Merge tag 'v1.144.0' into pmm-6401-read-prometheus-data-files

Release v1.144.0

.github/dependabot.yml

@@ -4,6 +4,8 @@ updates:
     directory: "/"
     schedule:
       interval: "daily"
+    cooldown:
+      default-days: 21
   - package-ecosystem: "gomod"
     directory: "/"
     schedule:
@@ -23,6 +25,8 @@ updates:
     directory: "/"
     schedule:
       interval: "daily"
+    cooldown:
+      default-days: 21
   - package-ecosystem: "npm"
     directory: "/app/vmui/packages/vmui"
     schedule:

.github/pull_request_template.md

@@ -1,10 +1,3 @@
-### Describe Your Changes
+**PLEASE REMOVE LINE BELOW BEFORE SUBMITTING**
 
-Please provide a brief description of the changes you made. Be as specific as possible to help others understand the purpose and impact of your modifications.
-
-### Checklist
-
-The following checks are **mandatory**:
-
-- [ ] My change adheres to [VictoriaMetrics contributing guidelines](https://docs.victoriametrics.com/victoriametrics/contributing/#pull-request-checklist).
-- [ ] My change adheres to [VictoriaMetrics development goals](https://docs.victoriametrics.com/victoriametrics/goals/).
+Before creating the PR, make sure you have read and followed the [VictoriaMetrics contributing guidelines](https://docs.victoriametrics.com/victoriametrics/contributing/#pull-request-checklist).

.github/workflows/build.yml

@@ -57,15 +57,17 @@ jobs:
             arch: amd64
           - os: openbsd
             arch: amd64
+          - os: netbsd
+            arch: amd64
           - os: windows
             arch: amd64
     steps:
       - name: Code checkout
-        uses: actions/checkout@v6
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
 
       - name: Setup Go
         id: go
-        uses: actions/setup-go@v6
+        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c  # v6.4.0
         with:
           cache-dependency-path: |
             go.sum

.github/workflows/changelog-linter.yml

@@ -9,7 +9,7 @@ jobs:
   tip-lint:
     runs-on: 'ubuntu-latest'
     steps:
-      - uses: 'actions/checkout@v6'
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
         with:
           # needed for proper diff
           fetch-depth: 0

.github/workflows/check-commit-signed.yml

@@ -8,7 +8,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout code
-        uses: actions/checkout@v6
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
         with:
           fetch-depth: 0  # we need full history for commit verification
 

.github/workflows/check-licenses.yml

@@ -15,11 +15,11 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@master
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
 
       - name: Setup Go
         id: go
-        uses: actions/setup-go@v6
+        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c  # v6.4.0
         with:
           go-version-file: 'go.mod'
           cache: false
@@ -27,7 +27,7 @@ jobs:
       - run: go version
 
       - name: Cache Go artifacts
-        uses: actions/cache@v4
+        uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae  # v5.0.5
         with:
           path: |
             ~/.cache/go-build

.github/workflows/codeql-analysis-go.yml

@@ -29,18 +29,18 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v6
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
 
       - name: Set up Go
         id: go
-        uses: actions/setup-go@v6
+        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c  # v6.4.0
         with:
           cache: false
           go-version-file: 'go.mod'
       - run: go version
 
       - name: Cache Go artifacts
-        uses: actions/cache@v4
+        uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae  # v5.0.5
         with:
           path: |
             ~/.cache/go-build
@@ -50,14 +50,14 @@ jobs:
           restore-keys: go-artifacts-${{ runner.os }}-codeql-analyze-${{ steps.go.outputs.go-version }}-
 
       - name: Initialize CodeQL
-        uses: github/codeql-action/init@v4
+        uses: github/codeql-action/init@e46ed2cbd01164d986452f91f178727624ae40d7  # v4.35.3
         with:
           languages: go
 
       - name: Autobuild
-        uses: github/codeql-action/autobuild@v4
+        uses: github/codeql-action/autobuild@e46ed2cbd01164d986452f91f178727624ae40d7  # v4.35.3
 
       - name: Perform CodeQL Analysis
-        uses: github/codeql-action/analyze@v4
+        uses: github/codeql-action/analyze@e46ed2cbd01164d986452f91f178727624ae40d7  # v4.35.3
         with:
           category: 'language:go'

.github/workflows/docs.yaml

@@ -16,19 +16,19 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@v6
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
         with:
           path: __vm
 
       - name: Checkout private code
-        uses: actions/checkout@v6
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
         with:
           repository: VictoriaMetrics/vmdocs
           token: ${{ secrets.VM_BOT_GH_TOKEN }}
           path: __vm-docs
 
       - name: Import GPG key
-        uses: crazy-max/ghaction-import-gpg@v7
+        uses: crazy-max/ghaction-import-gpg@2dc316deee8e90f13e1a351ab510b4d5bc0c82cd  # v7.0.0
         id: import-gpg
         with:
           gpg_private_key: ${{ secrets.VM_BOT_GPG_PRIVATE_KEY }}

.github/workflows/test.yml

@@ -32,11 +32,11 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@v6
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
 
       - name: Setup Go
         id: go
-        uses: actions/setup-go@v6
+        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c  # v6.4.0
         with:
           cache-dependency-path: |
             go.sum
@@ -47,7 +47,7 @@ jobs:
       - run: go version
 
       - name: Cache golangci-lint
-        uses: actions/cache@v4
+        uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae  # v5.0.5
         with:
           path: |
             ~/.cache/golangci-lint
@@ -66,17 +66,17 @@ jobs:
     strategy:
       matrix:
         scenario:
-          - 'test-full'
-          - 'test-full-386'
+          - 'test'
+          - 'test-386'
           - 'test-pure'
 
     steps:
       - name: Code checkout
-        uses: actions/checkout@v6
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
 
       - name: Setup Go
         id: go
-        uses: actions/setup-go@v6
+        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c  # v6.4.0
         with:
           cache-dependency-path: |
             go.sum
@@ -88,22 +88,17 @@ jobs:
       - name: Run tests
         run: make ${{ matrix.scenario}}
 
-      - name: Publish coverage
-        uses: codecov/codecov-action@v6
-        with:
-          files: ./coverage.txt
-
   apptest:
     name: apptest
     runs-on: apptest
 
     steps:
       - name: Code checkout
-        uses: actions/checkout@v6
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
 
       - name: Setup Go
         id: go
-        uses: actions/setup-go@v6
+        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c  # v6.4.0
         with:
           cache-dependency-path: |
             go.sum

.github/workflows/vmui.yml

@@ -32,11 +32,11 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@v6
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
 
       - name: Cache node_modules
         id: cache
-        uses: actions/cache@v5
+        uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae  # v5.0.5
         with:
           path: app/vmui/packages/vmui/node_modules
           key: vmui-deps-${{ runner.os }}-${{ hashFiles('app/vmui/packages/vmui/package-lock.json', 'app/vmui/Dockerfile-build') }}
@@ -69,7 +69,7 @@ jobs:
           VMUI_SKIP_INSTALL: true
 
       - name: Annotate Code Linting Results
-        uses: ataylorme/eslint-annotate-action@v3
+        uses: ataylorme/eslint-annotate-action@d57a1193d4c59cbfbf3f86c271f42612f9dbd9e9  # 3.0.0
         with:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           report-json: app/vmui/packages/vmui/vmui-lint-report.json

Makefile

@@ -457,6 +457,9 @@ test:
 test-race:
 	go test -tags 'synctest' -race ./lib/... ./app/...
 
+test-386:
+	GOARCH=386 go test -tags 'synctest' ./lib/... ./app/...
+
 test-pure:
 	CGO_ENABLED=0 go test -tags 'synctest' ./lib/... ./app/...
 
@@ -467,10 +470,10 @@ test-full-386:
 	GOARCH=386 go test -tags 'synctest' -coverprofile=coverage.txt -covermode=atomic ./lib/... ./app/...
 
 apptest:
-	$(MAKE) victoria-metrics vmagent vmalert vmauth vmctl vmbackup vmrestore
+	$(MAKE) victoria-metrics-race vmagent-race vmalert-race vmauth-race vmctl-race vmbackup-race vmrestore-race
 	go test ./apptest/... -skip="^Test(Cluster|Legacy).*"
 
-apptest-legacy: victoria-metrics vmbackup vmrestore
+apptest-legacy: victoria-metrics-race vmbackup-race vmrestore-race
 	OS=$$(uname | tr '[:upper:]' '[:lower:]'); \
 	ARCH=$$(uname -m | tr '[:upper:]' '[:lower:]' | sed 's/x86_64/amd64/'); \
 	VERSION=v1.132.0; \
@@ -532,6 +535,15 @@ remove-golangci-lint:
 govulncheck: install-govulncheck
 	govulncheck ./...
 
+govulncheck-docker:
+	docker run -w $(PWD) -v $(PWD):$(PWD) \
+		-v govulncheck-gomod-cache:/root/go/pkg/mod \
+		-v govulncheck-gobuild-cache:/root/.cache/go-build \
+		-v govulncheck-go-bin:/root/go/bin \
+		--env="GOCACHE=/root/.cache/go-build" \
+		--env="GOMODCACHE=/root/go/pkg/mod" \
+		"$(GO_BUILDER_IMAGE)" /bin/sh -c "which govulncheck || go install golang.org/x/vuln/cmd/govulncheck@latest && govulncheck ./..."
+
 install-govulncheck:
 	which govulncheck || go install golang.org/x/vuln/cmd/govulncheck@latest
 

README.md

@@ -4,7 +4,6 @@
 [![Docker Pulls](https://img.shields.io/docker/pulls/victoriametrics/victoria-metrics?label=&logo=docker&logoColor=white&labelColor=2496ED&color=2496ED&link=https%3A%2F%2Fhub.docker.com%2Fr%2Fvictoriametrics%2Fvictoria-metrics)](https://hub.docker.com/u/victoriametrics)
 [![Go Report](https://goreportcard.com/badge/github.com/VictoriaMetrics/VictoriaMetrics?link=https%3A%2F%2Fgoreportcard.com%2Freport%2Fgithub.com%2FVictoriaMetrics%2FVictoriaMetrics)](https://goreportcard.com/report/github.com/VictoriaMetrics/VictoriaMetrics)
 [![Build Status](https://github.com/VictoriaMetrics/VictoriaMetrics/actions/workflows/build.yml/badge.svg?branch=master&link=https%3A%2F%2Fgithub.com%2FVictoriaMetrics%2FVictoriaMetrics%2Factions)](https://github.com/VictoriaMetrics/VictoriaMetrics/actions/workflows/build.yml)
-[![codecov](https://codecov.io/gh/VictoriaMetrics/VictoriaMetrics/branch/master/graph/badge.svg?link=https%3A%2F%2Fcodecov.io%2Fgh%2FVictoriaMetrics%2FVictoriaMetrics)](https://app.codecov.io/gh/VictoriaMetrics/VictoriaMetrics)
 [![License](https://img.shields.io/github/license/VictoriaMetrics/VictoriaMetrics?labelColor=green&label=&link=https%3A%2F%2Fgithub.com%2FVictoriaMetrics%2FVictoriaMetrics%2Fblob%2Fmaster%2FLICENSE)](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/LICENSE)
 [![Join Slack](https://img.shields.io/badge/Join%20Slack-4A154B?logo=slack)](https://slack.victoriametrics.com)
 [![X](https://img.shields.io/twitter/follow/VictoriaMetrics?style=flat&label=Follow&color=black&logo=x&labelColor=black&link=https%3A%2F%2Fx.com%2FVictoriaMetrics)](https://x.com/VictoriaMetrics/)

SECURITY.md

@@ -1,42 +1,4 @@
 # Security Policy
 
-## Supported Versions
+You can find out about our security policy and VictoriaMetrics version support on the [security page](https://docs.victoriametrics.com/victoriametrics/#security) in the documentation.
 
-The following versions of VictoriaMetrics receive regular security fixes:
-
-| Version                                                                        | Supported          |
-|--------------------------------------------------------------------------------|--------------------|
-| [Latest release](https://docs.victoriametrics.com/victoriametrics/changelog/)  | :white_check_mark: |
-| [LTS releases](https://docs.victoriametrics.com/victoriametrics/lts-releases/) | :white_check_mark: |
-| other releases                                                                 | :x:                |
-
-See [this page](https://victoriametrics.com/security/) for more details.
-
-## Software Bill of Materials (SBOM)
-
-Every VictoriaMetrics container{{% available_from "#" %}} image published to
-[Docker Hub](https://hub.docker.com/u/victoriametrics)
-and [Quay.io](https://quay.io/organization/victoriametrics)
-includes an [SPDX](https://spdx.dev/) SBOM attestation
-generated automatically by BuildKit during
-`docker buildx build`.
-
-To inspect the SBOM for an image:
-
-```sh
-docker buildx imagetools inspect \
-  docker.io/victoriametrics/victoria-metrics:latest \
-  --format "{{ json .SBOM }}"
-```
-
-To scan an image using its SBOM attestation with
-[Trivy](https://github.com/aquasecurity/trivy):
-
-```sh
-trivy image --sbom-sources oci \
-  docker.io/victoriametrics/victoria-metrics:latest
-```
-
-## Reporting a Vulnerability
-
-Please report any security issues to <security@victoriametrics.com>

app/victoria-metrics/main.go

@@ -39,6 +39,8 @@ var (
 		"The saved data survives unclean shutdowns such as OOM crash, hardware reset, SIGKILL, etc. "+
 		"Bigger intervals may help increase the lifetime of flash storage with limited write cycles (e.g. Raspberry PI). "+
 		"Smaller intervals increase disk IO load. Minimum supported value is 1s")
+	downsamplingPeriods = flagutil.NewArrayString("downsampling.period", "Comma-separated downsampling periods in the format 'offset:period'. For example, '30d:10m' instructs "+
+		"to leave a single sample per 10 minutes for samples older than 30 days. See https://docs.victoriametrics.com/#downsampling for details")
 	maxIngestionRate = flag.Int("maxIngestionRate", 0, "The maximum number of samples vmsingle can receive per second. Data ingestion is paused when the limit is exceeded. "+
 		"By default there are no limits on samples ingestion rate.")
 	finalDedupScheduleInterval = flag.Duration("storage.finalDedupScheduleCheckInterval", time.Hour, "The interval for checking when final deduplication process should be started."+
@@ -47,6 +49,12 @@ var (
 		" See also https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#deduplication")
 )
 
+// custom api help links [["/api","doc"]] without http.pathPrefix.
+var customAPIPathList = [][]string{
+	{"/graph/explore", "explore metrics grafana page"},
+	{"/graph/d/prometheus-advanced/advanced-data-exploration", "PMM grafana dashboard"},
+}
+
 func main() {
 	// VictoriaMetrics is optimized for reduced memory allocations,
 	// so it can run with the reduced GOGC in order to reduce the used memory,
@@ -87,7 +95,10 @@ func main() {
 	}
 	logger.Infof("starting VictoriaMetrics at %q...", listenAddrs)
 	startTime := time.Now()
-	storage.SetDedupInterval(*minScrapeInterval)
+	err := storage.SetDownsamplingPeriods(*downsamplingPeriods, *minScrapeInterval)
+	if err != nil {
+		logger.Fatalf("cannot parse -downsampling.period: %s", err)
+	}
 	storage.SetDataFlushInterval(*inmemoryDataFlushInterval)
 	if *finalDedupScheduleInterval < time.Hour {
 		logger.Fatalf("-dedup.finalDedupScheduleCheckInterval cannot be smaller than 1 hour; got %s", *finalDedupScheduleInterval)
@@ -118,8 +129,8 @@ func main() {
 		logger.Fatalf("cannot stop the webservice: %s", err)
 	}
 	logger.Infof("successfully shut down the webservice in %.3f seconds", time.Since(startTime).Seconds())
-	vminsert.Stop()
 	vminsertcommon.StopIngestionRateLimiter()
+	vminsert.Stop()
 
 	vmstorage.Stop()
 	vmselect.Stop()
@@ -152,6 +163,10 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 			{"api/v1/status/active_queries", "active queries"},
 			{"-/reload", "reload configuration"},
 		})
+		for _, p := range customAPIPathList {
+			p, doc := p[0], p[1]
+			fmt.Fprintf(w, "<a href=%q>%s</a> - %s<br/>", p, p, doc)
+		}
 		return true
 	}
 	if vminsert.RequestHandler(w, r) {

app/vmagent/main.go

@@ -83,6 +83,9 @@ var (
 	maxLabelsPerTimeseries = flag.Int("maxLabelsPerTimeseries", 0, "The maximum number of labels per time series to be accepted. Series with superfluous labels are ignored. In this case the vm_rows_ignored_total{reason=\"too_many_labels\"} metric at /metrics page is incremented")
 	maxLabelNameLen        = flag.Int("maxLabelNameLen", 0, "The maximum length of label names in the accepted time series. Series with longer label name are ignored. In this case the vm_rows_ignored_total{reason=\"too_long_label_name\"} metric at /metrics page is incremented")
 	maxLabelValueLen       = flag.Int("maxLabelValueLen", 0, "The maximum length of label values in the accepted time series. Series with longer label value are ignored. In this case the vm_rows_ignored_total{reason=\"too_long_label_value\"} metric at /metrics page is incremented")
+
+	enableMultitenancyViaHeaders = flag.Bool("enableMultitenancyViaHeaders", false, "Enables multitenancy via HTTP headers. "+
+		"See https://docs.victoriametrics.com/victoriametrics/vmagent/#multitenancy")
 )
 
 var (
@@ -216,7 +219,7 @@ func getOpenTSDBHTTPInsertHandler() func(req *http.Request) error {
 	}
 	return func(req *http.Request) error {
 		path := strings.ReplaceAll(req.URL.Path, "//", "/")
-		at, err := getAuthTokenFromPath(path)
+		at, err := getAuthTokenFromPath(path, req.Header)
 		if err != nil {
 			return fmt.Errorf("cannot obtain auth token from path %q: %w", path, err)
 		}
@@ -224,8 +227,15 @@ func getOpenTSDBHTTPInsertHandler() func(req *http.Request) error {
 	}
 }
 
-func getAuthTokenFromPath(path string) (*auth.Token, error) {
-	p, err := httpserver.ParsePath(path)
+func parsePath(path string, header http.Header) (*httpserver.Path, error) {
+	if *enableMultitenancyViaHeaders {
+		return httpserver.ParsePathAndHeaders(path, header)
+	}
+	return httpserver.ParsePath(path)
+}
+
+func getAuthTokenFromPath(path string, header http.Header) (*auth.Token, error) {
+	p, err := parsePath(path, header)
 	if err != nil {
 		return nil, fmt.Errorf("cannot parse multitenant path: %w", err)
 	}
@@ -559,14 +569,15 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 }
 
 func processMultitenantRequest(w http.ResponseWriter, r *http.Request, path string) bool {
-	p, err := httpserver.ParsePath(path)
+	p, err := parsePath(path, r.Header)
 	if err != nil {
 		// Cannot parse multitenant path. Skip it - probably it will be parsed later.
 		return false
 	}
 	if p.Prefix != "insert" {
-		httpserver.Errorf(w, r, `unsupported multitenant prefix: %q; expected "insert"`, p.Prefix)
-		return true
+		// processMultitenantRequest is called for all unmatched path variants,
+		// but we should try parsing only /insert prefixed to avoid catching all possible paths.
+		return false
 	}
 	at, err := auth.NewTokenPossibleMultitenant(p.AuthToken)
 	if err != nil {

app/vmagent/opentelemetry/request_handler.go

@@ -77,16 +77,6 @@ func insertRows(at *auth.Token, tss []prompb.TimeSeries, mms []prompb.MetricMeta
 
 	var metadataTotal int
 	if prommetadata.IsEnabled() {
-		var accountID, projectID uint32
-		if at != nil {
-			accountID = at.AccountID
-			projectID = at.ProjectID
-			for i := range mms {
-				mm := &mms[i]
-				mm.AccountID = accountID
-				mm.ProjectID = projectID
-			}
-		}
 		ctx.WriteRequest.Metadata = mms
 		metadataTotal = len(mms)
 	}

app/vmagent/prometheusimport/request_handler.go

@@ -75,11 +75,6 @@ func insertRows(at *auth.Token, rows []prometheus.Row, mms []prometheus.Metadata
 			Samples: samples[len(samples)-1:],
 		})
 	}
-	var accountID, projectID uint32
-	if at != nil {
-		accountID = at.AccountID
-		projectID = at.ProjectID
-	}
 	for i := range mms {
 		mm := &mms[i]
 		mmsDst = append(mmsDst, prompb.MetricMetadata{
@@ -88,8 +83,6 @@ func insertRows(at *auth.Token, rows []prometheus.Row, mms []prometheus.Metadata
 			Type:             mm.Type,
 			// there is no unit in Prometheus exposition formats
 
-			AccountID: accountID,
-			ProjectID: projectID,
 		})
 	}
 	ctx.WriteRequest.Timeseries = tssDst

app/vmagent/promremotewrite/request_handler.go

@@ -72,11 +72,6 @@ func insertRows(at *auth.Token, timeseries []prompb.TimeSeries, mms []prompb.Met
 
 	var metadataTotal int
 	if prommetadata.IsEnabled() {
-		var accountID, projectID uint32
-		if at != nil {
-			accountID = at.AccountID
-			projectID = at.ProjectID
-		}
 		for i := range mms {
 			mm := &mms[i]
 			mmsDst = append(mmsDst, prompb.MetricMetadata{
@@ -85,8 +80,8 @@ func insertRows(at *auth.Token, timeseries []prompb.TimeSeries, mms []prompb.Met
 				Type:             mm.Type,
 				Unit:             mm.Unit,
 
-				AccountID: accountID,
-				ProjectID: projectID,
+				AccountID: mm.AccountID,
+				ProjectID: mm.ProjectID,
 			})
 		}
 		ctx.WriteRequest.Metadata = mmsDst

app/vmagent/remotewrite/client.go

@@ -2,6 +2,7 @@ package remotewrite
 
 import (
 	"bytes"
+	"context"
 	"errors"
 	"fmt"
 	"io"
@@ -13,6 +14,9 @@ import (
 	"sync/atomic"
 	"time"
 
+	"github.com/VictoriaMetrics/metrics"
+	"github.com/golang/snappy"
+
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/awsapi"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
@@ -21,10 +25,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/persistentqueue"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/ratelimiter"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timerpool"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeutil"
-	"github.com/VictoriaMetrics/metrics"
-	"github.com/golang/snappy"
 )
 
 var (
@@ -59,6 +60,8 @@ var (
 		"Multiple headers must be delimited by '^^': -remoteWrite.headers='header1:value1^^header2:value2'")
 
 	basicAuthUsername     = flagutil.NewArrayString("remoteWrite.basicAuth.username", "Optional basic auth username to use for the corresponding -remoteWrite.url")
+	basicAuthUsernameFile = flagutil.NewArrayString("remoteWrite.basicAuth.usernameFile", "Optional path to basic auth username to use for the corresponding -remoteWrite.url. "+
+		"The file is re-read every second")
 	basicAuthPassword     = flagutil.NewArrayString("remoteWrite.basicAuth.password", "Optional basic auth password to use for the corresponding -remoteWrite.url")
 	basicAuthPasswordFile = flagutil.NewArrayString("remoteWrite.basicAuth.passwordFile", "Optional path to basic auth password to use for the corresponding -remoteWrite.url. "+
 		"The file is re-read every second")
@@ -223,12 +226,14 @@ func getAuthConfig(argIdx int) (*promauth.Config, error) {
 		hdrs = strings.Split(headersValue, "^^")
 	}
 	username := basicAuthUsername.GetOptionalArg(argIdx)
+	usernameFile := basicAuthUsernameFile.GetOptionalArg(argIdx)
 	password := basicAuthPassword.GetOptionalArg(argIdx)
 	passwordFile := basicAuthPasswordFile.GetOptionalArg(argIdx)
 	var basicAuthCfg *promauth.BasicAuthConfig
-	if username != "" || password != "" || passwordFile != "" {
+	if username != "" || usernameFile != "" || password != "" || passwordFile != "" {
 		basicAuthCfg = &promauth.BasicAuthConfig{
 			Username:     username,
+			UsernameFile: usernameFile,
 			Password:     promauth.NewSecret(password),
 			PasswordFile: passwordFile,
 		}
@@ -290,7 +295,7 @@ func getAWSAPIConfig(argIdx int) (*awsapi.Config, error) {
 	accessKey := awsAccessKey.GetOptionalArg(argIdx)
 	secretKey := awsSecretKey.GetOptionalArg(argIdx)
 	service := awsService.GetOptionalArg(argIdx)
-	cfg, err := awsapi.NewConfig(ec2Endpoint, stsEndpoint, region, roleARN, accessKey, secretKey, service)
+	cfg, err := awsapi.NewConfig(ec2Endpoint, stsEndpoint, region, roleARN, accessKey, secretKey, service, "")
 	if err != nil {
 		return nil, err
 	}
@@ -306,11 +311,6 @@ func (c *client) runWorker() {
 		if !ok {
 			return
 		}
-		if len(block) == 0 {
-			// skip empty data blocks from sending
-			// see https://github.com/VictoriaMetrics/VictoriaMetrics/pull/6241
-			continue
-		}
 		go func() {
 			startTime := time.Now()
 			ch <- c.sendBlock(block)
@@ -326,15 +326,20 @@ func (c *client) runWorker() {
 			c.fq.MustWriteBlockIgnoreDisabledPQ(block)
 			return
 		case <-c.stopCh:
-			// c must be stopped. Wait for a while in the hope the block will be sent.
-			graceDuration := 5 * time.Second
+			// c must be stopped. Wait up to 5 seconds for the in-flight request to complete.
+			// If it succeeds, drain the remaining in-memory queue before returning.
+			stopCtx, cancel := context.WithTimeout(context.Background(), time.Second*5)
+			defer cancel()
+
 			select {
 			case ok := <-ch:
 				if !ok {
 					// Return unsent block to the queue.
 					c.fq.MustWriteBlockIgnoreDisabledPQ(block)
+				} else {
+					c.drainInMemoryQueue(stopCtx, block[:0])
 				}
-			case <-time.After(graceDuration):
+			case <-stopCtx.Done():
 				// Return unsent block to the queue.
 				c.fq.MustWriteBlockIgnoreDisabledPQ(block)
 			}
@@ -405,8 +410,7 @@ func (c *client) newRequest(url string, body []byte) (*http.Request, error) {
 // Otherwise, it tries sending the block to remote storage indefinitely.
 func (c *client) sendBlockHTTP(block []byte) bool {
 	c.rl.Register(len(block))
-	maxRetryDuration := timeutil.AddJitterToDuration(c.retryMaxInterval)
-	retryDuration := timeutil.AddJitterToDuration(c.retryMinInterval)
+	bt := timeutil.NewBackoffTimer(c.retryMinInterval, c.retryMaxInterval)
 	retriesCount := 0
 
 again:
@@ -415,19 +419,10 @@ again:
 	c.requestDuration.UpdateDuration(startTime)
 	if err != nil {
 		c.errorsCount.Inc()
-		retryDuration *= 2
-		if retryDuration > maxRetryDuration {
-			retryDuration = maxRetryDuration
-		}
-		remoteWriteRetryLogger.Warnf("couldn't send a block with size %d bytes to %q: %s; re-sending the block in %.3f seconds",
-			len(block), c.sanitizedURL, err, retryDuration.Seconds())
-		t := timerpool.Get(retryDuration)
-		select {
-		case <-c.stopCh:
-			timerpool.Put(t)
+		remoteWriteRetryLogger.Warnf("couldn't send a block with size %d bytes to %q: %s; re-sending the block in %s",
+			len(block), c.sanitizedURL, err, bt.CurrentDelay())
+		if !bt.Wait(c.stopCh) {
 			return false
-		case <-t.C:
-			timerpool.Put(t)
 		}
 		c.retriesCount.Inc()
 		goto again
@@ -476,7 +471,7 @@ again:
 				goto again
 			}
 
-			logger.Warnf("failed to repack zstd block (%s bytes) to snappy: %s; The block will be rejected. "+
+			logger.Warnf("failed to repack zstd block (%d bytes) to snappy: %s; The block will be rejected. "+
 				"Possible cause: ungraceful shutdown leading to persisted queue corruption.",
 				zstdBlockLen, err)
 		}
@@ -493,7 +488,10 @@ again:
 	// Unexpected status code returned
 	retriesCount++
 	retryAfterHeader := parseRetryAfterHeader(resp.Header.Get("Retry-After"))
-	retryDuration = getRetryDuration(retryAfterHeader, retryDuration, maxRetryDuration)
+	// retryAfterDuration has the highest priority duration
+	if retryAfterHeader > 0 {
+		bt.SetDelay(retryAfterHeader)
+	}
 
 	// Handle response
 	body, err := io.ReadAll(resp.Body)
@@ -502,44 +500,44 @@ again:
 		logger.Errorf("cannot read response body from %q during retry #%d: %s", c.sanitizedURL, retriesCount, err)
 	} else {
 		logger.Errorf("unexpected status code received after sending a block with size %d bytes to %q during retry #%d: %d; response body=%q; "+
-			"re-sending the block in %.3f seconds", len(block), c.sanitizedURL, retriesCount, statusCode, body, retryDuration.Seconds())
+			"re-sending the block in %s", len(block), c.sanitizedURL, retriesCount, statusCode, body, bt.CurrentDelay())
 	}
-	t := timerpool.Get(retryDuration)
-	select {
-	case <-c.stopCh:
-		timerpool.Put(t)
+	if !bt.Wait(c.stopCh) {
 		return false
-	case <-t.C:
-		timerpool.Put(t)
 	}
 	c.retriesCount.Inc()
 	goto again
 }
 
+func (c *client) drainInMemoryQueue(stopCtx context.Context, block []byte) {
+	var ok bool
+	for {
+		select {
+		case <-stopCtx.Done():
+			return
+		default:
+		}
+
+		block, ok = c.fq.MustReadInMemoryBlock(block[:0])
+		if !ok {
+			// The in memory queue has already been drained,
+			// or persisted queue is being used.
+			// In this case it is guaranteed that fq will be empty
+			return
+		}
+
+		// at this stage c.stopCh should be closed
+		// so sendBlock function should not perform retries
+		if ok := c.sendBlock(block); !ok {
+			c.fq.MustWriteBlockIgnoreDisabledPQ(block)
+			return
+		}
+	}
+}
+
 var remoteWriteRejectedLogger = logger.WithThrottler("remoteWriteRejected", 5*time.Second)
 var remoteWriteRetryLogger = logger.WithThrottler("remoteWriteRetry", 5*time.Second)
 
-// getRetryDuration returns retry duration.
-// retryAfterDuration has the highest priority.
-// If retryAfterDuration is not specified, retryDuration gets doubled.
-// retryDuration can't exceed maxRetryDuration.
-//
-// Also see: https://github.com/VictoriaMetrics/VictoriaMetrics/issues/6097
-func getRetryDuration(retryAfterDuration, retryDuration, maxRetryDuration time.Duration) time.Duration {
-	// retryAfterDuration has the highest priority duration
-	if retryAfterDuration > 0 {
-		return timeutil.AddJitterToDuration(retryAfterDuration)
-	}
-
-	// default backoff retry policy
-	retryDuration *= 2
-	if retryDuration > maxRetryDuration {
-		retryDuration = maxRetryDuration
-	}
-
-	return retryDuration
-}
-
 // repackBlockFromZstdToSnappy repacks the given zstd-compressed block to snappy-compressed block.
 //
 // The input block may be corrupted, for example, if vmagent was shut down ungracefully and
@@ -570,24 +568,20 @@ func logBlockRejected(block []byte, sanitizedURL string, resp *http.Response) {
 }
 
 // parseRetryAfterHeader parses `Retry-After` value retrieved from HTTP response header.
-// retryAfterString should be in either HTTP-date or a number of seconds.
-// It will return time.Duration(0) if `retryAfterString` does not follow RFC 7231.
-func parseRetryAfterHeader(retryAfterString string) (retryAfterDuration time.Duration) {
-	if retryAfterString == "" {
-		return retryAfterDuration
+//
+// s should be in either HTTP-date or a number of seconds.
+// It returns time.Duration(0) if s does not follow RFC 7231.
+func parseRetryAfterHeader(s string) time.Duration {
+	if s == "" {
+		return 0
 	}
 
-	defer func() {
-		v := retryAfterDuration.Seconds()
-		logger.Infof("'Retry-After: %s' parsed into %.2f second(s)", retryAfterString, v)
-	}()
-
 	// Retry-After could be in "Mon, 02 Jan 2006 15:04:05 GMT" format.
-	if parsedTime, err := time.Parse(http.TimeFormat, retryAfterString); err == nil {
+	if parsedTime, err := time.Parse(http.TimeFormat, s); err == nil {
 		return time.Duration(time.Until(parsedTime).Seconds()) * time.Second
 	}
 	// Retry-After could be in seconds.
-	if seconds, err := strconv.Atoi(retryAfterString); err == nil {
+	if seconds, err := strconv.Atoi(s); err == nil {
 		return time.Duration(seconds) * time.Second
 	}
 

app/vmagent/remotewrite/client_test.go

@@ -6,66 +6,12 @@ import (
 	"testing"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
 	"github.com/golang/snappy"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 )
 
-func TestCalculateRetryDuration(t *testing.T) {
-	// `testFunc` call `calculateRetryDuration` for `n` times
-	// and evaluate if the result of `calculateRetryDuration` is
-	// 1. >= expectMinDuration
-	// 2. <= expectMinDuration + 10% (see timeutil.AddJitterToDuration)
-	f := func(retryAfterDuration, retryDuration time.Duration, n int, expectMinDuration time.Duration) {
-		t.Helper()
-
-		for range n {
-			retryDuration = getRetryDuration(retryAfterDuration, retryDuration, time.Minute)
-		}
-
-		expectMaxDuration := helper(expectMinDuration)
-		expectMinDuration = expectMinDuration - (1000 * time.Millisecond) // Avoid edge case when calculating time.Until(now)
-
-		if retryDuration < expectMinDuration || retryDuration > expectMaxDuration {
-			t.Fatalf(
-				"incorrect retry duration, want (ms): [%d, %d], got (ms): %d",
-				expectMinDuration.Milliseconds(), expectMaxDuration.Milliseconds(),
-				retryDuration.Milliseconds(),
-			)
-		}
-	}
-
-	// Call calculateRetryDuration for 1 time.
-	{
-		// default backoff policy
-		f(0, time.Second, 1, 2*time.Second)
-		// default backoff policy exceed max limit"
-		f(0, 10*time.Minute, 1, time.Minute)
-
-		// retry after > default backoff policy
-		f(10*time.Second, 1*time.Second, 1, 10*time.Second)
-		// retry after < default backoff policy
-		f(1*time.Second, 10*time.Second, 1, 1*time.Second)
-		// retry after invalid and < default backoff policy
-		f(0, time.Second, 1, 2*time.Second)
-
-	}
-
-	// Call calculateRetryDuration for multiple times.
-	{
-		// default backoff policy 2 times
-		f(0, time.Second, 2, 4*time.Second)
-		// default backoff policy 3 times
-		f(0, time.Second, 3, 8*time.Second)
-		// default backoff policy N times exceed max limit
-		f(0, time.Second, 10, time.Minute)
-
-		// retry after 120s 1 times
-		f(120*time.Second, time.Second, 1, 120*time.Second)
-		// retry after 120s 2 times
-		f(120*time.Second, time.Second, 2, 120*time.Second)
-	}
-}
-
 func TestParseRetryAfterHeader(t *testing.T) {
 	f := func(retryAfterString string, expectResult time.Duration) {
 		t.Helper()
@@ -91,11 +37,38 @@ func TestParseRetryAfterHeader(t *testing.T) {
 	f(time.Now().Add(10*time.Second).Format("Mon, 02 Jan 2006 15:04:05 FAKETZ"), 0)
 }
 
-// helper calculate the max possible time duration calculated by timeutil.AddJitterToDuration.
-func helper(d time.Duration) time.Duration {
-	dv := min(d/10, 10*time.Second)
+func TestInitSecretFlags(t *testing.T) {
+	showRemoteWriteURLOrig := *showRemoteWriteURL
+	defer func() {
+		*showRemoteWriteURL = showRemoteWriteURLOrig
+		flagutil.UnregisterAllSecretFlags()
+	}()
 
-	return d + dv
+	flagutil.UnregisterAllSecretFlags()
+	*showRemoteWriteURL = false
+	InitSecretFlags()
+	if !flagutil.IsSecretFlag("remotewrite.url") {
+		t.Fatalf("expecting remoteWrite.url to be secret")
+	}
+	if !flagutil.IsSecretFlag("remotewrite.headers") {
+		t.Fatalf("expecting remoteWrite.headers to be secret")
+	}
+	if !flagutil.IsSecretFlag("remotewrite.proxyurl") {
+		t.Fatalf("expecting remoteWrite.proxyURL to be secret")
+	}
+
+	flagutil.UnregisterAllSecretFlags()
+	*showRemoteWriteURL = true
+	InitSecretFlags()
+	if flagutil.IsSecretFlag("remotewrite.url") {
+		t.Fatalf("remoteWrite.url must remain visible when -remoteWrite.showURL is set")
+	}
+	if !flagutil.IsSecretFlag("remotewrite.headers") {
+		t.Fatalf("expecting remoteWrite.headers to remain secret")
+	}
+	if !flagutil.IsSecretFlag("remotewrite.proxyurl") {
+		t.Fatalf("expecting remoteWrite.proxyURL to remain secret")
+	}
 }
 
 func TestRepackBlockFromZstdToSnappy(t *testing.T) {

app/vmagent/remotewrite/pendingseries.go

@@ -211,6 +211,9 @@ func (wr *writeRequest) copyMetadata(dst, src *prompb.MetricMetadata) {
 	dst.Type = src.Type
 	dst.Unit = src.Unit
 
+	dst.AccountID = src.AccountID
+	dst.ProjectID = src.ProjectID
+
 	// Pre-allocate memory for all string fields.
 	neededBufLen := len(src.MetricFamilyName) + len(src.Help)
 	bufLen := len(wr.metadatabuf)

app/vmagent/remotewrite/remotewrite.go

@@ -3,6 +3,7 @@ package remotewrite
 import (
 	"flag"
 	"fmt"
+	"math"
 	"net/http"
 	"net/url"
 	"path/filepath"
@@ -11,6 +12,10 @@ import (
 	"sync/atomic"
 	"time"
 
+	"github.com/cespare/xxhash/v2"
+
+	"github.com/VictoriaMetrics/metrics"
+
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bloomfilter"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
@@ -23,6 +28,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/memory"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/persistentqueue"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/procutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prommetadata"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
@@ -30,8 +36,6 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/slicesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/streamaggr"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeserieslimits"
-	"github.com/VictoriaMetrics/metrics"
-	"github.com/cespare/xxhash/v2"
 )
 
 var (
@@ -80,10 +84,14 @@ var (
 		`This may be needed for reducing memory usage at remote storage when the order of labels in incoming samples is random. `+
 		`For example, if m{k1="v1",k2="v2"} may be sent as m{k2="v2",k1="v1"}`+
 		`Enabled sorting for labels can slow down ingestion performance a bit`)
-	maxHourlySeries = flag.Int("remoteWrite.maxHourlySeries", 0, "The maximum number of unique series vmagent can send to remote storage systems during the last hour. "+
-		"Excess series are logged and dropped. This can be useful for limiting series cardinality. See https://docs.victoriametrics.com/victoriametrics/vmagent/#cardinality-limiter")
-	maxDailySeries = flag.Int("remoteWrite.maxDailySeries", 0, "The maximum number of unique series vmagent can send to remote storage systems during the last 24 hours. "+
-		"Excess series are logged and dropped. This can be useful for limiting series churn rate. See https://docs.victoriametrics.com/victoriametrics/vmagent/#cardinality-limiter")
+	maxHourlySeries = flag.Int64("remoteWrite.maxHourlySeries", 0, "The maximum number of unique series vmagent can send to remote storage systems during the last hour. "+
+		"Excess series are logged and dropped. This can be useful for limiting series cardinality. "+
+		fmt.Sprintf("Setting this flag to '-1' sets limit to maximum possible value (%d) which is useful in order to enable series tracking without enforcing limits. ", math.MaxInt32)+
+		"See https://docs.victoriametrics.com/victoriametrics/vmagent/#cardinality-limiter")
+	maxDailySeries = flag.Int64("remoteWrite.maxDailySeries", 0, "The maximum number of unique series vmagent can send to remote storage systems during the last 24 hours. "+
+		"Excess series are logged and dropped. This can be useful for limiting series churn rate. "+
+		fmt.Sprintf("Setting this flag to '-1' sets limit to maximum possible value (%d) which is useful in order to enable series tracking without enforcing limits. ", math.MaxInt32)+
+		"See https://docs.victoriametrics.com/victoriametrics/vmagent/#cardinality-limiter")
 	maxIngestionRate = flag.Int("maxIngestionRate", 0, "The maximum number of samples vmagent can receive per second. Data ingestion is paused when the limit is exceeded. "+
 		"By default there are no limits on samples ingestion rate. See also -remoteWrite.rateLimit")
 
@@ -92,6 +100,8 @@ var (
 		"See https://docs.victoriametrics.com/victoriametrics/vmagent/#disabling-on-disk-persistence . See also -remoteWrite.dropSamplesOnOverload")
 	dropSamplesOnOverload = flag.Bool("remoteWrite.dropSamplesOnOverload", false, "Whether to drop samples when -remoteWrite.disableOnDiskQueue is set and if the samples "+
 		"cannot be pushed into the configured -remoteWrite.url systems in a timely manner. See https://docs.victoriametrics.com/victoriametrics/vmagent/#disabling-on-disk-persistence")
+	disableMetadataPerURL = flagutil.NewArrayBool("remoteWrite.disableMetadata", "Whether to disable sending metadata to the corresponding -remoteWrite.url. "+
+		"By default, metadata sending is controlled by the global -enableMetadata flag")
 )
 
 var (
@@ -141,6 +151,10 @@ func InitSecretFlags() {
 		// remoteWrite.url can contain authentication codes, so hide it at `/metrics` output.
 		flagutil.RegisterSecretFlag("remoteWrite.url")
 	}
+	// remoteWrite.proxyURL can contain authentication codes.
+	flagutil.RegisterSecretFlag("remoteWrite.proxyURL")
+	// remoteWrite.headers can contain auth headers such as Authorization and API keys.
+	flagutil.RegisterSecretFlag("remoteWrite.headers")
 }
 
 var (
@@ -157,8 +171,20 @@ func Init() {
 	if len(*remoteWriteURLs) == 0 {
 		logger.Fatalf("at least one `-remoteWrite.url` command-line flag must be set")
 	}
-	if *maxHourlySeries > 0 {
-		hourlySeriesLimiter = bloomfilter.NewLimiter(*maxHourlySeries, time.Hour)
+	if *shardByURL && len(*disableOnDiskQueue) > 1 {
+		disableOnDiskQueues := *disableOnDiskQueue
+
+		firstValue := disableOnDiskQueues[0]
+		for _, v := range disableOnDiskQueues[1:] {
+			if firstValue != v {
+				logger.Fatalf("all -remoteWrite.url targets must have the same -remoteWrite.disableOnDiskQueue setting when -remoteWrite.shardByURL is enabled; " +
+					"either enable or disable -remoteWrite.disableOnDiskQueue for all targets")
+			}
+		}
+	}
+
+	if limit := getMaxHourlySeries(); limit > 0 {
+		hourlySeriesLimiter = bloomfilter.NewLimiter(limit, time.Hour)
 		_ = metrics.NewGauge(`vmagent_hourly_series_limit_max_series`, func() float64 {
 			return float64(hourlySeriesLimiter.MaxItems())
 		})
@@ -166,8 +192,8 @@ func Init() {
 			return float64(hourlySeriesLimiter.CurrentItems())
 		})
 	}
-	if *maxDailySeries > 0 {
-		dailySeriesLimiter = bloomfilter.NewLimiter(*maxDailySeries, 24*time.Hour)
+	if limit := getMaxDailySeries(); limit > 0 {
+		dailySeriesLimiter = bloomfilter.NewLimiter(limit, 24*time.Hour)
 		_ = metrics.NewGauge(`vmagent_daily_series_limit_max_series`, func() float64 {
 			return float64(dailySeriesLimiter.MaxItems())
 		})
@@ -275,6 +301,7 @@ func initRemoteWriteCtxs(urls []string) {
 		rwctxs[i] = newRemoteWriteCtx(i, remoteWriteURL, sanitizedURL)
 		rwctxIdx[i] = i
 	}
+	fs.RegisterPathFsMetrics(*tmpDataPath)
 
 	if *shardByURL {
 		consistentHashNodes := make([]string, 0, len(urls))
@@ -388,7 +415,7 @@ func tryPush(at *auth.Token, wr *prompb.WriteRequest, forceDropSamplesOnFailure
 
 	// Push metadata separately from time series, since it doesn't need sharding,
 	// relabeling, stream aggregation, deduplication, etc.
-	if !tryPushMetadataToRemoteStorages(rwctxs, mms, forceDropSamplesOnFailure) {
+	if !tryPushMetadataToRemoteStorages(at, rwctxs, mms, forceDropSamplesOnFailure) {
 		return false
 	}
 
@@ -488,7 +515,9 @@ func tryPush(at *auth.Token, wr *prompb.WriteRequest, forceDropSamplesOnFailure
 //
 // calculateHealthyRwctxIdx will rely on the order of rwctx to be in ascending order.
 func getEligibleRemoteWriteCtxs(tss []prompb.TimeSeries, forceDropSamplesOnFailure bool) ([]*remoteWriteCtx, bool) {
-	if !disableOnDiskQueueAny {
+	// When -remoteWrite.shardByURL=true always use all configured remote writes to preserve stable metrics distribution across shards.
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10507
+	if !disableOnDiskQueueAny || *shardByURL {
 		return rwctxsGlobal, true
 	}
 
@@ -503,12 +532,6 @@ func getEligibleRemoteWriteCtxs(tss []prompb.TimeSeries, forceDropSamplesOnFailu
 				return nil, false
 			}
 			rowsCount := getRowsCount(tss)
-			if *shardByURL {
-				// Todo: When shardByURL is enabled, the following metrics won't be 100% accurate. Because vmagent don't know
-				// which rwctx should data be pushed to yet. Let's consider the hashing algorithm fair and will distribute
-				// data to all rwctxs evenly.
-				rowsCount = rowsCount / len(rwctxsGlobal)
-			}
 			rwctx.rowsDroppedOnPushFailure.Add(rowsCount)
 		}
 	}
@@ -526,11 +549,18 @@ func pushTimeSeriesToRemoteStoragesTrackDropped(tss []prompb.TimeSeries) {
 	}
 }
 
-func tryPushMetadataToRemoteStorages(rwctxs []*remoteWriteCtx, mms []prompb.MetricMetadata, forceDropSamplesOnFailure bool) bool {
+func tryPushMetadataToRemoteStorages(at *auth.Token, rwctxs []*remoteWriteCtx, mms []prompb.MetricMetadata, forceDropSamplesOnFailure bool) bool {
 	if len(mms) == 0 {
 		// Nothing to push
 		return true
 	}
+	if at != nil {
+		for idx := range mms {
+			mm := &mms[idx]
+			mm.AccountID = at.AccountID
+			mm.ProjectID = at.ProjectID
+		}
+	}
 	// Do not shard metadata even if -remoteWrite.shardByURL is set, just replicate it among rwctxs.
 	// Since metadata is usually small and there is no guarantee that metadata can be sent to
 	// the same remote storage with the corresponding metrics.
@@ -540,6 +570,10 @@ func tryPushMetadataToRemoteStorages(rwctxs []*remoteWriteCtx, mms []prompb.Metr
 	var wg sync.WaitGroup
 	var anyPushFailed atomic.Bool
 	for _, rwctx := range rwctxs {
+		if !rwctx.enableMetadata {
+			// Skip remote storage with disabled metadata
+			continue
+		}
 		wg.Go(func() {
 			if !rwctx.tryPushMetadataInternal(mms) {
 				rwctx.pushFailures.Inc()
@@ -677,7 +711,7 @@ func shardAmountRemoteWriteCtx(tssBlock []prompb.TimeSeries, shards [][]prompb.T
 			}
 			tmpLabels.Labels = hashLabels
 		}
-		h := getLabelsHash(hashLabels)
+		h := getLabelsHashForShard(hashLabels)
 
 		// Get the rwctxIdx through consistent hashing and then map it to the index in shards.
 		// The rwctxIdx is not always equal to the shardIdx, for example, when some rwctx are not available.
@@ -768,11 +802,28 @@ var (
 	dailySeriesLimitRowsDropped  = metrics.NewCounter(`vmagent_daily_series_limit_rows_dropped_total`)
 )
 
+// getLabelsHashForShard is a separate function from getLabelsHash because
+// it omits the '=' separator between label name and value for backward compatibility.
+// Changing it would re-shard all series across remoteWrite targets.
+func getLabelsHashForShard(labels []prompb.Label) uint64 {
+	bb := labelsHashBufPool.Get()
+	b := bb.B[:0]
+	for _, label := range labels {
+		b = append(b, label.Name...)
+		b = append(b, label.Value...)
+	}
+	h := xxhash.Sum64(b)
+	bb.B = b
+	labelsHashBufPool.Put(bb)
+	return h
+}
+
 func getLabelsHash(labels []prompb.Label) uint64 {
 	bb := labelsHashBufPool.Get()
 	b := bb.B[:0]
 	for _, label := range labels {
 		b = append(b, label.Name...)
+		b = append(b, '=')
 		b = append(b, label.Value...)
 	}
 	h := xxhash.Sum64(b)
@@ -811,6 +862,11 @@ type remoteWriteCtx struct {
 	streamAggrKeepInput bool
 	streamAggrDropInput bool
 
+	// enableMetadata indicates whether metadata should be sent to this remote storage.
+	// It is determined by -remoteWrite.enableMetadata per-URL flag if set,
+	// otherwise by the global -enableMetadata flag.
+	enableMetadata bool
+
 	pss        []*pendingSeries
 	pssNextIdx atomic.Uint64
 
@@ -822,6 +878,18 @@ type remoteWriteCtx struct {
 	rowsDroppedOnPushFailure     *metrics.Counter
 }
 
+// isMetadataEnabledForURL returns true if metadata should be sent to the remote storage at argIdx.
+// It checks the per-URL -remoteWrite.disableMetadata flag first.
+// If not set, it falls back to the global -enableMetadata flag.
+func isMetadataEnabledForURL(argIdx int) bool {
+	if disableMetadataPerURL.GetOptionalArg(argIdx) {
+		// Metadata is explicitly disabled for this URL
+		return false
+	}
+	// Use global -enableMetadata value
+	return prommetadata.IsEnabled()
+}
+
 func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, sanitizedURL string) *remoteWriteCtx {
 	// strip query params, otherwise changing params resets pq
 	pqURL := *remoteWriteURL
@@ -892,10 +960,11 @@ func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, sanitizedURL string)
 	}
 
 	rwctx := &remoteWriteCtx{
-		idx: argIdx,
-		fq:  fq,
-		c:   c,
-		pss: pss,
+		idx:            argIdx,
+		fq:             fq,
+		c:              c,
+		pss:            pss,
+		enableMetadata: isMetadataEnabledForURL(argIdx),
 
 		rowsPushedAfterRelabel: metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_rows_pushed_after_relabel_total{path=%q,url=%q}`, queuePath, sanitizedURL)),
 		rowsDroppedByRelabel:   metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_relabel_metrics_dropped_total{path=%q,url=%q}`, queuePath, sanitizedURL)),
@@ -1116,3 +1185,21 @@ func newMapFromStrings(a []string) map[string]struct{} {
 	}
 	return m
 }
+
+func getMaxHourlySeries() int {
+	limit := *maxHourlySeries
+	if limit == -1 || limit > math.MaxInt32 {
+		return math.MaxInt32
+	}
+
+	return int(limit)
+}
+
+func getMaxDailySeries() int {
+	limit := *maxDailySeries
+	if limit == -1 || limit > math.MaxInt32 {
+		return math.MaxInt32
+	}
+
+	return int(limit)
+}

app/vmagent/remotewrite/remotewrite_test.go

@@ -25,7 +25,7 @@ func TestGetLabelsHash_Distribution(t *testing.T) {
 		t.Helper()
 
 		// Distribute itemsCount hashes returned by getLabelsHash() across bucketsCount buckets.
-		itemsCount := 1_000 * bucketsCount
+		itemsCount := 10_000 * bucketsCount
 		m := make([]int, bucketsCount)
 		var labels []prompb.Label
 		for i := range itemsCount {
@@ -44,10 +44,12 @@ func TestGetLabelsHash_Distribution(t *testing.T) {
 		}
 
 		// Verify that the distribution is even
-		expectedItemsPerBucket := itemsCount / bucketsCount
+		expectedItemsPerBucket := float64(itemsCount / bucketsCount)
+		allowedDeviation := math.Round(float64(expectedItemsPerBucket) * 0.04)
 		for _, n := range m {
-			if math.Abs(1-float64(n)/float64(expectedItemsPerBucket)) > 0.04 {
-				t.Fatalf("unexpected items in the bucket for %d buckets; got %d; want around %d", bucketsCount, n, expectedItemsPerBucket)
+			if math.Abs(expectedItemsPerBucket-float64(n)) > allowedDeviation {
+				t.Fatalf("unexpected items in the bucket for %d buckets; got %d; want in range [%.0f, %.0f]",
+					bucketsCount, n, expectedItemsPerBucket-allowedDeviation, expectedItemsPerBucket+allowedDeviation)
 			}
 		}
 	}

app/vmalert-tool/unittest/unittest.go

@@ -61,7 +61,7 @@ func UnitTest(files []string, disableGroupLabel bool, externalLabels []string, e
 	}
 	eu, err := url.Parse(externalURL)
 	if err != nil {
-		logger.Fatalf("failed to parse external URL: %w", err)
+		logger.Fatalf("failed to parse external URL: %s", err)
 	}
 	if err := templates.Load([]string{}, *eu); err != nil {
 		logger.Fatalf("failed to load template: %v", err)

app/vmalert/config/config.go

@@ -222,6 +222,9 @@ func (r *Rule) Validate() error {
 	if r.Expr == "" {
 		return fmt.Errorf("expression can't be empty")
 	}
+	if _, ok := r.Labels["__name__"]; ok {
+		return fmt.Errorf("invalid rule label __name__")
+	}
 	return checkOverflow(r.XXX, "rule")
 }
 

app/vmalert/config/config_test.go

@@ -136,6 +136,9 @@ func TestRuleValidate(t *testing.T) {
 	if err := (&Rule{Alert: "alert"}).Validate(); err == nil {
 		t.Fatalf("expected empty expr error")
 	}
+	if err := (&Rule{Record: "record", Expr: "sum(test)", Labels: map[string]string{"__name__": "test"}}).Validate(); err == nil {
+		t.Fatalf("invalid rule label; got %s", err)
+	}
 	if err := (&Rule{Alert: "alert", Expr: "test>0"}).Validate(); err != nil {
 		t.Fatalf("expected valid rule; got %s", err)
 	}

app/vmalert/datasource/client_test.go

@@ -772,7 +772,7 @@ func TestHeaders(t *testing.T) {
 
 	// basic auth
 	f(func() *Client {
-		cfg, err := vmalertutil.AuthConfig(vmalertutil.WithBasicAuth("foo", "bar", ""))
+		cfg, err := vmalertutil.AuthConfig(vmalertutil.WithBasicAuth("foo", "", "bar", ""))
 		if err != nil {
 			t.Fatalf("Error get auth config: %s", err)
 		}
@@ -817,7 +817,7 @@ func TestHeaders(t *testing.T) {
 
 	// custom header overrides basic auth
 	f(func() *Client {
-		cfg, err := vmalertutil.AuthConfig(vmalertutil.WithBasicAuth("foo", "bar", ""))
+		cfg, err := vmalertutil.AuthConfig(vmalertutil.WithBasicAuth("foo", "", "bar", ""))
 		if err != nil {
 			t.Fatalf("Error get auth config: %s", err)
 		}

app/vmalert/datasource/datasource.go

@@ -87,6 +87,7 @@ func (m *Metric) DelLabel(key string) {
 	for i, l := range m.Labels {
 		if l.Name == key {
 			m.Labels = append(m.Labels[:i], m.Labels[i+1:]...)
+			break
 		}
 	}
 }

app/vmalert/datasource/init.go

@@ -27,6 +27,7 @@ var (
 		"Multiple headers must be delimited by '^^': -datasource.headers='header1:value1^^header2:value2'")
 
 	basicAuthUsername     = flag.String("datasource.basicAuth.username", "", "Optional basic auth username for -datasource.url")
+	basicAuthUsernameFile = flag.String("datasource.basicAuth.usernameFile", "", "Optional path to basic auth username to use for -datasource.url")
 	basicAuthPassword     = flag.String("datasource.basicAuth.password", "", "Optional basic auth password for -datasource.url")
 	basicAuthPasswordFile = flag.String("datasource.basicAuth.passwordFile", "", "Optional path to basic auth password to use for -datasource.url")
 
@@ -63,6 +64,7 @@ func InitSecretFlags() {
 	if !*showDatasourceURL {
 		flagutil.RegisterSecretFlag("datasource.url")
 	}
+	flagutil.RegisterSecretFlag("datasource.headers")
 }
 
 // ShowDatasourceURL whether to show -datasource.url with sensitive information
@@ -105,7 +107,7 @@ func Init(extraParams url.Values) (QuerierBuilder, error) {
 		return nil, fmt.Errorf("cannot parse JSON for -datasource.oauth2.endpointParams=%s: %w", *oauth2EndpointParams, err)
 	}
 	authCfg, err := vmalertutil.AuthConfig(
-		vmalertutil.WithBasicAuth(*basicAuthUsername, *basicAuthPassword, *basicAuthPasswordFile),
+		vmalertutil.WithBasicAuth(*basicAuthUsername, *basicAuthUsernameFile, *basicAuthPassword, *basicAuthPasswordFile),
 		vmalertutil.WithBearer(*bearerToken, *bearerTokenFile),
 		vmalertutil.WithOAuth(*oauth2ClientID, *oauth2ClientSecret, *oauth2ClientSecretFile, *oauth2TokenURL, *oauth2Scopes, endpointParams),
 		vmalertutil.WithHeaders(*headers))

app/vmalert/notifier/alertmanager.go

@@ -191,7 +191,7 @@ func NewAlertManager(alertManagerURL string, fn AlertURLGenerator, authCfg proma
 	}
 
 	aCfg, err := vmalertutil.AuthConfig(
-		vmalertutil.WithBasicAuth(ba.Username, ba.Password.String(), ba.PasswordFile),
+		vmalertutil.WithBasicAuth(ba.Username, ba.UsernameFile, ba.Password.String(), ba.PasswordFile),
 		vmalertutil.WithBearer(authCfg.BearerToken.String(), authCfg.BearerTokenFile),
 		vmalertutil.WithOAuth(oauth.ClientID, oauth.ClientSecret.String(), oauth.ClientSecretFile, oauth.TokenURL, strings.Join(oauth.Scopes, ";"), oauth.EndpointParams),
 		vmalertutil.WithHeaders(strings.Join(authCfg.Headers, "^^")),

app/vmalert/notifier/config_watcher.go

@@ -105,7 +105,7 @@ func (cw *configWatcher) add(typeK TargetType, interval time.Duration, targetsFn
 			}
 			targetMetadata, errors := getTargetMetadata(targetsFn, cw.cfg)
 			for _, err := range errors {
-				logger.Errorf("failed to init notifier for %q: %w", typeK, err)
+				logger.Errorf("failed to init notifier for %q: %s", typeK, err)
 			}
 			cw.updateTargets(typeK, targetMetadata, cw.cfg, cw.genFn)
 		}
@@ -274,7 +274,7 @@ func (cw *configWatcher) updateTargets(key TargetType, targetMts map[string]targ
 	for addr, metadata := range targetMts {
 		am, err := NewAlertManager(addr, genFn, cfg.HTTPClientConfig, metadata.alertRelabelConfigs, cfg.Timeout.Duration())
 		if err != nil {
-			logger.Errorf("failed to init %s notifier with addr %q: %w", key, addr, err)
+			logger.Errorf("failed to init %s notifier with addr %q: %s", key, addr, err)
 			continue
 		}
 		updatedTargets = append(updatedTargets, Target{

app/vmalert/notifier/init.go

@@ -36,6 +36,7 @@ var (
 		"For example, -remoteWrite.headers='My-Auth:foobar' would send 'My-Auth: foobar' HTTP header with every request to the corresponding -notifier.url. "+
 		"Multiple headers must be delimited by '^^': -notifier.headers='header1:value1^^header2:value2,header3:value3'")
 	basicAuthUsername     = flagutil.NewArrayString("notifier.basicAuth.username", "Optional basic auth username for -notifier.url")
+	basicAuthUsernameFile = flagutil.NewArrayString("notifier.basicAuth.usernameFile", "Optional path to basic auth username file for -notifier.url")
 	basicAuthPassword     = flagutil.NewArrayString("notifier.basicAuth.password", "Optional basic auth password for -notifier.url")
 	basicAuthPasswordFile = flagutil.NewArrayString("notifier.basicAuth.passwordFile", "Optional path to basic auth password file for -notifier.url")
 
@@ -193,6 +194,7 @@ func InitSecretFlags() {
 	if !*showNotifierURL {
 		flagutil.RegisterSecretFlag("notifier.url")
 	}
+	flagutil.RegisterSecretFlag("notifier.headers")
 }
 
 func notifiersFromFlags(gen AlertURLGenerator) ([]Notifier, error) {
@@ -213,6 +215,7 @@ func notifiersFromFlags(gen AlertURLGenerator) ([]Notifier, error) {
 			},
 			BasicAuth: &promauth.BasicAuthConfig{
 				Username:     basicAuthUsername.GetOptionalArg(i),
+				UsernameFile: basicAuthUsernameFile.GetOptionalArg(i),
 				Password:     promauth.NewSecret(basicAuthPassword.GetOptionalArg(i)),
 				PasswordFile: basicAuthPasswordFile.GetOptionalArg(i),
 			},

app/vmalert/notifier/notifier.go

@@ -14,7 +14,7 @@ type Notifier interface {
 	Send(ctx context.Context, alerts []Alert, alertLabels [][]prompb.Label, notifierHeaders map[string]string) error
 	// Addr returns address where alerts are sent.
 	Addr() string
-	// LastError returns error, that occured during last attempt to send data
+	// LastError returns error, that occurred during last attempt to send data
 	LastError() string
 	// Close is a destructor for the Notifier
 	Close()

app/vmalert/remoteread/init.go

@@ -28,6 +28,7 @@ var (
 		"Multiple headers must be delimited by '^^': -remoteRead.headers='header1:value1^^header2:value2'")
 
 	basicAuthUsername     = flag.String("remoteRead.basicAuth.username", "", "Optional basic auth username for -remoteRead.url")
+	basicAuthUsernameFile = flag.String("remoteRead.basicAuth.usernameFile", "", "Optional path to basic auth username to use for -remoteRead.url")
 	basicAuthPassword     = flag.String("remoteRead.basicAuth.password", "", "Optional basic auth password for -remoteRead.url")
 	basicAuthPasswordFile = flag.String("remoteRead.basicAuth.passwordFile", "", "Optional path to basic auth password to use for -remoteRead.url")
 
@@ -58,6 +59,7 @@ func InitSecretFlags() {
 	if !*showRemoteReadURL {
 		flagutil.RegisterSecretFlag("remoteRead.url")
 	}
+	flagutil.RegisterSecretFlag("remoteRead.headers")
 }
 
 // Init creates a Querier from provided flag values.
@@ -80,7 +82,7 @@ func Init() (datasource.QuerierBuilder, error) {
 		return nil, fmt.Errorf("cannot parse JSON for -remoteRead.oauth2.endpointParams=%s: %w", *oauth2EndpointParams, err)
 	}
 	authCfg, err := vmalertutil.AuthConfig(
-		vmalertutil.WithBasicAuth(*basicAuthUsername, *basicAuthPassword, *basicAuthPasswordFile),
+		vmalertutil.WithBasicAuth(*basicAuthUsername, *basicAuthUsernameFile, *basicAuthPassword, *basicAuthPasswordFile),
 		vmalertutil.WithBearer(*bearerToken, *bearerTokenFile),
 		vmalertutil.WithOAuth(*oauth2ClientID, *oauth2ClientSecret, *oauth2ClientSecretFile, *oauth2TokenURL, *oauth2Scopes, endpointParams),
 		vmalertutil.WithHeaders(*headers))

app/vmalert/remotewrite/client.go

@@ -11,16 +11,23 @@ import (
 	"path"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
+	"github.com/cespare/xxhash/v2"
 	"github.com/golang/snappy"
 
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding/zstd"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/netutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeutil"
+
 	"github.com/VictoriaMetrics/metrics"
 )
 
@@ -53,6 +60,11 @@ type Client struct {
 
 	wg     sync.WaitGroup
 	doneCh chan struct{}
+
+	// Whether to encode the write request with VictoriaMetrics remote write protocol.
+	// It is set to true by default, and will be switched to false if the client
+	// receives specific errors indicating that the remote storage doesn't support VictoriaMetrics remote write protocol.
+	isVMRemoteWrite atomic.Bool
 }
 
 // Config is config for remote write client.
@@ -112,9 +124,12 @@ func NewClient(ctx context.Context, cfg Config) (*Client, error) {
 		doneCh:        make(chan struct{}),
 		input:         make(chan prompb.TimeSeries, cfg.MaxQueueSize),
 	}
+	c.isVMRemoteWrite.Store(true)
 
-	for range cc {
-		c.run(ctx)
+	for i := 0; i < cc; i++ {
+		c.wg.Go(func() {
+			c.run(ctx, i)
+		})
 	}
 	return c, nil
 }
@@ -156,8 +171,7 @@ func (c *Client) Close() error {
 	return nil
 }
 
-func (c *Client) run(ctx context.Context) {
-	ticker := time.NewTicker(c.flushInterval)
+func (c *Client) run(ctx context.Context, id int) {
 	wr := &prompb.WriteRequest{}
 	shutdown := func() {
 		lastCtx, cancel := context.WithTimeout(context.Background(), defaultWriteTimeout)
@@ -174,45 +188,72 @@ func (c *Client) run(ctx context.Context) {
 		cancel()
 	}
 
-	c.wg.Go(func() {
-		defer ticker.Stop()
-		for {
+	// add jitter to spread remote write flushes over the flush interval to avoid congestion at the remote write destination
+	h := xxhash.Sum64(bytesutil.ToUnsafeBytes(fmt.Sprintf("%d", id)))
+	randJitter := uint64(float64(c.flushInterval) * (float64(h) / (1 << 64)))
+	timer := time.NewTimer(time.Duration(randJitter))
+addJitter:
+	for {
+		select {
+		case <-c.doneCh:
+			timer.Stop()
+			shutdown()
+			return
+		case <-ctx.Done():
+			timer.Stop()
+			shutdown()
+			return
+		case <-timer.C:
+			break addJitter
+		}
+	}
+
+	ticker := time.NewTicker(c.flushInterval)
+	defer ticker.Stop()
+	for {
+		select {
+		case <-c.doneCh:
+			shutdown()
+			return
+		case <-ctx.Done():
+			shutdown()
+			return
+		case <-ticker.C:
+			c.flush(ctx, wr)
+			// drain the potential stale tick to avoid small or empty flushes after a slow flush.
 			select {
-			case <-c.doneCh:
-				shutdown()
-				return
-			case <-ctx.Done():
-				shutdown()
-				return
 			case <-ticker.C:
+			default:
+			}
+		case ts, ok := <-c.input:
+			if !ok {
+				continue
+			}
+			wr.Timeseries = append(wr.Timeseries, ts)
+			if len(wr.Timeseries) >= c.maxBatchSize {
 				c.flush(ctx, wr)
-				// drain the potential stale tick to avoid small or empty flushes after a slow flush.
-				select {
-				case <-ticker.C:
-				default:
-				}
-			case ts, ok := <-c.input:
-				if !ok {
-					continue
-				}
-				wr.Timeseries = append(wr.Timeseries, ts)
-				if len(wr.Timeseries) >= c.maxBatchSize {
-					c.flush(ctx, wr)
-				}
 			}
 		}
-	})
+	}
 }
 
 var (
 	rwErrors = metrics.NewCounter(`vmalert_remotewrite_errors_total`)
 	rwTotal  = metrics.NewCounter(`vmalert_remotewrite_total`)
 
-	sentRows            = metrics.NewCounter(`vmalert_remotewrite_sent_rows_total`)
-	sentBytes           = metrics.NewCounter(`vmalert_remotewrite_sent_bytes_total`)
-	droppedRows         = metrics.NewCounter(`vmalert_remotewrite_dropped_rows_total`)
-	sendDuration        = metrics.NewFloatCounter(`vmalert_remotewrite_send_duration_seconds_total`)
-	bufferFlushDuration = metrics.NewHistogram(`vmalert_remotewrite_flush_duration_seconds`)
+	// sentRows and sentBytes are historical counters that can now be replaced by flushedRows and flushedBytes histograms. They may be deprecated in the future after the new histograms have been adopted for some time.
+	sentRows             = metrics.NewCounter(`vmalert_remotewrite_sent_rows_total`)
+	sentBytes            = metrics.NewCounter(`vmalert_remotewrite_sent_bytes_total`)
+	flushedRows          = metrics.NewHistogram(`vmalert_remotewrite_sent_rows`)
+	flushedBytes         = metrics.NewHistogram(`vmalert_remotewrite_sent_bytes`)
+	droppedRows          = metrics.NewCounter(`vmalert_remotewrite_dropped_rows_total`)
+	sendDuration         = metrics.NewFloatCounter(`vmalert_remotewrite_send_duration_seconds_total`)
+	bufferFlushDuration  = metrics.NewHistogram(`vmalert_remotewrite_flush_duration_seconds`)
+	remoteWriteQueueSize = metrics.NewHistogram(`vmalert_remotewrite_queue_size`)
+
+	_ = metrics.NewGauge(`vmalert_remotewrite_queue_capacity`, func() float64 {
+		return float64(*maxQueueSize)
+	})
 
 	_ = metrics.NewGauge(`vmalert_remotewrite_concurrency`, func() float64 {
 		return float64(*concurrency)
@@ -226,19 +267,26 @@ func GetDroppedRows() int { return int(droppedRows.Get()) }
 // it to remote-write endpoint. Flush performs limited amount of retries
 // if request fails.
 func (c *Client) flush(ctx context.Context, wr *prompb.WriteRequest) {
+	remoteWriteQueueSize.Update(float64(len(c.input)))
 	if len(wr.Timeseries) < 1 {
 		return
 	}
 	defer wr.Reset()
 	defer bufferFlushDuration.UpdateDuration(time.Now())
 
-	data := wr.MarshalProtobuf(nil)
-	b := snappy.Encode(nil, data)
-
-	retryInterval, maxRetryInterval := *retryMinInterval, *retryMaxTime
-	if retryInterval > maxRetryInterval {
-		retryInterval = maxRetryInterval
+	bb := writeRequestBufPool.Get()
+	bb.B = wr.MarshalProtobuf(bb.B[:0])
+	zb := compressBufPool.Get()
+	defer compressBufPool.Put(zb)
+	if c.isVMRemoteWrite.Load() {
+		zb.B = zstd.CompressLevel(zb.B[:0], bb.B, 0)
+	} else {
+		zb.B = snappy.Encode(zb.B[:cap(zb.B)], bb.B)
 	}
+	writeRequestBufPool.Put(bb)
+
+	maxRetryInterval := *retryMaxTime
+	bt := timeutil.NewBackoffTimer(*retryMinInterval, maxRetryInterval)
 	timeStart := time.Now()
 	defer func() {
 		sendDuration.Add(time.Since(timeStart).Seconds())
@@ -247,15 +295,17 @@ func (c *Client) flush(ctx context.Context, wr *prompb.WriteRequest) {
 	attempts := 0
 L:
 	for {
-		err := c.send(ctx, b)
+		err := c.send(ctx, zb.B)
 		if err != nil && (errors.Is(err, io.EOF) || netutil.IsTrivialNetworkError(err)) {
 			// Something in the middle between client and destination might be closing
 			// the connection. So we do a one more attempt in hope request will succeed.
-			err = c.send(ctx, b)
+			err = c.send(ctx, zb.B)
 		}
 		if err == nil {
 			sentRows.Add(len(wr.Timeseries))
-			sentBytes.Add(len(b))
+			sentBytes.Add(len(zb.B))
+			flushedRows.Update(float64(len(wr.Timeseries)))
+			flushedBytes.Update(float64(len(zb.B)))
 			return
 		}
 
@@ -281,12 +331,11 @@ L:
 			break
 		}
 
-		if retryInterval > timeLeftForRetries {
-			retryInterval = timeLeftForRetries
+		if bt.CurrentDelay() > timeLeftForRetries {
+			bt.SetDelay(timeLeftForRetries)
 		}
 		// sleeping to prevent remote db hammering
-		time.Sleep(retryInterval)
-		retryInterval *= 2
+		bt.Wait(ctx.Done())
 
 		attempts++
 	}
@@ -308,12 +357,16 @@ func (c *Client) send(ctx context.Context, data []byte) error {
 		return fmt.Errorf("failed to create new HTTP request: %w", err)
 	}
 
-	// RFC standard compliant headers
-	req.Header.Set("Content-Encoding", "snappy")
+	req.Header.Set("User-Agent", "vmalert")
 	req.Header.Set("Content-Type", "application/x-protobuf")
 
-	// Prometheus compliant headers
-	req.Header.Set("X-Prometheus-Remote-Write-Version", "0.1.0")
+	if encoding.IsZstd(data) {
+		req.Header.Set("Content-Encoding", "zstd")
+		req.Header.Set("X-VictoriaMetrics-Remote-Write-Version", "1")
+	} else {
+		req.Header.Set("Content-Encoding", "snappy")
+		req.Header.Set("X-Prometheus-Remote-Write-Version", "0.1.0")
+	}
 
 	if c.authCfg != nil {
 		err = c.authCfg.SetHeaders(req, true)
@@ -342,6 +395,29 @@ func (c *Client) send(ctx context.Context, data []byte) error {
 		// respond with HTTP 2xx status code when write is successful.
 		return nil
 	case 4:
+		// - Remote Write v1 specification implicitly expects a `400 Bad Request` when the encoding is not supported.
+		// - Remote Write v2 specification explicitly specifies a `415 Unsupported Media Type` for unsupported encodings.
+		// - Real-world implementations of v1 use both 400 and 415 status codes.
+		// See more in research: https://github.com/VictoriaMetrics/VictoriaMetrics/pull/8462#issuecomment-2786918054
+		if resp.StatusCode == http.StatusUnsupportedMediaType || resp.StatusCode == http.StatusBadRequest {
+			if encoding.IsZstd(data) {
+				logger.Infof("received unsupported media type or bad request from remote storage at %q. Re-packing the block to Prometheus remote write and retrying."+
+					"See https://docs.victoriametrics.com/victoriametrics/vmagent/#victoriametrics-remote-write-protocol", req.URL.Redacted())
+				zstdBlockLen := len(data)
+				data, err = repackBlockFromZstdToSnappy(data)
+				if err == nil {
+					logger.Infof("received unsupported media type or bad request from remote storage at %q. Downgrading protocol from VictoriaMetrics to Prometheus remote write for all future requests. "+
+						"See https://docs.victoriametrics.com/victoriametrics/vmagent/#victoriametrics-remote-write-protocol", req.URL.Redacted())
+					c.isVMRemoteWrite.Store(false)
+					return c.send(ctx, data)
+				}
+
+				logger.Warnf("failed to repack zstd block (%d bytes) to snappy: %s; The block will be rejected. "+
+					"Possible cause: ungraceful shutdown leading to persisted queue corruption.",
+					zstdBlockLen, err)
+			}
+		}
+
 		if resp.StatusCode != http.StatusTooManyRequests {
 			// MUST NOT retry write requests on HTTP 4xx responses other than 429
 			return &nonRetriableError{
@@ -362,3 +438,19 @@ type nonRetriableError struct {
 func (e *nonRetriableError) Error() string {
 	return e.err.Error()
 }
+
+var (
+	writeRequestBufPool bytesutil.ByteBufferPool
+	compressBufPool     bytesutil.ByteBufferPool
+)
+
+// repackBlockFromZstdToSnappy repacks the given zstd-compressed block to snappy-compressed block.
+func repackBlockFromZstdToSnappy(zstdBlock []byte) ([]byte, error) {
+	plainBlock := make([]byte, 0, len(zstdBlock)*2)
+	plainBlock, err := encoding.DecompressZSTD(plainBlock, zstdBlock)
+	if err != nil {
+		return nil, err
+	}
+
+	return snappy.Encode(nil, plainBlock), nil
+}

app/vmalert/remotewrite/client_test.go

@@ -12,8 +12,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/golang/snappy"
-
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding/zstd"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
 )
 
@@ -103,7 +102,10 @@ func TestClient_run_maxBatchSizeDuringShutdown(t *testing.T) {
 
 		// push time series to the client.
 		for range pushCnt {
-			if err = rwClient.Push(prompb.TimeSeries{}); err != nil {
+			if err = rwClient.Push(prompb.TimeSeries{
+				Labels:  []prompb.Label{{Name: "__name__", Value: "m"}},
+				Samples: []prompb.Sample{{Value: 1, Timestamp: 1000}},
+			}); err != nil {
 				t.Fatalf("cannot time series to the client: %s", err)
 			}
 		}
@@ -156,8 +158,8 @@ func (rw *rwServer) handler(w http.ResponseWriter, r *http.Request) {
 	}
 
 	h := r.Header.Get("Content-Encoding")
-	if h != "snappy" {
-		rw.err(w, fmt.Errorf("header read error: Content-Encoding is not snappy (%q)", h))
+	if h != "zstd" {
+		rw.err(w, fmt.Errorf("header read error: Content-Encoding is not zstd (%q)", h))
 	}
 
 	h = r.Header.Get("Content-Type")
@@ -165,9 +167,9 @@ func (rw *rwServer) handler(w http.ResponseWriter, r *http.Request) {
 		rw.err(w, fmt.Errorf("header read error: Content-Type is not x-protobuf (%q)", h))
 	}
 
-	h = r.Header.Get("X-Prometheus-Remote-Write-Version")
-	if h != "0.1.0" {
-		rw.err(w, fmt.Errorf("header read error: X-Prometheus-Remote-Write-Version is not 0.1.0 (%q)", h))
+	h = r.Header.Get("X-VictoriaMetrics-Remote-Write-Version")
+	if h != "1" {
+		rw.err(w, fmt.Errorf("header read error: X-VictoriaMetrics-Remote-Write-Version is not 1 (%q)", h))
 	}
 
 	data, err := io.ReadAll(r.Body)
@@ -177,7 +179,7 @@ func (rw *rwServer) handler(w http.ResponseWriter, r *http.Request) {
 	}
 	defer func() { _ = r.Body.Close() }()
 
-	b, err := snappy.Decode(nil, data)
+	b, err := zstd.Decompress(nil, data)
 	if err != nil {
 		rw.err(w, fmt.Errorf("decode err: %w", err))
 		return

app/vmalert/remotewrite/debug_client.go

@@ -9,8 +9,7 @@ import (
 	"strings"
 	"sync"
 
-	"github.com/golang/snappy"
-
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding/zstd"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
@@ -64,19 +63,17 @@ func (c *DebugClient) Close() error {
 }
 
 func (c *DebugClient) send(data []byte) error {
-	b := snappy.Encode(nil, data)
+	b := zstd.CompressLevel(nil, data, 0)
 	r := bytes.NewReader(b)
 	req, err := http.NewRequest(http.MethodPost, c.addr, r)
 	if err != nil {
 		return fmt.Errorf("failed to create new HTTP request: %w", err)
 	}
 
-	// RFC standard compliant headers
-	req.Header.Set("Content-Encoding", "snappy")
+	req.Header.Set("Content-Encoding", "zstd")
 	req.Header.Set("Content-Type", "application/x-protobuf")
 
-	// Prometheus compliant headers
-	req.Header.Set("X-Prometheus-Remote-Write-Version", "0.1.0")
+	req.Header.Set("X-VictoriaMetrics-Remote-Write-Version", "1")
 
 	if !*disablePathAppend {
 		req.URL.Path = path.Join(req.URL.Path, "/api/v1/write")

app/vmalert/remotewrite/init.go

@@ -13,8 +13,8 @@ import (
 )
 
 var (
-	addr = flag.String("remoteWrite.url", "", "Optional URL to VictoriaMetrics or vminsert where to persist alerts state "+
-		"and recording rules results in form of timeseries. "+
+	addr = flag.String("remoteWrite.url", "", "Optional URL to persist alerts state and recording rules results in form of timeseries. "+
+		"It must support either VictoriaMetrics remote write protocol or Prometheus remote_write protocol. "+
 		"Supports address in the form of IP address with a port (e.g., http://127.0.0.1:8428) or DNS SRV record. "+
 		"For example, if -remoteWrite.url=http://127.0.0.1:8428 is specified, "+
 		"then the alerts state will be written to http://127.0.0.1:8428/api/v1/write . See also -remoteWrite.disablePathAppend, '-remoteWrite.showURL'.")
@@ -26,6 +26,7 @@ var (
 		"Multiple headers must be delimited by '^^': -remoteWrite.headers='header1:value1^^header2:value2'")
 
 	basicAuthUsername     = flag.String("remoteWrite.basicAuth.username", "", "Optional basic auth username for -remoteWrite.url")
+	basicAuthUsernameFile = flag.String("remoteWrite.basicAuth.usernameFile", "", "Optional path to basic auth username to use for -remoteWrite.url")
 	basicAuthPassword     = flag.String("remoteWrite.basicAuth.password", "", "Optional basic auth password for -remoteWrite.url")
 	basicAuthPasswordFile = flag.String("remoteWrite.basicAuth.passwordFile", "", "Optional path to basic auth password to use for -remoteWrite.url")
 
@@ -61,6 +62,7 @@ func InitSecretFlags() {
 	if !*showRemoteWriteURL {
 		flagutil.RegisterSecretFlag("remoteWrite.url")
 	}
+	flagutil.RegisterSecretFlag("remoteWrite.headers")
 }
 
 // Init creates Client object from given flags.
@@ -83,7 +85,7 @@ func Init(ctx context.Context) (*Client, error) {
 		return nil, fmt.Errorf("cannot parse JSON for -remoteWrite.oauth2.endpointParams=%s: %w", *oauth2EndpointParams, err)
 	}
 	authCfg, err := vmalertutil.AuthConfig(
-		vmalertutil.WithBasicAuth(*basicAuthUsername, *basicAuthPassword, *basicAuthPasswordFile),
+		vmalertutil.WithBasicAuth(*basicAuthUsername, *basicAuthUsernameFile, *basicAuthPassword, *basicAuthPasswordFile),
 		vmalertutil.WithBearer(*bearerToken, *bearerTokenFile),
 		vmalertutil.WithOAuth(*oauth2ClientID, *oauth2ClientSecret, *oauth2ClientSecretFile, *oauth2TokenURL, *oauth2Scopes, endpointParams),
 		vmalertutil.WithHeaders(*headers))

app/vmalert/rule/alerting.go

@@ -312,9 +312,11 @@ type labelSet struct {
 // On k conflicts in origin set, the original value is preferred and copied
 // to processed with `exported_%k` key. The copy happens only if passed v isn't equal to origin[k] value.
 func (ls *labelSet) add(k, v string) {
-	// do not add label with empty value, since it has no meaning.
-	// see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/9984
+	// do not add label with empty value to the result, as it has no meaning:
+	// if the label already exists in the original query result, remove it to preserve compatibility with relabeling, see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10766.
+	// otherwise, ignore the label, see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/9984.
 	if v == "" {
+		delete(ls.processed, k)
 		return
 	}
 	ls.processed[k] = v

app/vmalert/rule/alerting_test.go

@@ -1363,6 +1363,7 @@ func TestAlertingRule_ToLabels(t *testing.T) {
 			{Name: "instance", Value: "0.0.0.0:8800"},
 			{Name: "group", Value: "vmalert"},
 			{Name: "alertname", Value: "ConfigurationReloadFailure"},
+			{Name: "pod", Value: "vmalert-0"},
 		},
 		Values:     []float64{1},
 		Timestamps: []int64{time.Now().UnixNano()},
@@ -1374,6 +1375,7 @@ func TestAlertingRule_ToLabels(t *testing.T) {
 			"group":         "vmalert",  // this shouldn't have effect since value in metric is equal
 			"invalid_label": "{{ .Values.mustRuntimeFail }}",
 			"empty_label":   "", // this should be dropped
+			"pod":           "", // this should remove the pod label from query result
 		},
 		Expr:      "sum(vmalert_alerting_rules_error) by(instance, group, alertname) > 0",
 		Name:      "AlertingRulesError",
@@ -1385,6 +1387,7 @@ func TestAlertingRule_ToLabels(t *testing.T) {
 		"group":         "vmalert",
 		"alertname":     "ConfigurationReloadFailure",
 		"alertgroup":    "vmalert",
+		"pod":           "vmalert-0",
 		"invalid_label": `error evaluating template: template: :1:298: executing "" at <.Values.mustRuntimeFail>: can't evaluate field Values in type notifier.tplData`,
 	}
 

app/vmalert/rule/group.go

@@ -8,6 +8,7 @@ import (
 	"hash/fnv"
 	"maps"
 	"net/url"
+	"path"
 	"sync"
 	"time"
 
@@ -42,6 +43,9 @@ var (
 		"For example, if lookback=1h then range from now() to now()-1h will be scanned.")
 	maxStartDelay = flag.Duration("group.maxStartDelay", 5*time.Minute, "Defines the max delay before starting the group evaluation. Group's start is artificially delayed for random duration on interval"+
 		" [0..min(--group.maxStartDelay, group.interval)]. This helps smoothing out the load on the configured datasource, so evaluations aren't executed too close to each other.")
+	ruleStripFilePath = flag.Bool("rule.stripFilePath", false, "Whether to strip rule file paths in logs and all API responses, including /metrics. "+
+		"For example, file path '/path/to/tenant_id/rules.yml' will be stripped to 'groupHashID/rules.yml'. "+
+		"This flag may be useful for hiding sensitive information in file paths, such as S3 bucket details.")
 )
 
 // Group is an entity for grouping rules
@@ -147,6 +151,12 @@ func NewGroup(cfg config.Group, qb datasource.QuerierBuilder, defaultInterval ti
 		g.EvalDelay = &cfg.EvalDelay.D
 	}
 	g.id = g.CreateID()
+	// strip file path from group.File after generated group ID when ruleStripFilePath is set,
+	// so it won't be exposed in logs and api responses
+	if *ruleStripFilePath {
+		_, filename := path.Split(g.File)
+		g.File = fmt.Sprintf("%d/%s", g.id, filename)
+	}
 	for _, h := range cfg.Headers {
 		g.Headers[h.Key] = h.Value
 	}
@@ -381,7 +391,9 @@ func (g *Group) Start(ctx context.Context, rw remotewrite.RWClient, rr datasourc
 
 		if len(g.Rules) < 1 {
 			g.metrics.iterationDuration.UpdateDuration(start)
+			g.mu.Lock()
 			g.LastEvaluation = start
+			g.mu.Unlock()
 			return ts
 		}
 
@@ -395,7 +407,9 @@ func (g *Group) Start(ctx context.Context, rw remotewrite.RWClient, rr datasourc
 			}
 		}
 		g.metrics.iterationDuration.UpdateDuration(start)
+		g.mu.Lock()
 		g.LastEvaluation = start
+		g.mu.Unlock()
 		return ts
 	}
 
@@ -405,11 +419,14 @@ func (g *Group) Start(ctx context.Context, rw remotewrite.RWClient, rr datasourc
 	g.mu.Unlock()
 	defer g.evalCancel()
 
-	realEvalTS := eval(evalCtx, evalTS)
-
+	// start the interval ticker before the first evaluation,
+	// so that the evaluation timestamps of groups with the `eval_offset` option are also aligned,
+	// see https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10773
 	t := time.NewTicker(g.Interval)
 	defer t.Stop()
 
+	realEvalTS := eval(evalCtx, evalTS)
+
 	// restore the rules state after the first evaluation
 	// so only active alerts can be restored.
 	if rr != nil {

app/vmalert/rule/group_test.go

@@ -742,3 +742,64 @@ func parseTime(t *testing.T, s string) time.Time {
 	}
 	return tt
 }
+
+func TestRuleStripFilePath(t *testing.T) {
+	configG := config.Group{
+		Name:        "group",
+		File:        "/var/local/test/rules.yaml",
+		Type:        config.NewRawType("prometheus"),
+		Concurrency: 1,
+		Rules: []config.Rule{
+			{
+				ID:    0,
+				Alert: "alert",
+			},
+			{
+				ID:     1,
+				Record: "record",
+			},
+		}}
+	qb := &datasource.FakeQuerier{}
+	g := NewGroup(configG, qb, 1*time.Minute, nil)
+
+	gID := g.id
+	if g.File != "/var/local/test/rules.yaml" {
+		t.Fatalf("expected file path to be unchanged; got %q instead", g.File)
+	}
+
+	for _, r := range g.Rules {
+		if ar, ok := r.(*AlertingRule); ok {
+			if ar.File != "/var/local/test/rules.yaml" {
+				t.Fatalf("expected rule file path to be unchanged; got %q instead", ar.File)
+			}
+		}
+		if rr, ok := r.(*RecordingRule); ok {
+			if rr.File != "/var/local/test/rules.yaml" {
+				t.Fatalf("expected rule file path to be unchanged; got %q instead", rr.File)
+			}
+		}
+	}
+
+	oldRuleStripFilePath := *ruleStripFilePath
+	*ruleStripFilePath = true
+	defer func() {
+		*ruleStripFilePath = oldRuleStripFilePath
+	}()
+	g = NewGroup(configG, qb, 1*time.Minute, nil)
+
+	if g.File != fmt.Sprintf("%d/rules.yaml", gID) {
+		t.Fatalf("expected file path to be stripped to %q; got %q instead", fmt.Sprintf("%d/rules.yaml", gID), g.File)
+	}
+	for _, r := range g.Rules {
+		if ar, ok := r.(*AlertingRule); ok {
+			if ar.File != fmt.Sprintf("%d/rules.yaml", gID) {
+				t.Fatalf("expected rule file path to be unchanged; got %q instead", ar.File)
+			}
+		}
+		if rr, ok := r.(*RecordingRule); ok {
+			if rr.File != fmt.Sprintf("%d/rules.yaml", gID) {
+				t.Fatalf("expected rule file path to be unchanged; got %q instead", rr.File)
+			}
+		}
+	}
+}

app/vmalert/rule/recording.go

@@ -293,9 +293,11 @@ func (rr *RecordingRule) toTimeSeries(m datasource.Metric) prompb.TimeSeries {
 	}
 	// add extra labels configured by user
 	for k := range rr.Labels {
-		// do not add label with empty value, since it has no meaning.
-		// see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/9984
+		// do not add label with empty value to the result, as it has no meaning:
+		// if the label already exists in the original query result, remove it to preserve compatibility with relabeling, see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10766.
+		// otherwise, ignore the label, see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/9984.
 		if rr.Labels[k] == "" {
+			m.DelLabel(k)
 			continue
 		}
 		existingLabel := promrelabel.GetLabelByName(m.Labels, k)

app/vmalert/rule/recording_test.go

@@ -163,11 +163,13 @@ func TestRecordingRule_Exec(t *testing.T) {
 	f(&RecordingRule{
 		Name: "job:foo",
 		Labels: map[string]string{
-			"source": "test",
+			"source":      "test",
+			"empty_label": "", // this should be dropped
+			"pod":         "", // this should remove the pod label from query result
 		},
 	}, [][]datasource.Metric{{
-		metricWithValueAndLabels(t, 2, "__name__", "foo", "job", "foo"),
-		metricWithValueAndLabels(t, 1, "__name__", "bar", "job", "bar", "source", "origin"),
+		metricWithValueAndLabels(t, 2, "__name__", "foo", "job", "foo", "pod", "vmalert-0"),
+		metricWithValueAndLabels(t, 1, "__name__", "bar", "job", "bar", "source", "origin", "pod", "vmalert-1"),
 		metricWithValueAndLabels(t, 1, "__name__", "baz", "job", "baz", "source", "test"),
 	}}, [][]prompb.TimeSeries{{
 		newTimeSeries([]float64{2}, []int64{ts.UnixNano()}, []prompb.Label{

app/vmalert/rule/web.go

@@ -252,6 +252,9 @@ func (r *ApiRule) ExtendState() {
 
 // ToAPI returns ApiGroup representation of g
 func (g *Group) ToAPI() *ApiGroup {
+	if g == nil {
+		return &ApiGroup{}
+	}
 	g.mu.RLock()
 	defer g.mu.RUnlock()
 	ag := ApiGroup{

app/vmalert/templates/template.go

@@ -402,6 +402,20 @@ func templateFuncs() textTpl.FuncMap {
 			return t, nil
 		},
 
+		// formatTime formats the given Unix timestamp with the provided layout.
+		// For example: {{ now | formatTime "2006-01-02T15:04:05Z07:00" }}
+		"formatTime": func(layout string, i any) (string, error) {
+			v, err := toFloat64(i)
+			if err != nil {
+				return "", fmt.Errorf("formatTime: %w", err)
+			}
+			if math.IsNaN(v) || math.IsInf(v, 0) {
+				return "", fmt.Errorf("formatTime: cannot convert %v to time", v)
+			}
+			t := timeFromUnixTimestamp(v).Time().UTC()
+			return t.Format(layout), nil
+		},
+
 		/* URLs */
 
 		// externalURL returns value of `external.url` flag

app/vmalert/templates/template_test.go

@@ -6,6 +6,7 @@ import (
 	"strings"
 	"testing"
 	textTpl "text/template"
+	"time"
 )
 
 func TestTemplateFuncs_StringConversion(t *testing.T) {
@@ -103,6 +104,26 @@ func TestTemplateFuncs_Formatting(t *testing.T) {
 	f("humanizeTimestamp", 1679055557, "2023-03-17 12:19:17 +0000 UTC")
 }
 
+func TestTemplateFuncs_FormatTime(t *testing.T) {
+	funcs := templateFuncs()
+	formatTime := funcs["formatTime"].(func(layout string, i any) (string, error))
+
+	f := func(layout string, input any, expected string) {
+		t.Helper()
+		result, err := formatTime(layout, input)
+		if err != nil {
+			t.Fatalf("unexpected error for formatTime(%q, %v): %s", layout, input, err)
+		}
+		if result != expected {
+			t.Fatalf("unexpected result for formatTime(%q, %v); got\n%s\nwant\n%s", layout, input, result, expected)
+		}
+	}
+
+	f(time.RFC3339, float64(1679055557), "2023-03-17T12:19:17Z")
+	f("2006-01-02T15:04:05", int64(1679055557), "2023-03-17T12:19:17")
+	f(time.RFC822, int(1679055557), "17 Mar 23 12:19 UTC")
+}
+
 func mkTemplate(current, replacement any) textTemplate {
 	tmpl := textTemplate{}
 	if current != nil {

app/vmalert/vmalertutil/auth.go

@@ -20,11 +20,12 @@ func AuthConfig(filterOptions ...AuthConfigOptions) (*promauth.Config, error) {
 }
 
 // WithBasicAuth returns AuthConfigOptions and initialized promauth.BasicAuthConfig based on given params
-func WithBasicAuth(username, password, passwordFile string) AuthConfigOptions {
+func WithBasicAuth(username, usernameFile, password, passwordFile string) AuthConfigOptions {
 	return func(config *promauth.HTTPClientConfig) {
-		if username != "" || password != "" || passwordFile != "" {
+		if username != "" || usernameFile != "" || password != "" || passwordFile != "" {
 			config.BasicAuth = &promauth.BasicAuthConfig{
 				Username:     username,
+				UsernameFile: usernameFile,
 				Password:     promauth.NewSecret(password),
 				PasswordFile: passwordFile,
 			}

app/vmalert/web.go

@@ -52,7 +52,13 @@ var (
 		"alert":  rule.TypeAlerting,
 		"record": rule.TypeRecording,
 	}
-	ruleStates = []string{"ok", "nomatch", "inactive", "firing", "pending", "unhealthy"}
+
+	// The "recovering", "noData", "normal", "error" states are used by Grafana.
+	// Ignore "recovering" since it is not currently acknowledged by vmalert,
+	// treat "noData" as an alias for "nomatch",
+	// treat "normal" as an alias for "inactive",
+	// treat "error" as an alias for "unhealthy"
+	ruleStates = []string{"ok", "nomatch", "inactive", "firing", "pending", "unhealthy", "recovering", "noData", "normal", "error"}
 )
 
 type requestHandler struct {
@@ -363,6 +369,15 @@ func newRulesFilter(r *http.Request) (*rulesFilter, *httpserver.ErrorWithStatusC
 			if !slices.Contains(ruleStates, v) {
 				return nil, errResponse(fmt.Errorf(`invalid parameter "state": contains not supported value %q`, v), http.StatusBadRequest)
 			}
+			// Replace grafana states with supported internal states
+			switch v {
+			case "noData":
+				v = "nomatch"
+			case "normal":
+				v = "inactive"
+			case "error":
+				v = "unhealthy"
+			}
 			rf.states = append(rf.states, v)
 		}
 	}

app/vmauth/auth_config.go

@@ -362,40 +362,62 @@ func (up *URLPrefix) setLoadBalancingPolicy(loadBalancingPolicy string) error {
 }
 
 type backendURLs struct {
-	healthChecksContext context.Context
-	healthChecksCancel  func()
-	healthChecksWG      sync.WaitGroup
-
+	bhc backendHealthCheck
 	bus []*backendURL
 }
 
+type backendHealthCheck struct {
+	ctx context.Context
+	// mu protects fields below
+	cancel    func()
+	mu        sync.Mutex
+	isStopped bool
+	wg        sync.WaitGroup
+}
+
+func (bhc *backendHealthCheck) run(hc func()) {
+	bhc.mu.Lock()
+	defer bhc.mu.Unlock()
+	if bhc.isStopped {
+		return
+	}
+	bhc.wg.Go(hc)
+}
+
+func (bhc *backendHealthCheck) stop() {
+	bhc.mu.Lock()
+	bhc.cancel()
+	bhc.isStopped = true
+	bhc.mu.Unlock()
+	bhc.wg.Wait()
+}
+
 func newBackendURLs() *backendURLs {
 	ctx, cancel := context.WithCancel(context.Background())
 	return &backendURLs{
-		healthChecksContext: ctx,
-		healthChecksCancel:  cancel,
+		bhc: backendHealthCheck{
+			ctx:    ctx,
+			cancel: cancel,
+		},
 	}
 }
 
 func (bus *backendURLs) add(u *url.URL) {
 	bus.bus = append(bus.bus, &backendURL{
-		url:                u,
-		healthCheckContext: bus.healthChecksContext,
-		healthCheckWG:      &bus.healthChecksWG,
-		hasPlaceHolders:    hasAnyPlaceholders(u),
+		url:             u,
+		bhc:             &bus.bhc,
+		hasPlaceHolders: hasAnyPlaceholders(u),
 	})
 }
 
 func (bus *backendURLs) stopHealthChecks() {
-	bus.healthChecksCancel()
-	bus.healthChecksWG.Wait()
+	bus.bhc.stop()
 }
 
 type backendURL struct {
 	broken atomic.Bool
 
-	healthCheckContext context.Context
-	healthCheckWG      *sync.WaitGroup
+	bhc *backendHealthCheck
 
 	concurrentRequests atomic.Int32
 
@@ -410,7 +432,7 @@ func (bu *backendURL) isBroken() bool {
 
 func (bu *backendURL) setBroken() {
 	if bu.broken.CompareAndSwap(false, true) {
-		bu.healthCheckWG.Go(func() {
+		bu.bhc.run(func() {
 			bu.runHealthCheck()
 			bu.broken.Store(false)
 		})
@@ -432,11 +454,11 @@ func (bu *backendURL) runHealthCheck() {
 		case <-t.C:
 			// Verify network connectivity via TCP dial before marking backend healthy.
 			// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/9997
-			ctx, cancel := context.WithTimeout(bu.healthCheckContext, time.Second)
+			ctx, cancel := context.WithTimeout(bu.bhc.ctx, time.Second)
 			c, err := netutil.Dialer.DialContext(ctx, "tcp", addr)
 			cancel()
 			if err != nil {
-				if errors.Is(bu.healthCheckContext.Err(), context.Canceled) {
+				if errors.Is(bu.bhc.ctx.Err(), context.Canceled) {
 					return
 				}
 				logger.Warnf("ignoring the backend at %s for %s because of dial error: %s", addr, *failTimeout, err)
@@ -445,7 +467,7 @@ func (bu *backendURL) runHealthCheck() {
 
 			_ = c.Close()
 			return
-		case <-bu.healthCheckContext.Done():
+		case <-bu.bhc.ctx.Done():
 			return
 		}
 	}
@@ -588,6 +610,7 @@ func areEqualBackendURLs(a, b []*backendURL) bool {
 }
 
 // getFirstAvailableBackendURL returns the first available backendURL, which isn't broken.
+// If all backendURLs are broken, then returns the first backendURL.
 //
 // backendURL.put() must be called on the returned backendURL after the request is complete.
 func getFirstAvailableBackendURL(bus []*backendURL) *backendURL {
@@ -606,21 +629,22 @@ func getFirstAvailableBackendURL(bus []*backendURL) *backendURL {
 			return bu
 		}
 	}
-	return nil
+
+	// All backend urls are unavailable, then returning a first one, it could help increase the success rate of the requests。
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10837#issuecomment-4307050980.
+	bu.get()
+	return bu
 }
 
 // getLeastLoadedBackendURL returns a non-broken backendURL with the lowest number of concurrent requests.
+// If all backendURLs are broken, then returns the first backendURL.
 //
 // backendURL.put() must be called on the returned backendURL after the request is complete.
 func getLeastLoadedBackendURL(bus []*backendURL, atomicCounter *atomic.Uint32) *backendURL {
+	firstBu := bus[0]
 	if len(bus) == 1 {
-		// Fast path - return the only backend url.
-		bu := bus[0]
-		if bu.isBroken() {
-			return nil
-		}
-		bu.get()
-		return bu
+		firstBu.get()
+		return firstBu
 	}
 
 	// Slow path - select other backend urls.
@@ -658,7 +682,10 @@ func getLeastLoadedBackendURL(bus []*backendURL, atomicCounter *atomic.Uint32) *
 	}
 	buMin := bus[buMinIdx]
 	if buMin.isBroken() {
-		return nil
+		// If all backendURLs are broken, then returns the first backendURL.
+		// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10837#issuecomment-4307050980.
+		firstBu.get()
+		return firstBu
 	}
 	buMin.get()
 	atomicCounter.CompareAndSwap(n+1, buMinIdx+1)

app/vmauth/auth_config_test.go

@@ -1031,6 +1031,33 @@ func TestLogRequest(t *testing.T) {
 	f("foo", 404, 10*time.Millisecond, `access_log request_host="localhost:8080" request_uri="" status_code=404 remote_addr="" user_agent="" referer="" duration_ms=10 username="foo"`)
 }
 
+func TestGetFirstAvailableBackend(t *testing.T) {
+	f := func(broken []bool, expectedIdx int) {
+		t.Helper()
+		bus := make([]*backendURL, len(broken))
+		for i := range broken {
+			bus[i] = &backendURL{
+				url: &url.URL{Host: fmt.Sprintf("server-%d", i)},
+			}
+			bus[i].broken.Store(broken[i])
+		}
+		bu := getFirstAvailableBackendURL(bus)
+		if bu == nil {
+			t.Fatalf("unexpected nil backend")
+		}
+		if bu.url.Host != fmt.Sprintf("server-%d", expectedIdx) {
+			t.Fatalf("unexpected backend, expected server-%d, got %s", expectedIdx, bu.url.Host)
+		}
+	}
+
+	f([]bool{false, false, false}, 0)
+	f([]bool{true, true, false}, 2)
+	// all backend are broken, then return the first one.
+	f([]bool{true, true, true}, 0)
+	f([]bool{true}, 0)
+
+}
+
 func getRegexs(paths []string) []*Regex {
 	var sps []*Regex
 	for _, path := range paths {

app/vmauth/example_config.yml

@@ -130,6 +130,16 @@ users:
   - "http://vmselect1:8481/select/{{.MetricsTenant}}/prometheus"
   - "http://vmselect2:8481/select/{{.MetricsTenant}}/prometheus"
 
+# JWT-based routing using header-based tenant identification (VictoriaMetrics cluster)
+# The AccountID and ProjectID from JWT vm_access claims are injected as HTTP headers.
+- name: jwt-header-tenant
+  jwt:
+    skip_verify: true
+  headers:
+  - "AccountID: {{.MetricsAccountID}}"
+  - "ProjectID: {{.MetricsProjectID}}"
+  url_prefix: "http://vminsert:8480/insert/prometheus"
+
 # Requests without Authorization header are proxied according to `unauthorized_user` section.
 # Requests are proxied in round-robin fashion between `url_prefix` backends.
 # The deny_partial_response query arg is added to all the proxied requests.

app/vmauth/jwt.go

@@ -17,6 +17,8 @@ import (
 
 const (
 	metricsTenantPlaceholder       = `{{.MetricsTenant}}`
+	metricsAccountIDPlaceholder    = `{{.MetricsAccountID}}`
+	metricsProjectIDPlaceholder    = `{{.MetricsProjectID}}`
 	metricsExtraLabelsPlaceholder  = `{{.MetricsExtraLabels}}`
 	metricsExtraFiltersPlaceholder = `{{.MetricsExtraFilters}}`
 
@@ -30,6 +32,8 @@ const (
 
 var allPlaceholders = []string{
 	metricsTenantPlaceholder,
+	metricsAccountIDPlaceholder,
+	metricsProjectIDPlaceholder,
 	metricsExtraLabelsPlaceholder,
 	metricsExtraFiltersPlaceholder,
 	logsAccountIDPlaceholder,
@@ -40,6 +44,8 @@ var allPlaceholders = []string{
 
 var urlPathPlaceHolders = []string{
 	metricsTenantPlaceholder,
+	metricsAccountIDPlaceholder,
+	metricsProjectIDPlaceholder,
 	logsAccountIDPlaceholder,
 	logsProjectIDPlaceholder,
 }
@@ -371,6 +377,8 @@ func jwtClaimsData(vma *jwt.VMAccessClaim) map[string][]string {
 	data := map[string][]string{
 		// TODO: optimize at parsing stage
 		metricsTenantPlaceholder:       {fmt.Sprintf("%d:%d", vma.MetricsAccountID, vma.MetricsProjectID)},
+		metricsAccountIDPlaceholder:    {fmt.Sprintf("%d", vma.MetricsAccountID)},
+		metricsProjectIDPlaceholder:    {fmt.Sprintf("%d", vma.MetricsProjectID)},
 		metricsExtraLabelsPlaceholder:  vma.MetricsExtraLabels,
 		metricsExtraFiltersPlaceholder: vma.MetricsExtraFilters,
 

app/vmauth/jwt_test.go

@@ -170,13 +170,13 @@ users:
   url_prefix: http://foo.bar
 `, "cannot parse public key from file \""+publicKeyFile+"\": failed to parse key \"invalidPEM\": failed to decode PEM block containing public key")
 
-	// unsupported placeholder in a header
+	// unsupported placeholder in a URL path
 	f(`
 users:
 - jwt: 
     skip_verify: true
   url_prefix: http://foo.bar/{{.UnsupportedPlaceholder}}/foo`,
-		"invalid placeholder found in URL request path: \"/{{.UnsupportedPlaceholder}}/foo\", supported values are: {{.MetricsTenant}}, {{.MetricsExtraLabels}}, {{.MetricsExtraFilters}}, {{.LogsAccountID}}, {{.LogsProjectID}}, {{.LogsExtraFilters}}, {{.LogsExtraStreamFilters}}",
+		"invalid placeholder found in URL request path: \"/{{.UnsupportedPlaceholder}}/foo\", supported values are: {{.MetricsTenant}}, {{.MetricsAccountID}}, {{.MetricsProjectID}}, {{.MetricsExtraLabels}}, {{.MetricsExtraFilters}}, {{.LogsAccountID}}, {{.LogsProjectID}}, {{.LogsExtraFilters}}, {{.LogsExtraStreamFilters}}",
 	)
 	// unsupported placeholder in a header
 	f(`
@@ -187,7 +187,7 @@ users:
     - "AccountID: {{.UnsupportedPlaceholder}}"
   url_prefix: http://foo.bar
 `,
-		"request header: \"AccountID\" has unsupported placeholder: \"{{.UnsupportedPlaceholder}}\", supported values are: {{.MetricsTenant}}, {{.MetricsExtraLabels}}, {{.MetricsExtraFilters}}, {{.LogsAccountID}}, {{.LogsProjectID}}, {{.LogsExtraFilters}}, {{.LogsExtraStreamFilters}}",
+		"request header: \"AccountID\" has unsupported placeholder: \"{{.UnsupportedPlaceholder}}\", supported values are: {{.MetricsTenant}}, {{.MetricsAccountID}}, {{.MetricsProjectID}}, {{.MetricsExtraLabels}}, {{.MetricsExtraFilters}}, {{.LogsAccountID}}, {{.LogsProjectID}}, {{.LogsExtraFilters}}, {{.LogsExtraStreamFilters}}",
 	)
 
 	// spaces in templating not allowed
@@ -199,7 +199,19 @@ users:
     - "AccountID: {{ .LogsAccountID }}"
   url_prefix: http://foo.bar
 `,
-		"request header: \"AccountID\" has unsupported placeholder: \"{{ .LogsAccountID }}\", supported values are: {{.MetricsTenant}}, {{.MetricsExtraLabels}}, {{.MetricsExtraFilters}}, {{.LogsAccountID}}, {{.LogsProjectID}}, {{.LogsExtraFilters}}, {{.LogsExtraStreamFilters}}",
+		"request header: \"AccountID\" has unsupported placeholder: \"{{ .LogsAccountID }}\", supported values are: {{.MetricsTenant}}, {{.MetricsAccountID}}, {{.MetricsProjectID}}, {{.MetricsExtraLabels}}, {{.MetricsExtraFilters}}, {{.LogsAccountID}}, {{.LogsProjectID}}, {{.LogsExtraFilters}}, {{.LogsExtraStreamFilters}}",
+	)
+
+	// placeholder must match the entire header value
+	f(`
+users:
+- jwt: 
+    skip_verify: true
+  headers:
+    - "AccountID: foo {{.MetricsAccountID}}"
+  url_prefix: http://foo.bar
+`,
+		"request header: \"AccountID\" has unsupported placeholder: \"foo {{.MetricsAccountID}}\", supported values are: {{.MetricsTenant}}, {{.MetricsAccountID}}, {{.MetricsProjectID}}, {{.MetricsExtraLabels}}, {{.MetricsExtraFilters}}, {{.LogsAccountID}}, {{.LogsProjectID}}, {{.LogsExtraFilters}}, {{.LogsExtraStreamFilters}}",
 	)
 
 	// oidc is not an object
@@ -364,10 +376,25 @@ users:
   url_prefix: http://foo.bar
 `, validRSAPublicKey, validECDSAPublicKey))
 
+	// metrics header placeholders
 	f(`
 users:
 - jwt:
     skip_verify: true
+  headers:
+  - "MetricsAccountID: {{.MetricsAccountID}}"
+  - "MetricsProjectID: {{.MetricsProjectID}}"
+  url_prefix: http://foo.bar
+`)
+
+	// logs header placeholders
+	f(`
+users:
+- jwt:
+    skip_verify: true
+  headers:
+  - "LogsAccountID: {{.LogsAccountID}}"
+  - "LogsProjectID: {{.LogsProjectID}}"
   url_prefix: http://foo.bar
 `)
 

app/vmauth/main.go

@@ -51,7 +51,7 @@ var (
 		"This allows reducing the consumption of backend resources when processing requests from clients connected via slow networks. "+
 		"Set to 0 to disable request buffering. See https://docs.victoriametrics.com/victoriametrics/vmauth/#request-body-buffering")
 	maxRequestBodySizeToRetry = flagutil.NewBytes("maxRequestBodySizeToRetry", 16*1024, "The maximum request body size to buffer in memory for potential retries at other backends. "+
-		"Request bodies larger than this size cannot be retried if the backend fails. Zero or negative value disables request body buffering and retries. "+
+		"Request bodies larger than this size cannot be retried if the backend fails. Zero or negative value disables retries. "+
 		"See also -requestBufferSize")
 
 	maxConcurrentRequests = flag.Int("maxConcurrentRequests", 1000, "The maximum number of concurrent requests vmauth can process simultaneously. "+
@@ -357,6 +357,7 @@ func bufferRequestBody(ctx context.Context, r io.ReadCloser, userName string) (i
 
 	maxBufSize := max(requestBufferSize.IntN(), maxRequestBodySizeToRetry.IntN())
 	if maxBufSize <= 0 {
+		// Request buffering is disabled.
 		return r, nil
 	}
 
@@ -480,6 +481,9 @@ func tryProcessingRequest(w http.ResponseWriter, r *http.Request, targetURL *url
 	canRetry := !bbOK || bb.canRetry()
 
 	res, err := ui.rt.RoundTrip(req)
+	if err == nil {
+		defer func() { _ = res.Body.Close() }()
+	}
 
 	if errors.Is(r.Context().Err(), context.Canceled) {
 		// Do not retry canceled requests.
@@ -549,7 +553,6 @@ func tryProcessingRequest(w http.ResponseWriter, r *http.Request, targetURL *url
 	w.WriteHeader(res.StatusCode)
 
 	err = copyStreamToClient(w, res.Body)
-	_ = res.Body.Close()
 
 	if errors.Is(r.Context().Err(), context.Canceled) {
 		// Do not retry canceled requests.
@@ -763,7 +766,7 @@ var concurrentRequestsLimitReached = metrics.NewCounter("vmauth_concurrent_reque
 
 func usage() {
 	const s = `
-vmauth authenticates and authorizes incoming requests and proxies them to VictoriaMetrics.
+vmauth authenticates and authorizes incoming requests and proxies them to VictoriaMetrics components or any other HTTP backends.
 
 See the docs at https://docs.victoriametrics.com/victoriametrics/vmauth/ .
 `
@@ -792,10 +795,11 @@ func handleConcurrencyLimitError(w http.ResponseWriter, r *http.Request, err err
 }
 
 // bufferedBody serves two purposes:
-//  1. Enables request retries when the body size does not exceed maxBodySize
-//     by fully buffering the body in memory.
-//  2. Prevents slow clients from reducing effective server capacity by
-//     buffering the request body before acquiring a per-user concurrency slot.
+//
+//  1. It enables request retries when the request body size does not exceed maxBufSize
+//     by fully buffering the request body in memory.
+//  2. It prevents slow clients from reducing effective server capacity
+//     by buffering the request body before acquiring a per-user concurrency slot.
 //
 // See bufferRequestBody for details on how bufferedBody is used.
 type bufferedBody struct {
@@ -819,7 +823,7 @@ func newBufferedBody(r io.ReadCloser, buf []byte, maxBufSize int) *bufferedBody
 	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8051
 
 	if len(buf) < maxBufSize {
-		// Read the full request body into buf.
+		// The full request body has been already read into buf.
 		r = nil
 	}
 
@@ -832,7 +836,7 @@ func newBufferedBody(r io.ReadCloser, buf []byte, maxBufSize int) *bufferedBody
 // Read implements io.Reader interface.
 func (bb *bufferedBody) Read(p []byte) (int, error) {
 	if bb.cannotRetry {
-		return 0, fmt.Errorf("cannot read already closed body")
+		return 0, fmt.Errorf("cannot read already closed request body")
 	}
 	if bb.bufOffset < len(bb.buf) {
 		n := copy(p, bb.buf[bb.bufOffset:])
@@ -846,14 +850,18 @@ func (bb *bufferedBody) Read(p []byte) (int, error) {
 }
 
 func (bb *bufferedBody) canRetry() bool {
-	return bb.r == nil
+	if bb.r != nil {
+		return false
+	}
+	maxRetrySize := maxRequestBodySizeToRetry.IntN()
+	return len(bb.buf) == 0 || (maxRetrySize > 0 && len(bb.buf) <= maxRetrySize)
 }
 
 // Close implements io.Closer interface.
 func (bb *bufferedBody) Close() error {
 	bb.resetReader()
+	bb.cannotRetry = !bb.canRetry()
 	if bb.r != nil {
-		bb.cannotRetry = true
 		return bb.r.Close()
 	}
 	return nil

app/vmauth/main_test.go

@@ -19,6 +19,7 @@ import (
 	"os"
 	"path/filepath"
 	"sort"
+	"strconv"
 	"strings"
 	"sync/atomic"
 	"testing"
@@ -850,6 +851,30 @@ users:
 		responseExpected,
 	)
 
+	// test header injection and URL templating with individual placeholders
+	request = httptest.NewRequest(`GET`, "http://some-host.com/api/v1/query", nil)
+	request.Header.Set(`Authorization`, `Bearer `+fullToken)
+	responseExpected = `
+statusCode=200
+path: /select/123/234/api/v1/query
+query:
+headers:
+    AccountID=123
+    ProjectID=234`
+	f(fmt.Sprintf(
+		`
+users:
+- jwt:
+    public_keys:
+    - %q
+  url_prefix: {BACKEND}/select/{{.MetricsAccountID}}/{{.MetricsProjectID}}
+  headers:
+  - "AccountID: {{.MetricsAccountID}}"
+  - "ProjectID: {{.MetricsProjectID}}"`, string(publicKeyPEM)),
+		request,
+		responseExpected,
+	)
+
 	// extra_label and extra_filters from vm_access claim merged with statically defined
 	request = httptest.NewRequest(`GET`, "http://some-host.com/api/v1/query", nil)
 	request.Header.Set(`Authorization`, `Bearer `+fullToken)
@@ -1831,7 +1856,7 @@ func (r *mockBody) Read(p []byte) (n int, err error) {
 }
 
 func TestBufferedBody_RetrySuccess(t *testing.T) {
-	f := func(s string, maxBodySize int) {
+	f := func(s string, maxSizeToRetry, bufferSize int) {
 		t.Helper()
 
 		defaultRequestBufferSize := requestBufferSize.String()
@@ -1840,7 +1865,7 @@ func TestBufferedBody_RetrySuccess(t *testing.T) {
 				t.Fatalf("cannot reset requestBufferSize: %s", err)
 			}
 		}()
-		if err := requestBufferSize.Set(fmt.Sprintf("%d", maxBodySize)); err != nil {
+		if err := requestBufferSize.Set(strconv.Itoa(bufferSize)); err != nil {
 			t.Fatalf("cannot set requestBufferSize: %s", err)
 		}
 
@@ -1850,7 +1875,7 @@ func TestBufferedBody_RetrySuccess(t *testing.T) {
 				t.Fatalf("cannot reset maxRequestBodySizeToRetry: %s", err)
 			}
 		}()
-		if err := maxRequestBodySizeToRetry.Set("0"); err != nil {
+		if err := maxRequestBodySizeToRetry.Set(strconv.Itoa(maxSizeToRetry)); err != nil {
 			t.Fatalf("cannot set maxRequestBodySizeToRetry: %s", err)
 		}
 
@@ -1879,16 +1904,20 @@ func TestBufferedBody_RetrySuccess(t *testing.T) {
 		}
 	}
 
-	f("", 0)
-	f("", -1)
-	f("", 100)
-	f("foo", 100)
-	f("foobar", 100)
-	f(newTestString(1000), 1001)
+	f("", 0, 2000)
+	f("", 0, 0)
+	f("", -1, 2000)
+	f("", 100, 2000)
+	f("foo", 100, 2000)
+	f("foobar", 100, 2000)
+	f("foobar", 100, 0)
+	f("foobar", 100, -1)
+	f(newTestString(1000), 1001, 2000)
+	f(newTestString(1000), 1001, 500)
 }
 
 func TestBufferedBody_RetrySuccessPartialRead(t *testing.T) {
-	f := func(s string, maxBodySize int) {
+	f := func(s string, maxSizeToRetry, bufferSize int) {
 		t.Helper()
 
 		// Check the case with partial read
@@ -1898,7 +1927,7 @@ func TestBufferedBody_RetrySuccessPartialRead(t *testing.T) {
 				t.Fatalf("cannot reset requestBufferSize: %s", err)
 			}
 		}()
-		if err := requestBufferSize.Set(fmt.Sprintf("%d", maxBodySize)); err != nil {
+		if err := requestBufferSize.Set(strconv.Itoa(bufferSize)); err != nil {
 			t.Fatalf("cannot set requestBufferSize: %s", err)
 		}
 
@@ -1908,7 +1937,7 @@ func TestBufferedBody_RetrySuccessPartialRead(t *testing.T) {
 				t.Fatalf("cannot reset maxRequestBodySizeToRetry: %s", err)
 			}
 		}()
-		if err := maxRequestBodySizeToRetry.Set("0"); err != nil {
+		if err := maxRequestBodySizeToRetry.Set(strconv.Itoa(maxSizeToRetry)); err != nil {
 			t.Fatalf("cannot set maxRequestBodySizeToRetry: %s", err)
 		}
 
@@ -1952,16 +1981,20 @@ func TestBufferedBody_RetrySuccessPartialRead(t *testing.T) {
 		}
 	}
 
-	f("", 0)
-	f("", -1)
-	f("", 100)
-	f("foo", 100)
-	f("foobar", 100)
-	f(newTestString(1000), 1001)
+	f("", 0, 2000)
+	f("", 0, 0)
+	f("", -1, 2000)
+	f("", 100, 2000)
+	f("foo", 100, 2000)
+	f("foobar", 100, 2000)
+	f("foobar", 100, 0)
+	f("foobar", 100, -1)
+	f(newTestString(1000), 1001, 2000)
+	f(newTestString(1000), 1001, 500)
 }
 
 func TestBufferedBody_RetryFailureTooBigBody(t *testing.T) {
-	f := func(s string, maxBodySize int) {
+	f := func(s string, maxSizeToRetry, bufferSize int) {
 		t.Helper()
 
 		defaultRequestBufferSize := requestBufferSize.String()
@@ -1970,7 +2003,7 @@ func TestBufferedBody_RetryFailureTooBigBody(t *testing.T) {
 				t.Fatalf("cannot reset requestBufferSize: %s", err)
 			}
 		}()
-		if err := requestBufferSize.Set("0"); err != nil {
+		if err := requestBufferSize.Set(strconv.Itoa(bufferSize)); err != nil {
 			t.Fatalf("cannot set requestBufferSize: %s", err)
 		}
 
@@ -1980,7 +2013,7 @@ func TestBufferedBody_RetryFailureTooBigBody(t *testing.T) {
 				t.Fatalf("cannot reset maxRequestBodySizeToRetry: %s", err)
 			}
 		}()
-		if err := maxRequestBodySizeToRetry.Set(fmt.Sprintf("%d", maxBodySize)); err != nil {
+		if err := maxRequestBodySizeToRetry.Set(strconv.Itoa(maxSizeToRetry)); err != nil {
 			t.Fatalf("cannot set maxRequestBodySizeToRetry: %s", err)
 		}
 
@@ -2025,12 +2058,17 @@ func TestBufferedBody_RetryFailureTooBigBody(t *testing.T) {
 	}
 
 	const maxBodySize = 1000
-	f(newTestString(maxBodySize+1), maxBodySize)
-	f(newTestString(2*maxBodySize), maxBodySize)
+	f(newTestString(maxBodySize+1), 0, 2*maxBodySize)
+	f(newTestString(maxBodySize+1), -1, 2*maxBodySize)
+	f(newTestString(maxBodySize+1), maxBodySize, 0)
+	f(newTestString(maxBodySize+1), maxBodySize, -1)
+	f(newTestString(maxBodySize+1), maxBodySize, maxBodySize)
+	f(newTestString(maxBodySize+1), maxBodySize, 2*maxBodySize)
+	f(newTestString(2*maxBodySize), maxBodySize, 0)
 }
 
-func TestBufferedBody_RetryFailureZeroOrNegativeMaxBodySize(t *testing.T) {
-	f := func(s string, maxBodySize int) {
+func TestBufferedBody_RetryDisabledByMaxRequestBodySizeToRetry(t *testing.T) {
+	f := func(s string, maxSizeToRetry, bufferSize int) {
 		t.Helper()
 
 		defaultRequestBufferSize := requestBufferSize.String()
@@ -2039,10 +2077,20 @@ func TestBufferedBody_RetryFailureZeroOrNegativeMaxBodySize(t *testing.T) {
 				t.Fatalf("cannot reset requestBufferSize: %s", err)
 			}
 		}()
-		if err := requestBufferSize.Set(fmt.Sprintf("%d", maxBodySize)); err != nil {
+		if err := requestBufferSize.Set(strconv.Itoa(bufferSize)); err != nil {
 			t.Fatalf("cannot set requestBufferSize: %s", err)
 		}
 
+		defaultMaxRequestBodySizeToRetry := maxRequestBodySizeToRetry.String()
+		defer func() {
+			if err := maxRequestBodySizeToRetry.Set(defaultMaxRequestBodySizeToRetry); err != nil {
+				t.Fatalf("cannot reset maxRequestBodySizeToRetry: %s", err)
+			}
+		}()
+		if err := maxRequestBodySizeToRetry.Set(strconv.Itoa(maxSizeToRetry)); err != nil {
+			t.Fatalf("cannot set maxRequestBodySizeToRetry: %s", err)
+		}
+
 		ctx := context.Background()
 		rb, err := bufferRequestBody(ctx, io.NopCloser(bytes.NewBufferString(s)), "foo")
 		if err != nil {
@@ -2051,8 +2099,8 @@ func TestBufferedBody_RetryFailureZeroOrNegativeMaxBodySize(t *testing.T) {
 		bb, ok := rb.(*bufferedBody)
 		canRetry := !ok || bb.canRetry()
 
-		if !canRetry {
-			t.Fatalf("canRetry() must return true before reading anything")
+		if canRetry {
+			t.Fatalf("canRetry() must return false before reading anything")
 		}
 		data, err := io.ReadAll(rb)
 		if err != nil {
@@ -2066,19 +2114,19 @@ func TestBufferedBody_RetryFailureZeroOrNegativeMaxBodySize(t *testing.T) {
 		}
 
 		data, err = io.ReadAll(rb)
-		if err != nil {
-			t.Fatalf("unexpected error in io.ReadAll: %s", err)
+		if err == nil {
+			t.Fatalf("expecting non-nil error")
 		}
-		if string(data) != s {
-			t.Fatalf("unexpected data read\ngot\n%s\nwant\n%s", data, s)
+		if len(data) != 0 {
+			t.Fatalf("unexpected non-empty data read: %q", data)
 		}
 	}
 
-	f("foobar", 0)
-	f(newTestString(1000), 0)
+	f("foobar", 0, 2048)
+	f(newTestString(1000), 0, 2048)
 
-	f("foobar", -1)
-	f(newTestString(1000), -1)
+	f("foobar", -1, 2048)
+	f(newTestString(1000), -1, 2048)
 }
 
 func newTestString(sLen int) string {

app/vmbackup/main.go

@@ -21,6 +21,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/procutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/pushmetrics"
+
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/snapshot"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/snapshot/snapshotutil"
 )

app/vmctl/flags.go

@@ -416,6 +416,16 @@ const (
 	promTemporaryDirPath = "prom-tmp-dir-path"
 )
 
+const (
+	thanosSnapshot         = "thanos-snapshot"
+	thanosConcurrency      = "thanos-concurrency"
+	thanosFilterTimeStart  = "thanos-filter-time-start"
+	thanosFilterTimeEnd    = "thanos-filter-time-end"
+	thanosFilterLabel      = "thanos-filter-label"
+	thanosFilterLabelValue = "thanos-filter-label-value"
+	thanosAggrTypes        = "thanos-aggr-types"
+)
+
 var (
 	promFlags = []cli.Flag{
 		&cli.StringFlag{
@@ -451,6 +461,43 @@ var (
 			Value: os.TempDir(),
 		},
 	}
+
+	thanosFlags = []cli.Flag{
+		&cli.StringFlag{
+			Name:     thanosSnapshot,
+			Usage:    "Path to Thanos snapshot directory containing raw and/or downsampled blocks.",
+			Required: true,
+		},
+		&cli.IntFlag{
+			Name:  thanosConcurrency,
+			Usage: "Number of concurrently running snapshot readers",
+			Value: 1,
+		},
+		&cli.StringFlag{
+			Name:  thanosFilterTimeStart,
+			Usage: "The time filter in RFC3339 format to select timeseries with timestamp equal or higher than provided value. E.g. '2020-01-01T20:07:00Z'",
+		},
+		&cli.StringFlag{
+			Name:  thanosFilterTimeEnd,
+			Usage: "The time filter in RFC3339 format to select timeseries with timestamp equal or lower than provided value. E.g. '2020-01-01T20:07:00Z'",
+		},
+		&cli.StringFlag{
+			Name:  thanosFilterLabel,
+			Usage: "Thanos label name to filter timeseries by. E.g. '__name__' will filter timeseries by name.",
+		},
+		&cli.StringFlag{
+			Name:  thanosFilterLabelValue,
+			Usage: fmt.Sprintf("Thanos regular expression to filter label from %q flag.", thanosFilterLabel),
+			Value: ".*",
+		},
+		&cli.StringSliceFlag{
+			Name: thanosAggrTypes,
+			Usage: "Aggregate types to import from Thanos downsampled blocks. Supported values: count, sum, min, max, counter. " +
+				"Each aggregate will be imported as a separate metric with the aggregate type as suffix (e.g., metric_name:5m:count). " +
+				"If not specified, all aggregate types will be imported from downsampled blocks.",
+			Value: nil,
+		},
+	}
 )
 
 const (

app/vmctl/main.go

@@ -27,6 +27,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmctl/influx"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmctl/opentsdb"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmctl/prometheus"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmctl/thanos"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmctl/vm"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/buildinfo"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
@@ -285,6 +286,7 @@ func main() {
 					if err != nil {
 						return fmt.Errorf("failed to create prometheus client: %s", err)
 					}
+
 					pp := prometheusProcessor{
 						cl:        cl,
 						im:        importer,
@@ -294,6 +296,59 @@ func main() {
 					return pp.run(ctx)
 				},
 			},
+			{
+				Name:   "thanos",
+				Usage:  "Migrate time series from Thanos blocks (supports raw and downsampled data)",
+				Flags:  mergeFlags(globalFlags, thanosFlags, vmFlags),
+				Before: beforeFn,
+				Action: func(c *cli.Context) error {
+					fmt.Println("Thanos import mode")
+
+					vmCfg, err := initConfigVM(c)
+					if err != nil {
+						return fmt.Errorf("failed to init VM configuration: %s", err)
+					}
+
+					importer, err = vm.NewImporter(ctx, vmCfg)
+					if err != nil {
+						return fmt.Errorf("failed to create VM importer: %s", err)
+					}
+
+					thanosCfg := thanos.Config{
+						Snapshot: c.String(thanosSnapshot),
+						Filter: thanos.Filter{
+							TimeMin:    c.String(thanosFilterTimeStart),
+							TimeMax:    c.String(thanosFilterTimeEnd),
+							Label:      c.String(thanosFilterLabel),
+							LabelValue: c.String(thanosFilterLabelValue),
+						},
+					}
+					cl, err := thanos.NewClient(thanosCfg)
+					if err != nil {
+						return fmt.Errorf("failed to create thanos client: %s", err)
+					}
+
+					var aggrTypes []thanos.AggrType
+					if aggrTypesStr := c.StringSlice(thanosAggrTypes); len(aggrTypesStr) > 0 {
+						for _, typeStr := range aggrTypesStr {
+							aggrType, err := thanos.ParseAggrType(typeStr)
+							if err != nil {
+								return fmt.Errorf("failed to parse aggregate type %q: %s", typeStr, err)
+							}
+							aggrTypes = append(aggrTypes, aggrType)
+						}
+					}
+
+					tp := thanosProcessor{
+						cl:        cl,
+						im:        importer,
+						cc:        c.Int(thanosConcurrency),
+						isVerbose: c.Bool(globalVerbose),
+						aggrTypes: aggrTypes,
+					}
+					return tp.run(ctx)
+				},
+			},
 			{
 				Name:   "vm-native",
 				Usage:  "Migrate time series between VictoriaMetrics installations",

app/vmctl/opentsdb.go

@@ -8,10 +8,10 @@ import (
 	"time"
 
 	vmetrics "github.com/VictoriaMetrics/metrics"
+	"github.com/cheggaaa/pb/v3"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmctl/opentsdb"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmctl/vm"
-	"github.com/cheggaaa/pb/v3"
 )
 
 type otsdbProcessor struct {
@@ -89,9 +89,6 @@ func (op *otsdbProcessor) run(ctx context.Context) error {
 		// we're going to make serieslist * queryRanges queries, so we should represent that in the progress bar
 		otsdbSeriesTotal.Add(len(serieslist) * queryRanges)
 		bar := pb.StartNew(len(serieslist) * queryRanges)
-		defer func(bar *pb.ProgressBar) {
-			bar.Finish()
-		}(bar)
 		var wg sync.WaitGroup
 		for range op.otsdbcc {
 			wg.Go(func() {
@@ -106,41 +103,22 @@ func (op *otsdbProcessor) run(ctx context.Context) error {
 				}
 			})
 		}
-		/*
-			Loop through all series for this metric, processing all retentions and time ranges
-			requested. This loop is our primary "collect data from OpenTSDB loop" and should
-			be async, sending data to VictoriaMetrics over time.
+		runErr := op.sendQueries(ctx, serieslist, seriesCh, errCh, startTime)
 
-			The idea with having the select at the inner-most loop is to ensure quick
-			short-circuiting on error.
-		*/
-		for _, series := range serieslist {
-			for _, rt := range op.oc.Retentions {
-				for _, tr := range rt.QueryRanges {
-					select {
-					case otsdbErr := <-errCh:
-						return fmt.Errorf("opentsdb error: %s", otsdbErr)
-					case vmErr := <-op.im.Errors():
-						otsdbErrorsTotal.Inc()
-						return fmt.Errorf("import process failed: %s", wrapErr(vmErr, op.isVerbose))
-					case seriesCh <- queryObj{
-						Tr: tr, StartTime: startTime,
-						Series: series, Rt: opentsdb.RetentionMeta{
-							FirstOrder: rt.FirstOrder, SecondOrder: rt.SecondOrder, AggTime: rt.AggTime}}:
-					}
-				}
-			}
-		}
-
-		// Drain channels per metric
+		// Always drain channels and wait for workers to prevent goroutine leaks
 		close(seriesCh)
 		wg.Wait()
 		close(errCh)
 		// check for any lingering errors on the query side
 		for otsdbErr := range errCh {
-			return fmt.Errorf("import process failed: \n%s", otsdbErr)
+			if runErr == nil {
+				runErr = fmt.Errorf("import process failed: \n%s", otsdbErr)
+			}
 		}
 		bar.Finish()
+		if runErr != nil {
+			return runErr
+		}
 		log.Print(op.im.Stats())
 	}
 	op.im.Close()
@@ -155,6 +133,34 @@ func (op *otsdbProcessor) run(ctx context.Context) error {
 	return nil
 }
 
+// sendQueries iterates over all series and retention ranges, sending queries to workers.
+// It returns early if ctx is canceled or an error is received.
+func (op *otsdbProcessor) sendQueries(ctx context.Context, serieslist []opentsdb.Meta, seriesCh chan<- queryObj, errCh <-chan error, startTime int64) error {
+	for _, series := range serieslist {
+		for _, rt := range op.oc.Retentions {
+			for _, tr := range rt.QueryRanges {
+				select {
+				case <-ctx.Done():
+					return fmt.Errorf("context canceled: %s", ctx.Err())
+				case otsdbErr := <-errCh:
+					otsdbErrorsTotal.Inc()
+					return fmt.Errorf("opentsdb error: %s", otsdbErr)
+				case vmErr := <-op.im.Errors():
+					return fmt.Errorf("import process failed: %s", wrapErr(vmErr, op.isVerbose))
+				case seriesCh <- queryObj{
+					Tr: tr, StartTime: startTime,
+					Series: series, Rt: opentsdb.RetentionMeta{
+						FirstOrder:  rt.FirstOrder,
+						SecondOrder: rt.SecondOrder,
+						AggTime:     rt.AggTime,
+					}}:
+				}
+			}
+		}
+	}
+	return nil
+}
+
 func (op *otsdbProcessor) do(s queryObj) error {
 	start := s.StartTime - s.Tr.Start
 	end := s.StartTime - s.Tr.End
@@ -163,6 +169,7 @@ func (op *otsdbProcessor) do(s queryObj) error {
 		return fmt.Errorf("failed to collect data for %v in %v:%v :: %v", s.Series, s.Rt, s.Tr, err)
 	}
 	if len(data.Timestamps) < 1 || len(data.Values) < 1 {
+		log.Printf("no data found for %v in %v:%v...skipping", s.Series, s.Rt, s.Tr)
 		return nil
 	}
 	labels := make([]vm.LabelPair, 0, len(data.Tags))

app/vmctl/opentsdb/opentsdb.go

@@ -108,10 +108,10 @@ func (c Client) FindMetrics(q string) ([]string, error) {
 	if err != nil {
 		return nil, fmt.Errorf("failed to send GET request to %q: %s", q, err)
 	}
+	defer func() { _ = resp.Body.Close() }()
 	if resp.StatusCode != 200 {
 		return nil, fmt.Errorf("bad return from OpenTSDB: %d: %v", resp.StatusCode, resp)
 	}
-	defer func() { _ = resp.Body.Close() }()
 	body, err := io.ReadAll(resp.Body)
 	if err != nil {
 		return nil, fmt.Errorf("could not retrieve metric data from %q: %s", q, err)
@@ -130,12 +130,12 @@ func (c Client) FindSeries(metric string) ([]Meta, error) {
 	q := fmt.Sprintf("%s/api/search/lookup?m=%s&limit=%d", c.Addr, metric, c.Limit)
 	resp, err := c.c.Get(q)
 	if err != nil {
-		return nil, fmt.Errorf("failed to set GET request to %q: %s", q, err)
+		return nil, fmt.Errorf("failed to send GET request to %q: %s", q, err)
 	}
+	defer func() { _ = resp.Body.Close() }()
 	if resp.StatusCode != 200 {
 		return nil, fmt.Errorf("bad return from OpenTSDB: %d: %v", resp.StatusCode, resp)
 	}
-	defer func() { _ = resp.Body.Close() }()
 	body, err := io.ReadAll(resp.Body)
 	if err != nil {
 		return nil, fmt.Errorf("could not retrieve series data from %q: %s", q, err)
@@ -185,6 +185,7 @@ func (c Client) GetData(series Meta, rt RetentionMeta, start int64, end int64, m
 	if err != nil {
 		return Metric{}, fmt.Errorf("failed to send GET request to %q: %s", q, err)
 	}
+	defer func() { _ = resp.Body.Close() }()
 	/*
 		There are three potential failures here, none of which should kill the entire
 		migration run:
@@ -196,7 +197,6 @@ func (c Client) GetData(series Meta, rt RetentionMeta, start int64, end int64, m
 		log.Printf("bad response code from OpenTSDB query %v for %q...skipping", resp.StatusCode, q)
 		return Metric{}, nil
 	}
-	defer func() { _ = resp.Body.Close() }()
 	body, err := io.ReadAll(resp.Body)
 	if err != nil {
 		log.Println("couldn't read response body from OpenTSDB query...skipping")
@@ -239,27 +239,20 @@ func (c Client) GetData(series Meta, rt RetentionMeta, start int64, end int64, m
 		In all "bad" cases, we don't end the migration, we just don't process that particular message
 	*/
 	if len(output) < 1 {
-		// no results returned...return an empty object without error
 		return Metric{}, nil
 	}
 	if len(output) > 1 {
-		// multiple series returned for a single query. We can't process this right, so...
-		return Metric{}, nil
+		return Metric{}, fmt.Errorf("unexpected number of series returned: %d for query %q; expected 1", len(output), q)
 	}
 	if len(output[0].AggregateTags) > 0 {
-		// This failure means we've suppressed potential series somehow...
-		return Metric{}, nil
+		return Metric{}, fmt.Errorf("aggregate tags %v present in response for query %q; series may be suppressed", output[0].AggregateTags, q)
 	}
 	data := Metric{}
 	data.Metric = output[0].Metric
 	data.Tags = output[0].Tags
-	/*
-		We evaluate data for correctness before formatting the actual values
-		to skip a little bit of time if the series has invalid formatting
-	*/
 	data, err = modifyData(data, c.Normalize)
 	if err != nil {
-		return Metric{}, nil
+		return Metric{}, fmt.Errorf("failed to convert metric data for query %q: %w", q, err)
 	}
 
 	/*

app/vmctl/opentsdb/parser.go

@@ -32,7 +32,7 @@ func convertDuration(duration string) (time.Duration, error) {
 	var err error
 	var timeValue int
 	if strings.HasSuffix(duration, "y") {
-		timeValue, err = strconv.Atoi(strings.Trim(duration, "y"))
+		timeValue, err = strconv.Atoi(strings.TrimSuffix(duration, "y"))
 		if err != nil {
 			return 0, fmt.Errorf("invalid time range: %q", duration)
 		}
@@ -42,7 +42,7 @@ func convertDuration(duration string) (time.Duration, error) {
 			return 0, fmt.Errorf("invalid time range: %q", duration)
 		}
 	} else if strings.HasSuffix(duration, "w") {
-		timeValue, err = strconv.Atoi(strings.Trim(duration, "w"))
+		timeValue, err = strconv.Atoi(strings.TrimSuffix(duration, "w"))
 		if err != nil {
 			return 0, fmt.Errorf("invalid time range: %q", duration)
 		}
@@ -52,7 +52,7 @@ func convertDuration(duration string) (time.Duration, error) {
 			return 0, fmt.Errorf("invalid time range: %q", duration)
 		}
 	} else if strings.HasSuffix(duration, "d") {
-		timeValue, err = strconv.Atoi(strings.Trim(duration, "d"))
+		timeValue, err = strconv.Atoi(strings.TrimSuffix(duration, "d"))
 		if err != nil {
 			return 0, fmt.Errorf("invalid time range: %q", duration)
 		}
@@ -95,6 +95,9 @@ func convertRetention(retention string, offset int64, msecTime bool) (Retention,
 	if !msecTime {
 		queryLength = queryLength / 1000
 	}
+	if queryLength <= 0 {
+		return Retention{}, fmt.Errorf("ttl %q resolves to non-positive query range %d; use a larger duration", chunks[2], queryLength)
+	}
 	queryRange := queryLength
 	// bump by the offset so we don't look at empty ranges any time offset > ttl
 	queryLength += offset
@@ -138,16 +141,29 @@ func convertRetention(retention string, offset int64, msecTime bool) (Retention,
 			2. we discover the actual size of each "chunk"
 			   This is second division step
 		*/
-		querySize = int64(queryRange / (queryRange / (rowLength * 4)))
+		divisor := queryRange / (rowLength * 4)
+		if divisor == 0 {
+			querySize = queryRange
+		} else {
+			querySize = queryRange / divisor
+		}
 	} else {
 		/*
 			Unless the aggTime (how long a range of data we're requesting per individual point)
 			is greater than the row size. Then we'll need to use that to determine
 			how big each individual query should be
 		*/
-		querySize = int64(queryRange / (queryRange / (aggTime * 4)))
+		divisor := queryRange / (aggTime * 4)
+		if divisor == 0 {
+			querySize = queryRange
+		} else {
+			querySize = queryRange / divisor
+		}
 	}
 
+	if querySize <= 0 {
+		return Retention{}, fmt.Errorf("computed non-positive querySize=%d for retention %q; check parameters", querySize, retention)
+	}
 	var timeChunks []TimeRange
 	var i int64
 	for i = offset; i <= queryLength; i = i + querySize {

app/vmctl/thanos/aggr_chunk.go

@@ -0,0 +1,233 @@
+package thanos
+
+import (
+	"encoding/binary"
+	"errors"
+	"fmt"
+
+	"github.com/prometheus/prometheus/tsdb/chunkenc"
+)
+
+// ChunkEncAggr is the top level encoding byte for the AggrChunk.
+// It is defined by Thanos as 0xff to prevent collisions with Prometheus encodings.
+const ChunkEncAggr = chunkenc.Encoding(0xff)
+
+// AggrType represents an aggregation type in Thanos downsampled blocks.
+type AggrType uint8
+
+// AggrTypeNone indicates raw blocks with no aggregation.
+// It is used as a sentinel to distinguish raw block processing from downsampled.
+const AggrTypeNone AggrType = 255
+
+// Valid aggregation types matching Thanos definitions.
+const (
+	AggrCount AggrType = iota
+	AggrSum
+	AggrMin
+	AggrMax
+	AggrCounter
+)
+
+// AllAggrTypes contains all supported aggregation types.
+var AllAggrTypes = []AggrType{AggrCount, AggrSum, AggrMin, AggrMax, AggrCounter}
+
+func (t AggrType) String() string {
+	switch t {
+	case AggrCount:
+		return "count"
+	case AggrSum:
+		return "sum"
+	case AggrMin:
+		return "min"
+	case AggrMax:
+		return "max"
+	case AggrCounter:
+		return "counter"
+	}
+	return "<unknown>"
+}
+
+// ParseAggrType parses aggregate type from string.
+func ParseAggrType(s string) (AggrType, error) {
+	switch s {
+	case "count":
+		return AggrCount, nil
+	case "sum":
+		return AggrSum, nil
+	case "min":
+		return AggrMin, nil
+	case "max":
+		return AggrMax, nil
+	case "counter":
+		return AggrCounter, nil
+	}
+	return 0, fmt.Errorf("unknown aggregate type: %q", s)
+}
+
+// ErrAggrNotExist is returned if a requested aggregation is not present in an AggrChunk.
+var ErrAggrNotExist = errors.New("aggregate does not exist")
+
+// AggrChunk is a chunk that is composed of a set of aggregates for the same underlying data.
+// Not all aggregates must be present.
+// This is a read-only implementation for decoding Thanos downsampled blocks.
+type AggrChunk []byte
+
+// IsAggrChunk checks if the encoding byte indicates this is an AggrChunk.
+func IsAggrChunk(enc chunkenc.Encoding) bool {
+	return enc == ChunkEncAggr
+}
+
+// Get returns the sub-chunk for the given aggregate type if it exists.
+func (c AggrChunk) Get(t AggrType) (chunkenc.Chunk, error) {
+	b := c[:]
+	var x []byte
+
+	for i := AggrType(0); i <= t; i++ {
+		l, n := binary.Uvarint(b)
+		if n < 1 {
+			return nil, errors.New("invalid size: failed to read uvarint")
+		}
+		if l > uint64(len(b[n:])) || l+1 > uint64(len(b[n:])) {
+			if l > 0 {
+				return nil, errors.New("invalid size: not enough bytes")
+			}
+		}
+		b = b[n:]
+		// If length is set to zero explicitly, that means the aggregate is unset.
+		if l == 0 {
+			if i == t {
+				return nil, ErrAggrNotExist
+			}
+			continue
+		}
+		chunkLen := int(l) + 1
+		x = b[:chunkLen]
+		b = b[chunkLen:]
+	}
+	if len(x) == 0 {
+		return nil, ErrAggrNotExist
+	}
+	return chunkenc.FromData(chunkenc.Encoding(x[0]), x[1:])
+}
+
+// Encoding returns the encoding type for AggrChunk.
+func (c AggrChunk) Encoding() chunkenc.Encoding {
+	return ChunkEncAggr
+}
+
+// errIterator wraps a nop iterator but reports an error via Err().
+// It embeds chunkenc.Iterator to inherit all methods (including Seek)
+// which avoids go vet stdmethods warning about Seek signature.
+type errIterator struct {
+	chunkenc.Iterator
+	err error
+}
+
+// Err returns the underlying error.
+func (it *errIterator) Err() error {
+	return it.err
+}
+
+// newAggrChunkIterator creates a new iterator for the specified aggregate type.
+// If the aggregate is not present in the chunk (ErrAggrNotExist), a nop iterator
+// is returned without error — the caller will simply see zero samples.
+// Real decoding/corruption errors are reported via the iterator's Err() method.
+func newAggrChunkIterator(data []byte, aggrType AggrType) chunkenc.Iterator {
+	chunk := AggrChunk(data)
+	subChunk, err := chunk.Get(aggrType)
+	if err != nil {
+		if errors.Is(err, ErrAggrNotExist) {
+			return chunkenc.NewNopIterator()
+		}
+		return &errIterator{
+			Iterator: chunkenc.NewNopIterator(),
+			err:      err,
+		}
+	}
+	return subChunk.Iterator(nil)
+}
+
+// AggrChunkWrapper wraps AggrChunk to implement chunkenc.Chunk interface.
+// It delegates iteration to a specific aggregate type.
+type AggrChunkWrapper struct {
+	data     []byte
+	aggrType AggrType
+}
+
+// NewAggrChunkWrapper creates a new AggrChunk wrapper for the specified aggregate type.
+func NewAggrChunkWrapper(data []byte, aggrType AggrType) *AggrChunkWrapper {
+	return &AggrChunkWrapper{
+		data:     data,
+		aggrType: aggrType,
+	}
+}
+
+// Bytes returns the underlying byte slice.
+func (c *AggrChunkWrapper) Bytes() []byte {
+	return c.data
+}
+
+// Encoding returns the AggrChunk encoding.
+func (c *AggrChunkWrapper) Encoding() chunkenc.Encoding {
+	return ChunkEncAggr
+}
+
+// Appender returns an error since AggrChunk is read-only.
+func (c *AggrChunkWrapper) Appender() (chunkenc.Appender, error) {
+	return nil, errors.New("AggrChunk is read-only")
+}
+
+// Iterator returns an iterator for the specified aggregate type.
+func (c *AggrChunkWrapper) Iterator(it chunkenc.Iterator) chunkenc.Iterator {
+	return newAggrChunkIterator(c.data, c.aggrType)
+}
+
+// NumSamples returns the number of samples in the aggregate.
+func (c *AggrChunkWrapper) NumSamples() int {
+	chunk := AggrChunk(c.data)
+	subChunk, err := chunk.Get(c.aggrType)
+	if err != nil {
+		return 0
+	}
+	return subChunk.NumSamples()
+}
+
+// Compact is a no-op for read-only AggrChunk.
+func (c *AggrChunkWrapper) Compact() {}
+
+// Reset resets the chunk with new data.
+func (c *AggrChunkWrapper) Reset(stream []byte) {
+	c.data = stream
+}
+
+// AggrChunkPool is a custom Pool that understands AggrChunk encoding (0xff).
+// It delegates standard encodings to the default pool and handles AggrChunk specially.
+type AggrChunkPool struct {
+	defaultPool chunkenc.Pool
+	aggrType    AggrType
+}
+
+// NewAggrChunkPool creates a new pool that handles AggrChunk encoding.
+func NewAggrChunkPool(aggrType AggrType) *AggrChunkPool {
+	return &AggrChunkPool{
+		defaultPool: chunkenc.NewPool(),
+		aggrType:    aggrType,
+	}
+}
+
+// Get returns a chunk for the given encoding and data.
+func (p *AggrChunkPool) Get(e chunkenc.Encoding, b []byte) (chunkenc.Chunk, error) {
+	if e == ChunkEncAggr {
+		return NewAggrChunkWrapper(b, p.aggrType), nil
+	}
+	return p.defaultPool.Get(e, b)
+}
+
+// Put returns a chunk to the pool.
+func (p *AggrChunkPool) Put(c chunkenc.Chunk) error {
+	if c.Encoding() == ChunkEncAggr {
+		// AggrChunk wrappers are not pooled
+		return nil
+	}
+	return p.defaultPool.Put(c)
+}

app/vmctl/thanos/block_meta.go

@@ -0,0 +1,110 @@
+package thanos
+
+import (
+	"encoding/json"
+	"os"
+	"path/filepath"
+)
+
+// BlockMeta extends Prometheus BlockMeta with Thanos-specific fields.
+type BlockMeta struct {
+	// Thanos-specific metadata
+	Thanos ThanosMeta `json:"thanos,omitempty"`
+}
+
+// ThanosMeta contains Thanos-specific block metadata.
+type ThanosMeta struct {
+	// Labels are external labels identifying the producer.
+	Labels map[string]string `json:"labels,omitempty"`
+
+	// Downsample contains downsampling information.
+	Downsample ThanosDownsample `json:"downsample,omitempty"`
+
+	// Source indicates where the block came from.
+	Source string `json:"source,omitempty"`
+
+	// SegmentFiles contains list of segment files in the block.
+	SegmentFiles []string `json:"segment_files,omitempty"`
+
+	// Files contains metadata about files in the block.
+	Files []ThanosFile `json:"files,omitempty"`
+}
+
+// ThanosDownsample contains downsampling resolution info.
+type ThanosDownsample struct {
+	// Resolution is the downsampling resolution in milliseconds.
+	// 0 means raw data (no downsampling).
+	// 300000 (5 minutes) or 3600000 (1 hour) for downsampled data.
+	Resolution int64 `json:"resolution"`
+}
+
+// ThanosFile contains metadata about a file in the block.
+type ThanosFile struct {
+	RelPath   string `json:"rel_path"`
+	SizeBytes int64  `json:"size_bytes,omitempty"`
+}
+
+// ResolutionLevel represents the downsampling resolution.
+type ResolutionLevel int64
+
+const (
+	// ResolutionRaw is for raw, non-downsampled data.
+	ResolutionRaw ResolutionLevel = 0
+	// Resolution5m is for 5-minute downsampled data (300000 ms).
+	Resolution5m ResolutionLevel = 300000
+	// Resolution1h is for 1-hour downsampled data (3600000 ms).
+	Resolution1h ResolutionLevel = 3600000
+)
+
+// String returns human-readable resolution string.
+func (r ResolutionLevel) String() string {
+	switch r {
+	case ResolutionRaw:
+		return "raw"
+	case Resolution5m:
+		return "5m"
+	case Resolution1h:
+		return "1h"
+	default:
+		return "unknown"
+	}
+}
+
+// ReadBlockMeta reads Thanos-extended block metadata from meta.json.
+func ReadBlockMeta(blockDir string) (*BlockMeta, error) {
+	metaPath := filepath.Join(blockDir, "meta.json")
+	data, err := os.ReadFile(metaPath)
+	if err != nil {
+		return nil, err
+	}
+
+	var meta BlockMeta
+	if err := json.Unmarshal(data, &meta); err != nil {
+		return nil, err
+	}
+
+	return &meta, nil
+}
+
+// IsDownsampled returns true if the block contains downsampled data.
+func (m *BlockMeta) IsDownsampled() bool {
+	return m.Thanos.Downsample.Resolution > 0
+}
+
+// Resolution returns the block's downsampling resolution.
+func (m *BlockMeta) Resolution() ResolutionLevel {
+	return ResolutionLevel(m.Thanos.Downsample.Resolution)
+}
+
+// ResolutionSuffix returns a suffix string for metric names based on resolution.
+// For example: ":5m" or ":1h" for downsampled data, empty for raw data.
+func (m *BlockMeta) ResolutionSuffix() string {
+	switch m.Resolution() {
+	case Resolution5m:
+		return ":5m"
+	case Resolution1h:
+		return ":1h"
+	default:
+		return ""
+	}
+}

app/vmctl/thanos/block_reader.go

@@ -0,0 +1,83 @@
+package thanos
+
+import (
+	"fmt"
+	"io"
+	"os"
+	"path/filepath"
+
+	"github.com/prometheus/prometheus/tsdb"
+	"github.com/prometheus/prometheus/tsdb/chunkenc"
+)
+
+// BlockInfo contains information about a block including Thanos metadata.
+type BlockInfo struct {
+	Block      tsdb.BlockReader
+	Resolution ResolutionLevel
+	IsThanos   bool
+	// Closer releases the block's resources (file descriptors, mmap).
+	// Must be called only after all queriers on this block have been closed.
+	Closer io.Closer
+}
+
+// OpenBlocksWithInfo opens all blocks and returns them with their metadata.
+// snapshotDir must be a snapshot directory containing block directories.
+func OpenBlocksWithInfo(snapshotDir string, aggrType AggrType) ([]BlockInfo, error) {
+	entries, err := os.ReadDir(snapshotDir)
+	if err != nil {
+		return nil, fmt.Errorf("failed to read snapshot directory: %w", err)
+	}
+
+	var blocks []BlockInfo
+	for _, entry := range entries {
+		if !entry.IsDir() {
+			continue
+		}
+
+		blockDir := filepath.Join(snapshotDir, entry.Name())
+		metaPath := filepath.Join(blockDir, "meta.json")
+
+		// Check if this is a valid block directory (has meta.json)
+		if _, err := os.Stat(metaPath); os.IsNotExist(err) {
+			continue
+		}
+
+		meta, err := ReadBlockMeta(blockDir)
+		if err != nil {
+			CloseBlocks(blocks)
+			return nil, fmt.Errorf("failed to read Thanos metadata for block %s: %w", blockDir, err)
+		}
+
+		var pool chunkenc.Pool
+		if meta.IsDownsampled() {
+			// Use AggrChunkPool for downsampled blocks
+			pool = NewAggrChunkPool(aggrType)
+		}
+
+		block, err := tsdb.OpenBlock(nil, blockDir, pool, nil)
+		if err != nil {
+			// Close previously opened blocks before returning error
+			CloseBlocks(blocks)
+			return nil, fmt.Errorf("failed to open block %s: %w", blockDir, err)
+		}
+
+		blocks = append(blocks, BlockInfo{
+			Block:      block,
+			Resolution: meta.Resolution(),
+			IsThanos:   true,
+			Closer:     block,
+		})
+	}
+
+	return blocks, nil
+}
+
+// CloseBlocks closes all blocks in the slice.
+// Must be called only after all queriers on these blocks have been closed.
+func CloseBlocks(blocks []BlockInfo) {
+	for _, bi := range blocks {
+		if bi.Closer != nil {
+			_ = bi.Closer.Close()
+		}
+	}
+}

app/vmctl/thanos/client.go

@@ -0,0 +1,198 @@
+package thanos
+
+import (
+	"context"
+	"fmt"
+	"time"
+
+	"github.com/prometheus/prometheus/model/labels"
+	"github.com/prometheus/prometheus/storage"
+	"github.com/prometheus/prometheus/tsdb"
+)
+
+// Config contains parameters for reading Thanos snapshots.
+type Config struct {
+	Snapshot string
+	Filter   Filter
+}
+
+// Filter contains configuration for filtering the timeseries.
+type Filter struct {
+	TimeMin    string
+	TimeMax    string
+	Label      string
+	LabelValue string
+}
+
+// Client reads Thanos snapshot blocks, including downsampled blocks with AggrChunk encoding.
+type Client struct {
+	snapshotPath string
+	filter       filter
+	statsPrinted bool
+}
+
+type filter struct {
+	min, max   int64
+	label      string
+	labelValue string
+}
+
+func (f filter) inRange(minV, maxV int64) bool {
+	fmin, fmax := f.min, f.max
+	if fmin == 0 {
+		fmin = minV
+	}
+	if fmax == 0 {
+		fmax = maxV
+	}
+	return minV <= fmax && fmin <= maxV
+}
+
+// NewClient creates a new Thanos snapshot client.
+func NewClient(cfg Config) (*Client, error) {
+	minTime, maxTime, err := parseTime(cfg.Filter.TimeMin, cfg.Filter.TimeMax)
+	if err != nil {
+		return nil, fmt.Errorf("failed to parse time in filter: %s", err)
+	}
+	return &Client{
+		snapshotPath: cfg.Snapshot,
+		filter: filter{
+			min:        minTime,
+			max:        maxTime,
+			label:      cfg.Filter.Label,
+			labelValue: cfg.Filter.LabelValue,
+		},
+	}, nil
+}
+
+// Explore fetches all available blocks from the snapshot with support for
+// Thanos AggrChunk (downsampled blocks). It opens blocks with a custom pool
+// that can decode AggrChunk encoding (0xff).
+func (c *Client) Explore(aggrType AggrType) ([]BlockInfo, error) {
+	blockInfos, err := OpenBlocksWithInfo(c.snapshotPath, aggrType)
+	if err != nil {
+		return nil, fmt.Errorf("failed to open blocks: %w", err)
+	}
+
+	s := &Stats{
+		Filtered: c.filter.min != 0 || c.filter.max != 0 || c.filter.label != "",
+		Blocks:   len(blockInfos),
+	}
+
+	var blocksToImport []BlockInfo
+	for _, bi := range blockInfos {
+		meta := bi.Block.Meta()
+
+		if s.MinTime == 0 || meta.MinTime < s.MinTime {
+			s.MinTime = meta.MinTime
+		}
+		if s.MaxTime == 0 || meta.MaxTime > s.MaxTime {
+			s.MaxTime = meta.MaxTime
+		}
+
+		if !c.filter.inRange(meta.MinTime, meta.MaxTime) {
+			s.SkippedBlocks++
+			if bi.Closer != nil {
+				_ = bi.Closer.Close()
+			}
+			continue
+		}
+
+		s.Samples += meta.Stats.NumSamples
+		s.Series += meta.Stats.NumSeries
+		blocksToImport = append(blocksToImport, bi)
+	}
+	if !c.statsPrinted {
+		fmt.Println(s)
+		c.statsPrinted = true
+	}
+	return blocksToImport, nil
+}
+
+// querierSeriesSet wraps a SeriesSet and its underlying Querier, ensuring
+// the querier is closed once the SeriesSet has been fully consumed.
+// This releases the querier's read reference on the block, which is required
+// for Block.Close() to complete without hanging.
+type querierSeriesSet struct {
+	storage.SeriesSet
+	q      storage.Querier
+	closed bool
+}
+
+// Next advances the iterator. When the underlying SeriesSet is exhausted,
+// it closes the querier to release resources.
+func (s *querierSeriesSet) Next() bool {
+	if s.SeriesSet.Next() {
+		return true
+	}
+	if !s.closed {
+		_ = s.q.Close()
+		s.closed = true
+	}
+	return false
+}
+
+// Close explicitly closes the underlying querier.
+// This must be called if iteration is stopped early (before Next returns false)
+// to release block read references and prevent Block.Close() from hanging.
+func (s *querierSeriesSet) Close() {
+	if !s.closed {
+		_ = s.q.Close()
+		s.closed = true
+	}
+}
+
+// ClosableSeriesSet extends storage.SeriesSet with a Close method for explicit cleanup.
+type ClosableSeriesSet interface {
+	storage.SeriesSet
+	Close()
+}
+
+// Read reads the given BlockInfo according to configured time and label filters.
+// The returned ClosableSeriesSet automatically closes the underlying querier when fully consumed,
+// but Close() should be called explicitly (e.g., via defer) to handle early returns.
+func (c *Client) Read(bi BlockInfo) (ClosableSeriesSet, error) {
+	minTime, maxTime := bi.Block.Meta().MinTime, bi.Block.Meta().MaxTime
+	if c.filter.min != 0 {
+		minTime = c.filter.min
+	}
+	if c.filter.max != 0 {
+		maxTime = c.filter.max
+	}
+	q, err := tsdb.NewBlockQuerier(bi.Block, minTime, maxTime)
+	if err != nil {
+		return nil, err
+	}
+	ss := q.Select(
+		context.Background(),
+		false,
+		nil,
+		labels.MustNewMatcher(labels.MatchRegexp, c.filter.label, c.filter.labelValue),
+	)
+	return &querierSeriesSet{
+		SeriesSet: ss,
+		q:         q,
+	}, nil
+}
+
+func parseTime(start, end string) (int64, int64, error) {
+	var s, e int64
+	if start == "" && end == "" {
+		return 0, 0, nil
+	}
+	if start != "" {
+		v, err := time.Parse(time.RFC3339, start)
+		if err != nil {
+			return 0, 0, fmt.Errorf("failed to parse %q: %s", start, err)
+		}
+		s = v.UnixNano() / int64(time.Millisecond)
+	}
+	if end != "" {
+		v, err := time.Parse(time.RFC3339, end)
+		if err != nil {
+			return 0, 0, fmt.Errorf("failed to parse %q: %s", end, err)
+		}
+		e = v.UnixNano() / int64(time.Millisecond)
+	}
+	return s, e, nil
+}

app/vmctl/thanos/stats.go

@@ -0,0 +1,38 @@
+package thanos
+
+import (
+	"fmt"
+	"time"
+)
+
+// Stats represents data migration stats for Thanos blocks.
+type Stats struct {
+	Filtered      bool
+	MinTime       int64
+	MaxTime       int64
+	Samples       uint64
+	Series        uint64
+	Blocks        int
+	SkippedBlocks int
+}
+
+// String returns string representation for s.
+func (s Stats) String() string {
+	str := fmt.Sprintf("Thanos snapshot stats:\n"+
+		"  blocks found: %d;\n"+
+		"  blocks skipped by time filter: %d;\n"+
+		"  min time: %d (%v);\n"+
+		"  max time: %d (%v);\n"+
+		"  samples: %d;\n"+
+		"  series: %d.",
+		s.Blocks, s.SkippedBlocks,
+		s.MinTime, time.Unix(s.MinTime/1e3, 0).Format(time.RFC3339),
+		s.MaxTime, time.Unix(s.MaxTime/1e3, 0).Format(time.RFC3339),
+		s.Samples, s.Series)
+
+	if s.Filtered {
+		str += "\n* Stats numbers are based on blocks meta info and don't account for applied filters."
+	}
+
+	return str
+}

app/vmctl/thanos_processor.go

@@ -0,0 +1,309 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"log"
+	"strings"
+	"sync"
+
+	"github.com/prometheus/prometheus/model/labels"
+	"github.com/prometheus/prometheus/tsdb/chunkenc"
+
+	"github.com/VictoriaMetrics/metrics"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmctl/barpool"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmctl/thanos"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmctl/vm"
+)
+
+type thanosProcessor struct {
+	cl *thanos.Client
+	im *vm.Importer
+	cc int
+
+	isVerbose bool
+	aggrTypes []thanos.AggrType
+}
+
+func (tp *thanosProcessor) run(ctx context.Context) error {
+	if len(tp.aggrTypes) == 0 {
+		tp.aggrTypes = thanos.AllAggrTypes
+	}
+
+	log.Printf("Processing blocks with aggregate types: %v", tp.aggrTypes)
+
+	// Use the first aggregate type to explore blocks (block list is the same for all types)
+	blocks, err := tp.cl.Explore(tp.aggrTypes[0])
+	if err != nil {
+		return fmt.Errorf("explore failed: %s", err)
+	}
+	if len(blocks) < 1 {
+		return fmt.Errorf("found no blocks to import")
+	}
+
+	// Separate blocks into raw (resolution=0) and downsampled (resolution>0)
+	var rawBlocks, downsampledBlocks []thanos.BlockInfo
+	for _, block := range blocks {
+		if block.Resolution == thanos.ResolutionRaw {
+			rawBlocks = append(rawBlocks, block)
+		} else {
+			downsampledBlocks = append(downsampledBlocks, block)
+		}
+	}
+
+	log.Printf("Found %d raw blocks and %d downsampled blocks", len(rawBlocks), len(downsampledBlocks))
+
+	question := fmt.Sprintf("Found %d blocks to import (%d raw + %d downsampled with %d aggregate types). Continue?",
+		len(blocks), len(rawBlocks), len(downsampledBlocks), len(tp.aggrTypes))
+	if !prompt(ctx, question) {
+		return nil
+	}
+
+	// Calculate total number of block processing passes for the progress bar:
+	// raw blocks are processed once, downsampled blocks are processed once per aggregate type.
+	totalPasses := len(rawBlocks) + len(downsampledBlocks)*len(tp.aggrTypes)
+	thanosBlocksTotal.Add(totalPasses)
+	bar := barpool.AddWithTemplate(fmt.Sprintf(barTpl, "Processing blocks"), totalPasses)
+	if err := barpool.Start(); err != nil {
+		return err
+	}
+	defer barpool.Stop()
+
+	tp.im.ResetStats()
+
+	type phaseStats struct {
+		name    string
+		series  uint64
+		samples uint64
+	}
+	var phases []phaseStats
+
+	// Process raw blocks first (no aggregate suffix)
+	if len(rawBlocks) > 0 {
+		log.Println("Processing raw blocks (resolution=0)...")
+		stats, err := tp.processBlocks(rawBlocks, thanos.AggrTypeNone, bar)
+		if err != nil {
+			return fmt.Errorf("migration failed for raw blocks: %s", err)
+		}
+		phases = append(phases, phaseStats{
+			name:    "raw",
+			series:  stats.series,
+			samples: stats.samples,
+		})
+	}
+
+	// Close blocks from the initial Explore. The querierSeriesSet wrapper
+	// has already released all querier read references, so Close won't hang.
+	thanos.CloseBlocks(blocks)
+
+	// Process downsampled blocks for each aggregate type.
+	// Each type needs its own AggrChunkPool, so we reopen blocks per type.
+	for _, aggrType := range tp.aggrTypes {
+		if len(downsampledBlocks) < 1 {
+			break
+		}
+
+		log.Printf("Processing downsampled blocks with aggregate type: %s", aggrType)
+
+		aggrBlocks, err := tp.cl.Explore(aggrType)
+		if err != nil {
+			return fmt.Errorf("explore failed for aggr type %s: %s", aggrType, err)
+		}
+
+		var downsampledOnly []thanos.BlockInfo
+		for _, block := range aggrBlocks {
+			if block.Resolution != thanos.ResolutionRaw {
+				downsampledOnly = append(downsampledOnly, block)
+			}
+		}
+
+		if len(downsampledOnly) < 1 {
+			log.Printf("No downsampled blocks found for aggregate type %s, skipping", aggrType)
+			thanos.CloseBlocks(aggrBlocks)
+			continue
+		}
+
+		log.Printf("Processing %d blocks for aggregate type: %s", len(downsampledOnly), aggrType)
+		stats, err := tp.processBlocks(downsampledOnly, aggrType, bar)
+		thanos.CloseBlocks(aggrBlocks)
+		if err != nil {
+			return fmt.Errorf("migration failed for aggr type %s: %s", aggrType, err)
+		}
+		phases = append(phases, phaseStats{
+			name:    aggrType.String(),
+			series:  stats.series,
+			samples: stats.samples,
+		})
+	}
+
+	// Print per-phase and total statistics
+	var totalSeries, totalSamples uint64
+	log.Printf("Migration statistics (%d raw blocks, %d downsampled blocks):", len(rawBlocks), len(downsampledBlocks))
+	for _, p := range phases {
+		log.Printf("  %s: %d series, %d samples", p.name, p.series, p.samples)
+		totalSeries += p.series
+		totalSamples += p.samples
+	}
+	log.Printf("  total: %d series, %d samples", totalSeries, totalSamples)
+
+	// Wait for all buffers to flush
+	tp.im.Close()
+	// Drain import errors channel
+	for vmErr := range tp.im.Errors() {
+		if vmErr.Err != nil {
+			thanosErrorsTotal.Inc()
+			return fmt.Errorf("import process failed: %s", wrapErr(vmErr, tp.isVerbose))
+		}
+	}
+
+	log.Println("Import finished!")
+	log.Println(tp.im.Stats())
+	return nil
+}
+
+// processBlocksStats holds statistics collected during block processing.
+type processBlocksStats struct {
+	blocks  uint64
+	series  uint64
+	samples uint64
+}
+
+func (tp *thanosProcessor) processBlocks(blocks []thanos.BlockInfo, aggrType thanos.AggrType, bar barpool.Bar) (processBlocksStats, error) {
+	blockReadersCh := make(chan thanos.BlockInfo)
+	errCh := make(chan error, tp.cc)
+
+	var processedBlocks, totalSeries, totalSamples uint64
+	var mu sync.Mutex
+
+	var wg sync.WaitGroup
+	for i := range tp.cc {
+		workerID := i
+		wg.Go(func() {
+			for bi := range blockReadersCh {
+				seriesCount, samplesCount, err := tp.do(bi, aggrType)
+				if err != nil {
+					thanosErrorsTotal.Inc()
+					errCh <- fmt.Errorf("read failed for block %q with aggr %s: %s", bi.Block.Meta().ULID, aggrType, err)
+					return
+				}
+
+				mu.Lock()
+				processedBlocks++
+				totalSeries += seriesCount
+				totalSamples += samplesCount
+				log.Printf("[Worker %d] Block %s: %d series, %d samples | Total: %d/%d blocks, %d series, %d samples",
+					workerID, bi.Block.Meta().ULID.String()[:8], seriesCount, samplesCount,
+					processedBlocks, len(blocks), totalSeries, totalSamples)
+				mu.Unlock()
+
+				thanosBlocksProcessed.Inc()
+				bar.Increment()
+			}
+		})
+	}
+
+	// any error breaks the import
+	for _, bi := range blocks {
+		select {
+		case thanosErr := <-errCh:
+			close(blockReadersCh)
+			wg.Wait()
+			return processBlocksStats{}, fmt.Errorf("thanos error: %s", thanosErr)
+		case vmErr := <-tp.im.Errors():
+			close(blockReadersCh)
+			wg.Wait()
+			thanosErrorsTotal.Inc()
+			return processBlocksStats{}, fmt.Errorf("import process failed: %s", wrapErr(vmErr, tp.isVerbose))
+		case blockReadersCh <- bi:
+		}
+	}
+
+	close(blockReadersCh)
+	wg.Wait()
+	close(errCh)
+	for err := range errCh {
+		return processBlocksStats{}, fmt.Errorf("import process failed: %s", err)
+	}
+
+	return processBlocksStats{
+		blocks:  processedBlocks,
+		series:  totalSeries,
+		samples: totalSamples,
+	}, nil
+}
+
+func (tp *thanosProcessor) do(bi thanos.BlockInfo, aggrType thanos.AggrType) (uint64, uint64, error) {
+	ss, err := tp.cl.Read(bi)
+	if err != nil {
+		return 0, 0, fmt.Errorf("failed to read block: %s", err)
+	}
+	defer ss.Close() // Ensure querier is closed even on early returns
+
+	var it chunkenc.Iterator
+	var seriesCount, samplesCount uint64
+
+	for ss.Next() {
+		var name string
+		var labelPairs []vm.LabelPair
+		series := ss.At()
+
+		series.Labels().Range(func(label labels.Label) {
+			if label.Name == "__name__" {
+				name = label.Value
+				return
+			}
+			labelPairs = append(labelPairs, vm.LabelPair{
+				Name:  strings.Clone(label.Name),
+				Value: strings.Clone(label.Value),
+			})
+		})
+		if name == "" {
+			return seriesCount, samplesCount, fmt.Errorf("failed to find `__name__` label in labelset for block %v", bi.Block.Meta().ULID)
+		}
+
+		// Add resolution and aggregate type suffix to metric name for downsampled blocks
+		if bi.Resolution != thanos.ResolutionRaw && aggrType != thanos.AggrTypeNone {
+			name = fmt.Sprintf("%s:%s:%s", name, bi.Resolution.String(), aggrType.String())
+		}
+
+		var timestamps []int64
+		var values []float64
+		it = series.Iterator(it)
+		for {
+			typ := it.Next()
+			if typ == chunkenc.ValNone {
+				break
+			}
+			if typ != chunkenc.ValFloat {
+				continue
+			}
+			t, v := it.At()
+			timestamps = append(timestamps, t)
+			values = append(values, v)
+		}
+		if err := it.Err(); err != nil {
+			return seriesCount, samplesCount, err
+		}
+
+		samplesCount += uint64(len(timestamps))
+		seriesCount++
+
+		ts := vm.TimeSeries{
+			Name:       name,
+			LabelPairs: labelPairs,
+			Timestamps: timestamps,
+			Values:     values,
+		}
+		if err := tp.im.Input(&ts); err != nil {
+			return seriesCount, samplesCount, err
+		}
+	}
+	return seriesCount, samplesCount, ss.Err()
+}
+
+var (
+	thanosBlocksTotal     = metrics.NewCounter(`vmctl_thanos_migration_blocks_total`)
+	thanosBlocksProcessed = metrics.NewCounter(`vmctl_thanos_migration_blocks_processed`)
+	thanosErrorsTotal     = metrics.NewCounter(`vmctl_thanos_migration_errors_total`)
+)

app/vmctl/vmctlutil/time_test.go

@@ -20,6 +20,9 @@ func TestGetTime_Failure(t *testing.T) {
 
 	// negative time
 	f("-292273086-05-16T16:47:06Z")
+
+	// relative duration that resolves to a timestamp before 1970
+	f("-9223372036.855")
 }
 
 func TestGetTime_Success(t *testing.T) {
@@ -77,9 +80,6 @@ func TestGetTime_Success(t *testing.T) {
 	// float timestamp representation",
 	f("1562529662.324", time.Date(2019, 7, 7, 20, 01, 02, 324e6, time.UTC))
 
-	// negative timestamp
-	f("-9223372036.855", time.Date(1970, 01, 01, 00, 00, 00, 00, time.UTC))
-
 	// big timestamp
 	f("1223372036855", time.Date(2008, 10, 7, 9, 33, 56, 855e6, time.UTC))
 

app/vmselect/main.go

@@ -69,6 +69,8 @@ func Init() {
 	concurrencyLimitCh = make(chan struct{}, *maxConcurrentRequests)
 	initVMUIConfig()
 	initVMAlertProxy()
+
+	flagutil.RegisterSecretFlag("vmalert.proxyURL")
 }
 
 // Stop stops vmselect
@@ -262,6 +264,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		return true
 	case "/api/v1/export":
 		exportRequests.Inc()
+		httpserver.EnableCORS(w, r)
 		if err := prometheus.ExportHandler(startTime, w, r); err != nil {
 			exportErrors.Inc()
 			httpserver.Errorf(w, r, "%s", err)
@@ -270,6 +273,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		return true
 	case "/api/v1/export/csv":
 		exportCSVRequests.Inc()
+		httpserver.EnableCORS(w, r)
 		if err := prometheus.ExportCSVHandler(startTime, w, r); err != nil {
 			exportCSVErrors.Inc()
 			httpserver.Errorf(w, r, "%s", err)
@@ -278,6 +282,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		return true
 	case "/api/v1/export/native":
 		exportNativeRequests.Inc()
+		httpserver.EnableCORS(w, r)
 		if err := prometheus.ExportNativeHandler(startTime, w, r); err != nil {
 			exportNativeErrors.Inc()
 			httpserver.Errorf(w, r, "%s", err)

app/vmselect/netstorage/netstorage.go

@@ -17,11 +17,13 @@ import (
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/searchutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmstorage/promdb"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricnamestats"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricsmetadata"
 )
 
@@ -323,14 +325,21 @@ func (rss *Results) runParallel(qt *querytracer.Tracer, f func(rs *Result, worke
 }
 
 var (
-	rowsReadPerSeries  = metrics.NewHistogram(`vm_rows_read_per_series`)
-	rowsReadPerQuery   = metrics.NewHistogram(`vm_rows_read_per_query`)
-	seriesReadPerQuery = metrics.NewHistogram(`vm_series_read_per_query`)
+	rowsReadPerSeries            = metrics.NewHistogram(`vm_rows_read_per_series`)
+	rowsReadPerQuery             = metrics.NewHistogram(`vm_rows_read_per_query`)
+	seriesReadPerQuery           = metrics.NewHistogram(`vm_series_read_per_query`)
+	seriesPrometheusReadPerQuery = metrics.NewHistogram(`vm_promdb_series_read_per_query`)
 )
 
 type packedTimeseries struct {
 	metricName string
 	brs        []blockRef
+	pd         *promData
+}
+
+type promData struct {
+	values     []float64
+	timestamps []int64
 }
 
 type unpackWork struct {
@@ -434,9 +443,21 @@ func (pts *packedTimeseries) Unpack(dst *Result, tbf *tmpBlocksFile, tr storage.
 		putSortBlocksHeap(sbh)
 		return err
 	}
-	dedupInterval := storage.GetDedupInterval()
+	if pts.pd != nil {
+		// Add data from Prometheus to dst.
+		// It usually has smaller timestamps than the data from sbs, so put it first.
+		dst.Values = append(dst.Values, pts.pd.values...)
+		dst.Timestamps = append(dst.Timestamps, pts.pd.timestamps...)
+	}
+	dedupInterval := storage.GetDedupInterval(tr.MinTimestamp)
 	mergeSortBlocks(dst, sbh, dedupInterval)
 	putSortBlocksHeap(sbh)
+	if pts.pd != nil {
+		if !sort.IsSorted(dst) {
+			sort.Sort(dst)
+		}
+		pts.pd = nil
+	}
 	return nil
 }
 
@@ -545,6 +566,27 @@ func (pts *packedTimeseries) unpackTo(dst []*sortBlock, tbf *tmpBlocksFile, tr s
 	return dst, firstErr
 }
 
+// sort.Interface implementation for Result
+
+// Len implements sort.Interface
+func (r *Result) Len() int {
+	return len(r.Timestamps)
+}
+
+// Less implements sort.Interface
+func (r *Result) Less(i, j int) bool {
+	timestamps := r.Timestamps
+	return timestamps[i] < timestamps[j]
+}
+
+// Swap implements sort.Interface
+func (r *Result) Swap(i, j int) {
+	timestamps := r.Timestamps
+	values := r.Values
+	timestamps[i], timestamps[j] = timestamps[j], timestamps[i]
+	values[i], values[j] = values[j], values[i]
+}
+
 func getSortBlock() *sortBlock {
 	v := sbPool.Get()
 	if v == nil {
@@ -784,6 +826,15 @@ func LabelNames(qt *querytracer.Tracer, sq *storage.SearchQuery, maxLabelNames i
 	if err != nil {
 		return nil, fmt.Errorf("error during labels search on time range: %w", err)
 	}
+
+	// Merge labels obtained from Prometheus storage.
+	promLabels, err := promdb.GetLabelNamesOnTimeRange(tr, deadline)
+	if err != nil {
+		return nil, fmt.Errorf("cannot obtain labels from Prometheus storage: %w", err)
+	}
+	qt.Printf("get %d label names from Prometheus storage", len(promLabels))
+	labels = mergeStrings(labels, promLabels)
+
 	// Sort labels like Prometheus does
 	sort.Strings(labels)
 	qt.Printf("sort %d labels", len(labels))
@@ -850,14 +901,44 @@ func LabelValues(qt *querytracer.Tracer, labelName string, sq *storage.SearchQue
 	}
 	labelValues, err := vmstorage.SearchLabelValues(qt, labelName, tfss, tr, maxLabelValues, sq.MaxMetrics, deadline.Deadline())
 	if err != nil {
-		return nil, fmt.Errorf("error during label values search on time range for labelName=%q: %w", labelName, err)
+		return nil, fmt.Errorf("error during label values search on time range: %w", err)
 	}
+
+	// Merge label values obtained from Prometheus storage.
+	promLabelValues, err := promdb.GetLabelValuesOnTimeRange(labelName, tr, deadline)
+	if err != nil {
+		return nil, fmt.Errorf("cannot obtain label values on time range for %q from Prometheus storage: %w", labelName, err)
+	}
+	qt.Printf("get %d label values from Prometheus storage", len(promLabelValues))
+	labelValues = mergeStrings(labelValues, promLabelValues)
+
 	// Sort labelValues like Prometheus does
 	sort.Strings(labelValues)
 	qt.Printf("sort %d label values", len(labelValues))
 	return labelValues, nil
 }
 
+func mergeStrings(a, b []string) []string {
+	if len(a) == 0 {
+		return b
+	}
+	if len(b) == 0 {
+		return a
+	}
+	m := make(map[string]struct{}, len(a)+len(b))
+	for _, s := range a {
+		m[s] = struct{}{}
+	}
+	for _, s := range b {
+		m[s] = struct{}{}
+	}
+	result := make([]string, 0, len(m))
+	for s := range m {
+		result = append(result, s)
+	}
+	return result
+}
+
 // GetMetricsMetadata returns time series metric names metadata for the given args
 func GetMetricsMetadata(qt *querytracer.Tracer, limit int, metricName string) ([]*metricsmetadata.Row, error) {
 	qt = qt.NewChild("get metrics metadata: limit=%d, metric_name=%q", limit, metricName)
@@ -989,9 +1070,6 @@ func ExportBlocks(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline sear
 		return fmt.Errorf("timeout exceeded before starting data export: %s", deadline.String())
 	}
 	tr := sq.GetTimeRange()
-	if err := vmstorage.CheckTimeRange(tr); err != nil {
-		return err
-	}
 	tfss, err := setupTfss(qt, tr, sq.TagFilterss, sq.MaxMetrics, deadline)
 	if err != nil {
 		return err
@@ -1097,9 +1175,6 @@ func SearchMetricNames(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline
 
 	// Setup search.
 	tr := sq.GetTimeRange()
-	if err := vmstorage.CheckTimeRange(tr); err != nil {
-		return nil, err
-	}
 	tfss, err := setupTfss(qt, tr, sq.TagFilterss, sq.MaxMetrics, deadline)
 	if err != nil {
 		return nil, err
@@ -1126,9 +1201,6 @@ func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadlin
 
 	// Setup search.
 	tr := sq.GetTimeRange()
-	if err := vmstorage.CheckTimeRange(tr); err != nil {
-		return nil, err
-	}
 	tfss, err := setupTfss(qt, tr, sq.TagFilterss, sq.MaxMetrics, deadline)
 	if err != nil {
 		return nil, err
@@ -1268,6 +1340,26 @@ func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadlin
 	}
 	qt.Printf("fetch unique series=%d, blocks=%d, samples=%d, bytes=%d", len(m), blocksRead, samples, tbf.Len())
 
+	// Fetch data from promdb.
+	pm := make(map[string]*promData)
+	err = promdb.VisitSeries(sq, deadline, func(metricName []byte, values []float64, timestamps []int64) {
+		pd := pm[string(metricName)]
+		if pd == nil {
+			if _, ok := m[string(metricName)]; !ok {
+				orderedMetricNames = append(orderedMetricNames, string(metricName))
+			}
+			pd = &promData{}
+			pm[string(metricName)] = pd
+		}
+		pd.values = append(pd.values, values...)
+		pd.timestamps = append(pd.timestamps, timestamps...)
+	})
+	if err != nil {
+		putTmpBlocksFile(tbf)
+		putStorageSearch(sr)
+		return nil, fmt.Errorf("error when searching in Prometheus data: %w", err)
+	}
+	seriesPrometheusReadPerQuery.Update(float64(len(pm)))
 	var rss Results
 	rss.tr = tr
 	rss.deadline = deadline
@@ -1276,6 +1368,7 @@ func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadlin
 		pts[i] = packedTimeseries{
 			metricName: metricName,
 			brs:        brssPool[m[metricName]].brs,
+			pd:         pm[metricName],
 		}
 	}
 	rss.packedTimeseries = pts
@@ -1362,7 +1455,7 @@ func applyGraphiteRegexpFilter(filter string, ss []string) ([]string, error) {
 const maxFastAllocBlockSize = 32 * 1024
 
 // GetMetricNamesStats returns statistic for timeseries metric names usage.
-func GetMetricNamesStats(qt *querytracer.Tracer, limit, le int, matchPattern string) (storage.MetricNamesStatsResponse, error) {
+func GetMetricNamesStats(qt *querytracer.Tracer, limit, le int, matchPattern string) (metricnamestats.StatsResult, error) {
 	qt = qt.NewChild("get metric names usage statistics with limit: %d, less or equal to: %d, match pattern=%q", limit, le, matchPattern)
 	defer qt.Done()
 	return vmstorage.GetMetricNamesStats(qt, limit, le, matchPattern)

app/vmselect/prometheus/export.qtpl

@@ -11,6 +11,16 @@
 
 {% stripspace %}
 
+{% func ExportCSVHeader(fieldNames []string) %}
+	{% if len(fieldNames) == 0 %}{% return %}{% endif %}
+	{%s= fieldNames[0] %}
+	{% for _, fieldName := range fieldNames[1:] %}
+		,
+		{%s= fieldName %}
+	{% endfor %}
+	{% newline %}
+{% endfunc %}
+
 {% func ExportCSVLine(xb *exportBlock, fieldNames []string) %}
 	{% if len(xb.timestamps) == 0 || len(fieldNames) == 0 %}{% return %}{% endif %}
 	{% for i, timestamp := range xb.timestamps %}

app/vmselect/prometheus/export_test.go

@@ -0,0 +1,132 @@
+package prometheus
+
+import (
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+)
+
+func TestExportCSVHeader(t *testing.T) {
+	f := func(fieldNames []string, expected string) {
+		t.Helper()
+		got := ExportCSVHeader(fieldNames)
+		if got != expected {
+			t.Fatalf("ExportCSVHeader(%v): got %q; want %q", fieldNames, got, expected)
+		}
+	}
+
+	f(nil, "")
+	f([]string{}, "")
+
+	f([]string{"__value__"}, "__value__\n")
+	f([]string{"__timestamp__"}, "__timestamp__\n")
+	f([]string{"__timestamp__:rfc3339"}, "__timestamp__:rfc3339\n")
+	f([]string{"__name__"}, "__name__\n")
+	f([]string{"job"}, "job\n")
+
+	f([]string{"__timestamp__:rfc3339", "__value__"}, "__timestamp__:rfc3339,__value__\n")
+	f([]string{"__value__", "__timestamp__"}, "__value__,__timestamp__\n")
+	f([]string{"job", "instance"}, "job,instance\n")
+
+	f([]string{"__name__", "__value__", "__timestamp__:unix_s"}, "__name__,__value__,__timestamp__:unix_s\n")
+	f([]string{"job", "instance", "__value__", "__timestamp__:unix_ms"}, "job,instance,__value__,__timestamp__:unix_ms\n")
+	f([]string{"__timestamp__:custom:2006-01-02", "__value__", "host", "dc", "env"},
+		"__timestamp__:custom:2006-01-02,__value__,host,dc,env\n")
+
+	// duplicate fields
+	f([]string{"__value__", "__value__"}, "__value__,__value__\n")
+	f([]string{"__timestamp__", "__timestamp__:rfc3339"}, "__timestamp__,__timestamp__:rfc3339\n")
+}
+
+func TestExportCSVLine(t *testing.T) {
+	localBak := time.Local
+	time.Local = time.UTC
+	defer func() { time.Local = localBak }()
+
+	f := func(mn *storage.MetricName, timestamps []int64, values []float64, fieldNames []string, expected string) {
+		t.Helper()
+		xb := &exportBlock{
+			mn:         mn,
+			timestamps: timestamps,
+			values:     values,
+		}
+		got := ExportCSVLine(xb, fieldNames)
+		if got != expected {
+			t.Fatalf("ExportCSVLine: got %q; want %q", got, expected)
+		}
+	}
+
+	mn := &storage.MetricName{
+		MetricGroup: []byte("cpu_usage"),
+		Tags: []storage.Tag{
+			{Key: []byte("job"), Value: []byte("node")},
+			{Key: []byte("instance"), Value: []byte("localhost:9090")},
+		},
+	}
+
+	// empty inputs
+	f(mn, nil, nil, []string{"__value__"}, "")
+	f(mn, []int64{}, []float64{}, []string{"__value__"}, "")
+	f(mn, []int64{1000}, []float64{1.5}, nil, "")
+	f(mn, []int64{1000}, []float64{1.5}, []string{}, "")
+
+	f(mn, []int64{1000}, []float64{42.5}, []string{"__value__"}, "42.5\n")
+	f(mn, []int64{1704067200000}, []float64{1}, []string{"__timestamp__"}, "1704067200000\n")
+	f(mn, []int64{1704067200000}, []float64{1}, []string{"__timestamp__:unix_s"}, "1704067200\n")
+	f(mn, []int64{1704067200000}, []float64{1}, []string{"__timestamp__:unix_ms"}, "1704067200000\n")
+	f(mn, []int64{1704067200000}, []float64{1}, []string{"__timestamp__:unix_ns"}, "1704067200000000000\n")
+	f(mn, []int64{1704067200000}, []float64{1}, []string{"__timestamp__:rfc3339"}, "2024-01-01T00:00:00Z\n")
+
+	f(mn, []int64{1000}, []float64{1}, []string{"__name__"}, "cpu_usage\n")
+	f(mn, []int64{1000}, []float64{1}, []string{"job"}, "node\n")
+	f(mn, []int64{1000}, []float64{1}, []string{"instance"}, "localhost:9090\n")
+	f(mn, []int64{1000}, []float64{1}, []string{"missing_label"}, "\n")
+
+	// multiple fields
+	f(mn, []int64{1704067200000}, []float64{99.9},
+		[]string{"__timestamp__:unix_s", "__value__", "job"},
+		"1704067200,99.9,node\n")
+
+	// multiple rows
+	f(mn, []int64{1000, 2000}, []float64{1.1, 2.2},
+		[]string{"__value__", "__timestamp__"},
+		"1.1,1000\n2.2,2000\n")
+	f(mn, []int64{1000, 2000, 3000}, []float64{10, 20, 30},
+		[]string{"__timestamp__:unix_s", "__value__"},
+		"1,10\n2,20\n3,30\n")
+
+	// escaping for special characters in tag values
+	f(&storage.MetricName{
+		MetricGroup: []byte("m"),
+		Tags:        []storage.Tag{{Key: []byte("desc"), Value: []byte("a,b")}},
+	}, []int64{1000}, []float64{1}, []string{"desc"}, "\"a,b\"\n")
+
+	f(&storage.MetricName{
+		MetricGroup: []byte("m"),
+		Tags:        []storage.Tag{{Key: []byte("desc"), Value: []byte(`say "hello"`)}},
+	}, []int64{1000}, []float64{1}, []string{"desc"}, "\"say \\\"hello\\\"\"\n")
+
+	f(&storage.MetricName{
+		MetricGroup: []byte("m"),
+		Tags:        []storage.Tag{{Key: []byte("desc"), Value: []byte("line1\nline2")}},
+	}, []int64{1000}, []float64{1}, []string{"desc"}, "\"line1\\nline2\"\n")
+
+	// header and data line field counts must match
+	fieldNames := []string{"__name__", "job", "instance", "__value__", "__timestamp__:unix_s"}
+	header := ExportCSVHeader(fieldNames)
+	line := ExportCSVLine(&exportBlock{
+		mn:         mn,
+		timestamps: []int64{1704067200000},
+		values:     []float64{99.9},
+	}, fieldNames)
+	headerCommas := strings.Count(header, ",")
+	lineCommas := strings.Count(line, ",")
+	if headerCommas != lineCommas {
+		t.Fatalf("header has %d commas, data line has %d commas", headerCommas, lineCommas)
+	}
+	if headerCommas != len(fieldNames)-1 {
+		t.Fatalf("expected %d commas in header, got %d", len(fieldNames)-1, headerCommas)
+	}
+}

app/vmselect/prometheus/prometheus.go

@@ -175,6 +175,7 @@ func ExportCSVHandler(startTime time.Time, w http.ResponseWriter, r *http.Reques
 	w.Header().Set("Content-Type", "text/csv; charset=utf-8")
 	bw := bufferedwriter.Get(w)
 	defer bufferedwriter.Put(bw)
+	WriteExportCSVHeader(bw, fieldNames)
 	sw := newScalableWriter(bw)
 	writeCSVLine := func(xb *exportBlock, workerID uint) error {
 		if len(xb.timestamps) == 0 {
@@ -1222,11 +1223,7 @@ func getCommonParamsInternal(r *http.Request, startTime time.Time, requireNonEmp
 	if err != nil {
 		return nil, err
 	}
-	// Limit the `end` arg to the current time +2 days in the same way
-	// as it is limited during data ingestion.
-	// See https://github.com/VictoriaMetrics/VictoriaMetrics/blob/ea06d2fd3ccbbb6aa4480ab3b04f7b671408be2a/lib/storage/table.go#L378
-	// This should fix possible timestamp overflow - see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2669
-	maxTS := startTime.UnixNano()/1e6 + 2*24*3600*1000
+	maxTS := int64(math.MaxInt64 / 1_000_000)
 	if end > maxTS {
 		end = maxTS
 	}

app/vmselect/prometheus/tsdb_status_response.qtpl

@@ -1,6 +1,7 @@
 {% import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricnamestats"
 ) %}
 
 {% stripspace %}
@@ -34,9 +35,9 @@ TSDBStatusResponse generates response for /api/v1/status/tsdb .
 ]
 {% endfunc %}
 
-{% func tsdbStatusMetricNameEntries(a []storage.TopHeapEntry, queryStats []storage.MetricNamesStatsRecord) %}
+{% func tsdbStatusMetricNameEntries(a []storage.TopHeapEntry, queryStats []metricnamestats.StatRecord) %}
 {% code
-  queryStatsByMetricName := make(map[string]storage.MetricNamesStatsRecord,len(queryStats))
+  queryStatsByMetricName := make(map[string]metricnamestats.StatRecord,len(queryStats))
   for _, record := range queryStats{
       queryStatsByMetricName[record.MetricName] = record
     }

app/vmselect/prometheus/tsdb_status_response.qtpl.go

@@ -8,228 +8,229 @@ package prometheus
 import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricnamestats"
 )
 
 // TSDBStatusResponse generates response for /api/v1/status/tsdb .
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:8
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:9
 import (
 	qtio422016 "io"
 
 	qt422016 "github.com/valyala/quicktemplate"
 )
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:8
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:9
 var (
 	_ = qtio422016.Copy
 	_ = qt422016.AcquireByteBuffer
 )
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:8
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:9
 func StreamTSDBStatusResponse(qw422016 *qt422016.Writer, status *storage.TSDBStatus, qt *querytracer.Tracer) {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:8
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:9
 	qw422016.N().S(`{"status":"success","data":{"totalSeries":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:12
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:13
 	qw422016.N().DUL(status.TotalSeries)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:12
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:13
 	qw422016.N().S(`,"totalLabelValuePairs":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:13
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:14
 	qw422016.N().DUL(status.TotalLabelValuePairs)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:13
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:14
 	qw422016.N().S(`,"seriesCountByMetricName":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:14
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:15
 	streamtsdbStatusMetricNameEntries(qw422016, status.SeriesCountByMetricName, status.SeriesQueryStatsByMetricName)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:14
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:15
 	qw422016.N().S(`,"seriesCountByLabelName":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:15
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:16
 	streamtsdbStatusEntries(qw422016, status.SeriesCountByLabelName)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:15
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:16
 	qw422016.N().S(`,"seriesCountByFocusLabelValue":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:16
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:17
 	streamtsdbStatusEntries(qw422016, status.SeriesCountByFocusLabelValue)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:16
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:17
 	qw422016.N().S(`,"seriesCountByLabelValuePair":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:17
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:18
 	streamtsdbStatusEntries(qw422016, status.SeriesCountByLabelValuePair)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:17
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:18
 	qw422016.N().S(`,"labelValueCountByLabelName":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:18
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:19
 	streamtsdbStatusEntries(qw422016, status.LabelValueCountByLabelName)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:18
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:19
 	qw422016.N().S(`}`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:20
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:21
 	qt.Done()
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:21
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:22
 	streamdumpQueryTrace(qw422016, qt)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:21
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:22
 	qw422016.N().S(`}`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 }
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 func WriteTSDBStatusResponse(qq422016 qtio422016.Writer, status *storage.TSDBStatus, qt *querytracer.Tracer) {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 	qw422016 := qt422016.AcquireWriter(qq422016)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 	StreamTSDBStatusResponse(qw422016, status, qt)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 	qt422016.ReleaseWriter(qw422016)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 }
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 func TSDBStatusResponse(status *storage.TSDBStatus, qt *querytracer.Tracer) string {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 	qb422016 := qt422016.AcquireByteBuffer()
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 	WriteTSDBStatusResponse(qb422016, status, qt)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 	qs422016 := string(qb422016.B)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 	qt422016.ReleaseByteBuffer(qb422016)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 	return qs422016
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:23
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:24
 }
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:25
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:26
 func streamtsdbStatusEntries(qw422016 *qt422016.Writer, a []storage.TopHeapEntry) {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:25
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:26
 	qw422016.N().S(`[`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:27
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:28
 	for i, e := range a {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:27
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:28
 		qw422016.N().S(`{"name":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:29
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:30
 		qw422016.N().Q(e.Name)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:29
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:30
 		qw422016.N().S(`,"value":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:30
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:31
 		qw422016.N().D(int(e.Count))
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:30
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:31
 		qw422016.N().S(`}`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:32
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:33
 		if i+1 < len(a) {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:32
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:33
 			qw422016.N().S(`,`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:32
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:33
 		}
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:33
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:34
 	}
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:33
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:34
 	qw422016.N().S(`]`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 }
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 func writetsdbStatusEntries(qq422016 qtio422016.Writer, a []storage.TopHeapEntry) {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 	qw422016 := qt422016.AcquireWriter(qq422016)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 	streamtsdbStatusEntries(qw422016, a)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 	qt422016.ReleaseWriter(qw422016)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 }
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 func tsdbStatusEntries(a []storage.TopHeapEntry) string {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 	qb422016 := qt422016.AcquireByteBuffer()
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 	writetsdbStatusEntries(qb422016, a)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 	qs422016 := string(qb422016.B)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 	qt422016.ReleaseByteBuffer(qb422016)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 	return qs422016
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:35
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:36
 }
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:37
-func streamtsdbStatusMetricNameEntries(qw422016 *qt422016.Writer, a []storage.TopHeapEntry, queryStats []storage.MetricNamesStatsRecord) {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:39
-	queryStatsByMetricName := make(map[string]storage.MetricNamesStatsRecord, len(queryStats))
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:38
+func streamtsdbStatusMetricNameEntries(qw422016 *qt422016.Writer, a []storage.TopHeapEntry, queryStats []metricnamestats.StatRecord) {
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:40
+	queryStatsByMetricName := make(map[string]metricnamestats.StatRecord, len(queryStats))
 	for _, record := range queryStats {
 		queryStatsByMetricName[record.MetricName] = record
 	}
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:43
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:44
 	qw422016.N().S(`[`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:45
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:46
 	for i, e := range a {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:45
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:46
 		qw422016.N().S(`{`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:48
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:49
 		entry, ok := queryStatsByMetricName[e.Name]
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:49
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:50
 		qw422016.N().S(`"name":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:50
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:51
 		qw422016.N().Q(e.Name)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:50
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:51
 		qw422016.N().S(`,`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:51
-		if !ok {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:51
-			qw422016.N().S(`"value":`)
 //line app/vmselect/prometheus/tsdb_status_response.qtpl:52
-			qw422016.N().D(int(e.Count))
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:53
-		} else {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:53
+		if !ok {
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:52
 			qw422016.N().S(`"value":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:54
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:53
 			qw422016.N().D(int(e.Count))
 //line app/vmselect/prometheus/tsdb_status_response.qtpl:54
+		} else {
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:54
+			qw422016.N().S(`"value":`)
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:55
+			qw422016.N().D(int(e.Count))
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:55
 			qw422016.N().S(`,"requestsCount":`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:55
-			qw422016.N().D(int(entry.RequestsCount))
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:55
-			qw422016.N().S(`,"lastRequestTimestamp":`)
 //line app/vmselect/prometheus/tsdb_status_response.qtpl:56
+			qw422016.N().D(int(entry.RequestsCount))
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:56
+			qw422016.N().S(`,"lastRequestTimestamp":`)
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:57
 			qw422016.N().D(int(entry.LastRequestTs))
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:57
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:58
 		}
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:57
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:58
 		qw422016.N().S(`}`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:59
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:60
 		if i+1 < len(a) {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:59
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:60
 			qw422016.N().S(`,`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:59
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:60
 		}
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:60
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:61
 	}
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:60
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:61
 	qw422016.N().S(`]`)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 }
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
-func writetsdbStatusMetricNameEntries(qq422016 qtio422016.Writer, a []storage.TopHeapEntry, queryStats []storage.MetricNamesStatsRecord) {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
+func writetsdbStatusMetricNameEntries(qq422016 qtio422016.Writer, a []storage.TopHeapEntry, queryStats []metricnamestats.StatRecord) {
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 	qw422016 := qt422016.AcquireWriter(qq422016)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 	streamtsdbStatusMetricNameEntries(qw422016, a, queryStats)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 	qt422016.ReleaseWriter(qw422016)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 }
 
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
-func tsdbStatusMetricNameEntries(a []storage.TopHeapEntry, queryStats []storage.MetricNamesStatsRecord) string {
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
+func tsdbStatusMetricNameEntries(a []storage.TopHeapEntry, queryStats []metricnamestats.StatRecord) string {
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 	qb422016 := qt422016.AcquireByteBuffer()
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 	writetsdbStatusMetricNameEntries(qb422016, a, queryStats)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 	qs422016 := string(qb422016.B)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 	qt422016.ReleaseByteBuffer(qb422016)
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 	return qs422016
-//line app/vmselect/prometheus/tsdb_status_response.qtpl:62
+//line app/vmselect/prometheus/tsdb_status_response.qtpl:63
 }

app/vmselect/searchutil/searchutil.go

@@ -132,9 +132,20 @@ func (d *Deadline) String() string {
 //
 //	{env="prod",team="devops",t1="v1",t2="v2"}
 //	{env=~"dev|staging",team!="devops",t1="v1",t2="v2"}
+//
+//	Query args from URL path have precedence over post form args.
 func GetExtraTagFilters(r *http.Request) ([][]storage.TagFilter, error) {
 	var tagFilters []storage.TagFilter
-	for _, match := range r.Form["extra_label"] {
+	urlQueryValues := r.URL.Query()
+	getRequestParam := func(key string) []string {
+		// query request param must always take precedence over form values
+		// in order to simplify security enforcement policy for extra_label and extra_filters
+		if uv, ok := urlQueryValues[key]; ok {
+			return uv
+		}
+		return r.Form[key]
+	}
+	for _, match := range getRequestParam("extra_label") {
 		tmp := strings.SplitN(match, "=", 2)
 		if len(tmp) != 2 {
 			return nil, fmt.Errorf("`extra_label` query arg must have the format `name=value`; got %q", match)
@@ -148,8 +159,8 @@ func GetExtraTagFilters(r *http.Request) ([][]storage.TagFilter, error) {
 			Value: []byte(tmp[1]),
 		})
 	}
-	extraFilters := append([]string{}, r.Form["extra_filters"]...)
-	extraFilters = append(extraFilters, r.Form["extra_filters[]"]...)
+	extraFilters := append([]string{}, getRequestParam("extra_filters")...)
+	extraFilters = append(extraFilters, getRequestParam("extra_filters[]")...)
 	if len(extraFilters) == 0 {
 		if len(tagFilters) == 0 {
 			return nil, nil

app/vmselect/searchutil/searchutil_test.go

@@ -20,6 +20,7 @@ func TestGetExtraTagFilters(t *testing.T) {
 		}
 		return &http.Request{
 			Form: q,
+			URL:  &url.URL{RawQuery: q.Encode()},
 		}
 	}
 	f := func(t *testing.T, r *http.Request, want []string, wantErr bool) {
@@ -79,6 +80,24 @@ func TestGetExtraTagFilters(t *testing.T) {
 		nil,
 		false,
 	)
+
+	formValues, err := url.ParseQuery(`extra_label=env=prod&extra_label=job=vmsingle&extra_label=tenant=prod&extra_filters[]={foo="bar"}&extra_filters[]={tenant="prod"}`)
+	if err != nil {
+		t.Fatalf("BUG: cannot parse query: %s", err)
+	}
+	urlValues, err := url.ParseQuery(`extra_label=job=vmagent&extra_label=env=dev&extra_filters[]={tenant="dev"}`)
+	if err != nil {
+		t.Fatalf("BUG: cannot parse query: %s", err)
+	}
+	httpReqWithBothFormAndURLParams := &http.Request{
+		Form: formValues,
+		URL: &url.URL{
+			RawQuery: urlValues.Encode(),
+		},
+	}
+	f(t, httpReqWithBothFormAndURLParams,
+		[]string{`{tenant="dev",job="vmagent",env="dev"}`},
+		false)
 }
 
 func TestParseMetricSelectorSuccess(t *testing.T) {

app/vmselect/stats/metric_names_usage_response.qtpl

@@ -1,11 +1,11 @@
 {% import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricnamestats"
 ) %}
 
 {% stripspace %}
 MetricNamesStatsResponse generates response for /api/v1/status/metric_names_stats .
-{% func MetricNamesStatsResponse(stats *storage.MetricNamesStatsResponse, qt *querytracer.Tracer) %}
+{% func MetricNamesStatsResponse(stats *metricnamestats.StatsResult, qt *querytracer.Tracer) %}
 {
 	"status":"success",
   "statsCollectedSince": {%dul= stats.CollectedSinceTs %},

app/vmselect/stats/metric_names_usage_response.qtpl.go

@@ -7,7 +7,7 @@ package stats
 //line app/vmselect/stats/metric_names_usage_response.qtpl:1
 import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricnamestats"
 )
 
 // MetricNamesStatsResponse generates response for /api/v1/status/metric_names_stats .
@@ -26,7 +26,7 @@ var (
 )
 
 //line app/vmselect/stats/metric_names_usage_response.qtpl:8
-func StreamMetricNamesStatsResponse(qw422016 *qt422016.Writer, stats *storage.MetricNamesStatsResponse, qt *querytracer.Tracer) {
+func StreamMetricNamesStatsResponse(qw422016 *qt422016.Writer, stats *metricnamestats.StatsResult, qt *querytracer.Tracer) {
 //line app/vmselect/stats/metric_names_usage_response.qtpl:8
 	qw422016.N().S(`{"status":"success","statsCollectedSince":`)
 //line app/vmselect/stats/metric_names_usage_response.qtpl:11
@@ -91,7 +91,7 @@ func StreamMetricNamesStatsResponse(qw422016 *qt422016.Writer, stats *storage.Me
 }
 
 //line app/vmselect/stats/metric_names_usage_response.qtpl:31
-func WriteMetricNamesStatsResponse(qq422016 qtio422016.Writer, stats *storage.MetricNamesStatsResponse, qt *querytracer.Tracer) {
+func WriteMetricNamesStatsResponse(qq422016 qtio422016.Writer, stats *metricnamestats.StatsResult, qt *querytracer.Tracer) {
 //line app/vmselect/stats/metric_names_usage_response.qtpl:31
 	qw422016 := qt422016.AcquireWriter(qq422016)
 //line app/vmselect/stats/metric_names_usage_response.qtpl:31
@@ -102,7 +102,7 @@ func WriteMetricNamesStatsResponse(qq422016 qtio422016.Writer, stats *storage.Me
 }
 
 //line app/vmselect/stats/metric_names_usage_response.qtpl:31
-func MetricNamesStatsResponse(stats *storage.MetricNamesStatsResponse, qt *querytracer.Tracer) string {
+func MetricNamesStatsResponse(stats *metricnamestats.StatsResult, qt *querytracer.Tracer) string {
 //line app/vmselect/stats/metric_names_usage_response.qtpl:31
 	qb422016 := qt422016.AcquireByteBuffer()
 //line app/vmselect/stats/metric_names_usage_response.qtpl:31

app/vmselect/vmui/index.html

@@ -37,11 +37,11 @@
   <meta property="og:title" content="UI for VictoriaMetrics">
   <meta property="og:url" content="https://victoriametrics.com/">
   <meta property="og:description" content="Explore and troubleshoot your VictoriaMetrics data">
-  <script type="module" crossorigin src="./assets/index-KEOgEEMl.js"></script>
+  <script type="module" crossorigin src="./assets/index-BjJ7fDL7.js"></script>
   <link rel="modulepreload" crossorigin href="./assets/rolldown-runtime-COnpUsM8.js">
-  <link rel="modulepreload" crossorigin href="./assets/vendor-Mr0bmX1E.js">
+  <link rel="modulepreload" crossorigin href="./assets/vendor-C8Kwp93_.js">
   <link rel="stylesheet" crossorigin href="./assets/vendor-CnsZ1jie.css">
-  <link rel="stylesheet" crossorigin href="./assets/index-D2OEy8Ra.css">
+  <link rel="stylesheet" crossorigin href="./assets/index-BL7jEFBa.css">
 </head>
 <body>
 <noscript>You need to enable JavaScript to run this app.</noscript>

app/vmstorage/main.go

@@ -5,6 +5,7 @@ import (
 	"flag"
 	"fmt"
 	"io"
+	"math"
 	"net/http"
 	"strconv"
 	"strings"
@@ -13,6 +14,7 @@ import (
 
 	"github.com/VictoriaMetrics/metrics"
 
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmstorage/promdb"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
@@ -22,6 +24,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/mergeset"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricnamestats"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricsmetadata"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/stringsutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/syncwg"
@@ -31,6 +34,8 @@ import (
 var (
 	retentionPeriod = flagutil.NewRetentionDuration("retentionPeriod", "1M", "Data with timestamps outside the retentionPeriod is automatically deleted. The minimum retentionPeriod is 24h or 1d. "+
 		"See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#retention. See also -retentionFilter")
+	futureRetention = flagutil.NewRetentionDuration("futureRetention", "2d", "Data with timestamps bigger than now+futureRetention is automatically deleted. "+
+		"The minimum futureRetention is 2 days. See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#retention")
 	snapshotAuthKey   = flagutil.NewPassword("snapshotAuthKey", "authKey, which must be passed in query string to /snapshot* pages. It overrides -httpAuth.*")
 	forceMergeAuthKey = flagutil.NewPassword("forceMergeAuthKey", "authKey, which must be passed in query string to /internal/force_merge pages. It overrides -httpAuth.*")
 	forceFlushAuthKey = flagutil.NewPassword("forceFlushAuthKey", "authKey, which must be passed in query string to /internal/force_flush pages. It overrides -httpAuth.*")
@@ -52,14 +57,16 @@ var (
 
 	logNewSeries = flag.Bool("logNewSeries", false, "Whether to log new series. This option is for debug purposes only. It can lead to performance issues "+
 		"when big number of new series are ingested into VictoriaMetrics")
-	denyQueriesOutsideRetention = flag.Bool("denyQueriesOutsideRetention", false, "Whether to deny queries outside the configured -retentionPeriod. "+
-		"When set, then /api/v1/query_range would return '503 Service Unavailable' error for queries with 'from' value outside -retentionPeriod. "+
+	denyQueriesOutsideRetention = flag.Bool("denyQueriesOutsideRetention", false, "Whether to deny queries outside the configured -retentionPeriod and -futureRetention. "+
+		"When set, then /api/v1/query_range will return an error for queries with 'from' value outside -retentionPeriod or 'to' value beyond -futureRetention. "+
 		"This may be useful when multiple data sources with distinct retentions are hidden behind query-tee")
-	maxHourlySeries = flag.Int("storage.maxHourlySeries", 0, "The maximum number of unique series can be added to the storage during the last hour. "+
+	maxHourlySeries = flag.Int64("storage.maxHourlySeries", 0, "The maximum number of unique series can be added to the storage during the last hour. "+
 		"Excess series are logged and dropped. This can be useful for limiting series cardinality. See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#cardinality-limiter . "+
+		fmt.Sprintf("Setting this flag to '-1' sets limit to maximum possible value (%d) which is useful in order to enable series tracking without enforcing limits. ", math.MaxInt32)+
 		"See also -storage.maxDailySeries")
-	maxDailySeries = flag.Int("storage.maxDailySeries", 0, "The maximum number of unique series can be added to the storage during the last 24 hours. "+
+	maxDailySeries = flag.Int64("storage.maxDailySeries", 0, "The maximum number of unique series can be added to the storage during the last 24 hours. "+
 		"Excess series are logged and dropped. This can be useful for limiting series churn rate. See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#cardinality-limiter . "+
+		fmt.Sprintf("Setting this flag to '-1' sets limit to maximum possible value (%d) which is useful in order to enable series tracking without enforcing limits. ", math.MaxInt32)+
 		"See also -storage.maxHourlySeries")
 
 	minFreeDiskSpaceBytes = flagutil.NewBytes("storage.minFreeDiskSpaceBytes", 100e6, "The minimum free disk space at -storageDataPath after which the storage stops accepting new data")
@@ -97,21 +104,6 @@ var (
 		"If set to 0 or a negative value, defaults to 1% of allowed memory.")
 )
 
-// CheckTimeRange returns true if the given tr is denied for querying.
-func CheckTimeRange(tr storage.TimeRange) error {
-	if !*denyQueriesOutsideRetention {
-		return nil
-	}
-	minAllowedTimestamp := int64(fasttime.UnixTimestamp()*1000) - retentionPeriod.Milliseconds()
-	if tr.MinTimestamp > minAllowedTimestamp {
-		return nil
-	}
-	return &httpserver.ErrorWithStatusCode{
-		Err:        fmt.Errorf("the given time range %s is outside the allowed -retentionPeriod=%s according to -denyQueriesOutsideRetention", &tr, retentionPeriod),
-		StatusCode: http.StatusServiceUnavailable,
-	}
-}
-
 // Init initializes vmstorage.
 func Init(resetCacheIfNeeded func(mrs []storage.MetricRow)) {
 	if err := encoding.CheckPrecisionBits(uint8(*precisionBits)); err != nil {
@@ -131,7 +123,12 @@ func Init(resetCacheIfNeeded func(mrs []storage.MetricRow)) {
 	mergeset.SetDataBlocksSparseCacheSize(cacheSizeIndexDBDataBlocksSparse.IntN())
 
 	if retentionPeriod.Duration() < 24*time.Hour {
-		logger.Fatalf("-retentionPeriod cannot be smaller than a day; got %s", retentionPeriod)
+		logger.Fatalf("-retentionPeriod cannot be smaller than a day; got %s. "+
+			"See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#retention", retentionPeriod)
+	}
+	if futureRetention.Duration() < 2*24*time.Hour {
+		logger.Fatalf("-futureRetention cannot be smaller than 2 days; got %s. "+
+			"See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#retention", futureRetention)
 	}
 	if *idbPrefillStart > 23*time.Hour {
 		logger.Panicf("-storage.idbPrefillStart cannot exceed 23 hours; got %s", idbPrefillStart)
@@ -140,13 +137,15 @@ func Init(resetCacheIfNeeded func(mrs []storage.MetricRow)) {
 	startTime := time.Now()
 	WG = syncwg.WaitGroup{}
 	opts := storage.OpenOptions{
-		Retention:             retentionPeriod.Duration(),
-		MaxHourlySeries:       *maxHourlySeries,
-		MaxDailySeries:        *maxDailySeries,
-		DisablePerDayIndex:    *disablePerDayIndex,
-		TrackMetricNamesStats: *trackMetricNamesStats,
-		IDBPrefillStart:       *idbPrefillStart,
-		LogNewSeries:          *logNewSeries,
+		Retention:                   retentionPeriod.Duration(),
+		FutureRetention:             futureRetention.Duration(),
+		DenyQueriesOutsideRetention: *denyQueriesOutsideRetention,
+		MaxHourlySeries:             getMaxHourlySeries(),
+		MaxDailySeries:              getMaxDailySeries(),
+		DisablePerDayIndex:          *disablePerDayIndex,
+		TrackMetricNamesStats:       *trackMetricNamesStats,
+		IDBPrefillStart:             *idbPrefillStart,
+		LogNewSeries:                *logNewSeries,
 	}
 	strg := storage.MustOpenStorage(*DataPath, opts)
 	Storage = strg
@@ -165,9 +164,12 @@ func Init(resetCacheIfNeeded func(mrs []storage.MetricRow)) {
 	// register storage metrics
 	storageMetrics = metrics.NewSet()
 	storageMetrics.RegisterMetricsWriter(func(w io.Writer) {
-		writeStorageMetrics(w, strg)
+		writeStorageMetrics(w, Storage)
 	})
 	metrics.RegisterSet(storageMetrics)
+
+	promdb.Init(retentionPeriod.Milliseconds())
+	fs.RegisterPathFsMetrics(*DataPath)
 }
 
 var storageMetrics *metrics.Set
@@ -233,7 +235,7 @@ func DeleteSeries(qt *querytracer.Tracer, tfss []*storage.TagFilters, maxMetrics
 }
 
 // GetMetricNamesStats returns metric names usage stats with give limit and lte predicate
-func GetMetricNamesStats(qt *querytracer.Tracer, limit, le int, matchPattern string) (storage.MetricNamesStatsResponse, error) {
+func GetMetricNamesStats(qt *querytracer.Tracer, limit, le int, matchPattern string) (metricnamestats.StatsResult, error) {
 	WG.Add(1)
 	r := Storage.GetMetricNamesStats(qt, limit, le, matchPattern)
 	WG.Done()
@@ -315,6 +317,7 @@ func Stop() {
 	logger.Infof("gracefully closing the storage at %s", *DataPath)
 	startTime := time.Now()
 	WG.WaitAndBlock()
+	promdb.MustClose()
 	stopStaleSnapshotsRemover()
 	Storage.MustClose()
 	logger.Infof("successfully closed the storage in %.3f seconds", time.Since(startTime).Seconds())
@@ -602,10 +605,10 @@ func writeStorageMetrics(w io.Writer, strg *storage.Storage) {
 	metrics.WriteCounterUint64(w, `vm_rows_ignored_total{reason="big_timestamp"}`, m.TooBigTimestampRows)
 	metrics.WriteCounterUint64(w, `vm_rows_ignored_total{reason="small_timestamp"}`, m.TooSmallTimestampRows)
 	metrics.WriteCounterUint64(w, `vm_rows_ignored_total{reason="invalid_raw_metric_name"}`, m.InvalidRawMetricNames)
-	if *maxHourlySeries > 0 {
+	if getMaxHourlySeries() > 0 {
 		metrics.WriteCounterUint64(w, `vm_rows_ignored_total{reason="hourly_limit_exceeded"}`, m.HourlySeriesLimitRowsDropped)
 	}
-	if *maxDailySeries > 0 {
+	if getMaxDailySeries() > 0 {
 		metrics.WriteCounterUint64(w, `vm_rows_ignored_total{reason="daily_limit_exceeded"}`, m.DailySeriesLimitRowsDropped)
 	}
 
@@ -615,13 +618,13 @@ func writeStorageMetrics(w io.Writer, strg *storage.Storage) {
 	metrics.WriteCounterUint64(w, `vm_slow_row_inserts_total`, m.SlowRowInserts)
 	metrics.WriteCounterUint64(w, `vm_slow_per_day_index_inserts_total`, m.SlowPerDayIndexInserts)
 
-	if *maxHourlySeries > 0 {
+	if getMaxHourlySeries() > 0 {
 		metrics.WriteGaugeUint64(w, `vm_hourly_series_limit_current_series`, m.HourlySeriesLimitCurrentSeries)
 		metrics.WriteGaugeUint64(w, `vm_hourly_series_limit_max_series`, m.HourlySeriesLimitMaxSeries)
 		metrics.WriteCounterUint64(w, `vm_hourly_series_limit_rows_dropped_total`, m.HourlySeriesLimitRowsDropped)
 	}
 
-	if *maxDailySeries > 0 {
+	if getMaxDailySeries() > 0 {
 		metrics.WriteGaugeUint64(w, `vm_daily_series_limit_current_series`, m.DailySeriesLimitCurrentSeries)
 		metrics.WriteGaugeUint64(w, `vm_daily_series_limit_max_series`, m.DailySeriesLimitMaxSeries)
 		metrics.WriteCounterUint64(w, `vm_daily_series_limit_rows_dropped_total`, m.DailySeriesLimitRowsDropped)
@@ -746,3 +749,21 @@ func jsonResponseError(w http.ResponseWriter, err error) {
 	errStr := err.Error()
 	fmt.Fprintf(w, `{"status":"error","msg":%s}`, stringsutil.JSONString(errStr))
 }
+
+func getMaxHourlySeries() int {
+	limit := *maxHourlySeries
+	if limit == -1 || limit > math.MaxInt32 {
+		return math.MaxInt32
+	}
+
+	return int(limit)
+}
+
+func getMaxDailySeries() int {
+	limit := *maxDailySeries
+	if limit == -1 || limit > math.MaxInt32 {
+		return math.MaxInt32
+	}
+
+	return int(limit)
+}

app/vmstorage/promdb/promdb.go

@@ -0,0 +1,268 @@
+package promdb
+
+import (
+	"context"
+	"flag"
+	"fmt"
+	"log/slog"
+	"time"
+
+	"github.com/oklog/ulid/v2"
+	"github.com/prometheus/prometheus/model/labels"
+	promstorage "github.com/prometheus/prometheus/storage"
+	"github.com/prometheus/prometheus/tsdb"
+	"github.com/prometheus/prometheus/tsdb/chunkenc"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/searchutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+)
+
+var prometheusDataPath = flag.String("prometheusDataPath", "", "Optional path to readonly historical Prometheus data")
+
+var prometheusRetentionMsecs int64
+
+// Init must be called after flag.Parse and before using the package.
+//
+// See also MustClose.
+func Init(retentionMsecs int64) {
+	if promDB != nil {
+		logger.Fatalf("BUG: promdb.Init is called multiple times without promdb.MustClose call")
+	}
+	prometheusRetentionMsecs = retentionMsecs
+	if *prometheusDataPath == "" {
+		return
+	}
+	l := slog.New(slog.Default().Handler())
+	opts := tsdb.DefaultOptions()
+	opts.RetentionDuration = retentionMsecs
+
+	// Set max block duration to 10% of retention period or 31 days
+	// according to https://prometheus.io/docs/prometheus/latest/storage/#compaction
+	maxBlockDuration := int64((31 * 24 * time.Hour) / time.Millisecond)
+	if maxBlockDuration > retentionMsecs/10 {
+		maxBlockDuration = retentionMsecs / 10
+	}
+	if maxBlockDuration < opts.MinBlockDuration {
+		maxBlockDuration = opts.MinBlockDuration
+	}
+	opts.MaxBlockDuration = maxBlockDuration
+
+	// Custom delete function is needed, because Prometheus by default doesn't delete
+	// blocks outside the retention if no new blocks are created with samples with the current timestamps.
+	// See https://github.com/prometheus/prometheus/blob/997bb7134fcfd7279f250e183e78681e48a56aff/tsdb/db.go#L1116
+	opts.BlocksToDelete = func(blocks []*tsdb.Block) map[ulid.ULID]struct{} {
+		m := make(map[ulid.ULID]struct{})
+		minRetentionTime := time.Now().Unix()*1000 - retentionMsecs
+		for _, block := range blocks {
+			meta := block.Meta()
+			// delete block marked for deletion by compaction code.
+			if meta.Compaction.Deletable {
+				m[meta.ULID] = struct{}{}
+				continue
+			}
+			if block.MaxTime() < minRetentionTime {
+				m[meta.ULID] = struct{}{}
+			}
+		}
+		return m
+	}
+	pdb, err := tsdb.Open(*prometheusDataPath, l, nil, opts, nil)
+	if err != nil {
+		logger.Panicf("FATAL: cannot open Prometheus data at -prometheusDataPath=%q: %s", *prometheusDataPath, err)
+	}
+	promDB = pdb
+	logger.Infof("successfully opened historical Prometheus data at -prometheusDataPath=%q with retentionMsecs=%d", *prometheusDataPath, retentionMsecs)
+}
+
+// MustClose must be called on graceful shutdown.
+//
+// Package functionality cannot be used after this call.
+func MustClose() {
+	if *prometheusDataPath == "" {
+		return
+	}
+	if promDB == nil {
+		logger.Panicf("BUG: promdb.MustClose is called without promdb.Init call")
+	}
+	if err := promDB.Close(); err != nil {
+		logger.Panicf("FATAL: cannot close promDB: %s", err)
+	}
+	promDB = nil
+	logger.Infof("successfully closed historical Prometheus data at -prometheusDataPath=%q", *prometheusDataPath)
+}
+
+var promDB *tsdb.DB
+
+// GetLabelNamesOnTimeRange returns label names.
+func GetLabelNamesOnTimeRange(tr storage.TimeRange, deadline searchutil.Deadline) ([]string, error) {
+	if *prometheusDataPath == "" {
+		return nil, nil
+	}
+	d := time.Unix(int64(deadline.Deadline()), 0)
+	ctx, cancel := context.WithDeadline(context.Background(), d)
+	defer cancel()
+	q, err := promDB.Querier(tr.MinTimestamp, tr.MaxTimestamp)
+	if err != nil {
+		return nil, err
+	}
+	defer mustCloseQuerier(q)
+
+	names, _, err := q.LabelNames(ctx, nil)
+	// Make full copy of names, since they cannot be used after q is closed.
+	names = copyStringsWithMemory(names)
+	return names, err
+}
+
+// GetLabelValuesOnTimeRange returns values for the given labelName on the given tr.
+func GetLabelValuesOnTimeRange(labelName string, tr storage.TimeRange, deadline searchutil.Deadline) ([]string, error) {
+	if *prometheusDataPath == "" {
+		return nil, nil
+	}
+	d := time.Unix(int64(deadline.Deadline()), 0)
+	ctx, cancel := context.WithDeadline(context.Background(), d)
+	defer cancel()
+	q, err := promDB.Querier(tr.MinTimestamp, tr.MaxTimestamp)
+	if err != nil {
+		return nil, err
+	}
+	defer mustCloseQuerier(q)
+
+	values, _, err := q.LabelValues(ctx, labelName, nil)
+	// Make full copy of values, since they cannot be used after q is closed.
+	values = copyStringsWithMemory(values)
+	return values, err
+}
+
+func copyStringsWithMemory(a []string) []string {
+	result := make([]string, len(a))
+	for i, s := range a {
+		result[i] = string(append([]byte{}, s...))
+	}
+	return result
+}
+
+// SeriesVisitor is called by VisitSeries for each matching time series.
+//
+// The caller shouldn't hold references to metricName, values and timestamps after returning.
+type SeriesVisitor func(metricName []byte, values []float64, timestamps []int64)
+
+// VisitSeries calls f for each series found in the pdb.
+func VisitSeries(sq *storage.SearchQuery, deadline searchutil.Deadline, f SeriesVisitor) error {
+	if *prometheusDataPath == "" {
+		return nil
+	}
+	d := time.Unix(int64(deadline.Deadline()), 0)
+	ctx, cancel := context.WithDeadline(context.Background(), d)
+	defer cancel()
+	minTime, maxTime := getSearchTimeRange(sq)
+	q, err := promDB.Querier(minTime, maxTime)
+	if err != nil {
+		return err
+	}
+	defer mustCloseQuerier(q)
+	var seriesSet []promstorage.SeriesSet
+	for _, tf := range sq.TagFilterss {
+		ms, err := convertTagFiltersToMatchers(tf)
+		if err != nil {
+			return fmt.Errorf("cannot convert tag filters to matchers: %w", err)
+		}
+		s := q.Select(ctx, false, nil, ms...)
+		seriesSet = append(seriesSet, s)
+	}
+	ss := promstorage.NewMergeSeriesSet(seriesSet, 0, promstorage.ChainedSeriesMerge)
+	var (
+		mn         storage.MetricName
+		metricName []byte
+		values     []float64
+		timestamps []int64
+	)
+	var it chunkenc.Iterator
+	for ss.Next() {
+		s := ss.At()
+		convertPromLabelsToMetricName(&mn, s.Labels())
+		metricName = mn.SortAndMarshal(metricName[:0])
+		values = values[:0]
+		timestamps = timestamps[:0]
+		it = s.Iterator(it)
+		for {
+			typ := it.Next()
+			if typ == chunkenc.ValNone {
+				break
+			}
+			if typ != chunkenc.ValFloat {
+				// Skip unsupported values
+				continue
+			}
+			ts, v := it.At()
+			values = append(values, v)
+			timestamps = append(timestamps, ts)
+		}
+		if err := it.Err(); err != nil {
+			return fmt.Errorf("error when iterating Prometheus series: %w", err)
+		}
+		f(metricName, values, timestamps)
+	}
+	return ss.Err()
+}
+
+func getSearchTimeRange(sq *storage.SearchQuery) (int64, int64) {
+	maxTime := sq.MaxTimestamp
+	minTime := sq.MinTimestamp
+	minRetentionTime := time.Now().Unix()*1000 - prometheusRetentionMsecs
+	if maxTime < minRetentionTime {
+		maxTime = minRetentionTime
+	}
+	if minTime < minRetentionTime {
+		minTime = minRetentionTime
+	}
+	return minTime, maxTime
+}
+
+func convertPromLabelsToMetricName(dst *storage.MetricName, ls labels.Labels) {
+	dst.Reset()
+	ls.Range(func(label labels.Label) {
+		if label.Name == "__name__" {
+			dst.MetricGroup = append(dst.MetricGroup[:0], label.Value...)
+		} else {
+			dst.AddTag(label.Name, label.Value)
+		}
+	})
+}
+
+func convertTagFiltersToMatchers(tfs []storage.TagFilter) ([]*labels.Matcher, error) {
+	ms := make([]*labels.Matcher, 0, len(tfs))
+	for _, tf := range tfs {
+		var mt labels.MatchType
+		if tf.IsNegative {
+			if tf.IsRegexp {
+				mt = labels.MatchNotRegexp
+			} else {
+				mt = labels.MatchNotEqual
+			}
+		} else {
+			if tf.IsRegexp {
+				mt = labels.MatchRegexp
+			} else {
+				mt = labels.MatchEqual
+			}
+		}
+		key := string(tf.Key)
+		if key == "" {
+			key = "__name__"
+		}
+		value := string(tf.Value)
+		m, err := labels.NewMatcher(mt, key, value)
+		if err != nil {
+			return nil, err
+		}
+		ms = append(ms, m)
+	}
+	return ms, nil
+}
+
+func mustCloseQuerier(q promstorage.Querier) {
+	if err := q.Close(); err != nil {
+		logger.Panicf("FATAL: cannot close querier: %s", err)
+	}
+}

app/vmui/Dockerfile-web

@@ -1,4 +1,4 @@
-FROM golang:1.26.1 AS build-web-stage
+FROM golang:1.26.3 AS build-web-stage
 COPY build /build
 
 WORKDIR /build
@@ -6,7 +6,7 @@ COPY web/ /build/
 RUN GOOS=linux GOARCH=amd64 CGO_ENABLED=0 go build -o web-amd64 github.com/VictoriMetrics/vmui/ && \
     GOOS=windows GOARCH=amd64 CGO_ENABLED=0 go build -o web-windows github.com/VictoriMetrics/vmui/
 
-FROM alpine:3.23.3
+FROM alpine:3.23.4
 USER root
 
 COPY --from=build-web-stage /build/web-amd64 /app/web

app/vmui/packages/vmui/package.json

@@ -23,14 +23,14 @@
     "classnames": "^2.5.1",
     "dayjs": "^1.11.20",
     "lodash.debounce": "^4.0.8",
-    "marked": "^17.0.5",
-    "preact": "^10.29.0",
-    "qs": "^6.15.0",
+    "marked": "^18.0.2",
+    "preact": "^10.29.1",
+    "qs": "^6.15.1",
     "react-input-mask": "^2.0.4",
-    "react-router-dom": "^7.13.2",
+    "react-router-dom": "^7.14.1",
     "uplot": "^1.6.32",
-    "vite": "^8.0.2",
-    "web-vitals": "^5.1.0"
+    "vite": "^8.0.8",
+    "web-vitals": "^5.2.0"
   },
   "devDependencies": {
     "@eslint/eslintrc": "^3.3.5",
@@ -39,24 +39,24 @@
     "@testing-library/jest-dom": "^6.9.1",
     "@testing-library/preact": "^3.2.4",
     "@types/lodash.debounce": "^4.0.9",
-    "@types/node": "^25.5.0",
+    "@types/node": "^25.6.0",
     "@types/qs": "^6.15.0",
     "@types/react": "^19.2.14",
     "@types/react-input-mask": "^3.0.6",
     "@types/react-router-dom": "^5.3.3",
-    "@typescript-eslint/eslint-plugin": "^8.57.2",
-    "@typescript-eslint/parser": "^8.57.2",
+    "@typescript-eslint/eslint-plugin": "^8.58.2",
+    "@typescript-eslint/parser": "^8.58.2",
     "cross-env": "^10.1.0",
     "eslint": "^9.39.2",
     "eslint-plugin-react": "^7.37.5",
     "eslint-plugin-unused-imports": "^4.4.1",
-    "globals": "^17.4.0",
+    "globals": "^17.5.0",
     "http-proxy-middleware": "^3.0.5",
-    "jsdom": "^29.0.1",
-    "postcss": "^8.5.8",
-    "sass-embedded": "^1.98.0",
-    "typescript": "^5.9.3",
-    "vitest": "^4.1.1"
+    "jsdom": "^29.0.2",
+    "postcss": "^8.5.10",
+    "sass-embedded": "^1.99.0",
+    "typescript": "^6.0.2",
+    "vitest": "^4.1.4"
   },
   "browserslist": {
     "production": [

app/vmui/packages/vmui/src/api/query-range.ts

@@ -16,23 +16,29 @@ export const getExportDataUrl = (server: string, query: string, period: TimePara
   return `${server}/api/v1/export?${params}`;
 };
 
-export const getExportCSVDataUrl = (server: string, query: string[], period: TimeParams, reduceMemUsage: boolean): string => {
+const getBaseParams = (period: TimeParams, query: string[]): URLSearchParams => {
   const params = new URLSearchParams({
     start: period.start.toString(),
     end: period.end.toString(),
-    format: "__name__,__value__,__timestamp__:unix_ms",
   });
   query.forEach((q => params.append("match[]", q)));
+  return params;
+};
+
+export const getLabelsUrl = (server: string, query: string[], period: TimeParams): string => {
+  const params = getBaseParams(period, query);
+  return `${server}/api/v1/labels?${params}`;
+};
+
+export const getExportCSVDataUrl = (server: string, query: string[], period: TimeParams, reduceMemUsage: boolean, format: string): string => {
+  const params = getBaseParams(period, query);
+  params.set("format", format);
   if (reduceMemUsage) params.set("reduce_mem_usage", "1");
   return `${server}/api/v1/export/csv?${params}`;
 };
 
 export const getExportJSONDataUrl = (server: string, query: string[], period: TimeParams, reduceMemUsage: boolean): string => {
-  const params = new URLSearchParams({
-    start: period.start.toString(),
-    end: period.end.toString(),
-  });
-  query.forEach((q => params.append("match[]", q)));
+  const params = getBaseParams(period, query);
   if (reduceMemUsage) params.set("reduce_mem_usage", "1");
   return `${server}/api/v1/export?${params}`;
 };

app/vmui/packages/vmui/src/api/raw-query.test.ts

@@ -0,0 +1,29 @@
+import { describe, expect, it, vi } from "vitest";
+import { fetchRawQueryCSVExport } from "./raw-query";
+
+describe("fetchRawQueryCSVExport", () => {
+  it.skip("requests all label columns before exporting CSV data", async () => {
+    const fetchMock = vi.fn()
+      .mockResolvedValueOnce({
+        ok: true,
+        json: async () => ({ data: ["job", "__name__", "instance"] }),
+      })
+      .mockResolvedValueOnce({
+        ok: true,
+        text: async () => "up,localhost:9100,node_exporter,1,1710000000000",
+      });
+
+    const result = await fetchRawQueryCSVExport(
+      "http://localhost:8428",
+      ["up"],
+      { start: 1710000000, end: 1710000300, step: "15s", date: "2024-03-09T16:05:00Z" },
+      false,
+      fetchMock as unknown as typeof fetch,
+    );
+
+    expect(fetchMock).toHaveBeenCalledTimes(2);
+    expect(fetchMock.mock.calls[0][0]).toBe("http://localhost:8428/api/v1/labels?start=1710000000&end=1710000300&match%5B%5D=up");
+    expect(fetchMock.mock.calls[1][0]).toBe("http://localhost:8428/api/v1/export/csv?start=1710000000&end=1710000300&match%5B%5D=up&format=__name__%2Cinstance%2Cjob%2C__value__%2C__timestamp__%3Aunix_ms");
+    expect(result).toBe("up,localhost:9100,node_exporter,1,1710000000000");
+  });
+});

app/vmui/packages/vmui/src/api/raw-query.ts

@@ -0,0 +1,31 @@
+import { getExportCSVDataUrl, getLabelsUrl } from "./query-range";
+import { TimeParams } from "../types";
+import { getCSVExportColumns } from "../utils/csv";
+
+interface LabelsResponse {
+  data?: string[];
+}
+
+export const fetchRawQueryCSVExport = async (
+  serverUrl: string,
+  query: string[],
+  period: TimeParams,
+  reduceMemUsage: boolean,
+  fetchFn: typeof fetch = fetch,
+): Promise<string> => {
+  const labelsResponse = await fetchFn(getLabelsUrl(serverUrl, query, period));
+  if (!labelsResponse.ok) {
+    throw new Error(await labelsResponse.text());
+  }
+
+  const { data = [] } = (await labelsResponse.json()) as LabelsResponse;
+  const columns = getCSVExportColumns(data);
+  const format = columns.join(",");
+
+  const response = await fetchFn(getExportCSVDataUrl(serverUrl, query, period, reduceMemUsage, format));
+  if (!response.ok) {
+    throw new Error(await response.text());
+  }
+
+  return await response.text();
+};

app/vmui/packages/vmui/src/api/types.ts

@@ -3,6 +3,7 @@ export interface MetricBase {
   metric: {
     [key: string]: string;
   };
+  nullTimestamps?: number[];
 }
 
 export interface MetricResult extends MetricBase {

app/vmui/packages/vmui/src/components/Chart/ChartTooltip/ChartTooltip.tsx

@@ -16,6 +16,7 @@ export interface ChartTooltipProps {
   point: { top: number, left: number };
   unit?: string;
   statsFormatted?: SeriesItemStatsFormatted;
+  description?: ReactNode;
   isSticky?: boolean;
   info?: ReactNode;
   marker?: string;
@@ -34,6 +35,7 @@ const ChartTooltip: FC<ChartTooltipProps> = ({
   unit = "",
   info,
   statsFormatted,
+  description,
   isSticky,
   marker,
   duplicateCount = 0,
@@ -173,6 +175,7 @@ const ChartTooltip: FC<ChartTooltipProps> = ({
           ))}
         </table>
       )}
+      {description && <p className="vm-chart-tooltip__description">{description}</p>}
       {info && <p className="vm-chart-tooltip__info">{info}</p>}
     </div>
   ), u.root);

app/vmui/packages/vmui/src/components/Chart/ChartTooltip/style.scss

@@ -143,4 +143,10 @@ $chart-tooltip-y: -1 * ($padding-global + $chart-tooltip-half-icon);
     word-break: break-all;
     white-space: pre-wrap;
   }
+
+  &__description {
+    word-break: break-word;
+    white-space: pre-wrap;
+    opacity: 0.85;
+  }
 }