuagoSzA8v9mc=
 github.com/google/pprof v0.0.0-20191218002539-d4f498aebedc/go.mod h1:ZgVRPoUq/hfqzAqh7sHMqb3I9Rq5C59dIz2SbBwJ4eM=
@@ -140,13 +128,17 @@ github.com/google/pprof v0.0.0-20200229191704-1ebb73c60ed3/go.mod h1:ZgVRPoUq/hf
 github.com/google/pprof v0.0.0-20200430221834-fc25d7d30c6d/go.mod h1:ZgVRPoUq/hfqzAqh7sHMqb3I9Rq5C59dIz2SbBwJ4eM=
 github.com/google/pprof v0.0.0-20200708004538-1a94d8640e99/go.mod h1:ZgVRPoUq/hfqzAqh7sHMqb3I9Rq5C59dIz2SbBwJ4eM=
 github.com/google/pprof v0.0.0-20200905233945-acf8798be1f7/go.mod h1:ZgVRPoUq/hfqzAqh7sHMqb3I9Rq5C59dIz2SbBwJ4eM=
+github.com/google/pprof v0.0.0-20201023163331-3e6fc7fc9c4c/go.mod h1:kpwsk12EmLew5upagYY7GY0pfYCcupk39gWOCRROcvE=
+github.com/google/pprof v0.0.0-20201117184057-ae444373da19/go.mod h1:kpwsk12EmLew5upagYY7GY0pfYCcupk39gWOCRROcvE=
 github.com/google/renameio v0.1.0/go.mod h1:KWCgfxg9yswjAJkECMjeO8J8rahYeXnNhOm40UhjYkI=
+github.com/google/uuid v1.1.2/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/googleapis/gax-go/v2 v2.0.4/go.mod h1:0Wqv26UfaUD9n4G6kQubkQ+KchISgw+vpHVxEJEs9eg=
 github.com/googleapis/gax-go/v2 v2.0.5 h1:sjZBwGj9Jlw33ImPtvFviGYvseOtDM7hkSKB7+Tv3SM=
 github.com/googleapis/gax-go/v2 v2.0.5/go.mod h1:DWXyrwAJ9X0FpwwEdw+IPEYBICEFu5mhpdKc/us6bOk=
 github.com/hashicorp/golang-lru v0.5.0/go.mod h1:/m3WP610KZHVQ1SGc6re/UDhFvYD7pJ4Ao+sR/qLZy8=
 github.com/hashicorp/golang-lru v0.5.1/go.mod h1:/m3WP610KZHVQ1SGc6re/UDhFvYD7pJ4Ao+sR/qLZy8=
 github.com/ianlancetaylor/demangle v0.0.0-20181102032728-5e5cf60278f6/go.mod h1:aSSvb/t6k1mPoxDqO4vJh6VOCGPwU4O0C2/Eqndh1Sc=
+github.com/ianlancetaylor/demangle v0.0.0-20200824232613-28f6c0f3b639/go.mod h1:aSSvb/t6k1mPoxDqO4vJh6VOCGPwU4O0C2/Eqndh1Sc=
 github.com/jmespath/go-jmespath v0.4.0 h1:BEgLn5cpjn8UN1mAw4NjwDrS35OdebyEtFe+9YPoQUg=
 github.com/jmespath/go-jmespath v0.4.0/go.mod h1:T8mJZnbsbmF+m6zOOFylbeCJqk5+pHWvzYPziyZiYoo=
 github.com/jmespath/go-jmespath/internal/testify v1.5.1 h1:shLQSRRSCCPj3f2gpwzGwWFoC7ycTf1rcQZHOlsJ6N8=
@@ -156,10 +148,9 @@ github.com/jstemmer/go-junit-report v0.9.1 h1:6QPYqodiu3GuPL+7mfx+NwDdp2eTkp9IfE
 github.com/jstemmer/go-junit-report v0.9.1/go.mod h1:Brl9GWCQeLvo8nXZwPNNblvFj/XSXhF0NWZEnDohbsk=
 github.com/kisielk/gotool v1.0.0/go.mod h1:XhKaO+MFFWcvkIS/tQcRk01m1F5IRFswLeQ+oQHNcck=
 github.com/klauspost/compress v1.10.7/go.mod h1:aoV0uJVorq1K+umq18yTdKaF57EivdYsUV+/s2qKfXs=
-github.com/klauspost/compress v1.11.0 h1:wJbzvpYMVGG9iTI9VxpnNZfd4DzMPoCWze3GgSqz8yg=
 github.com/klauspost/compress v1.11.0/go.mod h1:aoV0uJVorq1K+umq18yTdKaF57EivdYsUV+/s2qKfXs=
-github.com/klauspost/compress v1.11.1 h1:bPb7nMRdOZYDrpPMTA3EInUQrdgoBinqUuSwlGdKDdE=
-github.com/klauspost/compress v1.11.1/go.mod h1:aoV0uJVorq1K+umq18yTdKaF57EivdYsUV+/s2qKfXs=
+github.com/klauspost/compress v1.11.3 h1:dB4Bn0tN3wdCzQxnS8r06kV74qN/TAfaIS0bVE8h3jc=
+github.com/klauspost/compress v1.11.3/go.mod h1:aoV0uJVorq1K+umq18yTdKaF57EivdYsUV+/s2qKfXs=
 github.com/kr/pretty v0.1.0 h1:L/CwN0zerZDmRFUapSPitk6f+Q3+0za1rQkzVuMiMFI=
 github.com/kr/pretty v0.1.0/go.mod h1:dAy3ld7l9f0ibDNOQOHHMYYIIbhfbHSm3C4ZsoJORNo=
 github.com/kr/pty v1.1.1/go.mod h1:pFQYn66WHrOpPYNljwOMqo10TkYh1fy3cYio2l3bCsQ=
@@ -178,8 +169,8 @@ github.com/stretchr/testify v1.5.1/go.mod h1:5W2xD1RspED5o8YsWQXVCued0rvSQ+mT+I5
 github.com/valyala/bytebufferpool v1.0.0 h1:GqA5TC/0021Y/b9FG4Oi9Mr3q7XYx6KllzawFIhcdPw=
 github.com/valyala/bytebufferpool v1.0.0/go.mod h1:6bBcMArwyJ5K/AmCkWv1jt77kVWyCJ6HpOuEn7z0Csc=
 github.com/valyala/fasthttp v1.16.0/go.mod h1:YOKImeEosDdBPnxc0gy7INqi3m1zK6A+xl6TwOBhHCA=
-github.com/valyala/fastjson v1.6.1 h1:qJs/Kz/HebWzk8LmhOrSm7kdOyJBr1XB+zSkYtEEfQE=
-github.com/valyala/fastjson v1.6.1/go.mod h1:CLCAqky6SMuOcxStkYQvblddUtoRxhYMGLrsQns1aXY=
+github.com/valyala/fastjson v1.6.3 h1:tAKFnnwmeMGPbwJ7IwxcTPCNr3uIzoIj3/Fh90ra4xc=
+github.com/valyala/fastjson v1.6.3/go.mod h1:CLCAqky6SMuOcxStkYQvblddUtoRxhYMGLrsQns1aXY=
 github.com/valyala/fastrand v1.0.0 h1:LUKT9aKer2dVQNUi3waewTbKV+7H17kvWFNKs2ObdkI=
 github.com/valyala/fastrand v1.0.0/go.mod h1:HWqCzkrkg6QXT8V2EXWvXCoow7vLwOFN002oeRzjapQ=
 github.com/valyala/fasttemplate v1.2.1 h1:TVEnxayobAdVkhQfrfes2IzOB6o+z4roRkPF52WA1u4=
@@ -199,7 +190,6 @@ go.opencensus.io v0.21.0/go.mod h1:mSImk1erAIZhrmZN+AvHh14ztQfjbGwt4TtuofqLduU=
 go.opencensus.io v0.22.0/go.mod h1:+kGneAE2xo2IficOXnaByMWTGM9T73dGwxeWcUqIpI8=
 go.opencensus.io v0.22.2/go.mod h1:yxeiOL68Rb0Xd1ddK5vPZ/oVn4vY4Ynel7k9FzqtOIw=
 go.opencensus.io v0.22.3/go.mod h1:yxeiOL68Rb0Xd1ddK5vPZ/oVn4vY4Ynel7k9FzqtOIw=
-go.opencensus.io v0.22.4 h1:LYy1Hy3MJdrCdMwwzxA/dRok4ejH+RwNGbuoD9fCjto=
 go.opencensus.io v0.22.4/go.mod h1:yxeiOL68Rb0Xd1ddK5vPZ/oVn4vY4Ynel7k9FzqtOIw=
 go.opencensus.io v0.22.5 h1:dntmOdLpSpHlVqbW5Eay97DelsZHe+55D+xC6i0dDS0=
 go.opencensus.io v0.22.5/go.mod h1:5pWMHQbX5EPX2/62yrJeAkowc+lfs/XD7Uxpq3pI6kk=
@@ -238,8 +228,9 @@ golang.org/x/mod v0.1.0/go.mod h1:0QHyrYULN0/3qlju5TqG8bIK38QM8yzMo5ekMj3DlcY=
 golang.org/x/mod v0.1.1-0.20191105210325-c90efee705ee/go.mod h1:QqPTAvyqsEbceGzBzNggFXnrqF1CaUcvgkdR5Ot7KZg=
 golang.org/x/mod v0.1.1-0.20191107180719-034126e5016b/go.mod h1:QqPTAvyqsEbceGzBzNggFXnrqF1CaUcvgkdR5Ot7KZg=
 golang.org/x/mod v0.2.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
-golang.org/x/mod v0.3.0 h1:RM4zey1++hCTbCVQfnWeKs9/IEsaBLA8vTkd0WVtmH4=
 golang.org/x/mod v0.3.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
+golang.org/x/mod v0.4.0 h1:8pl+sMODzuvGJkmj2W4kZihvVb5mKm8pB/X44PIQHv8=
+golang.org/x/mod v0.4.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
 golang.org/x/net v0.0.0-20180724234803-3673e40ba225/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20180826012351-8a410e7b638d/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20190108225652-1e06a53dbb7e/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
@@ -265,24 +256,23 @@ golang.org/x/net v0.0.0-20200513185701-a91f0712d120/go.mod h1:qpuaurCH72eLCgpAm/
 golang.org/x/net v0.0.0-20200520182314-0ba52f642ac2/go.mod h1:qpuaurCH72eLCgpAm/N6yyVIVM9cpaDIP3A8BGJEC5A=
 golang.org/x/net v0.0.0-20200602114024-627f9648deb9/go.mod h1:qpuaurCH72eLCgpAm/N6yyVIVM9cpaDIP3A8BGJEC5A=
 golang.org/x/net v0.0.0-20200625001655-4c5254603344/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
-golang.org/x/net v0.0.0-20200707034311-ab3426394381 h1:VXak5I6aEWmAXeQjA+QSZzlgNrpq9mjcfDemuexIKsU=
 golang.org/x/net v0.0.0-20200707034311-ab3426394381/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
-golang.org/x/net v0.0.0-20200822124328-c89045814202 h1:VvcQYSHwXgi7W+TpUR6A9g6Up98WAHf3f/ulnJ62IyA=
 golang.org/x/net v0.0.0-20200822124328-c89045814202/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
-golang.org/x/net v0.0.0-20200904194848-62affa334b73 h1:MXfv8rhZWmFeqX3GNZRsd6vOLoaCHjYEX3qkRo3YBUA=
 golang.org/x/net v0.0.0-20200904194848-62affa334b73/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
-golang.org/x/net v0.0.0-20200927032502-5d4f70055728 h1:5wtQIAulKU5AbLQOkjxl32UufnIOqgBX72pS0AV14H0=
-golang.org/x/net v0.0.0-20200927032502-5d4f70055728/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
-golang.org/x/net v0.0.0-20201006153459-a7d1128ccaa0 h1:wBouT66WTYFXdxfVdz9sVWARVd/2vfGcmI45D2gj45M=
-golang.org/x/net v0.0.0-20201006153459-a7d1128ccaa0/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
+golang.org/x/net v0.0.0-20201021035429-f5854403a974/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
+golang.org/x/net v0.0.0-20201031054903-ff519b6c9102/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
+golang.org/x/net v0.0.0-20201110031124-69a78807bb2b/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
+golang.org/x/net v0.0.0-20201202161906-c7110b5ffcbb h1:eBmm0M9fYhWpKZLjQUUKka/LtIxf46G4fxeEz5KJr9U=
+golang.org/x/net v0.0.0-20201202161906-c7110b5ffcbb/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
 golang.org/x/oauth2 v0.0.0-20180821212333-d2e6202438be/go.mod h1:N/0e6XlmueqKjAGxoOufVs8QHGRruUQn6yWY3a++T0U=
 golang.org/x/oauth2 v0.0.0-20190226205417-e64efc72b421/go.mod h1:gOpvHmFTYa4IltrdGE7lF6nIHvwfUNPOp7c8zoXwtLw=
 golang.org/x/oauth2 v0.0.0-20190604053449-0f29369cfe45/go.mod h1:gOpvHmFTYa4IltrdGE7lF6nIHvwfUNPOp7c8zoXwtLw=
 golang.org/x/oauth2 v0.0.0-20191202225959-858c2ad4c8b6/go.mod h1:gOpvHmFTYa4IltrdGE7lF6nIHvwfUNPOp7c8zoXwtLw=
-golang.org/x/oauth2 v0.0.0-20200107190931-bf48bf16ab8d h1:TzXSXBo42m9gQenoE3b9BGiEpg5IG2JkU5FkPIawgtw=
 golang.org/x/oauth2 v0.0.0-20200107190931-bf48bf16ab8d/go.mod h1:gOpvHmFTYa4IltrdGE7lF6nIHvwfUNPOp7c8zoXwtLw=
-golang.org/x/oauth2 v0.0.0-20200902213428-5d25da1a8d43 h1:ld7aEMNHoBnnDAX15v1T6z31v8HwR2A9FYOuAhWqkwc=
 golang.org/x/oauth2 v0.0.0-20200902213428-5d25da1a8d43/go.mod h1:KelEdhl1UZF7XfJ4dDtk6s++YSgaE7mD/BuKKDLBl4A=
+golang.org/x/oauth2 v0.0.0-20201109201403-9fd604954f58/go.mod h1:KelEdhl1UZF7XfJ4dDtk6s++YSgaE7mD/BuKKDLBl4A=
+golang.org/x/oauth2 v0.0.0-20201203001011-0b49973bad19 h1:ZD+2Sd/BnevwJp8PSli8WgGAGzb9IZtxBsv1iZMYeEA=
+golang.org/x/oauth2 v0.0.0-20201203001011-0b49973bad19/go.mod h1:KelEdhl1UZF7XfJ4dDtk6s++YSgaE7mD/BuKKDLBl4A=
 golang.org/x/sync v0.0.0-20180314180146-1d60e4601c6f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20181108010431-42b317875d0f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20181221193216-37e7f081c4d4/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
@@ -291,6 +281,7 @@ golang.org/x/sync v0.0.0-20190423024810-112230192c58/go.mod h1:RxMgew5VJxzue5/jJ
 golang.org/x/sync v0.0.0-20190911185100-cd5d95a43a6e/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20200317015054-43a5402ce75a/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20200625203802-6e8e738ad208/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
+golang.org/x/sync v0.0.0-20201020160332-67f06af15bc9/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sys v0.0.0-20180830151530-49385e6e1522/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20190312061237-fead79001313/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
@@ -320,14 +311,16 @@ golang.org/x/sys v0.0.0-20200803210538-64077c9b5642/go.mod h1:h1NjWce9XRLGQEsW7w
 golang.org/x/sys v0.0.0-20200828194041-157a740278f4/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200905004654-be1d3432aa8f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
-golang.org/x/sys v0.0.0-20201008064518-c1f3e3309c71 h1:ZPX6UakxrJCxWiyGWpXtFY+fp86Esy7xJT/jJCG8bgU=
-golang.org/x/sys v0.0.0-20201008064518-c1f3e3309c71/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20201201145000-ef89a241ccb3/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20201204225414-ed752295db88 h1:KmZPnMocC93w341XZp26yTJg8Za7lhb2KhkYmixoeso=
+golang.org/x/sys v0.0.0-20201204225414-ed752295db88/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/text v0.0.0-20170915032832-14c0d48ead0c/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.1-0.20180807135948-17ff2d5776d2/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.2/go.mod h1:bEr9sfX3Q8Zfm5fL9x+3itogRgK3+ptLWKqgva+5dAk=
-golang.org/x/text v0.3.3 h1:cokOdA+Jmi5PJGXLlLllQSgYigAEfHXJAERHVMaCc2k=
 golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
+golang.org/x/text v0.3.4 h1:0YWbFKbhXG/wIiuHDSKpS0Iy7FSA+u45VtBMfQcFTTc=
+golang.org/x/text v0.3.4/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
 golang.org/x/time v0.0.0-20181108054448-85acf8d2951c/go.mod h1:tRJNPiyCQ0inRvYxbN9jk5I+vvW/OXSQhTDSoE431IQ=
 golang.org/x/time v0.0.0-20190308202827-9d24e82272b4/go.mod h1:tRJNPiyCQ0inRvYxbN9jk5I+vvW/OXSQhTDSoE431IQ=
 golang.org/x/time v0.0.0-20191024005414-555d28b269f0/go.mod h1:tRJNPiyCQ0inRvYxbN9jk5I+vvW/OXSQhTDSoE431IQ=
@@ -375,9 +368,11 @@ golang.org/x/tools v0.0.0-20200828161849-5deb26317202/go.mod h1:njjCfa9FT2d7l9Bc
 golang.org/x/tools v0.0.0-20200904185747-39188db58858/go.mod h1:Cj7w3i3Rnn0Xh82ur9kSqwfTHTeVxaDqrfMjpcNT6bE=
 golang.org/x/tools v0.0.0-20200915173823-2db8f0ff891c/go.mod h1:z6u4i615ZeAfBE4XtMziQW1fSVJXACjjbWkB/mvPzlU=
 golang.org/x/tools v0.0.0-20200918232735-d647fc253266/go.mod h1:z6u4i615ZeAfBE4XtMziQW1fSVJXACjjbWkB/mvPzlU=
-golang.org/x/tools v0.0.0-20201002184944-ecd9fd270d5d/go.mod h1:z6u4i615ZeAfBE4XtMziQW1fSVJXACjjbWkB/mvPzlU=
-golang.org/x/tools v0.0.0-20201008025239-9df69603baec h1:RY2OghEV/7X1MLaecgm1mwFd3sGvUddm5pGVSxQvX0c=
-golang.org/x/tools v0.0.0-20201008025239-9df69603baec/go.mod h1:z6u4i615ZeAfBE4XtMziQW1fSVJXACjjbWkB/mvPzlU=
+golang.org/x/tools v0.0.0-20201110124207-079ba7bd75cd/go.mod h1:emZCQorbCU4vsT4fOWvOPXz4eW1wZW4PmDk9uLelYpA=
+golang.org/x/tools v0.0.0-20201201161351-ac6f37ff4c2a/go.mod h1:emZCQorbCU4vsT4fOWvOPXz4eW1wZW4PmDk9uLelYpA=
+golang.org/x/tools v0.0.0-20201202200335-bef1c476418a/go.mod h1:emZCQorbCU4vsT4fOWvOPXz4eW1wZW4PmDk9uLelYpA=
+golang.org/x/tools v0.0.0-20201204222352-654352759326 h1:XKLw9EEEfGJFE2K5Ni4nXgtFBIfI+zKPIi2SlRYmIG4=
+golang.org/x/tools v0.0.0-20201204222352-654352759326/go.mod h1:emZCQorbCU4vsT4fOWvOPXz4eW1wZW4PmDk9uLelYpA=
 golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191011141410-1b5146add898/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
@@ -398,19 +393,20 @@ google.golang.org/api v0.22.0/go.mod h1:BwFmGc8tA3vsd7r/7kR8DY7iEEGSU04BFxCo5jP/
 google.golang.org/api v0.24.0/go.mod h1:lIXQywCXRcnZPGlsd8NbLnOjtAoL6em04bJ9+z0MncE=
 google.golang.org/api v0.28.0/go.mod h1:lIXQywCXRcnZPGlsd8NbLnOjtAoL6em04bJ9+z0MncE=
 google.golang.org/api v0.29.0/go.mod h1:Lcubydp8VUV7KeIHD9z2Bys/sm/vGKnG1UHuDBSrHWM=
-google.golang.org/api v0.30.0 h1:yfrXXP61wVuLb0vBcG6qaOoIoqYEzOQS8jum51jkv2w=
 google.golang.org/api v0.30.0/go.mod h1:QGmEvQ87FHZNiUVJkT14jQNYJ4ZJjdRF23ZXz5138Fc=
-google.golang.org/api v0.31.0 h1:1w5Sz/puhxFo9lTtip2n47k7toB/U2nCqOKNHd3Yrbo=
 google.golang.org/api v0.31.0/go.mod h1:CL+9IBCa2WWU6gRuBWaKqGWLFFwbEUXkfeMkHLQWYWo=
-google.golang.org/api v0.32.0 h1:Le77IccnTqEa8ryp9wIpX5W3zYm7Gf9LhOp9PHcwFts=
 google.golang.org/api v0.32.0/go.mod h1:/XrVsuzM0rZmrsbjJutiuftIzeuTQcEeaYcSk/mQ1dg=
+google.golang.org/api v0.35.0/go.mod h1:/XrVsuzM0rZmrsbjJutiuftIzeuTQcEeaYcSk/mQ1dg=
+google.golang.org/api v0.36.0 h1:l2Nfbl2GPXdWorv+dT2XfinX2jOOw4zv1VhLstx+6rE=
+google.golang.org/api v0.36.0/go.mod h1:+z5ficQTmoYpPn8LCUNVpK5I7hwkpjbcgqA7I34qYtE=
 google.golang.org/appengine v1.1.0/go.mod h1:EbEs0AVv82hx2wNQdGPgUI5lhzA/G0D9YwlJXL52JkM=
 google.golang.org/appengine v1.4.0/go.mod h1:xpcJRLb0r/rnEns0DIKYYv+WjYCduHsrkT7/EB5XEv4=
 google.golang.org/appengine v1.5.0/go.mod h1:xpcJRLb0r/rnEns0DIKYYv+WjYCduHsrkT7/EB5XEv4=
 google.golang.org/appengine v1.6.1/go.mod h1:i06prIuMbXzDqacNJfV5OdTW448YApPu5ww/cMBSeb0=
 google.golang.org/appengine v1.6.5/go.mod h1:8WjMMxjGQR8xUklV/ARdw2HLXBOI7O7uCIDZVag1xfc=
-google.golang.org/appengine v1.6.6 h1:lMO5rYAqUxkmaj76jAkRUvt5JZgFymx/+Q5Mzfivuhc=
 google.golang.org/appengine v1.6.6/go.mod h1:8WjMMxjGQR8xUklV/ARdw2HLXBOI7O7uCIDZVag1xfc=
+google.golang.org/appengine v1.6.7 h1:FZR1q0exgwxzPzp/aF+VccGrSfxfPpkBqjIIEq3ru6c=
+google.golang.org/appengine v1.6.7/go.mod h1:8WjMMxjGQR8xUklV/ARdw2HLXBOI7O7uCIDZVag1xfc=
 google.golang.org/genproto v0.0.0-20180817151627-c66870c02cf8/go.mod h1:JiN7NxoALGmiZfu7CAH4rXhgtRTLTxftemlI0sWmxmc=
 google.golang.org/genproto v0.0.0-20190307195333-5fe7a883aa19/go.mod h1:VzzqZJRnGkLBvHegQrXjBqPurQTc5/KpmUdxsrq26oE=
 google.golang.org/genproto v0.0.0-20190418145605-e7d98fc518a7/go.mod h1:VzzqZJRnGkLBvHegQrXjBqPurQTc5/KpmUdxsrq26oE=
@@ -441,14 +437,14 @@ google.golang.org/genproto v0.0.0-20200729003335-053ba62fc06f/go.mod h1:FWY/as6D
 google.golang.org/genproto v0.0.0-20200804131852-c06518451d9c/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
 google.golang.org/genproto v0.0.0-20200825200019-8632dd797987/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
 google.golang.org/genproto v0.0.0-20200831141814-d751682dd103/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
-google.golang.org/genproto v0.0.0-20200904004341-0bd0a958aa1d h1:92D1fum1bJLKSdr11OJ+54YeCMCGYIygTA7R/YZxH5M=
 google.golang.org/genproto v0.0.0-20200904004341-0bd0a958aa1d/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
 google.golang.org/genproto v0.0.0-20200914193844-75d14daec038/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
 google.golang.org/genproto v0.0.0-20200921151605-7abf4a1a14d5/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
-google.golang.org/genproto v0.0.0-20201002142447-3860012362da h1:DTQYk4u7nICKkkVZsBv0/0po0ChISxAJ5CTAfUhO0PQ=
-google.golang.org/genproto v0.0.0-20201002142447-3860012362da/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
-google.golang.org/genproto v0.0.0-20201007142714-5c0e72c5e71e h1:zYWvTjcAbEy4pVj21IfgFga4rAW107tKspqOi853hjI=
-google.golang.org/genproto v0.0.0-20201007142714-5c0e72c5e71e/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
+google.golang.org/genproto v0.0.0-20201109203340-2640f1f9cdfb/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
+google.golang.org/genproto v0.0.0-20201201144952-b05cb90ed32e/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
+google.golang.org/genproto v0.0.0-20201203001206-6486ece9c497/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
+google.golang.org/genproto v0.0.0-20201204160425-06b3db808446 h1:65ppmIPdaZE+BO34gntwqexoTYr30IRNGmS0OGOHu3A=
+google.golang.org/genproto v0.0.0-20201204160425-06b3db808446/go.mod h1:FWY/as6DDZQgahTzZj3fqbO1CbirC29ZNUFHwi0/+no=
 google.golang.org/grpc v1.19.0/go.mod h1:mqu4LbDTu4XGKhr4mRzUsmM4RtVoemTSY81AxZiDr8c=
 google.golang.org/grpc v1.20.1/go.mod h1:10oTOabMzJvdu6/UiuZezV6QK5dSlG84ov/aaiqXj38=
 google.golang.org/grpc v1.21.1/go.mod h1:oYelfM1adQP15Ek0mdvEgi9Df8B9CZIaU1084ijfRaM=
@@ -460,12 +456,12 @@ google.golang.org/grpc v1.27.1/go.mod h1:qbnxyOmOxrQa7FizSgH+ReBfzJrCY1pSN7KXBS8
 google.golang.org/grpc v1.28.0/go.mod h1:rpkK4SK4GF4Ach/+MFLZUBavHOvF2JJB5uozKKal+60=
 google.golang.org/grpc v1.29.1/go.mod h1:itym6AZVZYACWQqET3MqgPpjcuV5QH3BxFS3IjizoKk=
 google.golang.org/grpc v1.30.0/go.mod h1:N36X2cJ7JwdamYAgDz+s+rVMFjt3numwzf/HckM8pak=
-google.golang.org/grpc v1.31.0 h1:T7P4R73V3SSDPhH7WW7ATbfViLtmamH0DKrP3f9AuDI=
 google.golang.org/grpc v1.31.0/go.mod h1:N36X2cJ7JwdamYAgDz+s+rVMFjt3numwzf/HckM8pak=
-google.golang.org/grpc v1.31.1 h1:SfXqXS5hkufcdZ/mHtYCh53P2b+92WQq/DZcKLgsFRs=
 google.golang.org/grpc v1.31.1/go.mod h1:N36X2cJ7JwdamYAgDz+s+rVMFjt3numwzf/HckM8pak=
-google.golang.org/grpc v1.32.0 h1:zWTV+LMdc3kaiJMSTOFz2UgSBgx8RNQoTGiZu3fR9S0=
 google.golang.org/grpc v1.32.0/go.mod h1:N36X2cJ7JwdamYAgDz+s+rVMFjt3numwzf/HckM8pak=
+google.golang.org/grpc v1.33.2/go.mod h1:JMHMWHQWaTccqQQlmk3MJZS+GWXOdAesneDmEnv2fbc=
+google.golang.org/grpc v1.34.0 h1:raiipEjMOIC/TO2AvyTxP25XFdLxNIBwzDh3FM3XztI=
+google.golang.org/grpc v1.34.0/go.mod h1:WotjhfgOW/POjDeRt8vscBtXq+2VjORFy659qA51WJ8=
 google.golang.org/protobuf v0.0.0-20200109180630-ec00e32a8dfd/go.mod h1:DFci5gLYBciE7Vtevhsrf46CRTquxDuWsQurQQe4oz8=
 google.golang.org/protobuf v0.0.0-20200221191635-4d8936d0db64/go.mod h1:kwYJMbMJ01Woi6D6+Kah6886xMZcty6N08ah7+eCXa0=
 google.golang.org/protobuf v0.0.0-20200228230310-ab0ca4ff8a60/go.mod h1:cfTl7dwQJ+fmap5saPgwCLgHXTUD7jkjRqWcaiX5VyM=
@@ -483,8 +479,8 @@ gopkg.in/check.v1 v1.0.0-20180628173108-788fd7840127/go.mod h1:Co6ibVJAznAaIkqp8
 gopkg.in/errgo.v2 v2.1.0/go.mod h1:hNsd1EY+bozCKY1Ytp96fpM3vjJbqLJn88ws8XvfDNI=
 gopkg.in/yaml.v2 v2.2.2/go.mod h1:hI93XBmqTisBFMUTm0b8Fm+jr3Dg1NNxqwp+5A1VGuI=
 gopkg.in/yaml.v2 v2.2.8/go.mod h1:hI93XBmqTisBFMUTm0b8Fm+jr3Dg1NNxqwp+5A1VGuI=
-gopkg.in/yaml.v2 v2.3.0 h1:clyUAQHOM3G0M3f5vQj7LuJrETvjVot3Z5el9nffUtU=
-gopkg.in/yaml.v2 v2.3.0/go.mod h1:hI93XBmqTisBFMUTm0b8Fm+jr3Dg1NNxqwp+5A1VGuI=
+gopkg.in/yaml.v2 v2.4.0 h1:D8xgwECY7CYvx+Y2n4sBz93Jn9JRvxdiyyo8CTfuKaY=
+gopkg.in/yaml.v2 v2.4.0/go.mod h1:RDklbk79AGWmwhnvt/jBztapEOGDOx6ZbXqjP6csGnQ=
 honnef.co/go/tools v0.0.0-20190102054323-c2f93a96b099/go.mod h1:rf3lG4BRIbNafJWhAfAdb/ePZxsR/4RtNHQocxwk9r4=
 honnef.co/go/tools v0.0.0-20190106161140-3f1c8253044a/go.mod h1:rf3lG4BRIbNafJWhAfAdb/ePZxsR/4RtNHQocxwk9r4=
 honnef.co/go/tools v0.0.0-20190418001031-e561f6794a2a/go.mod h1:rf3lG4BRIbNafJWhAfAdb/ePZxsR/4RtNHQocxwk9r4=
--- a/lib/flagutil/duration.go
+++ b/lib/flagutil/duration.go
@@ -0,0 +1,69 @@
+package flagutil
+
+import (
+	"flag"
+	"fmt"
+	"strconv"
+	"strings"
+
+	"github.com/VictoriaMetrics/metricsql"
+)
+
+// NewDuration returns new `duration` flag with the given name, defaultValue and description.
+//
+// DefaultValue is in months.
+func NewDuration(name string, defaultValue float64, description string) *Duration {
+	description += "\nThe following optional suffixes are supported: h (hour), d (day), w (week), y (year). If suffix isn't set, then the duration is counted in months"
+	d := Duration{
+		Msecs:       int64(defaultValue * msecsPerMonth),
+		valueString: fmt.Sprintf("%g", defaultValue),
+	}
+	flag.Var(&d, name, description)
+	return &d
+}
+
+// Duration is a flag for holding duration.
+type Duration struct {
+	// Msecs contains parsed duration in milliseconds.
+	Msecs int64
+
+	valueString string
+}
+
+// String implements flag.Value interface
+func (d *Duration) String() string {
+	return d.valueString
+}
+
+// Set implements flag.Value interface
+func (d *Duration) Set(value string) error {
+	// An attempt to parse value in months.
+	months, err := strconv.ParseFloat(value, 64)
+	if err == nil {
+		if months > maxMonths {
+			return fmt.Errorf("duration months must be smaller than %d; got %g", maxMonths, months)
+		}
+		if months < 0 {
+			return fmt.Errorf("duration months cannot be negative; got %g", months)
+		}
+		d.Msecs = int64(months * msecsPerMonth)
+		d.valueString = value
+		return nil
+	}
+	// Parse duration.
+	value = strings.ToLower(value)
+	if strings.HasSuffix(value, "m") {
+		return fmt.Errorf("duration in months must be set without `m` suffix due to ambiguity with duration in minutes; got %s", value)
+	}
+	msecs, err := metricsql.PositiveDurationValue(value, 0)
+	if err != nil {
+		return err
+	}
+	d.Msecs = msecs
+	d.valueString = value
+	return nil
+}
+
+const maxMonths = 12 * 100
+
+const msecsPerMonth = 31 * 24 * 3600 * 1000
--- a/lib/flagutil/duration_test.go
+++ b/lib/flagutil/duration_test.go
@@ -0,0 +1,57 @@
+package flagutil
+
+import (
+	"strings"
+	"testing"
+)
+
+func TestDurationSetFailure(t *testing.T) {
+	f := func(value string) {
+		t.Helper()
+		var d Duration
+		if err := d.Set(value); err == nil {
+			t.Fatalf("expecting non-nil error in d.Set(%q)", value)
+		}
+	}
+	f("")
+	f("foobar")
+	f("5foobar")
+	f("ah")
+	f("134xd")
+	f("2.43sdfw")
+
+	// Too big value in months
+	f("12345")
+
+	// Negative duration
+	f("-1")
+	f("-34h")
+
+	// Duration in minutes is confused with duration in months
+	f("1m")
+}
+
+func TestDurationSetSuccess(t *testing.T) {
+	f := func(value string, expectedMsecs int64) {
+		t.Helper()
+		var d Duration
+		if err := d.Set(value); err != nil {
+			t.Fatalf("unexpected error in d.Set(%q): %s", value, err)
+		}
+		if d.Msecs != expectedMsecs {
+			t.Fatalf("unexpected result; got %d; want %d", d.Msecs, expectedMsecs)
+		}
+		valueString := d.String()
+		valueExpected := strings.ToLower(value)
+		if valueString != valueExpected {
+			t.Fatalf("unexpected valueString; got %q; want %q", valueString, valueExpected)
+		}
+	}
+	f("0", 0)
+	f("1", msecsPerMonth)
+	f("123.456", 123.456*msecsPerMonth)
+	f("1h", 3600*1000)
+	f("1.5d", 1.5*24*3600*1000)
+	f("2.3W", 2.3*7*24*3600*1000)
+	f("0.25y", 0.25*365*24*3600*1000)
+}
--- a/lib/flagutil/usage.go
+++ b/lib/flagutil/usage.go
@@ -0,0 +1,36 @@
+package flagutil
+
+import (
+	"flag"
+	"fmt"
+	"os"
+	"strings"
+)
+
+// Usage prints s and optional description for all the flags if -h or -help flag is passed to the app.
+func Usage(s string) {
+	f := flag.CommandLine.Output()
+	fmt.Fprintf(f, "%s\n", s)
+	if hasHelpFlag(os.Args[1:]) {
+		flag.PrintDefaults()
+	} else {
+		fmt.Fprintf(f, `Run "%s -help" in order to see the description for all the available flags`+"\n", os.Args[0])
+	}
+}
+
+func hasHelpFlag(args []string) bool {
+	for _, arg := range args {
+		if isHelpArg(arg) {
+			return true
+		}
+	}
+	return false
+}
+
+func isHelpArg(arg string) bool {
+	if !strings.HasPrefix(arg, "-") {
+		return false
+	}
+	arg = strings.TrimPrefix(arg[1:], "-")
+	return arg == "h" || arg == "help"
+}
--- a/lib/fs/fadvise_darwin.go
+++ b/lib/fs/fadvise_darwin.go
@@ -0,0 +1,10 @@
+package fs
+
+import (
+	"os"
+)
+
+func fadviseSequentialRead(f *os.File, prefetch bool) error {
+	// TODO: implement this properly
+	return nil
+}
--- a/lib/fs/fadvise_openbsd.go
+++ b/lib/fs/fadvise_openbsd.go
@@ -0,0 +1,10 @@
+package fs
+
+import (
+	"os"
+)
+
+func fadviseSequentialRead(f *os.File, prefetch bool) error {
+	// TODO: implement this properly
+	return nil
+}
--- a/lib/fs/fadvise_unix.go
+++ b/lib/fs/fadvise_unix.go
@@ -0,0 +1,22 @@
+// +build linux freebsd
+
+package fs
+
+import (
+	"fmt"
+	"os"
+
+	"golang.org/x/sys/unix"
+)
+
+func fadviseSequentialRead(f *os.File, prefetch bool) error {
+	fd := int(f.Fd())
+	mode := unix.FADV_SEQUENTIAL
+	if prefetch {
+		mode |= unix.FADV_WILLNEED
+	}
+	if err := unix.Fadvise(int(fd), 0, 0, mode); err != nil {
+		return fmt.Errorf("error returned from unix.Fadvise(%d): %w", mode, err)
+	}
+	return nil
+}
--- a/lib/fs/reader_at.go
+++ b/lib/fs/reader_at.go
@@ -149,13 +149,22 @@ func (r *ReaderAt) MustClose() {
 	readersCount.Dec()
 }

-// OpenReaderAt opens ReaderAt for reading from filename.
+// MustFadviseSequentialRead hints the OS that f is read mostly sequentially.
+//
+// if prefetch is set, then the OS is hinted to prefetch f data.
+func (r *ReaderAt) MustFadviseSequentialRead(prefetch bool) {
+	if err := fadviseSequentialRead(r.f, prefetch); err != nil {
+		logger.Panicf("FATAL: error in fadviseSequentialRead(%q, %v): %s", r.f.Name(), prefetch, err)
+	}
+}
+
+// MustOpenReaderAt opens ReaderAt for reading from filename.
 //
 // MustClose must be called on the returned ReaderAt when it is no longer needed.
-func OpenReaderAt(path string) (*ReaderAt, error) {
+func MustOpenReaderAt(path string) *ReaderAt {
 	f, err := os.Open(path)
 	if err != nil {
-		return nil, fmt.Errorf("cannot open file %q for reader: %w", path, err)
+		logger.Panicf("FATAL: cannot open file %q for reading: %s", path, err)
 	}
 	var r ReaderAt
 	r.f = f
@@ -163,7 +172,8 @@ func OpenReaderAt(path string) (*ReaderAt, error) {
 	if !*disableMmap {
 		fi, err := f.Stat()
 		if err != nil {
-			return nil, fmt.Errorf("error in stat: %w", err)
+			MustClose(f)
+			logger.Panicf("FATAL: error in fstat(%q): %s", path, err)
 		}
 		size := fi.Size()
 		bm := &pageCacheBitmap{
@@ -179,12 +189,12 @@ func OpenReaderAt(path string) (*ReaderAt, error) {
 		data, err := mmapFile(f, size)
 		if err != nil {
 			MustClose(f)
-			return nil, fmt.Errorf("cannot init reader for %q: %w", path, err)
+			logger.Panicf("FATAL: cannot mmap %q: %s", path, err)
 		}
 		r.mmapData = data
 	}
 	readersCount.Inc()
-	return &r, nil
+	return &r
 }

 func pageCacheBitmapCleaner(pcbm *atomic.Value, stopCh <-chan struct{}) {
--- a/lib/fs/reader_at_test.go
+++ b/lib/fs/reader_at_test.go
@@ -22,10 +22,7 @@ func testReaderAt(t *testing.T, bufSize int) {
 		t.Fatalf("cannot create %q: %s", path, err)
 	}
 	defer MustRemoveAll(path)
-	r, err := OpenReaderAt(path)
-	if err != nil {
-		t.Fatalf("error in OpenReaderAt(%q): %s", path, err)
-	}
+	r := MustOpenReaderAt(path)
 	defer r.MustClose()

 	buf := make([]byte, bufSize)
--- a/lib/fs/reader_at_timing_test.go
+++ b/lib/fs/reader_at_timing_test.go
@@ -29,10 +29,7 @@ func benchmarkReaderAtMustReadAt(b *testing.B, isMmap bool) {
 		b.Fatalf("cannot create %q: %s", path, err)
 	}
 	defer MustRemoveAll(path)
-	r, err := OpenReaderAt(path)
-	if err != nil {
-		b.Fatalf("error in OpenReaderAt(%q): %s", path, err)
-	}
+	r := MustOpenReaderAt(path)
 	defer r.MustClose()

 	b.ResetTimer()
--- a/lib/httpserver/httpserver.go
+++ b/lib/httpserver/httpserver.go
@@ -208,7 +208,7 @@ func handlerWrapper(s *server, w http.ResponseWriter, r *http.Request, rh Reques
 	r.URL.Path = path
 	switch r.URL.Path {
 	case "/health":
-		w.Header().Set("Content-Type", "text/plain")
+		w.Header().Set("Content-Type", "text/plain; charset=utf-8")
 		deadline := atomic.LoadInt64(&s.shutdownDelayDeadline)
 		if deadline <= 0 {
 			w.Write([]byte("OK"))
@@ -244,7 +244,7 @@ func handlerWrapper(s *server, w http.ResponseWriter, r *http.Request, rh Reques
 			return
 		}
 		startTime := time.Now()
-		w.Header().Set("Content-Type", "text/plain")
+		w.Header().Set("Content-Type", "text/plain; charset=utf-8")
 		WritePrometheusMetrics(w)
 		metricsHandlerDuration.UpdateDuration(startTime)
 		return
@@ -306,6 +306,9 @@ func maybeGzipResponseWriter(w http.ResponseWriter, r *http.Request) http.Respon
 	if *disableResponseCompression {
 		return w
 	}
+	if r.Header.Get("Connection") == "Upgrade" {
+		return w
+	}
 	ae := r.Header.Get("Accept-Encoding")
 	if ae == "" {
 		return w
@@ -392,7 +395,7 @@ func (zrw *gzipResponseWriter) Write(p []byte) (int, error) {
 			if h.Get("Content-Type") == "" {
 				// Disable auto-detection of content-type, since it
 				// is incorrectly detected after the compression.
-				h.Set("Content-Type", "text/html")
+				h.Set("Content-Type", "text/html; charset=utf-8")
 			}
 		}
 		zrw.writeHeader()
--- a/lib/httpserver/metrics.go
+++ b/lib/httpserver/metrics.go
@@ -4,6 +4,7 @@ import (
 	"flag"
 	"fmt"
 	"io"
+	"regexp"
 	"strings"
 	"time"

@@ -13,11 +14,14 @@ import (
 	"github.com/VictoriaMetrics/metrics"
 )

+var versionRe = regexp.MustCompile(`v\d+\.\d+\.\d+`)
+
 // WritePrometheusMetrics writes all the registered metrics to w in Prometheus exposition format.
 func WritePrometheusMetrics(w io.Writer) {
 	metrics.WritePrometheus(w, true)

-	fmt.Fprintf(w, "vm_app_version{version=%q} 1\n", buildinfo.Version)
+	fmt.Fprintf(w, "vm_app_version{version=%q, short_version=%q} 1\n", buildinfo.Version,
+		versionRe.FindString(buildinfo.Version))
 	fmt.Fprintf(w, "vm_allowed_memory_bytes %d\n", memory.Allowed())

 	// Export start time and uptime in seconds
--- a/lib/logger/logger.go
+++ b/lib/logger/logger.go
@@ -10,7 +10,6 @@ import (
 	"runtime"
 	"strings"
 	"sync"
-	"sync/atomic"
 	"time"

 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/buildinfo"
@@ -23,8 +22,10 @@ var (
 	loggerOutput      = flag.String("loggerOutput", "stderr", "Output for the logs. Supported values: stderr, stdout")
 	disableTimestamps = flag.Bool("loggerDisableTimestamps", false, "Whether to disable writing timestamps in logs")

-	errorsPerSecondLimit = flag.Int("loggerErrorsPerSecondLimit", 10, "Per-second limit on the number of ERROR messages. If more than the given number of errors "+
+	errorsPerSecondLimit = flag.Int("loggerErrorsPerSecondLimit", 0, "Per-second limit on the number of ERROR messages. If more than the given number of errors "+
 		"are emitted per second, then the remaining errors are suppressed. Zero value disables the rate limit")
+	warnsPerSecondLimit = flag.Int("loggerWarnsPerSecondLimit", 0, "Per-second limit on the number of WARN messages. If more than the given number of warns "+
+		"are emitted per second, then the remaining warns are suppressed. Zero value disables the rate limit")
 )

 // Init initializes the logger.
@@ -36,7 +37,7 @@ func Init() {
 	setLoggerOutput()
 	validateLoggerLevel()
 	validateLoggerFormat()
-	go errorsLoggedCleaner()
+	go logLimiterCleaner()
 	logAllFlags()
 }

@@ -125,14 +126,60 @@ func logLevelSkipframes(skipframes int, level, format string, args ...interface{
 	logMessage(level, msg, 3+skipframes)
 }

-func errorsLoggedCleaner() {
+func logLimiterCleaner() {
 	for {
 		time.Sleep(time.Second)
-		atomic.StoreUint64(&errorsLogged, 0)
+		logLimiter.reset()
 	}
 }

-var errorsLogged uint64
+var logLimiter = newLogLimit()
+
+func newLogLimit() *logLimit {
+	return &logLimit{
+		m: make(map[string]uint64),
+	}
+}
+
+type logLimit struct {
+	mu sync.Mutex
+	m  map[string]uint64
+}
+
+func (ll *logLimit) reset() {
+	ll.mu.Lock()
+	ll.m = make(map[string]uint64, len(ll.m))
+	ll.mu.Unlock()
+}
+
+// needSuppress checks if the number of calls for the given location exceeds the given limit.
+//
+// When the number of calls equals limit, log message prefix returned.
+func (ll *logLimit) needSuppress(location string, limit uint64) (bool, string) {
+	// fast path
+	var msg string
+	if limit == 0 {
+		return false, msg
+	}
+	ll.mu.Lock()
+	defer ll.mu.Unlock()
+
+	if n, ok := ll.m[location]; ok {
+		if n >= limit {
+			switch n {
+			// report only once
+			case limit:
+				msg = fmt.Sprintf("suppressing log message with rate limit=%d: ", limit)
+			default:
+				return true, msg
+			}
+		}
+		ll.m[location] = n + 1
+	} else {
+		ll.m[location] = 1
+	}
+	return false, msg
+}

 type logWriter struct {
 }
@@ -143,13 +190,6 @@ func (lw *logWriter) Write(p []byte) (int, error) {
 }

 func logMessage(level, msg string, skipframes int) {
-	// rate limit ERROR log messages
-	if level == "ERROR" {
-		if n := atomic.AddUint64(&errorsLogged, 1); *errorsPerSecondLimit > 0 && n > uint64(*errorsPerSecondLimit) {
-			return
-		}
-	}
-
 	timestamp := ""
 	if !*disableTimestamps {
 		timestamp = time.Now().UTC().Format("2006-01-02T15:04:05.000Z")
@@ -164,23 +204,39 @@ func logMessage(level, msg string, skipframes int) {
 		// Strip /VictoriaMetrics/ prefix
 		file = file[n+len("/VictoriaMetrics/"):]
 	}
+	location := fmt.Sprintf("%s:%d", file, line)
+
+	// rate limit ERROR and WARN log messages with given limit.
+	if level == "ERROR" || level == "WARN" {
+		limit := uint64(*errorsPerSecondLimit)
+		if level == "WARN" {
+			limit = uint64(*warnsPerSecondLimit)
+		}
+		ok, suppressMessage := logLimiter.needSuppress(location, limit)
+		if ok {
+			return
+		}
+		if len(suppressMessage) > 0 {
+			msg = suppressMessage + msg
+		}
+	}
+
 	for len(msg) > 0 && msg[len(msg)-1] == '\n' {
 		msg = msg[:len(msg)-1]
 	}
 	var logMsg string
 	switch *loggerFormat {
 	case "json":
-		caller := fmt.Sprintf("%s:%d", file, line)
 		if *disableTimestamps {
-			logMsg = fmt.Sprintf(`{"level":%q,"caller":%q,"msg":%q}`+"\n", levelLowercase, caller, msg)
+			logMsg = fmt.Sprintf(`{"level":%q,"caller":%q,"msg":%q}`+"\n", levelLowercase, location, msg)
 		} else {
-			logMsg = fmt.Sprintf(`{"ts":%q,"level":%q,"caller":%q,"msg":%q}`+"\n", timestamp, levelLowercase, caller, msg)
+			logMsg = fmt.Sprintf(`{"ts":%q,"level":%q,"caller":%q,"msg":%q}`+"\n", timestamp, levelLowercase, location, msg)
 		}
 	default:
 		if *disableTimestamps {
-			logMsg = fmt.Sprintf("%s\t%s:%d\t%s\n", levelLowercase, file, line, msg)
+			logMsg = fmt.Sprintf("%s\t%s\t%s\n", levelLowercase, location, msg)
 		} else {
-			logMsg = fmt.Sprintf("%s\t%s\t%s:%d\t%s\n", timestamp, levelLowercase, file, line, msg)
+			logMsg = fmt.Sprintf("%s\t%s\t%s\t%s\n", timestamp, levelLowercase, location, msg)
 		}
 	}

@@ -190,7 +246,6 @@ func logMessage(level, msg string, skipframes int) {
 	mu.Unlock()

 	// Increment vm_log_messages_total
-	location := fmt.Sprintf("%s:%d", file, line)
 	counterName := fmt.Sprintf(`vm_log_messages_total{app_version=%q, level=%q, location=%q}`, buildinfo.Version, levelLowercase, location)
 	metrics.GetOrCreateCounter(counterName).Inc()

--- a/lib/memory/memory.go
+++ b/lib/memory/memory.go
@@ -35,12 +35,12 @@ func initOnce() {
 	mem := sysTotalMemory()
 	if allowedBytes.N <= 0 {
 		if *allowedPercent < 1 || *allowedPercent > 200 {
-			logger.Panicf("FATAL: -memory.allowedPercent must be in the range [1...200]; got %f", *allowedPercent)
+			logger.Panicf("FATAL: -memory.allowedPercent must be in the range [1...200]; got %g", *allowedPercent)
 		}
 		percent := *allowedPercent / 100
 		allowedMemory = int(float64(mem) * percent)
 		remainingMemory = mem - allowedMemory
-		logger.Infof("limiting caches to %d bytes, leaving %d bytes to the OS according to -memory.allowedPercent=%f", allowedMemory, remainingMemory, *allowedPercent)
+		logger.Infof("limiting caches to %d bytes, leaving %d bytes to the OS according to -memory.allowedPercent=%g", allowedMemory, remainingMemory, *allowedPercent)
 	} else {
 		allowedMemory = allowedBytes.N
 		remainingMemory = mem - allowedMemory
--- a/lib/mergeset/encoding.go
+++ b/lib/mergeset/encoding.go
@@ -31,6 +31,12 @@ type inmemoryBlock struct {
 func (ib *inmemoryBlock) Reset() {
 	ib.commonPrefix = ib.commonPrefix[:0]
 	ib.data = ib.data[:0]
+
+	items := ib.items
+	for i := range items {
+		// Remove reference to by slice, so GC could free the byte slice.
+		items[i] = nil
+	}
 	ib.items = ib.items[:0]
 }

--- a/lib/mergeset/part.go
+++ b/lib/mergeset/part.go
@@ -78,30 +78,15 @@ func openFilePart(path string) (*part, error) {
 	metaindexSize := fs.MustFileSize(metaindexPath)

 	indexPath := path + "/index.bin"
-	indexFile, err := fs.OpenReaderAt(indexPath)
-	if err != nil {
-		metaindexFile.MustClose()
-		return nil, fmt.Errorf("cannot open %q: %w", indexPath, err)
-	}
+	indexFile := fs.MustOpenReaderAt(indexPath)
 	indexSize := fs.MustFileSize(indexPath)

 	itemsPath := path + "/items.bin"
-	itemsFile, err := fs.OpenReaderAt(itemsPath)
-	if err != nil {
-		metaindexFile.MustClose()
-		indexFile.MustClose()
-		return nil, fmt.Errorf("cannot open %q: %w", itemsPath, err)
-	}
+	itemsFile := fs.MustOpenReaderAt(itemsPath)
 	itemsSize := fs.MustFileSize(itemsPath)

 	lensPath := path + "/lens.bin"
-	lensFile, err := fs.OpenReaderAt(lensPath)
-	if err != nil {
-		metaindexFile.MustClose()
-		indexFile.MustClose()
-		itemsFile.MustClose()
-		return nil, fmt.Errorf("cannot open %q: %w", lensPath, err)
-	}
+	lensFile := fs.MustOpenReaderAt(lensPath)
 	lensSize := fs.MustFileSize(lensPath)

 	size := metaindexSize + indexSize + itemsSize + lensSize
@@ -231,8 +216,8 @@ func (idxbc *indexBlockCache) cleanByTimeout() {
 	currentTime := fasttime.UnixTimestamp()
 	idxbc.mu.Lock()
 	for k, idxbe := range idxbc.m {
-		// Delete items accessed more than 10 minutes ago.
-		if currentTime-atomic.LoadUint64(&idxbe.lastAccessTime) > 10*60 {
+		// Delete items accessed more than a minute ago.
+		if currentTime-atomic.LoadUint64(&idxbe.lastAccessTime) > 60 {
 			delete(idxbc.m, k)
 		}
 	}
@@ -260,7 +245,7 @@ func (idxbc *indexBlockCache) Get(k uint64) *indexBlock {
 func (idxbc *indexBlockCache) Put(k uint64, idxb *indexBlock) {
 	idxbc.mu.Lock()

-	// Remove superflouos entries.
+	// Remove superfluous entries.
 	if overflow := len(idxbc.m) - getMaxCachedIndexBlocksPerPart(); overflow > 0 {
 		// Remove 10% of items from the cache.
 		overflow = int(float64(len(idxbc.m)) * 0.1)
@@ -378,8 +363,8 @@ func (ibc *inmemoryBlockCache) cleanByTimeout() {
 	currentTime := fasttime.UnixTimestamp()
 	ibc.mu.Lock()
 	for k, ibe := range ibc.m {
-		// Delete items accessed more than 10 minutes ago.
-		if currentTime-atomic.LoadUint64(&ibe.lastAccessTime) > 10*60 {
+		// Delete items accessed more than a minute ago.
+		if currentTime-atomic.LoadUint64(&ibe.lastAccessTime) > 60 {
 			delete(ibc.m, k)
 		}
 	}
@@ -408,7 +393,7 @@ func (ibc *inmemoryBlockCache) Get(k inmemoryBlockCacheKey) *inmemoryBlock {
 func (ibc *inmemoryBlockCache) Put(k inmemoryBlockCacheKey, ib *inmemoryBlock) {
 	ibc.mu.Lock()

-	// Clean superflouos entries in cache.
+	// Clean superfluous entries in cache.
 	if overflow := len(ibc.m) - getMaxCachedInmemoryBlocksPerPart(); overflow > 0 {
 		// Remove 10% of items from the cache.
 		overflow = int(float64(len(ibc.m)) * 0.1)
--- a/lib/mergeset/table.go
+++ b/lib/mergeset/table.go
@@ -369,7 +369,7 @@ func (tb *Table) AddItems(items [][]byte) error {
 			tb.rawItemsBlocks = append(tb.rawItemsBlocks, ib)
 		}
 	}
-	if len(tb.rawItemsBlocks) >= 1024 {
+	if len(tb.rawItemsBlocks) >= 512 {
 		blocksToMerge = tb.rawItemsBlocks
 		tb.rawItemsBlocks = nil
 		tb.rawItemsLastFlushTime = fasttime.UnixTimestamp()
--- a/lib/mergeset/table_search_test.go
+++ b/lib/mergeset/table_search_test.go
@@ -136,7 +136,7 @@ func testTableSearchSerial(tb *Table, items []string) error {
 			n++
 		}
 		if ts.NextItem() {
-			return fmt.Errorf("superflouos item found at position %d when searching for %q: %q", n, key, ts.Item)
+			return fmt.Errorf("superfluous item found at position %d when searching for %q: %q", n, key, ts.Item)
 		}
 		if err := ts.Error(); err != nil {
 			return fmt.Errorf("unexpected error when searching for %q: %w", key, err)
--- a/lib/promauth/config.go
+++ b/lib/promauth/config.go
@@ -13,18 +13,18 @@ import (
 //
 // See https://prometheus.io/docs/prometheus/latest/configuration/configuration/#tls_config
 type TLSConfig struct {
-	CAFile             string `yaml:"ca_file"`
-	CertFile           string `yaml:"cert_file"`
-	KeyFile            string `yaml:"key_file"`
-	ServerName         string `yaml:"server_name"`
-	InsecureSkipVerify bool   `yaml:"insecure_skip_verify"`
+	CAFile             string `yaml:"ca_file,omitempty"`
+	CertFile           string `yaml:"cert_file,omitempty"`
+	KeyFile            string `yaml:"key_file,omitempty"`
+	ServerName         string `yaml:"server_name,omitempty"`
+	InsecureSkipVerify bool   `yaml:"insecure_skip_verify,omitempty"`
 }

 // BasicAuthConfig represents basic auth config.
 type BasicAuthConfig struct {
 	Username     string `yaml:"username"`
-	Password     string `yaml:"password"`
-	PasswordFile string `yaml:"password_file"`
+	Password     string `yaml:"password,omitempty"`
+	PasswordFile string `yaml:"password_file,omitempty"`
 }

 // Config is auth config.
--- a/lib/promrelabel/config.go
+++ b/lib/promrelabel/config.go
@@ -14,13 +14,13 @@ import (
 //
 // See https://prometheus.io/docs/prometheus/latest/configuration/configuration/#relabel_config
 type RelabelConfig struct {
-	SourceLabels []string `yaml:"source_labels"`
-	Separator    *string  `yaml:"separator"`
-	TargetLabel  string   `yaml:"target_label"`
-	Regex        *string  `yaml:"regex"`
-	Modulus      uint64   `yaml:"modulus"`
-	Replacement  *string  `yaml:"replacement"`
-	Action       string   `yaml:"action"`
+	SourceLabels []string `yaml:"source_labels,flow,omitempty"`
+	Separator    *string  `yaml:"separator,omitempty"`
+	TargetLabel  string   `yaml:"target_label,omitempty"`
+	Regex        *string  `yaml:"regex,omitempty"`
+	Modulus      uint64   `yaml:"modulus,omitempty"`
+	Replacement  *string  `yaml:"replacement,omitempty"`
+	Action       string   `yaml:"action,omitempty"`
 }

 // LoadRelabelConfigs loads relabel configs from the given path.
--- a/lib/promrelabel/relabel.go
+++ b/lib/promrelabel/relabel.go
@@ -336,3 +336,14 @@ func GetLabelValueByName(labels []prompbmarshal.Label, name string) string {
 	}
 	return label.Value
 }
+
+// CleanLabels sets label.Name and label.Value to an empty string for all the labels.
+//
+// This should help GC cleaning up label.Name and label.Value strings.
+func CleanLabels(labels []prompbmarshal.Label) {
+	for i := range labels {
+		label := &labels[i]
+		label.Name = ""
+		label.Value = ""
+	}
+}
--- a/lib/promscrape/client.go
+++ b/lib/promscrape/client.go
@@ -1,13 +1,19 @@
 package promscrape

 import (
+	"context"
 	"crypto/tls"
 	"flag"
 	"fmt"
+	"io"
+	"io/ioutil"
+	"net/http"
 	"strings"
 	"time"

+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/fasthttp"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -22,11 +28,19 @@ var (
 		"This may be useful when targets has no support for HTTP keep-alive connection. "+
 		"It is possible to set `disable_keepalive: true` individually per each 'scrape_config` section in '-promscrape.config' for fine grained control. "+
 		"Note that disabling HTTP keep-alive may increase load on both vmagent and scrape targets")
+	streamParse = flag.Bool("promscrape.streamParse", false, "Whether to enable stream parsing for metrics obtained from scrape targets. This may be useful "+
+		"for reducing memory usage when millions of metrics are exposed per each scrape target. "+
+		"It is posible to set `stream_parse: true` individually per each `scrape_config` section in `-promscrape.config` for fine grained control")
 )

 type client struct {
+	// hc is the default client optimized for common case of scraping targets with moderate number of metrics.
 	hc *fasthttp.HostClient

+	// sc (aka `stream client`) is used instead of hc if ScrapeWork.ParseStream is set.
+	// It may be useful for scraping targets with millions of metrics per target.
+	sc *http.Client
+
 	scrapeURL          string
 	host               string
 	requestURI         string
@@ -64,8 +78,23 @@ func newClient(sw *ScrapeWork) *client {
 		MaxResponseBodySize:          maxScrapeSize.N,
 		MaxIdempotentRequestAttempts: 1,
 	}
+	var sc *http.Client
+	if *streamParse || sw.StreamParse {
+		sc = &http.Client{
+			Transport: &http.Transport{
+				TLSClientConfig:     tlsCfg,
+				TLSHandshakeTimeout: 10 * time.Second,
+				IdleConnTimeout:     2 * sw.ScrapeInterval,
+				DisableCompression:  *disableCompression || sw.DisableCompression,
+				DisableKeepAlives:   *disableKeepAlive || sw.DisableKeepAlive,
+				DialContext:         statStdDial,
+			},
+			Timeout: sw.ScrapeTimeout,
+		}
+	}
 	return &client{
 		hc: hc,
+		sc: sc,

 		scrapeURL:          sw.ScrapeURL,
 		host:               host,
@@ -76,6 +105,43 @@ func newClient(sw *ScrapeWork) *client {
 	}
 }

+func (c *client) GetStreamReader() (*streamReader, error) {
+	deadline := time.Now().Add(c.hc.ReadTimeout)
+	ctx, cancel := context.WithDeadline(context.Background(), deadline)
+	req, err := http.NewRequestWithContext(ctx, "GET", c.scrapeURL, nil)
+	if err != nil {
+		cancel()
+		return nil, fmt.Errorf("cannot create request for %q: %w", c.scrapeURL, err)
+	}
+	// The following `Accept` header has been copied from Prometheus sources.
+	// See https://github.com/prometheus/prometheus/blob/f9d21f10ecd2a343a381044f131ea4e46381ce09/scrape/scrape.go#L532 .
+	// This is needed as a workaround for scraping stupid Java-based servers such as Spring Boot.
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/608 for details.
+	// Do not bloat the `Accept` header with OpenMetrics shit, since it looks like dead standard now.
+	req.Header.Set("Accept", "text/plain;version=0.0.4;q=1,*/*;q=0.1")
+	if c.authHeader != "" {
+		req.Header.Set("Authorization", c.authHeader)
+	}
+	resp, err := c.sc.Do(req)
+	if err != nil {
+		cancel()
+		return nil, fmt.Errorf("cannot scrape %q: %w", c.scrapeURL, err)
+	}
+	if resp.StatusCode != http.StatusOK {
+		metrics.GetOrCreateCounter(fmt.Sprintf(`vm_promscrape_scrapes_total{status_code="%d"}`, resp.StatusCode)).Inc()
+		respBody, _ := ioutil.ReadAll(resp.Body)
+		_ = resp.Body.Close()
+		cancel()
+		return nil, fmt.Errorf("unexpected status code returned when scraping %q: %d; expecting %d; response body: %q",
+			c.scrapeURL, resp.StatusCode, http.StatusOK, respBody)
+	}
+	scrapesOK.Inc()
+	return &streamReader{
+		r:      resp.Body,
+		cancel: cancel,
+	}, nil
+}
+
 func (c *client) ReadData(dst []byte) ([]byte, error) {
 	deadline := time.Now().Add(c.hc.ReadTimeout)
 	req := fasthttp.AcquireRequest()
@@ -87,7 +153,7 @@ func (c *client) ReadData(dst []byte) ([]byte, error) {
 	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/608 for details.
 	// Do not bloat the `Accept` header with OpenMetrics shit, since it looks like dead standard now.
 	req.Header.Set("Accept", "text/plain;version=0.0.4;q=1,*/*;q=0.1")
-	if !*disableCompression || c.disableCompression {
+	if !*disableCompression && !c.disableCompression {
 		req.Header.Set("Accept-Encoding", "gzip")
 	}
 	if *disableKeepAlive || c.disableKeepAlive {
@@ -104,9 +170,6 @@ func (c *client) ReadData(dst []byte) ([]byte, error) {
 		dst = resp.SwapBody(dst)
 	}
 	err := doRequestWithPossibleRetry(c.hc, req, resp, deadline)
-	if swapResponseBodies {
-		dst = resp.SwapBody(dst)
-	}
 	statusCode := resp.StatusCode()
 	if err == nil && (statusCode == fasthttp.StatusMovedPermanently || statusCode == fasthttp.StatusFound) {
 		// Allow a single redirect.
@@ -118,6 +181,9 @@ func (c *client) ReadData(dst []byte) ([]byte, error) {
 			statusCode = resp.StatusCode()
 		}
 	}
+	if swapResponseBodies {
+		dst = resp.SwapBody(dst)
+	}
 	fasthttp.ReleaseRequest(req)
 	if err != nil {
 		fasthttp.ReleaseResponse(resp)
@@ -131,17 +197,16 @@ func (c *client) ReadData(dst []byte) ([]byte, error) {
 		}
 		return dst, fmt.Errorf("error when scraping %q: %w", c.scrapeURL, err)
 	}
-	dstLen := len(dst)
 	if ce := resp.Header.Peek("Content-Encoding"); string(ce) == "gzip" {
 		var err error
-		var src []byte
 		if swapResponseBodies {
-			src = append(src, dst...)
-			dst = dst[:0]
+			zb := gunzipBufPool.Get()
+			zb.B, err = fasthttp.AppendGunzipBytes(zb.B[:0], dst)
+			dst = append(dst[:0], zb.B...)
+			gunzipBufPool.Put(zb)
 		} else {
-			src = resp.Body()
+			dst, err = fasthttp.AppendGunzipBytes(dst, resp.Body())
 		}
-		dst, err = fasthttp.AppendGunzipBytes(dst, src)
 		if err != nil {
 			fasthttp.ReleaseResponse(resp)
 			scrapesGunzipFailed.Inc()
@@ -151,16 +216,18 @@ func (c *client) ReadData(dst []byte) ([]byte, error) {
 	} else if !swapResponseBodies {
 		dst = append(dst, resp.Body()...)
 	}
+	fasthttp.ReleaseResponse(resp)
 	if statusCode != fasthttp.StatusOK {
 		metrics.GetOrCreateCounter(fmt.Sprintf(`vm_promscrape_scrapes_total{status_code="%d"}`, statusCode)).Inc()
 		return dst, fmt.Errorf("unexpected status code returned when scraping %q: %d; expecting %d; response body: %q",
-			c.scrapeURL, statusCode, fasthttp.StatusOK, dst[dstLen:])
+			c.scrapeURL, statusCode, fasthttp.StatusOK, dst)
 	}
 	scrapesOK.Inc()
-	fasthttp.ReleaseResponse(resp)
 	return dst, nil
 }

+var gunzipBufPool bytesutil.ByteBufferPool
+
 var (
 	scrapesTimedout     = metrics.NewCounter(`vm_promscrape_scrapes_timed_out_total`)
 	scrapesOK           = metrics.NewCounter(`vm_promscrape_scrapes_total{status_code="200"}`)
@@ -185,3 +252,22 @@ func doRequestWithPossibleRetry(hc *fasthttp.HostClient, req *fasthttp.Request,
 		}
 	}
 }
+
+type streamReader struct {
+	r         io.ReadCloser
+	cancel    context.CancelFunc
+	bytesRead int64
+}
+
+func (sr *streamReader) Read(p []byte) (int, error) {
+	n, err := sr.r.Read(p)
+	sr.bytesRead += int64(n)
+	return n, err
+}
+
+func (sr *streamReader) MustClose() {
+	sr.cancel()
+	if err := sr.r.Close(); err != nil {
+		logger.Errorf("cannot close reader: %s", err)
+	}
+}
--- a/lib/promscrape/config.go
+++ b/lib/promscrape/config.go
@@ -5,7 +5,6 @@ import (
 	"fmt"
 	"io/ioutil"
 	"net/url"
-	"os"
 	"path/filepath"
 	"strings"
 	"sync/atomic"
@@ -20,6 +19,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/dns"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/dockerswarm"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/ec2"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/eureka"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/gce"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/kubernetes"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/openstack"
@@ -32,6 +32,9 @@ var (
 	dryRun = flag.Bool("promscrape.config.dryRun", false, "Checks -promscrape.config file for errors and unsupported fields and then exits. "+
 		"Returns non-zero exit code on parsing errors and emits these errors to stderr. "+
 		"Pass -loggerLevel=ERROR if you don't need to see info messages in the output")
+	dropOriginalLabels = flag.Bool("promscrape.dropOriginalLabels", false, "Whether to drop original labels for scrape targets at /targets and /api/v1/targets pages. "+
+		"This may be needed for reducing memory usage when original labels for big number of scrape targets occupy big amounts of memory. "+
+		"Note that this reduces debuggability for improper per-target relabeling configs")
 )

 // Config represents essential parts from Prometheus config defined at https://prometheus.io/docs/prometheus/latest/configuration/configuration/
@@ -47,9 +50,9 @@ type Config struct {
 //
 // See https://prometheus.io/docs/prometheus/latest/configuration/configuration/
 type GlobalConfig struct {
-	ScrapeInterval time.Duration     `yaml:"scrape_interval"`
-	ScrapeTimeout  time.Duration     `yaml:"scrape_timeout"`
-	ExternalLabels map[string]string `yaml:"external_labels"`
+	ScrapeInterval time.Duration     `yaml:"scrape_interval,omitempty"`
+	ScrapeTimeout  time.Duration     `yaml:"scrape_timeout,omitempty"`
+	ExternalLabels map[string]string `yaml:"external_labels,omitempty"`
 }

 // ScrapeConfig represents essential parts for `scrape_config` section of Prometheus config.
@@ -57,33 +60,35 @@ type GlobalConfig struct {
 // See https://prometheus.io/docs/prometheus/latest/configuration/configuration/#scrape_config
 type ScrapeConfig struct {
 	JobName              string                      `yaml:"job_name"`
-	ScrapeInterval       time.Duration               `yaml:"scrape_interval"`
-	ScrapeTimeout        time.Duration               `yaml:"scrape_timeout"`
-	MetricsPath          string                      `yaml:"metrics_path"`
-	HonorLabels          bool                        `yaml:"honor_labels"`
-	HonorTimestamps      bool                        `yaml:"honor_timestamps"`
-	Scheme               string                      `yaml:"scheme"`
-	Params               map[string][]string         `yaml:"params"`
-	BasicAuth            *promauth.BasicAuthConfig   `yaml:"basic_auth"`
-	BearerToken          string                      `yaml:"bearer_token"`
-	BearerTokenFile      string                      `yaml:"bearer_token_file"`
-	TLSConfig            *promauth.TLSConfig         `yaml:"tls_config"`
-	StaticConfigs        []StaticConfig              `yaml:"static_configs"`
-	FileSDConfigs        []FileSDConfig              `yaml:"file_sd_configs"`
-	KubernetesSDConfigs  []kubernetes.SDConfig       `yaml:"kubernetes_sd_configs"`
-	OpenStackSDConfigs   []openstack.SDConfig        `yaml:"openstack_sd_configs"`
-	ConsulSDConfigs      []consul.SDConfig           `yaml:"consul_sd_configs"`
-	DockerSwarmConfigs   []dockerswarm.SDConfig      `yaml:"dockerswarm_sd_configs"`
-	DNSSDConfigs         []dns.SDConfig              `yaml:"dns_sd_configs"`
-	EC2SDConfigs         []ec2.SDConfig              `yaml:"ec2_sd_configs"`
-	GCESDConfigs         []gce.SDConfig              `yaml:"gce_sd_configs"`
-	RelabelConfigs       []promrelabel.RelabelConfig `yaml:"relabel_configs"`
-	MetricRelabelConfigs []promrelabel.RelabelConfig `yaml:"metric_relabel_configs"`
-	SampleLimit          int                         `yaml:"sample_limit"`
+	ScrapeInterval       time.Duration               `yaml:"scrape_interval,omitempty"`
+	ScrapeTimeout        time.Duration               `yaml:"scrape_timeout,omitempty"`
+	MetricsPath          string                      `yaml:"metrics_path,omitempty"`
+	HonorLabels          bool                        `yaml:"honor_labels,omitempty"`
+	HonorTimestamps      bool                        `yaml:"honor_timestamps,omitempty"`
+	Scheme               string                      `yaml:"scheme,omitempty"`
+	Params               map[string][]string         `yaml:"params,omitempty"`
+	BasicAuth            *promauth.BasicAuthConfig   `yaml:"basic_auth,omitempty"`
+	BearerToken          string                      `yaml:"bearer_token,omitempty"`
+	BearerTokenFile      string                      `yaml:"bearer_token_file,omitempty"`
+	TLSConfig            *promauth.TLSConfig         `yaml:"tls_config,omitempty"`
+	StaticConfigs        []StaticConfig              `yaml:"static_configs,omitempty"`
+	FileSDConfigs        []FileSDConfig              `yaml:"file_sd_configs,omitempty"`
+	KubernetesSDConfigs  []kubernetes.SDConfig       `yaml:"kubernetes_sd_configs,omitempty"`
+	OpenStackSDConfigs   []openstack.SDConfig        `yaml:"openstack_sd_configs,omitempty"`
+	ConsulSDConfigs      []consul.SDConfig           `yaml:"consul_sd_configs,omitempty"`
+	EurekaSDConfigs      []eureka.SDConfig           `yaml:"eureka_sd_configs,omitempty"`
+	DockerSwarmConfigs   []dockerswarm.SDConfig      `yaml:"dockerswarm_sd_configs,omitempty"`
+	DNSSDConfigs         []dns.SDConfig              `yaml:"dns_sd_configs,omitempty"`
+	EC2SDConfigs         []ec2.SDConfig              `yaml:"ec2_sd_configs,omitempty"`
+	GCESDConfigs         []gce.SDConfig              `yaml:"gce_sd_configs,omitempty"`
+	RelabelConfigs       []promrelabel.RelabelConfig `yaml:"relabel_configs,omitempty"`
+	MetricRelabelConfigs []promrelabel.RelabelConfig `yaml:"metric_relabel_configs,omitempty"`
+	SampleLimit          int                         `yaml:"sample_limit,omitempty"`

 	// These options are supported only by lib/promscrape.
-	DisableCompression bool `yaml:"disable_compression"`
-	DisableKeepAlive   bool `yaml:"disable_keepalive"`
+	DisableCompression bool `yaml:"disable_compression,omitempty"`
+	DisableKeepAlive   bool `yaml:"disable_keepalive,omitempty"`
+	StreamParse        bool `yaml:"stream_parse,omitempty"`

 	// This is set in loadConfig
 	swc *scrapeWorkConfig
@@ -102,7 +107,7 @@ type FileSDConfig struct {
 // See https://prometheus.io/docs/prometheus/latest/configuration/configuration/#static_config
 type StaticConfig struct {
 	Targets []string          `yaml:"targets"`
-	Labels  map[string]string `yaml:"labels"`
+	Labels  map[string]string `yaml:"labels,omitempty"`
 }

 func loadStaticConfigs(path string) ([]StaticConfig, error) {
@@ -128,16 +133,14 @@ func loadConfig(path string) (cfg *Config, data []byte, err error) {
 	if err := cfgObj.parse(data, path); err != nil {
 		return nil, nil, fmt.Errorf("cannot parse Prometheus config from %q: %w", path, err)
 	}
-	if *dryRun {
-		// This is a dirty hack for checking Prometheus config only.
-		// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/362
-		// and https://github.com/VictoriaMetrics/VictoriaMetrics/issues/508 for details.
-		logger.Infof("Success: the config at %q has no errors; exitting with 0 status code", path)
-		os.Exit(0)
-	}
 	return &cfgObj, data, nil
 }

+// IsDryRun returns true if -promscrape.config.dryRun command-line flag is set
+func IsDryRun() bool {
+	return *dryRun
+}
+
 func (cfg *Config) parse(data []byte, path string) error {
 	if err := unmarshalMaybeStrict(data, cfg); err != nil {
 		return fmt.Errorf("cannot unmarshal data: %w", err)
@@ -180,7 +183,7 @@ func getSWSByJob(sws []ScrapeWork) map[string][]ScrapeWork {
 // getKubernetesSDScrapeWork returns `kubernetes_sd_configs` ScrapeWork from cfg.
 func (cfg *Config) getKubernetesSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 	swsPrevByJob := getSWSByJob(prev)
-	var dst []ScrapeWork
+	dst := make([]ScrapeWork, 0, len(prev))
 	for i := range cfg.ScrapeConfigs {
 		sc := &cfg.ScrapeConfigs[i]
 		dstLen := len(dst)
@@ -208,7 +211,7 @@ func (cfg *Config) getKubernetesSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 // getOpenStackSDScrapeWork returns `openstack_sd_configs` ScrapeWork from cfg.
 func (cfg *Config) getOpenStackSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 	swsPrevByJob := getSWSByJob(prev)
-	var dst []ScrapeWork
+	dst := make([]ScrapeWork, 0, len(prev))
 	for i := range cfg.ScrapeConfigs {
 		sc := &cfg.ScrapeConfigs[i]
 		dstLen := len(dst)
@@ -236,7 +239,7 @@ func (cfg *Config) getOpenStackSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 // getDockerSwarmSDScrapeWork returns `dockerswarm_sd_configs` ScrapeWork from cfg.
 func (cfg *Config) getDockerSwarmSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 	swsPrevByJob := getSWSByJob(prev)
-	var dst []ScrapeWork
+	dst := make([]ScrapeWork, 0, len(prev))
 	for i := range cfg.ScrapeConfigs {
 		sc := &cfg.ScrapeConfigs[i]
 		dstLen := len(dst)
@@ -264,7 +267,7 @@ func (cfg *Config) getDockerSwarmSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 // getConsulSDScrapeWork returns `consul_sd_configs` ScrapeWork from cfg.
 func (cfg *Config) getConsulSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 	swsPrevByJob := getSWSByJob(prev)
-	var dst []ScrapeWork
+	dst := make([]ScrapeWork, 0, len(prev))
 	for i := range cfg.ScrapeConfigs {
 		sc := &cfg.ScrapeConfigs[i]
 		dstLen := len(dst)
@@ -289,10 +292,38 @@ func (cfg *Config) getConsulSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 	return dst
 }

+// getEurekaSDScrapeWork returns `eureka_sd_configs` ScrapeWork from cfg.
+func (cfg *Config) getEurekaSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
+	swsPrevByJob := getSWSByJob(prev)
+	dst := make([]ScrapeWork, 0, len(prev))
+	for i := range cfg.ScrapeConfigs {
+		sc := &cfg.ScrapeConfigs[i]
+		dstLen := len(dst)
+		ok := true
+		for j := range sc.EurekaSDConfigs {
+			sdc := &sc.EurekaSDConfigs[j]
+			var okLocal bool
+			dst, okLocal = appendEurekaScrapeWork(dst, sdc, cfg.baseDir, sc.swc)
+			if ok {
+				ok = okLocal
+			}
+		}
+		if ok {
+			continue
+		}
+		swsPrev := swsPrevByJob[sc.swc.jobName]
+		if len(swsPrev) > 0 {
+			logger.Errorf("there were errors when discovering eureka targets for job %q, so preserving the previous targets", sc.swc.jobName)
+			dst = append(dst[:dstLen], swsPrev...)
+		}
+	}
+	return dst
+}
+
 // getDNSSDScrapeWork returns `dns_sd_configs` ScrapeWork from cfg.
 func (cfg *Config) getDNSSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 	swsPrevByJob := getSWSByJob(prev)
-	var dst []ScrapeWork
+	dst := make([]ScrapeWork, 0, len(prev))
 	for i := range cfg.ScrapeConfigs {
 		sc := &cfg.ScrapeConfigs[i]
 		dstLen := len(dst)
@@ -320,7 +351,7 @@ func (cfg *Config) getDNSSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 // getEC2SDScrapeWork returns `ec2_sd_configs` ScrapeWork from cfg.
 func (cfg *Config) getEC2SDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 	swsPrevByJob := getSWSByJob(prev)
-	var dst []ScrapeWork
+	dst := make([]ScrapeWork, 0, len(prev))
 	for i := range cfg.ScrapeConfigs {
 		sc := &cfg.ScrapeConfigs[i]
 		dstLen := len(dst)
@@ -348,7 +379,7 @@ func (cfg *Config) getEC2SDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 // getGCESDScrapeWork returns `gce_sd_configs` ScrapeWork from cfg.
 func (cfg *Config) getGCESDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 	swsPrevByJob := getSWSByJob(prev)
-	var dst []ScrapeWork
+	dst := make([]ScrapeWork, 0, len(prev))
 	for i := range cfg.ScrapeConfigs {
 		sc := &cfg.ScrapeConfigs[i]
 		dstLen := len(dst)
@@ -386,7 +417,7 @@ func (cfg *Config) getFileSDScrapeWork(prev []ScrapeWork) []ScrapeWork {
 			swsMapPrev[filepath] = append(swsMapPrev[filepath], *sw)
 		}
 	}
-	var dst []ScrapeWork
+	dst := make([]ScrapeWork, 0, len(prev))
 	for i := range cfg.ScrapeConfigs {
 		sc := &cfg.ScrapeConfigs[i]
 		for j := range sc.FileSDConfigs {
@@ -473,6 +504,7 @@ func getScrapeWorkConfig(sc *ScrapeConfig, baseDir string, globalCfg *GlobalConf
 		sampleLimit:          sc.SampleLimit,
 		disableCompression:   sc.DisableCompression,
 		disableKeepAlive:     sc.DisableKeepAlive,
+		streamParse:          sc.StreamParse,
 	}
 	return swc, nil
 }
@@ -493,6 +525,7 @@ type scrapeWorkConfig struct {
 	sampleLimit          int
 	disableCompression   bool
 	disableKeepAlive     bool
+	streamParse          bool
 }

 func appendKubernetesScrapeWork(dst []ScrapeWork, sdc *kubernetes.SDConfig, baseDir string, swc *scrapeWorkConfig) ([]ScrapeWork, bool) {
@@ -531,6 +564,15 @@ func appendConsulScrapeWork(dst []ScrapeWork, sdc *consul.SDConfig, baseDir stri
 	return appendScrapeWorkForTargetLabels(dst, swc, targetLabels, "consul_sd_config"), true
 }

+func appendEurekaScrapeWork(dst []ScrapeWork, sdc *eureka.SDConfig, baseDir string, swc *scrapeWorkConfig) ([]ScrapeWork, bool) {
+	targetLabels, err := eureka.GetLabels(sdc, baseDir)
+	if err != nil {
+		logger.Errorf("error when discovering eureka targets for `job_name` %q: %s; skipping it", swc.jobName, err)
+		return dst, false
+	}
+	return appendScrapeWorkForTargetLabels(dst, swc, targetLabels, "eureka_sd_config"), true
+}
+
 func appendDNSScrapeWork(dst []ScrapeWork, sdc *dns.SDConfig, swc *scrapeWorkConfig) ([]ScrapeWork, bool) {
 	targetLabels, err := dns.GetLabels(sdc)
 	if err != nil {
@@ -636,12 +678,21 @@ func (stc *StaticConfig) appendScrapeWork(dst []ScrapeWork, swc *scrapeWorkConfi

 func appendScrapeWork(dst []ScrapeWork, swc *scrapeWorkConfig, target string, extraLabels, metaLabels map[string]string) ([]ScrapeWork, error) {
 	labels := mergeLabels(swc.jobName, swc.scheme, target, swc.metricsPath, extraLabels, swc.externalLabels, metaLabels, swc.params)
-	originalLabels := append([]prompbmarshal.Label{}, labels...)
-	promrelabel.SortLabels(originalLabels)
+	var originalLabels []prompbmarshal.Label
+	if !*dropOriginalLabels {
+		originalLabels = append([]prompbmarshal.Label{}, labels...)
+		promrelabel.SortLabels(originalLabels)
+	}
 	labels = promrelabel.ApplyRelabelConfigs(labels, 0, swc.relabelConfigs, false)
 	labels = promrelabel.RemoveMetaLabels(labels[:0], labels)
+	// Remove references to already deleted labels, so GC could clean strings for label name and label value past len(labels).
+	// This should reduce memory usage when relabeling creates big number of temporary labels with long names and/or values.
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/825 for details.
+	labels = append([]prompbmarshal.Label{}, labels...)
+
 	if len(labels) == 0 {
 		// Drop target without labels.
+		droppedTargetsMap.Register(originalLabels)
 		return dst, nil
 	}
 	// See https://www.robustperception.io/life-of-a-label
@@ -652,10 +703,12 @@ func appendScrapeWork(dst []ScrapeWork, swc *scrapeWorkConfig, target string, ex
 	addressRelabeled := promrelabel.GetLabelValueByName(labels, "__address__")
 	if len(addressRelabeled) == 0 {
 		// Drop target without scrape address.
+		droppedTargetsMap.Register(originalLabels)
 		return dst, nil
 	}
 	if strings.Contains(addressRelabeled, "/") {
 		// Drop target with '/'
+		droppedTargetsMap.Register(originalLabels)
 		return dst, nil
 	}
 	addressRelabeled = addMissingPort(schemeRelabeled, addressRelabeled)
@@ -663,6 +716,9 @@ func appendScrapeWork(dst []ScrapeWork, swc *scrapeWorkConfig, target string, ex
 	if metricsPathRelabeled == "" {
 		metricsPathRelabeled = "/metrics"
 	}
+	if !strings.HasPrefix(metricsPathRelabeled, "/") {
+		metricsPathRelabeled = "/" + metricsPathRelabeled
+	}
 	paramsRelabeled := getParamsFromLabels(labels, swc.params)
 	optionalQuestion := "?"
 	if len(paramsRelabeled) == 0 || strings.Contains(metricsPathRelabeled, "?") {
@@ -696,6 +752,7 @@ func appendScrapeWork(dst []ScrapeWork, swc *scrapeWorkConfig, target string, ex
 		SampleLimit:          swc.sampleLimit,
 		DisableCompression:   swc.disableCompression,
 		DisableKeepAlive:     swc.disableKeepAlive,
+		StreamParse:          swc.streamParse,

 		jobNameOriginal: swc.jobName,
 	})
--- a/lib/promscrape/config_test.go
+++ b/lib/promscrape/config_test.go
@@ -475,7 +475,7 @@ scrape_configs:
 - job_name: foo
  static_configs:
  - targets: ["xxx"]
-`, nil)
+`, []ScrapeWork{})
 	f(`
 scrape_configs:
 - job_name: foo
@@ -1276,6 +1276,7 @@ scrape_configs:
    sample_limit: 100
    disable_keepalive: true
    disable_compression: true
+    stream_parse: true
    static_configs:
      - targets:
        - 192.168.1.2  # SNMP device.
@@ -1328,9 +1329,49 @@ scrape_configs:
 			SampleLimit:        100,
 			DisableKeepAlive:   true,
 			DisableCompression: true,
+			StreamParse:        true,
 			jobNameOriginal:    "snmp",
 		},
 	})
+	f(`
+scrape_configs:
+- job_name: path wo slash
+  static_configs: 
+  - targets: ["foo.bar:1234"]
+  relabel_configs:
+  - replacement: metricspath
+    target_label: __metrics_path__
+`, []ScrapeWork{
+		{
+			ScrapeURL:      "http://foo.bar:1234/metricspath",
+			ScrapeInterval: defaultScrapeInterval,
+			ScrapeTimeout:  defaultScrapeTimeout,
+			Labels: []prompbmarshal.Label{
+				{
+					Name:  "__address__",
+					Value: "foo.bar:1234",
+				},
+				{
+					Name:  "__metrics_path__",
+					Value: "metricspath",
+				},
+				{
+					Name:  "__scheme__",
+					Value: "http",
+				},
+				{
+					Name:  "instance",
+					Value: "foo.bar:1234",
+				},
+				{
+					Name:  "job",
+					Value: "path wo slash",
+				},
+			},
+			jobNameOriginal: "path wo slash",
+			AuthConfig:      &promauth.Config{},
+		},
+	})
 }

 var defaultRegexForRelabelConfig = regexp.MustCompile("^(.*)$")
--- a/Show More
+++ b/Show More