make linter happy

In https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10607 vmagent
metrics relabling debug UI has been refined. There are a few changes
that make sence in VMUI too:
- Rename "Labels" to "A time series" to highlight that a single
Prometheus time series can be provided in text format.
- Rename "Relabel Configs" to "Config". It should be clear from the page
contex what config is meant. Also there are links to documention.
- Add placeholder. It hints what could be placed into "A time series"
textarea.

.github/workflows/build.yml

@@ -65,7 +65,7 @@ jobs:
             arch: amd64
     steps:
       - name: Code checkout
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           persist-credentials: false
 

.github/workflows/changelog-linter.yml

@@ -13,7 +13,7 @@ jobs:
       contents: read
     runs-on: 'ubuntu-latest'
     steps:
-      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           # needed for proper diff
           fetch-depth: 0

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

@@ -12,7 +12,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout code
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           fetch-depth: 0  # we need full history for commit verification
           persist-credentials: false

.github/workflows/check-licenses.yml

@@ -17,7 +17,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           persist-credentials: false
 

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

@@ -31,7 +31,7 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           persist-credentials: false
 
@@ -54,14 +54,14 @@ jobs:
           restore-keys: go-artifacts-${{ runner.os }}-codeql-analyze-${{ steps.go.outputs.go-version }}-
 
       - name: Initialize CodeQL
-        uses: github/codeql-action/init@e46ed2cbd01164d986452f91f178727624ae40d7  # v4.35.3
+        uses: github/codeql-action/init@54f647b7e1bb85c95cddabcd46b0c578ec92bc1a  # v4.36.3
         with:
           languages: go
 
       - name: Autobuild
-        uses: github/codeql-action/autobuild@e46ed2cbd01164d986452f91f178727624ae40d7  # v4.35.3
+        uses: github/codeql-action/autobuild@54f647b7e1bb85c95cddabcd46b0c578ec92bc1a  # v4.36.3
 
       - name: Perform CodeQL Analysis
-        uses: github/codeql-action/analyze@e46ed2cbd01164d986452f91f178727624ae40d7  # v4.35.3
+        uses: github/codeql-action/analyze@54f647b7e1bb85c95cddabcd46b0c578ec92bc1a  # v4.36.3
         with:
           category: 'language:go'

.github/workflows/docs.yaml

@@ -18,13 +18,13 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           path: __vm
           persist-credentials: false
 
       - name: Checkout private code
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           repository: VictoriaMetrics/vmdocs
           token: ${{ secrets.VM_BOT_GH_TOKEN }}

.github/workflows/test.yml

@@ -34,7 +34,7 @@ jobs:
     runs-on: 'vm-runner'
     steps:
       - name: Code checkout
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           persist-credentials: false
 
@@ -79,7 +79,7 @@ jobs:
 
     steps:
       - name: Code checkout
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           persist-credentials: false
 
@@ -106,7 +106,7 @@ jobs:
 
     steps:
       - name: Code checkout
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           persist-credentials: false
 

.github/workflows/vmui.yml

@@ -26,13 +26,11 @@ jobs:
   vmui-checks:
     name: VMUI Checks (lint, test, typecheck)
     permissions:
-      checks: write
       contents: read
-      pull-requests: read
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7.0.0
         with:
           persist-credentials: false
 
@@ -70,12 +68,6 @@ jobs:
         env:
           VMUI_SKIP_INSTALL: true
 
-      - name: Annotate Code Linting Results
-        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
-
       - name: Check overall status
         run: |
           echo "Lint status: ${{ steps.lint.outcome }}"

.golangci.yml

@@ -23,3 +23,4 @@ linters:
         text: 'SA(4003|1019|5011):'
     paths:
       - ^app/vmui/
+      - app/vmui/packages/vmui/node_modules/

Makefile

@@ -447,7 +447,7 @@ vet:
 	go vet ./app/...
 	go vet ./apptest/...
 
-check-all: fmt vet golangci-lint govulncheck
+check-all: fmt vet golangci-lint
 
 clean-checkers: remove-golangci-lint remove-govulncheck
 
@@ -471,8 +471,9 @@ test-full-386:
 
 apptest:
 	$(MAKE) victoria-metrics-race vmagent-race vmalert-race vmauth-race vmctl-race vmbackup-race vmrestore-race
-	go test ./apptest/... -skip="^Test(Cluster|Legacy).*"
+	go test ./apptest/... -skip="^Test(Cluster|Mixed|Legacy).*"
 
+# App tests for legacy indexDB
 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/'); \
@@ -489,6 +490,22 @@ apptest-legacy: victoria-metrics-race vmbackup-race vmrestore-race
 	VMSTORAGE_V1_132_0_PATH=$${DIR}/vmstorage-prod \
 	go test ./apptest/tests -run="^TestLegacySingle.*"
 
+# App tests for mixed setups where vmsingle and vmcluster coexist.
+apptest-mixed: victoria-metrics-race
+	OS=$$(uname | tr '[:upper:]' '[:lower:]'); \
+	ARCH=$$(uname -m | tr '[:upper:]' '[:lower:]' | sed 's/x86_64/amd64/'); \
+	VERSION=v1.147.0; \
+	VMSINGLE=victoria-metrics-$${OS}-$${ARCH}-$${VERSION}.tar.gz; \
+	VMCLUSTER=victoria-metrics-$${OS}-$${ARCH}-$${VERSION}-cluster.tar.gz; \
+	URL=https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/$${VERSION}; \
+	DIR=/tmp/$${VERSION}; \
+	test -d $${DIR} || (mkdir $${DIR} && \
+		curl --output-dir /tmp -LO $${URL}/$${VMSINGLE} && tar xzf /tmp/$${VMSINGLE} -C $${DIR} && \
+		curl --output-dir /tmp -LO $${URL}/$${VMCLUSTER} && tar xzf /tmp/$${VMCLUSTER} -C $${DIR} \
+	); \
+	VMSELECT_PATH=$${DIR}/vmselect-prod \
+	go test ./apptest/tests -run="^TestMixed.*"
+
 benchmark:
 	go test -run=NO_TESTS -bench=. ./lib/...
 	go test -run=NO_TESTS -bench=. ./app/...

app/victoria-metrics/main.go

@@ -89,7 +89,7 @@ func main() {
 	}
 	logger.Infof("starting VictoriaMetrics at %q...", listenAddrs)
 	startTime := time.Now()
-	vmstorage.Init(*vmselectMaxConcurrentRequests, promql.ResetRollupResultCacheIfNeeded)
+	vmstorage.Init(*vmselectMaxConcurrentRequests, *vmselectMaxQueueDuration, promql.ResetRollupResultCacheIfNeeded)
 	vmselect.Init(*vmselectMaxConcurrentRequests, *vmselectMaxQueueDuration)
 	vminsertcommon.StartIngestionRateLimiter(*maxIngestionRate)
 	vminsert.Init()

app/vmagent/influx/request_handler.go

@@ -51,7 +51,7 @@ func InsertHandlerForHTTP(at *auth.Token, req *http.Request) error {
 	}
 	q := req.URL.Query()
 	precision := q.Get("precision")
-	// Read db tag from https://docs.influxdata.com/influxdb/v1.7/tools/api/#write-http-endpoint
+	// Read db tag from https://docs.influxdata.com/influxdb/v1/api/write/#operation/PostWrite
 	db := q.Get("db")
 	encoding := req.Header.Get("Content-Encoding")
 	isStreamMode := req.Header.Get("Stream-Mode") == "1"

app/vmagent/main.go

@@ -462,7 +462,9 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 		return true
 	case "/prometheus/metric-relabel-debug", "/metric-relabel-debug":
 		promscrapeMetricRelabelDebugRequests.Inc()
-		promscrape.WriteMetricRelabelDebug(w, r)
+		rwGlobalRelabelConfigs := remotewrite.GetRemoteWriteRelabelConfigString()
+		rwURLRelabelConfigss := remotewrite.GetURLRelabelConfigString()
+		promscrape.WriteMetricRelabelDebug(w, r, rwGlobalRelabelConfigs, rwURLRelabelConfigss)
 		return true
 	case "/prometheus/target-relabel-debug", "/target-relabel-debug":
 		promscrapeTargetRelabelDebugRequests.Inc()

app/vmagent/opentelemetry/request_handler.go

@@ -63,6 +63,7 @@ func insertRows(at *auth.Token, tss []prompb.TimeSeries, mms []prompb.MetricMeta
 
 	rowsTotal := 0
 	tssDst := ctx.WriteRequest.Timeseries[:0]
+	mmsDst := ctx.WriteRequest.Metadata[:0]
 	labels := ctx.Labels[:0]
 	samples := ctx.Samples[:0]
 	for i := range tss {
@@ -82,7 +83,19 @@ func insertRows(at *auth.Token, tss []prompb.TimeSeries, mms []prompb.MetricMeta
 
 	var metadataTotal int
 	if prommetadata.IsEnabled() {
-		ctx.WriteRequest.Metadata = mms
+		for i := range mms {
+			mm := &mms[i]
+			mmsDst = append(mmsDst, prompb.MetricMetadata{
+				MetricFamilyName: mm.MetricFamilyName,
+				Help:             mm.Help,
+				Type:             mm.Type,
+				Unit:             mm.Unit,
+
+				AccountID: mm.AccountID,
+				ProjectID: mm.ProjectID,
+			})
+		}
+		ctx.WriteRequest.Metadata = mmsDst
 		metadataTotal = len(mms)
 	}
 

app/vmagent/promremotewrite/request_handler.go

@@ -35,6 +35,9 @@ func InsertHandler(at *auth.Token, req *http.Request) error {
 }
 
 func insertRows(at *auth.Token, timeseries []prompb.TimeSeries, mms []prompb.MetricMetadata, extraLabels []prompb.Label) error {
+	if len(extraLabels) == 0 && !prommetadata.IsEnabled() && at == nil {
+		return insertRowsFast(at, timeseries)
+	}
 	ctx := common.GetPushCtx()
 	defer common.PutPushCtx(ctx)
 
@@ -102,3 +105,17 @@ func insertRows(at *auth.Token, timeseries []prompb.TimeSeries, mms []prompb.Met
 	rowsPerInsert.Update(float64(rowsTotal))
 	return nil
 }
+
+func insertRowsFast(at *auth.Token, timeseries []prompb.TimeSeries) error {
+	rowsTotal := 0
+	for i := range timeseries {
+		rowsTotal += len(timeseries[i].Samples)
+	}
+	wr := &prompb.WriteRequest{Timeseries: timeseries}
+	if !remotewrite.TryPush(at, wr) {
+		return remotewrite.ErrQueueFullHTTPRetry
+	}
+	rowsInserted.Add(rowsTotal)
+	rowsPerInsert.Update(float64(rowsTotal))
+	return nil
+}

app/vmagent/remotewrite/client.go

@@ -151,17 +151,23 @@ func newHTTPClient(argIdx int, remoteWriteURL, sanitizedURL string, fq *persiste
 		}
 		tr.Proxy = http.ProxyURL(pu)
 	}
+
 	hc := &http.Client{
 		Transport: authCfg.NewRoundTripper(tr),
 		Timeout:   sendTimeout.GetOptionalArg(argIdx),
 	}
+	rwURL, err := url.Parse(remoteWriteURL)
+	if err != nil {
+		logger.Fatalf("BUG: cannot parse already parsed -remoteWrite.url=%q: %s", remoteWriteURL, err)
+	}
+	hc.Transport, rwURL = httputil.NewLoadBalancerTransport(hc.Transport, rwURL)
 	retryMaxIntervalFlag := retryMaxTime
 	if retryMaxInterval.String() != "" {
 		retryMaxIntervalFlag = retryMaxInterval
 	}
 	c := &client{
 		sanitizedURL:     sanitizedURL,
-		remoteWriteURL:   remoteWriteURL,
+		remoteWriteURL:   rwURL.String(),
 		authCfg:          authCfg,
 		awsCfg:           awsCfg,
 		fq:               fq,

app/vmagent/remotewrite/relabel.go

@@ -12,6 +12,7 @@ import (
 	"github.com/VictoriaMetrics/metrics"
 	"gopkg.in/yaml.v2"
 
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
@@ -82,6 +83,16 @@ func WriteRelabelConfigData(w io.Writer) {
 	_, _ = w.Write(*p)
 }
 
+// GetRemoteWriteRelabelConfigString returns -remoteWrite.relabelConfig contents in string
+func GetRemoteWriteRelabelConfigString() string {
+	var bb bytesutil.ByteBuffer
+	WriteRelabelConfigData(&bb)
+	if bb.Len() == 0 {
+		return ""
+	}
+	return string(bb.B)
+}
+
 // WriteURLRelabelConfigData writes -remoteWrite.urlRelabelConfig contents to w
 func WriteURLRelabelConfigData(w io.Writer) {
 	p := remoteWriteURLRelabelConfigData.Load()
@@ -108,6 +119,24 @@ func WriteURLRelabelConfigData(w io.Writer) {
 	_, _ = w.Write(d)
 }
 
+// GetURLRelabelConfigString returns -remoteWrite.urlRelabelConfig contents in []string
+func GetURLRelabelConfigString() []string {
+	p := remoteWriteURLRelabelConfigData.Load()
+	if p == nil {
+		return nil
+	}
+	var ss []string
+	for i := range *remoteWriteURLs {
+		cfgData := (*p)[i]
+		var cfgDataBytes []byte
+		if cfgData != nil {
+			cfgDataBytes, _ = yaml.Marshal(cfgData)
+		}
+		ss = append(ss, string(cfgDataBytes))
+	}
+	return ss
+}
+
 func reloadRelabelConfigs() {
 	rcs := allRelabelConfigs.Load()
 	if !rcs.isSet() {

app/vmagent/remotewrite/remotewrite.go

@@ -12,19 +12,18 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/cespare/xxhash/v2"
-
 	"github.com/VictoriaMetrics/metrics"
+	"github.com/cespare/xxhash/v2"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bloomfilter"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/consistenthash"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/mdx"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/memory"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/persistentqueue"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/procutil"
@@ -106,6 +105,9 @@ var (
 		"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")
+
+	enableMdx = flagutil.NewArrayBool("remoteWrite.mdx.enable", "Whether to only retain metrics from VictoriaMetrics services before sending them to the corresponding -remoteWrite.url. "+
+		"Please see https://docs.victoriametrics.com/victoriametrics/vmagent/#monitoring-data-exchange")
 )
 
 var (
@@ -162,8 +164,8 @@ func InitSecretFlags() {
 }
 
 var (
-	shardByURLLabelsMap       map[string]struct{}
-	shardByURLIgnoreLabelsMap map[string]struct{}
+	shardByURLLabelsFilter       []string
+	shardByURLIgnoreLabelsFilter []string
 )
 
 // Init initializes remotewrite.
@@ -210,8 +212,8 @@ func Init() {
 		logger.Fatalf("-remoteWrite.shardByURL.labels and -remoteWrite.shardByURL.ignoreLabels cannot be set simultaneously; " +
 			"see https://docs.victoriametrics.com/victoriametrics/vmagent/#sharding-among-remote-storages")
 	}
-	shardByURLLabelsMap = newMapFromStrings(*shardByURLLabels)
-	shardByURLIgnoreLabelsMap = newMapFromStrings(*shardByURLIgnoreLabels)
+	shardByURLLabelsFilter = slices.Clone(*shardByURLLabels)
+	shardByURLIgnoreLabelsFilter = slices.Clone(*shardByURLIgnoreLabels)
 
 	initLabelsGlobal()
 
@@ -307,6 +309,10 @@ func initRemoteWriteCtxs(urls []string) {
 	}
 	fs.RegisterPathFsMetrics(*tmpDataPath)
 
+	if slices.Contains(*enableMdx, true) && *shardByURL {
+		logger.Fatalf("-remoteWrite.mdx.enable and -remoteWrite.shardByURL cannot be set to true simultaneously.")
+	}
+
 	if *shardByURL {
 		consistentHashNodes := make([]string, 0, len(urls))
 		for i, url := range urls {
@@ -565,6 +571,14 @@ func tryPushMetadataToRemoteStorages(at *auth.Token, rwctxs []*remoteWriteCtx, m
 			mm.ProjectID = at.ProjectID
 		}
 	}
+	tmp := mms[:0]
+	for _, mm := range mms {
+		if timeserieslimits.IsMetricMetadataExceeding(&mm) {
+			continue
+		}
+		tmp = append(tmp, mm)
+	}
+	mms = tmp
 	// 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.
@@ -698,18 +712,18 @@ func shardAmountRemoteWriteCtx(tssBlock []prompb.TimeSeries, shards [][]prompb.T
 
 	for _, ts := range tssBlock {
 		hashLabels := ts.Labels
-		if len(shardByURLLabelsMap) > 0 {
+		if len(shardByURLLabelsFilter) > 0 {
 			hashLabels = tmpLabels.Labels[:0]
 			for _, label := range ts.Labels {
-				if _, ok := shardByURLLabelsMap[label.Name]; ok {
+				if slices.Contains(shardByURLLabelsFilter, label.Name) {
 					hashLabels = append(hashLabels, label)
 				}
 			}
 			tmpLabels.Labels = hashLabels
-		} else if len(shardByURLIgnoreLabelsMap) > 0 {
+		} else if len(shardByURLIgnoreLabelsFilter) > 0 {
 			hashLabels = tmpLabels.Labels[:0]
 			for _, label := range ts.Labels {
-				if _, ok := shardByURLIgnoreLabelsMap[label.Name]; !ok {
+				if !slices.Contains(shardByURLIgnoreLabelsFilter, label.Name) {
 					hashLabels = append(hashLabels, label)
 				}
 			}
@@ -810,34 +824,26 @@ var (
 // 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]
+	var d xxhash.Digest
+	d.Reset()
 	for _, label := range labels {
-		b = append(b, label.Name...)
-		b = append(b, label.Value...)
+		_, _ = d.WriteString(label.Name)
+		_, _ = d.WriteString(label.Value)
 	}
-	h := xxhash.Sum64(b)
-	bb.B = b
-	labelsHashBufPool.Put(bb)
-	return h
+	return d.Sum64()
 }
 
 func getLabelsHash(labels []prompb.Label) uint64 {
-	bb := labelsHashBufPool.Get()
-	b := bb.B[:0]
+	var d xxhash.Digest
+	d.Reset()
 	for _, label := range labels {
-		b = append(b, label.Name...)
-		b = append(b, '=')
-		b = append(b, label.Value...)
+		_, _ = d.WriteString(label.Name)
+		_, _ = d.WriteString("=")
+		_, _ = d.WriteString(label.Value)
 	}
-	h := xxhash.Sum64(b)
-	bb.B = b
-	labelsHashBufPool.Put(bb)
-	return h
+	return d.Sum64()
 }
 
-var labelsHashBufPool bytesutil.ByteBufferPool
-
 func logSkippedSeries(labels []prompb.Label, flagName string, flagValue int) {
 	select {
 	case <-logSkippedSeriesTicker.C:
@@ -862,6 +868,7 @@ type remoteWriteCtx struct {
 
 	sas          atomic.Pointer[streamaggr.Aggregators]
 	deduplicator *streamaggr.Deduplicator
+	mdxFilter    *mdx.Filter
 
 	streamAggrKeepInput bool
 	streamAggrDropInput bool
@@ -876,6 +883,7 @@ type remoteWriteCtx struct {
 
 	rowsPushedAfterRelabel *metrics.Counter
 	rowsDroppedByRelabel   *metrics.Counter
+	mdxRowsPreserved       *metrics.Counter
 
 	pushFailures                 *metrics.Counter
 	metadataDroppedOnPushFailure *metrics.Counter
@@ -972,7 +980,6 @@ func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, sanitizedURL string)
 	for i := range pss {
 		pss[i] = newPendingSeries(fq, &c.useVMProto, sf, rd)
 	}
-
 	rwctx := &remoteWriteCtx{
 		idx:            argIdx,
 		fq:             fq,
@@ -989,6 +996,16 @@ func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, sanitizedURL string)
 	}
 	rwctx.initStreamAggrConfig()
 
+	if enableMdx.GetOptionalArg(argIdx) {
+		mdxFilter := mdx.NewFilter()
+		rwctx.mdxFilter = mdxFilter
+		rwctx.mdxRowsPreserved = metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_mdx_rows_preserved_total{path=%q,url=%q}`, queuePath, sanitizedURL))
+		_ = metrics.NewGauge(fmt.Sprintf(`vmagent_remotewrite_mdx_tracked_instances{path=%q,url=%q}`, queuePath, sanitizedURL), func() float64 {
+			return float64(mdxFilter.VMInstancesCount())
+		})
+
+	}
+
 	return rwctx
 }
 
@@ -1002,6 +1019,11 @@ func (rwctx *remoteWriteCtx) MustStop() {
 		rwctx.deduplicator.MustStop()
 		rwctx.deduplicator = nil
 	}
+	if rwctx.mdxFilter != nil {
+		rwctx.mdxFilter.MustStop()
+		rwctx.mdxFilter = nil
+		rwctx.mdxRowsPreserved = nil
+	}
 
 	for _, ps := range rwctx.pss {
 		ps.MustStop()
@@ -1017,6 +1039,7 @@ func (rwctx *remoteWriteCtx) MustStop() {
 
 	rwctx.rowsPushedAfterRelabel = nil
 	rwctx.rowsDroppedByRelabel = nil
+
 }
 
 // TryPushTimeSeries sends tss series to the configured remote write endpoint
@@ -1024,16 +1047,41 @@ func (rwctx *remoteWriteCtx) MustStop() {
 // TryPushTimeSeries doesn't modify tss, so tss can be passed concurrently to TryPush across distinct rwctx instances.
 func (rwctx *remoteWriteCtx) TryPushTimeSeries(tss []prompb.TimeSeries, forceDropSamplesOnFailure bool) bool {
 	var rctx *relabelCtx
+	var mctx *mdx.Ctx
 	var v *[]prompb.TimeSeries
 	defer func() {
-		if rctx == nil {
-			return
+		if v != nil {
+			*v = prompb.ResetTimeSeries(tss)
+			tssPool.Put(v)
+		}
+		if rctx != nil {
+			putRelabelCtx(rctx)
+		}
+		if mctx != nil {
+			mdx.PutContext(mctx)
 		}
-		*v = prompb.ResetTimeSeries(tss)
-		tssPool.Put(v)
-		putRelabelCtx(rctx)
 	}()
 
+	copyTimeSeriesIfNeeded := func() {
+		if v == nil {
+			v = tssPool.Get().(*[]prompb.TimeSeries)
+			tss = append(*v, tss...)
+		}
+	}
+
+	if rwctx.mdxFilter != nil {
+		mctx = mdx.GetContext()
+		// Make a copy of tss before applying relabeling in order to prevent
+		// from affecting time series for other remoteWrite.mdx configs.
+		copyTimeSeriesIfNeeded()
+		tss = rwctx.mdxFilter.Filter(mctx, tss)
+		if len(tss) == 0 {
+			return true
+		}
+		rowsCount := getRowsCount(tss)
+		rwctx.mdxRowsPreserved.Add(rowsCount)
+	}
+
 	// Apply relabeling
 	rcs := allRelabelConfigs.Load()
 	pcs := rcs.perURL[rwctx.idx]
@@ -1043,8 +1091,7 @@ func (rwctx *remoteWriteCtx) TryPushTimeSeries(tss []prompb.TimeSeries, forceDro
 		// from affecting time series for other remoteWrite.url configs.
 		// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/467
 		// and https://github.com/VictoriaMetrics/VictoriaMetrics/issues/599
-		v = tssPool.Get().(*[]prompb.TimeSeries)
-		tss = append(*v, tss...)
+		copyTimeSeriesIfNeeded()
 		rowsCountBeforeRelabel := getRowsCount(tss)
 		tss = rctx.applyRelabeling(tss, pcs)
 		rowsCountAfterRelabel := getRowsCount(tss)
@@ -1062,8 +1109,7 @@ func (rwctx *remoteWriteCtx) TryPushTimeSeries(tss []prompb.TimeSeries, forceDro
 			if rctx == nil {
 				rctx = getRelabelCtx()
 				// Make a copy of tss before dropping aggregated series
-				v = tssPool.Get().(*[]prompb.TimeSeries)
-				tss = append(*v, tss...)
+				copyTimeSeriesIfNeeded()
 			}
 			tss = dropAggregatedSeries(tss, matchIdxs.B, rwctx.streamAggrDropInput)
 		} else if rwctx.streamAggrDropInput {
@@ -1071,8 +1117,7 @@ func (rwctx *remoteWriteCtx) TryPushTimeSeries(tss []prompb.TimeSeries, forceDro
 			if rctx == nil {
 				rctx = getRelabelCtx()
 				// Make a copy of tss before dropping aggregated series
-				v = tssPool.Get().(*[]prompb.TimeSeries)
-				tss = append(*v, tss...)
+				copyTimeSeriesIfNeeded()
 			}
 			tss = dropUnaggregatedSeries(tss, matchIdxs.B)
 		}
@@ -1191,15 +1236,6 @@ func getRowsCount(tss []prompb.TimeSeries) int {
 	}
 	return rowsCount
 }
-
-func newMapFromStrings(a []string) map[string]struct{} {
-	m := make(map[string]struct{}, len(a))
-	for _, s := range a {
-		m[s] = struct{}{}
-	}
-	return m
-}
-
 func getMaxHourlySeries() int {
 	limit := *maxHourlySeries
 	if limit == -1 || limit > math.MaxInt32 {

app/vmalert-tool/unittest/recording.go

@@ -6,7 +6,6 @@ import (
 	"net/url"
 	"reflect"
 	"sort"
-	"time"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/datasource"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
@@ -94,10 +93,3 @@ Outer:
 	}
 	return
 }
-
-func durationToTime(pd *promutil.Duration) time.Time {
-	if pd == nil {
-		return time.Time{}
-	}
-	return time.UnixMilli(pd.Duration().Milliseconds())
-}

app/vmalert-tool/unittest/unittest.go

@@ -44,12 +44,19 @@ import (
 var (
 	storagePath    string
 	httpListenAddr string
-	// insert series from 1970-01-01T00:00:00
-	testStartTime          = time.Unix(0, 0).UTC()
+	// Insert series from 2000-01-01T00:00:00.
+	testStartTime          = time.Date(2000, 1, 1, 0, 0, 0, 0, time.UTC)
 	testLogLevel           = "ERROR"
 	disableAlertgroupLabel bool
 )
 
+func durationToTime(pd *promutil.Duration) time.Time {
+	if pd == nil {
+		return testStartTime
+	}
+	return testStartTime.Add(pd.Duration())
+}
+
 const (
 	testStoragePath = "vmalert-unittest"
 )
@@ -282,7 +289,8 @@ func processFlags() {
 
 func setUp() {
 	const maxConcurrentRequests = 4
-	vmstorage.Init(maxConcurrentRequests, promql.ResetRollupResultCacheIfNeeded)
+	maxQueueDuration := 5 * time.Second
+	vmstorage.Init(maxConcurrentRequests, maxQueueDuration, promql.ResetRollupResultCacheIfNeeded)
 	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
 	defer cancel()
 	readyCheckFunc := func() bool {

app/vmalert/config/config_test.go

@@ -145,10 +145,10 @@ func TestRuleValidate(t *testing.T) {
 }
 
 func TestGroupValidate_Failure(t *testing.T) {
-	f := func(group *Group, validateExpressions bool, errStrExpected string) {
+	f := func(data []byte, validateExpressions bool, errStrExpected string) {
 		t.Helper()
 
-		err := group.Validate(nil, validateExpressions)
+		_, err := parse(map[string][]byte{"test.yaml": data}, nil, validateExpressions)
 		if err == nil {
 			t.Fatalf("expecting non-nil error")
 		}
@@ -158,275 +158,238 @@ func TestGroupValidate_Failure(t *testing.T) {
 		}
 	}
 
-	f(&Group{}, false, "group name must be set")
+	f([]byte(`
+groups:
+- name: ""
+`), false, "group name must be set")
 
-	f(&Group{
-		Name: "both record and alert are not set",
-		Rules: []Rule{
-			{
-				Expr: "sum(up == 0 ) by (host)",
-				For:  promutil.NewDuration(10 * time.Millisecond),
-			},
-			{
-				Expr: "sumSeries(time('foo.bar',10))",
-			},
-		},
-	}, false, "invalid rule")
+	f([]byte(`
+groups:
+- name: both record and alert are not set
+  rules:
+  - expr: "sum(up == 0 ) by (host)"
+    for: 10ms
+  - expr: "sumSeries(time('foo.bar',10))"
+`), false, "invalid rule")
 
-	f(&Group{
-		Name:     "negative interval",
-		Interval: promutil.NewDuration(-1),
-	}, false, "interval shouldn't be lower than 0")
+	f([]byte(`
+groups:
+- name: negative interval
+  interval: -1ms
+`), false, "interval shouldn't be lower than 0")
 
-	f(&Group{
-		Name:       "too big eval_offset",
-		Interval:   promutil.NewDuration(time.Minute),
-		EvalOffset: promutil.NewDuration(2 * time.Minute),
-	}, false, "eval_offset should be smaller than interval")
+	f([]byte(`
+groups:
+- name: too big eval_offset
+  interval: 1m
+  eval_offset: 2m
+`), false, "eval_offset should be smaller than interval")
 
-	f(&Group{
-		Name:       "too big negative eval_offset",
-		Interval:   promutil.NewDuration(time.Minute),
-		EvalOffset: promutil.NewDuration(-2 * time.Minute),
-	}, false, "eval_offset should be smaller than interval")
+	f([]byte(`
+groups:
+- name: too big negative eval_offset
+  interval: 1m
+  eval_offset: -2m
+`), false, "eval_offset should be smaller than interval")
 
-	limit := -1
-	f(&Group{
-		Name:  "wrong limit",
-		Limit: &limit,
-	}, false, "invalid limit")
+	f([]byte(`
+groups:
+- name: wrong limit
+  limit: -1
+`), false, "invalid limit")
 
-	f(&Group{
-		Name:        "wrong concurrency",
-		Concurrency: -1,
-	}, false, "invalid concurrency")
+	f([]byte(`
+groups:
+- name: wrong concurrency
+  concurrency: -1
+`), false, "invalid concurrency")
 
-	f(&Group{
-		Name: "test",
-		Rules: []Rule{
-			{
-				Alert: "alert",
-				Expr:  "up == 1",
-			},
-			{
-				Alert: "alert",
-				Expr:  "up == 1",
-			},
-		},
-	}, false, "duplicate")
+	f([]byte(`
+groups:
+- name: test
+  rules:
+  - alert: alert
+    expr: up == 1
+  - alert: alert
+    expr: up == 1
+`), false, "duplicate")
 
-	f(&Group{
-		Name: "test",
-		Rules: []Rule{
-			{Alert: "alert", Expr: "up == 1", Labels: map[string]string{
-				"summary": "{{ value|query }}",
-			}},
-			{Alert: "alert", Expr: "up == 1", Labels: map[string]string{
-				"summary": "{{ value|query }}",
-			}},
-		},
-	}, false, "duplicate")
+	f([]byte(`
+groups:
+- name: test
+  rules:
+  - alert: alert
+    expr: up == 1
+    labels:
+      summary: "{{ value|query }}"
+  - alert: alert
+    expr: up == 1
+    labels:
+      summary: "{{ value|query }}"
+`), false, "duplicate")
 
-	f(&Group{
-		Name: "test",
-		Rules: []Rule{
-			{Record: "record", Expr: "up == 1", Labels: map[string]string{
-				"summary": "{{ value|query }}",
-			}},
-			{Record: "record", Expr: "up == 1", Labels: map[string]string{
-				"summary": "{{ value|query }}",
-			}},
-		},
-	}, false, "duplicate")
+	f([]byte(`
+groups:
+- name: test
+  rules:
+  - record: record
+    expr: up == 1
+    labels:
+      summary: "{{ value|query }}"
+  - record: record
+    expr: up == 1
+    labels:
+      summary: "{{ value|query }}"
+`), false, "duplicate")
 
-	f(&Group{
-		Name: "test",
-		Rules: []Rule{
-			{Alert: "alert", Expr: "up == 1", Labels: map[string]string{
-				"summary": "{{ value|query }}",
-			}},
-			{Alert: "alert", Expr: "up == 1", Labels: map[string]string{
-				"description": "{{ value|query }}",
-			}},
-		},
-	}, false, "duplicate")
-
-	f(&Group{
-		Name: "test",
-		Rules: []Rule{
-			{Record: "alert", Expr: "up == 1", Labels: map[string]string{
-				"summary": "{{ value|query }}",
-			}},
-			{Alert: "alert", Expr: "up == 1", Labels: map[string]string{
-				"summary": "{{ value|query }}",
-			}},
-		},
-	}, false, "duplicate")
-
-	f(&Group{
-		Name: "test thanos",
-		Type: NewRawType("thanos"),
-		Rules: []Rule{
-			{Alert: "alert", Expr: "up == 1", Labels: map[string]string{
-				"description": "{{ value|query }}",
-			}},
-		},
-	}, true, "unknown datasource type")
+	f([]byte(`
+groups:
+- name: test thanos
+  type: thanos
+  rules:
+  - alert: alert
+    expr: up == 1
+    labels:
+      description: "{{ value|query }}"
+`), true, "unknown datasource type")
 
 	// validate expressions
-	f(&Group{
-		Name: "test prometheus expr",
-		Type: NewPrometheusType(),
-		Rules: []Rule{
-			{
-				Record: "record",
-				Expr:   "up | 0",
-			},
-		},
-	}, true, "bad MetricsQL expr")
+	f([]byte(`
+groups:
+- name: test prometheus expr
+  type: prometheus
+  rules:
+  - record: record
+    expr: "up | 0"
+`), true, "bad MetricsQL expr")
 
-	f(&Group{
-		Name: "test graphite expr",
-		Type: NewGraphiteType(),
-		Rules: []Rule{
-			{Alert: "alert", Expr: "up == 1", Labels: map[string]string{
-				"description": "some-description",
-			}},
-		},
-	}, true, "bad GraphiteQL expr")
+	f([]byte(`
+groups:
+- name: test graphite expr
+  type: graphite
+  rules:
+  - alert: alert
+    expr: up == 1
+    labels:
+      description: some-description
+`), true, "bad GraphiteQL expr")
 
-	f(&Group{
-		Name: "test vlogs expr",
-		Type: NewVLogsType(),
-		Rules: []Rule{
-			{Alert: "alert", Expr: "stats count(*) as requests"},
-		},
-	}, true, "bad LogsQL expr")
+	f([]byte(`
+groups:
+- name: test vlogs expr
+  type: vlogs
+  rules:
+  - alert: alert
+    expr: "stats count(*) as requests"
+`), true, "bad LogsQL expr")
 
-	f(&Group{
-		Name: "test vlogs expr",
-		Type: NewVLogsType(),
-		Rules: []Rule{
-			{Alert: "alert", Expr: "_time: 1m | stats by (path, _time: 1m) count(*) as requests"},
-		},
-	}, true, "bad LogsQL expr")
+	f([]byte(`
+groups:
+- name: test vlogs expr multipart
+  type: vlogs
+  rules:
+  - alert: alert
+    expr: "_time: 1m | stats by (path, _time: 1m) count(*) as requests"
+`), true, "bad LogsQL expr")
 
-	f(&Group{
-		Name: "test graphite with prometheus expr",
-		Type: NewGraphiteType(),
-		Rules: []Rule{
-			{
-				Record: "r1",
-				ID:     1,
-				Expr:   "sumSeries(time('foo.bar',10))",
-				For:    promutil.NewDuration(10 * time.Millisecond),
-			},
-			{
-				Record: "r2",
-				ID:     2,
-				Expr:   "sum(up == 0 ) by (host)",
-			},
-		},
-	}, true, "bad GraphiteQL expr")
+	f([]byte(`
+groups:
+- name: test graphite with prometheus expr
+  type: graphite
+  rules:
+  - record: r1
+    expr: "sumSeries(time('foo.bar',10))"
+    for: 10ms
+  - record: r2
+    expr: "sum(up == 0 ) by (host)"
+`), true, "bad GraphiteQL expr")
 
-	f(&Group{
-		Name: "test vlogs with prometheus exp",
-		Type: NewVLogsType(),
-		Rules: []Rule{
-			{
-				Record: "r1",
-				Expr:   "sum(up == 0 ) by (host)",
-				For:    promutil.NewDuration(10 * time.Millisecond),
-			},
-		},
-	}, true, "bad LogsQL expr")
+	f([]byte(`
+groups:
+- name: test vlogs with prometheus expr
+  type: vlogs
+  rules:
+  - record: r1
+    expr: "sum(up == 0 ) by (host)"
+    for: 10ms
+`), true, "bad LogsQL expr")
 
-	f(&Group{
-		Name: "test prometheus with vlogs exp",
-		Type: NewPrometheusType(),
-		Rules: []Rule{
-			{
-				Record: "r1",
-				Expr:   "* | stats by (path) count()",
-				For:    promutil.NewDuration(10 * time.Millisecond),
-			},
-		},
-	}, true, "bad MetricsQL expr")
+	f([]byte(`
+groups:
+- name: test prometheus with vlogs expr
+  type: prometheus
+  rules:
+  - record: r1
+    expr: "* | stats by (path) count()"
+    for: 10ms
+`), true, "bad MetricsQL expr")
 }
 
 func TestGroupValidate_Success(t *testing.T) {
-	f := func(group *Group, validateAnnotations, validateExpressions bool) {
+	f := func(data []byte, validateAnnotations, validateExpressions bool) {
 		t.Helper()
 
 		var validateTplFn ValidateTplFn
 		if validateAnnotations {
 			validateTplFn = notifier.ValidateTemplates
 		}
-		err := group.Validate(validateTplFn, validateExpressions)
+		_, err := parse(map[string][]byte{"test.yaml": data}, validateTplFn, validateExpressions)
 		if err != nil {
 			t.Fatalf("unexpected error: %s", err)
 		}
 	}
 
-	f(&Group{
-		Name: "test",
-		Rules: []Rule{
-			{
-				Record: "record",
-				Expr:   "up | 0",
-			},
-		},
-	}, false, false)
+	f([]byte(`
+groups:
+- name: test
+  rules:
+  - record: record
+    expr: "up | 0"
+`), false, false)
 
-	f(&Group{
-		Name: "test",
-		Rules: []Rule{
-			{
-				Alert: "alert",
-				Expr:  "up == 1",
-				Labels: map[string]string{
-					"summary": "{{ value|query }}",
-				},
-			},
-		},
-	}, false, false)
+	f([]byte(`
+groups:
+- name: test
+  rules:
+  - alert: alert
+    expr: up == 1
+    labels:
+      summary: "{{ value|query }}"
+`), false, false)
 
 	// validate annotations
-	f(&Group{
-		Name: "test",
-		Rules: []Rule{
-			{
-				Alert: "alert",
-				Expr:  "up == 1",
-				Labels: map[string]string{
-					"summary": `
-{{ with printf "node_memory_MemTotal{job='node',instance='%s'}" "localhost" | query }}
-  {{ . | first | value | humanize1024 }}B
-{{ end }}`,
-				},
-			},
-		},
-	}, true, false)
+	f([]byte(`
+groups:
+- name: test
+  rules:
+  - alert: alert
+    expr: up == 1
+    labels:
+      summary: "\n{{ with printf \"node_memory_MemTotal{job='node',instance='%s'}\" \"localhost\" | query }}\n  {{ . | first | value | humanize1024 }}B\n{{ end }}"
+`), true, false)
 
 	// validate expressions
-	f(&Group{
-		Name: "test prometheus",
-		Type: NewPrometheusType(),
-		Rules: []Rule{
-			{Alert: "alert", Expr: "up == 1", Labels: map[string]string{
-				"description": "{{ value|query }}",
-			}},
-		},
-	}, false, true)
-	f(&Group{
-		Name: "test victorialogs",
-		Type: NewVLogsType(),
-		Rules: []Rule{
-			{Alert: "alert", Expr: " _time: 1m | stats count(*) as requests", Labels: map[string]string{
-				"description": "{{ value|query }}",
-			}},
-		},
-	}, false, true)
+	f([]byte(`
+groups:
+- name: test prometheus
+  type: prometheus
+  rules:
+  - alert: alert
+    expr: up == 1
+    labels:
+      description: "{{ value|query }}"
+`), false, true)
+
+	f([]byte(`
+groups:
+- name: test victorialogs
+  type: vlogs
+  rules:
+  - alert: alert
+    expr: " _time: 1m | stats count(*) as requests"
+    labels:
+      description: "{{ value|query }}"
+`), false, true)
 }
 
 func TestHashRule_NotEqual(t *testing.T) {

app/vmalert/datasource/init.go

@@ -11,6 +11,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/vmalertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 )
 
@@ -94,6 +95,12 @@ func Init(extraParams url.Values) (QuerierBuilder, error) {
 		tr.MaxIdleConns = tr.MaxIdleConnsPerHost
 	}
 	tr.IdleConnTimeout = *idleConnectionTimeout
+	hc := &http.Client{Transport: tr}
+	datasourceURL, err := url.Parse(*addr)
+	if err != nil {
+		logger.Fatalf("BUG: cannot parse already parsed -datasource.url=%q: %s", *addr, err)
+	}
+	hc.Transport, datasourceURL = httputil.NewLoadBalancerTransport(tr, datasourceURL)
 
 	if extraParams == nil {
 		extraParams = url.Values{}
@@ -120,9 +127,9 @@ func Init(extraParams url.Values) (QuerierBuilder, error) {
 	}
 
 	return &Client{
-		c:                &http.Client{Transport: tr},
+		c:                hc,
 		authCfg:          authCfg,
-		datasourceURL:    strings.TrimSuffix(*addr, "/"),
+		datasourceURL:    strings.TrimSuffix(datasourceURL.String(), "/"),
 		appendTypePrefix: *appendTypePrefix,
 		queryStep:        *queryStep,
 		extraParams:      extraParams,

app/vmalert/remoteread/init.go

@@ -4,12 +4,14 @@ import (
 	"flag"
 	"fmt"
 	"net/http"
+	"net/url"
 	"time"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/datasource"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/vmalertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 )
 
@@ -76,7 +78,13 @@ func Init() (datasource.QuerierBuilder, error) {
 		return nil, fmt.Errorf("failed to create transport for -remoteRead.url=%q: %w", *addr, err)
 	}
 	tr.IdleConnTimeout = *idleConnectionTimeout
+	c := &http.Client{Transport: tr}
+	rrURL, err := url.Parse(*addr)
+	if err != nil {
+		logger.Fatalf("BUG: cannot parse already parsed -remoteRead.url=%q: %s", *addr, err)
+	}
 
+	c.Transport, rrURL = httputil.NewLoadBalancerTransport(tr, rrURL)
 	endpointParams, err := flagutil.ParseJSONMap(*oauth2EndpointParams)
 	if err != nil {
 		return nil, fmt.Errorf("cannot parse JSON for -remoteRead.oauth2.endpointParams=%s: %w", *oauth2EndpointParams, err)
@@ -89,6 +97,5 @@ func Init() (datasource.QuerierBuilder, error) {
 	if err != nil {
 		return nil, fmt.Errorf("failed to configure auth: %w", err)
 	}
-	c := &http.Client{Transport: tr}
-	return datasource.NewPrometheusClient(*addr, authCfg, false, c), nil
+	return datasource.NewPrometheusClient(rrURL.String(), authCfg, false, c), nil
 }

app/vmalert/remotewrite/client.go

@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"io"
 	"net/http"
+	"net/url"
 	"path"
 	"strings"
 	"sync"
@@ -111,12 +112,18 @@ func NewClient(ctx context.Context, cfg Config) (*Client, error) {
 	if cfg.Concurrency > 0 {
 		cc = cfg.Concurrency
 	}
+	hc := &http.Client{
+		Timeout:   *sendTimeout,
+		Transport: cfg.Transport,
+	}
+	rwURL, err := url.Parse(cfg.Addr)
+	if err != nil {
+		logger.Fatalf("cannot parse already parsed -remoteWrite.url=%q: %s", cfg.Addr, err)
+	}
+	hc.Transport, rwURL = httputil.NewLoadBalancerTransport(hc.Transport, rwURL)
 	c := &Client{
-		c: &http.Client{
-			Timeout:   *sendTimeout,
-			Transport: cfg.Transport,
-		},
-		addr:          strings.TrimSuffix(cfg.Addr, "/"),
+		c:             hc,
+		addr:          strings.TrimSuffix(rwURL.String(), "/"),
 		authCfg:       cfg.AuthCfg,
 		flushInterval: cfg.FlushInterval,
 		maxBatchSize:  cfg.MaxBatchSize,

app/vmalert/rule/group.go

@@ -97,6 +97,7 @@ type groupMetrics struct {
 	iterationMissed   *metrics.Counter
 	iterationReset    *metrics.Counter
 	iterationInterval *metrics.Gauge
+	iterationLimit    *metrics.Gauge
 }
 
 // merges group rule labels into result map
@@ -336,6 +337,12 @@ func (g *Group) Init() {
 		i := g.Interval.Seconds()
 		return i
 	})
+	g.metrics.iterationLimit = g.metrics.set.NewGauge(fmt.Sprintf(`vmalert_rule_group_results_limit{%s}`, labels), func() float64 {
+		g.mu.RLock()
+		limit := g.Limit
+		g.mu.RUnlock()
+		return float64(limit)
+	})
 	for i := range g.Rules {
 		g.Rules[i].registerMetrics(g.metrics.set)
 	}

app/vmalert/rule/recording_test.go

@@ -457,12 +457,10 @@ func TestSetIntervalAsTimeFilter(t *testing.T) {
 	f(`* | count()`, "vlogs", true)
 	f(`error OR _time:5m  | count()`, "vlogs", true)
 	f(`(_time: 5m AND error) OR (_time: 5m AND warn) | count()`, "vlogs", true)
-	f(`* | error OR _time:5m | count()`, "vlogs", true)
 
 	f(`_time:5m | count()`, "vlogs", false)
 	f(`_time:2023-04-25T22:45:59Z | count()`, "vlogs", false)
 	f(`error AND _time:5m | count()`, "vlogs", false)
-	f(`* | error AND _time:5m | count()`, "vlogs", false)
 }
 
 func TestRecordingRuleExec_Partial(t *testing.T) {

app/vmalert/rule/web.go

@@ -75,10 +75,13 @@ type GroupAlerts struct {
 // ApiRule represents a Rule for web view
 // see https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#get-apiv1rules
 type ApiRule struct {
-	// State must be one of these under following scenarios
-	//  "pending": at least 1 alert in the rule in pending state and no other alert in firing ruleState.
-	//  "firing": at least 1 alert in the rule in firing state.
-	//  "inactive": no alert in the rule in firing or pending state.
+	// Rule state must be one of these under following scenarios:
+	//  "pending": at least 1 alert in the rule in pending state and no other alert in firing state. (only for alerting rules)
+	//  "firing": at least 1 alert in the rule in firing state. (only for alerting rules)
+	//  "inactive": rule's last evaluation was successful but no alert in the rule in firing or pending state. (only for alerting rules)
+	//  "unhealthy": rule's last evaluation was failed with error. (for both alerting and recording rules)
+	// 	"nomatch": rule's last evaluation was successful but no time series matched the rule's expression. (for both alerting and recording rules)
+	//  "ok": the recording rule's last evaluation was successful. (only for recording rules)
 	State string `json:"state"`
 	Name  string `json:"name"`
 	// Query represents Rule's `expression` field
@@ -237,6 +240,7 @@ func NewAlertAPI(ar *AlertingRule, a *notifier.Alert) *ApiAlert {
 }
 
 func (r *ApiRule) ExtendState() {
+	// if alerting rule already has alerts, then state is already set to either "pending" or "firing" and we don't need to change it
 	if len(r.Alerts) > 0 {
 		return
 	}

app/vmalert/web.go

@@ -115,7 +115,7 @@ func (rh *requestHandler) handler(w http.ResponseWriter, r *http.Request) bool {
 		}
 		WriteRule(w, r, rule)
 		return true
-	// current used by old vmalert UI and Grafana Alerts
+	// used by old vmalert UI
 	case "/vmalert/groups", "/rules":
 		rf, err := newRulesFilter(r)
 		if err != nil {
@@ -128,6 +128,8 @@ func (rh *requestHandler) handler(w http.ResponseWriter, r *http.Request) bool {
 			state = rf.states[0]
 			rf.states = rf.states[:1]
 		}
+		// enable extendedStates by default for vmalert UI
+		rf.extendedStates = true
 		lr := rh.groups(rf)
 		WriteListGroups(w, r, lr.Data.Groups, state)
 		return true
@@ -543,6 +545,8 @@ func (rh *requestHandler) groups(rf *rulesFilter) *listGroupsResponse {
 			if !groupFound && !strings.Contains(strings.ToLower(rule.Name), rf.search) {
 				continue
 			}
+			// extendedStates is used by the vmalert UI to extend the rule state with values such as "nomatch" and "unhealthy".
+			// those states are not supported by Grafana and not part of the Prometheus API spec yet
 			if rf.extendedStates {
 				rule.ExtendState()
 			}

app/vmalert/web.qtpl

@@ -130,9 +130,12 @@
                             data-bs-target="#item-{%s g.ID %}"
                         >
                             <span class="d-flex gap-2">
-                                {% if g.States["unhealthy"] > 0 %}<span class="badge bg-danger" title="Number of rules with status Error">{%d g.States["unhealthy"] %}</span> {% endif %}
-                                {% if g.States["nomatch"] > 0 %}<span class="badge bg-warning" title="Number of rules with status NoMatch">{%d g.States["nomatch"] %}</span> {% endif %}
-                                <span class="badge bg-success" title="Number of rules with status Ok">{%d g.States["ok"] %}</span>
+                                {% if g.States["inactive"] > 0 %}<span class="badge bg-light text-success border border-success" title="None of the alert instances is in a pending or firing state">{%d g.States["inactive"] %} inactive</span> {% endif %}
+                                {% if g.States["pending"] > 0 %}<span class="badge bg-warning text-dark" title="At least one alert instance is pending">{%d g.States["pending"] %} pending</span> {% endif %}
+                                {% if g.States["firing"] > 0 %}<span class="badge bg-danger" title="At least one alert instance is firing">{%d g.States["firing"] %} firing</span> {% endif %}
+                                {% if g.States["ok"] > 0 %}<span class="badge bg-success" title="Recording rule last evaluation succeeded">{%d g.States["ok"] %} ok</span>{% endif %}
+                                {% if g.States["unhealthy"] > 0 %}<span class="badge bg-danger" title="Last evaluation failed with an error">{%d g.States["unhealthy"] %} unhealthy</span> {% endif %}
+                                {% if g.States["nomatch"] > 0 %}<span class="badge bg-warning" title="Rule expression matched no time series">{%d g.States["nomatch"] %} nomatch</span> {% endif %}
                             </span>
                         </span>
                     </span>
@@ -169,7 +172,7 @@
                             <thead>
                                 <tr>
                                     <th scope="col" class="w-60">Rule</th>
-                                    <th scope="col" class="w-20" class="text-center" title="How many series were produced by the rule">Series</th>
+                                    <th scope="col" class="w-20" class="text-center" title="How many series were produced by the rule">Series Returned</th>
                                     <th scope="col" class="w-20" class="text-center" title="How many seconds ago rule was executed">Updated</th>
                                 </tr>
                             </thead>
@@ -188,8 +191,21 @@
                                                     {% else %}
                                                         <b>record:</b> {%s r.Name %}
                                                     {% endif %}
-                                                    |
+                                                    {% if r.State == "inactive" %}
+                                                        <span><a class="badge bg-success">{%s r.State %}</a></span>
+                                                    {% elseif  r.State == "pending" %}
+                                                        <span><a class="badge bg-warning">{%s r.State %}</a></span>
+                                                    {% elseif  r.State == "firing" %}
+                                                        <span><a class="badge bg-danger">{%s r.State %}</a></span>
+                                                    {% elseif  r.State == "ok" %}
+                                                        <span><a class="badge bg-success">{%s r.State %}</a></span>
+                                                    {% elseif  r.State == "unhealthy" %}
+                                                        <span><a class="badge bg-danger">{%s r.State %}</a></span>
+                                                    {% elseif  r.State == "nomatch" %}
+                                                        <span><a class="badge bg-warning">{%s r.State %}</a></span>
+                                                    {% endif %}
                                                     {%= seriesFetchedWarn(prefix, &r) %}
+                                                    |
                                                     <span><a target="_blank" href="{%s prefix+r.WebLink() %}">Details</a></span>
                                                 </div>
                                                 <div class="col-12">

app/vmauth/auth_config.go

@@ -118,9 +118,10 @@ type AccessLogFilters struct {
 }
 
 func (ui *UserInfo) logRequest(r *http.Request, userName string, statusCode int, duration time.Duration) {
-	if ui.AccessLog == nil {
+	if ui == nil || ui.AccessLog == nil {
 		return
 	}
+
 	filters := ui.AccessLog.Filters
 	if filters != nil && len(filters.SkipStatusCodes) > 0 {
 		if slices.Contains(filters.SkipStatusCodes, statusCode) {
@@ -134,6 +135,17 @@ func (ui *UserInfo) logRequest(r *http.Request, userName string, statusCode int,
 		r.Host, requestURI, statusCode, remoteAddr, r.UserAgent(), r.Referer(), duration.Milliseconds(), userName)
 }
 
+// hasAnyURLs reports whether ui has at least one backend URL route configured.
+// It is used only for unauthorized_user config section, since other users
+// must always have either URLPrefix or URLMaps set.
+func (ui *UserInfo) hasAnyURLs() bool {
+	if ui == nil {
+		return false
+	}
+
+	return ui.URLPrefix != nil || len(ui.URLMaps) > 0 || ui.DefaultURL != nil
+}
+
 // HeadersConf represents config for request and response headers.
 type HeadersConf struct {
 	RequestHeaders     []*Header `yaml:"headers,omitempty"`
@@ -983,8 +995,11 @@ func parseAuthConfig(data []byte) (*AuthConfig, error) {
 		if err := parseJWTPlaceholdersForUserInfo(ui, false); err != nil {
 			return nil, err
 		}
-		if err := ui.initURLs(); err != nil {
-			return nil, err
+
+		if ui.hasAnyURLs() {
+			if err := ui.initURLs(); err != nil {
+				return nil, err
+			}
 		}
 
 		metricLabels, err := ui.getMetricLabels()

app/vmauth/example_config.yml

@@ -140,6 +140,18 @@ users:
   - "ProjectID: {{.MetricsProjectID}}"
   url_prefix: "http://vminsert:8480/insert/prometheus"
 
+# JWT-based routing that relies solely on custom claims.
+# The `vm_access` claim is missing, default value will be used.
+# e.g. {"role": "admin"}.
+- name: jwt-custom-claims
+  jwt:
+    skip_verify: true
+    vm_default_access_claim:
+        metrics_account_id: 1
+    match_claims:
+      role: admin
+  url_prefix: "http://vmselect-admin:8481/select/0/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

@@ -65,6 +65,8 @@ type JWTConfig struct {
 	MatchClaims       map[string]string `yaml:"match_claims,omitempty"`
 	parsedMatchClaims []*jwt.Claim
 
+	DefaultVMAccessClaim *jwt.VMAccessClaim `yaml:"default_vm_access_claim,omitempty"`
+
 	// verifierPool is used to verify JWT tokens.
 	// It is initialized from PublicKeys and/or PublicKeyFiles.
 	// In this case, it is initialized once at config reload and never updated until next reload
@@ -432,7 +434,6 @@ func validateJWTPlaceholdersForURL(up *URLPrefix, isAllowed bool) error {
 			}
 			if strings.Contains(p, placeholderPrefix) {
 				return fmt.Errorf("invalid placeholder found in URL request path: %q, supported values are: %s", bu.Path, strings.Join(allPlaceholders, ", "))
-
 			}
 		}
 		for param, values := range bu.Query() {
@@ -487,7 +488,6 @@ func hasAnyPlaceholders(u *url.URL) bool {
 				return true
 			}
 		}
-
 	}
 	return false
 }

app/vmauth/main.go

@@ -6,6 +6,7 @@ import (
 	"flag"
 	"fmt"
 	"io"
+	"math/rand/v2"
 	"net"
 	"net/http"
 	"net/textproto"
@@ -31,6 +32,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/procutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/pushmetrics"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timerpool"
 )
 
 var (
@@ -173,11 +175,12 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 	if len(ats) == 0 {
 		// Process requests for unauthorized users
 		ui := authConfig.Load().UnauthorizedUser
-		if ui != nil {
+		if ui.hasAnyURLs() {
 			processUserRequest(w, r, ui, nil)
 			return true
 		}
 
+		ui.logRequest(r, `unauthorized`, http.StatusUnauthorized, 0)
 		handleMissingAuthorizationError(w)
 		return true
 	}
@@ -191,17 +194,23 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 			logger.Panicf("BUG: unexpected nil jwt token for user %q", ui.name())
 		}
 		defer putToken(tkn)
-		processUserRequest(w, r, ui, tkn)
-		return true
+		// Call processUserRequest only if the token contains the vm_access claim
+		// or a default claim is configured; otherwise fall through to unauthorized_user.
+		if tkn.HasVMAccessClaim() || ui.JWT.DefaultVMAccessClaim != nil {
+			processUserRequest(w, r, ui, tkn)
+			return true
+		}
 	}
 
 	uu := authConfig.Load().UnauthorizedUser
-	if uu != nil {
+	if uu.hasAnyURLs() {
 		processUserRequest(w, r, uu, nil)
 		return true
 	}
 
 	invalidAuthTokenRequests.Inc()
+	slowdownUnauthorizedResponse(r)
+	uu.logRequest(r, `unauthorized`, http.StatusUnauthorized, 0)
 	if *logInvalidAuthTokens {
 		err := fmt.Errorf("cannot authorize request with auth tokens %q", ats)
 		err = &httpserver.ErrorWithStatusCode{
@@ -377,7 +386,7 @@ func bufferRequestBody(ctx context.Context, r io.ReadCloser, userName string) (i
 			return nil, &httpserver.ErrorWithStatusCode{
 				Err: fmt.Errorf("reject request from the user %s because the request body couldn't be read in -maxQueueDuration=%s; read %d bytes in %s",
 					userName, *maxQueueDuration, len(buf), d.Truncate(time.Second)),
-				StatusCode: http.StatusBadRequest,
+				StatusCode: http.StatusRequestTimeout,
 			}
 		}
 
@@ -424,8 +433,12 @@ func processRequest(w http.ResponseWriter, r *http.Request, ui *UserInfo, tkn *j
 		}
 		targetURL := bu.url
 		if tkn != nil {
+			vmac := tkn.VMAccess()
+			if !tkn.HasVMAccessClaim() {
+				vmac = ui.JWT.DefaultVMAccessClaim
+			}
 			// for security reasons allow templating only for configured url values and headers
-			targetURL, hc = replaceJWTPlaceholders(bu, hc, tkn.VMAccess())
+			targetURL, hc = replaceJWTPlaceholders(bu, hc, vmac)
 		}
 		if isDefault {
 			// Don't change path and add request_path query param for default route.
@@ -881,3 +894,20 @@ func debugInfo(u *url.URL, r *http.Request) string {
 	fmt.Fprint(s, ")")
 	return s.String()
 }
+
+// slowdownUnauthorizedResponse adds a random delay in the [2..3] seconds range before returning an unauthorized response.
+// This reduces the effectiveness of brute-force.
+//
+// Recommended by OWASP Top10:
+// https://owasp.org/Top10/2025/A07_2025-Authentication_Failures
+func slowdownUnauthorizedResponse(r *http.Request) {
+
+	d := 2*time.Second + time.Duration(rand.IntN(1000))*time.Millisecond
+	t := timerpool.Get(d)
+
+	select {
+	case <-t.C:
+	case <-r.Context().Done():
+	}
+	timerpool.Put(t)
+}

app/vmauth/main_synctest_test.go

@@ -0,0 +1,54 @@
+//go:build synctest
+
+package main
+
+import (
+	"context"
+	"net/http"
+	"testing"
+	"testing/synctest"
+	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
+)
+
+func TestBufferRequestBody_Timeout(t *testing.T) {
+	synctest.Test(t, func(t *testing.T) {
+
+		defaultMaxQueueDuration := *maxQueueDuration
+		defer func() {
+			*maxQueueDuration = defaultMaxQueueDuration
+		}()
+
+		*maxQueueDuration = 100 * time.Millisecond
+
+		ctx, cancel := context.WithTimeout(context.Background(), *maxQueueDuration)
+		defer cancel()
+
+		_, err := bufferRequestBody(ctx, &timeoutBody{delay: *maxQueueDuration + 100*time.Millisecond}, "foo")
+		if err == nil {
+			t.Fatalf("expecting non-nil error")
+		}
+
+		esc, ok := err.(*httpserver.ErrorWithStatusCode)
+		if !ok {
+			t.Fatalf("unexpected error type: %s", err)
+		}
+		if esc.StatusCode != http.StatusRequestTimeout {
+			t.Fatalf("read request body timeout meet unexpected status code; got %d; want %d", esc.StatusCode, http.StatusRequestTimeout)
+		}
+	})
+}
+
+type timeoutBody struct {
+	delay time.Duration
+}
+
+func (r *timeoutBody) Read(_ []byte) (int, error) {
+	time.Sleep(r.delay)
+	return 0, context.DeadlineExceeded
+}
+
+func (r *timeoutBody) Close() error {
+	return nil
+}

app/vmauth/main_test.go

@@ -739,6 +739,12 @@ users:
 		"vm_access": map[string]any{},
 	}, false)
 
+	// token without vm_access claim, but with a custom claim usable for routing
+	roleToken := genToken(t, map[string]any{
+		"exp":  time.Now().Add(10 * time.Minute).Unix(),
+		"role": "admin",
+	}, true)
+
 	fullToken := genToken(t, map[string]any{
 		"exp": time.Now().Add(10 * time.Minute).Unix(),
 		"vm_access": map[string]any{
@@ -779,6 +785,45 @@ statusCode=401
 Unauthorized`
 	f(simpleCfgStr, request, responseExpected)
 
+	// token without vm_access claim should fall through to unauthorized_user
+	request = httptest.NewRequest(`GET`, "http://some-host.com/abc", nil)
+	request.Header.Set(`Authorization`, `Bearer `+noVMAccessClaimToken)
+	responseExpected = `
+statusCode=200
+path: /bar/abc
+query:
+headers:`
+	f(fmt.Sprintf(`
+unauthorized_user:
+  url_prefix: {BACKEND}/bar
+users:
+- jwt:
+    public_keys:
+    - %q
+    match_claims:
+      role: admin
+  url_prefix: {BACKEND}/foo`, string(publicKeyPEM)), request, responseExpected)
+
+	// token without vm_access claim is accepted when default_vm_access_claim configured
+	request = httptest.NewRequest(`GET`, "http://some-host.com/abc", nil)
+	request.Header.Set(`Authorization`, `Bearer `+roleToken)
+	responseExpected = `
+statusCode=200
+path: /foo/abc
+query:
+headers:`
+	f(fmt.Sprintf(`
+users:
+- jwt:
+    public_keys:
+    - %q
+    default_vm_access_claim:
+      metrics_account_id: 10
+      metrics_project_id: 10
+    match_claims:
+      role: admin
+  url_prefix: {BACKEND}/foo`, string(publicKeyPEM)), request, responseExpected)
+
 	// expired token
 	request = httptest.NewRequest(`GET`, "http://some-host.com/abc", nil)
 	request.Header.Set(`Authorization`, `Bearer `+expiredToken)

app/vmctl/README.md

@@ -1,3 +1,3 @@
 See vmctl docs [here](https://docs.victoriametrics.com/victoriametrics/vmctl/).
 
-vmctl docs can be edited at [docs/vmctl.md](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/docs/victoriametrics/vmctl.md).
+vmctl docs can be edited at [docs/vmctl.md](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/docs/victoriametrics/vmctl/vmctl.md).

app/vmctl/vm/vm.go

@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"io"
 	"net/http"
+	"net/url"
 	"strings"
 	"sync"
 	"time"
@@ -111,7 +112,7 @@ func AddExtraLabelsToImportPath(path string, extraLabels []string) (string, erro
 		if strings.Contains(dst, "?") {
 			separator = "&"
 		}
-		dst += fmt.Sprintf("%sextra_label=%s", separator, extraLabel)
+		dst += fmt.Sprintf("%sextra_label=%s", separator, url.QueryEscape(extraLabel))
 	}
 	return dst, nil
 }

app/vmctl/vm/vm_test.go

@@ -33,11 +33,14 @@ func TestAddExtraLabelsToImportPath_Success(t *testing.T) {
 	f("/api/v1/import", nil, "/api/v1/import")
 
 	// ok one extra label
-	f("/api/v1/import", []string{"instance=host-1"}, "/api/v1/import?extra_label=instance=host-1")
+	f("/api/v1/import", []string{"instance=host-1"}, "/api/v1/import?extra_label=instance%3Dhost-1")
 
 	// ok two extra labels
-	f("/api/v1/import", []string{"instance=host-2", "job=vmagent"}, "/api/v1/import?extra_label=instance=host-2&extra_label=job=vmagent")
+	f("/api/v1/import", []string{"instance=host-2", "job=vmagent"}, "/api/v1/import?extra_label=instance%3Dhost-2&extra_label=job%3Dvmagent")
 
 	// ok two extra with exist param
-	f("/api/v1/import?timeout=50", []string{"instance=host-2", "job=vmagent"}, "/api/v1/import?timeout=50&extra_label=instance=host-2&extra_label=job=vmagent")
+	f("/api/v1/import?timeout=50", []string{"instance=host-2", "job=vmagent"}, "/api/v1/import?timeout=50&extra_label=instance%3Dhost-2&extra_label=job%3Dvmagent")
+
+	// ok special chars in label value
+	f("/api/v1/import", []string{"team=a&b"}, "/api/v1/import?extra_label=team%3Da%26b")
 }

app/vminsert/common/insert_ctx.go

@@ -175,13 +175,19 @@ func (ctx *InsertCtx) WriteMetadata(mmpbs []prompb.MetricMetadata) error {
 	}
 	mms := ctx.mms
 	mms = slicesutil.SetLength(mms, len(mmpbs))
-	for idx, mmpb := range mmpbs {
-		mm := &mms[idx]
+	var cnt int
+	for _, mmpb := range mmpbs {
+		if timeserieslimits.IsMetricMetadataExceeding(&mmpb) {
+			continue
+		}
+		mm := &mms[cnt]
 		mm.MetricFamilyName = bytesutil.ToUnsafeBytes(mmpb.MetricFamilyName)
 		mm.Help = bytesutil.ToUnsafeBytes(mmpb.Help)
 		mm.Type = mmpb.Type
 		mm.Unit = bytesutil.ToUnsafeBytes(mmpb.Unit)
+		cnt++
 	}
+	mms = mms[:cnt]
 	ctx.mms = mms
 
 	err := vmstorage.VMInsertAPI.WriteMetadata(mms)
@@ -201,14 +207,19 @@ func (ctx *InsertCtx) WritePromMetadata(mmps []prometheus.Metadata) error {
 	}
 	mms := ctx.mms
 	mms = slicesutil.SetLength(mms, len(mmps))
-	for idx, mmpb := range mmps {
-		mm := &mms[idx]
+	var cnt int
+	for _, mmpb := range mmps {
+		mm := &mms[cnt]
+		if timeserieslimits.IsPrometheusMetadataExceeding(&mmpb) {
+			continue
+		}
 		mm.MetricFamilyName = bytesutil.ToUnsafeBytes(mmpb.Metric)
 		mm.Help = bytesutil.ToUnsafeBytes(mmpb.Help)
 		mm.Type = mmpb.Type
+		cnt++
 	}
+	mms = mms[:cnt]
 	ctx.mms = mms
-
 	err := vmstorage.VMInsertAPI.WriteMetadata(mms)
 	if err != nil {
 		return &httpserver.ErrorWithStatusCode{

app/vminsert/influx/request_handler.go

@@ -49,7 +49,7 @@ func InsertHandlerForHTTP(req *http.Request) error {
 	}
 	q := req.URL.Query()
 	precision := q.Get("precision")
-	// Read db tag from https://docs.influxdata.com/influxdb/v1.7/tools/api/#write-http-endpoint
+	// Read db tag from https://docs.influxdata.com/influxdb/v1/api/write/#operation/PostWrite
 	db := q.Get("db")
 	encoding := req.Header.Get("Content-Encoding")
 	isStreamMode := req.Header.Get("Stream-Mode") == "1"

app/vminsert/main.go

@@ -68,9 +68,11 @@ var (
 		"at -opentsdbHTTPListenAddr . See https://www.haproxy.org/download/1.8/doc/proxy-protocol.txt")
 	configAuthKey          = flagutil.NewPassword("configAuthKey", "Authorization key for accessing /config page. It must be passed via authKey query arg. It overrides -httpAuth.*")
 	reloadAuthKey          = flagutil.NewPassword("reloadAuthKey", "Auth key for /-/reload http endpoint. It must be passed via authKey query arg. It overrides httpAuth.* settings.")
-	maxLabelsPerTimeseries = flag.Int("maxLabelsPerTimeseries", 40, "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", 256, "The maximum length of label name 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", 4*1024, "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")
+	maxLabelsPerTimeseries = flag.Int("maxLabelsPerTimeseries", 40, "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", 256, "The maximum length of label name 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. "+
+		"Value must be in range 1..65535.")
+	maxLabelValueLen = flag.Int("maxLabelValueLen", 4*1024, "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. "+
+		"Value must be in range 1..65535.")
 )
 
 var (
@@ -106,7 +108,7 @@ func Init() {
 	promscrape.Init(func(_ *auth.Token, wr *prompb.WriteRequest) {
 		prompush.Push(wr)
 	})
-	timeserieslimits.Init(*maxLabelsPerTimeseries, *maxLabelNameLen, *maxLabelValueLen)
+	timeserieslimits.MustInit(*maxLabelsPerTimeseries, *maxLabelNameLen, *maxLabelValueLen)
 }
 
 // Stop stops vminsert.

app/vmselect/main.go

@@ -541,7 +541,7 @@ func handleStaticAndSimpleRequests(w http.ResponseWriter, r *http.Request, path
 		return true
 	case "/metric-relabel-debug":
 		promscrapeMetricRelabelDebugRequests.Inc()
-		promscrape.WriteMetricRelabelDebug(w, r)
+		promscrape.WriteMetricRelabelDebug(w, r, "", nil)
 		return true
 	case "/target-relabel-debug":
 		promscrapeTargetRelabelDebugRequests.Inc()

app/vmselect/prometheus/prometheus.go

@@ -956,6 +956,7 @@ func queryRangeHandler(qt *querytracer.Tracer, startTime time.Time, w http.Respo
 	start, end, step int64, r *http.Request, ct int64, etfs [][]storage.TagFilter) error {
 	deadline := searchutil.GetDeadlineForQuery(r, startTime)
 	mayCache := !httputil.GetBool(r, "nocache")
+	optimizeRepeatedBinaryOpSubexprs := httputil.GetBool(r, "optimize_repeated_binary_op_subexprs")
 	lookbackDelta, err := getMaxLookback(r)
 	if err != nil {
 		return err
@@ -977,18 +978,19 @@ func queryRangeHandler(qt *querytracer.Tracer, startTime time.Time, w http.Respo
 	}
 
 	ec := &promql.EvalConfig{
-		Start:               start,
-		End:                 end,
-		Step:                step,
-		MaxPointsPerSeries:  *maxPointsPerTimeseries,
-		MaxSeries:           0, // let vmstorage use maxUniqueTimeseries by default
-		QuotedRemoteAddr:    httpserver.GetQuotedRemoteAddr(r),
-		Deadline:            deadline,
-		MayCache:            mayCache,
-		LookbackDelta:       lookbackDelta,
-		RoundDigits:         getRoundDigits(r),
-		EnforcedTagFilterss: etfs,
-		CacheTagFilters:     etfs,
+		Start:                            start,
+		End:                              end,
+		Step:                             step,
+		MaxPointsPerSeries:               *maxPointsPerTimeseries,
+		MaxSeries:                        0, // let vmstorage use maxUniqueTimeseries by default
+		QuotedRemoteAddr:                 httpserver.GetQuotedRemoteAddr(r),
+		Deadline:                         deadline,
+		MayCache:                         mayCache,
+		OptimizeRepeatedBinaryOpSubexprs: optimizeRepeatedBinaryOpSubexprs,
+		LookbackDelta:                    lookbackDelta,
+		RoundDigits:                      getRoundDigits(r),
+		EnforcedTagFilterss:              etfs,
+		CacheTagFilters:                  etfs,
 		GetRequestURI: func() string {
 			return httpserver.GetRequestURI(r)
 		},

app/vmselect/promql/binary_op.go

@@ -164,15 +164,21 @@ func newBinaryOpFunc(bf func(left, right float64, isBool bool) float64) binaryOp
 		left := bfa.left
 		right := bfa.right
 		op := bfa.be.Op
-		switch true {
-		case metricsql.IsBinaryOpCmp(op):
-			// Do not remove empty series for comparison operations,
-			// since this may lead to missing result.
-		default:
+		isCmpOp := metricsql.IsBinaryOpCmp(op)
+		if !isCmpOp {
+			// Do not remove empty series for comparison operations, since NaN can be an
+			// explicitly present sample value. In particular, `value != NaN` is true.
+			// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/150.
 			left = removeEmptySeries(left)
 			right = removeEmptySeries(right)
 		}
-		if len(left) == 0 || len(right) == 0 {
+		if len(left) == 0 && len(right) == 0 {
+			return nil, nil
+		}
+		if len(left) == 0 && bfa.be.FillLeft == nil {
+			return nil, nil
+		}
+		if len(right) == 0 && bfa.be.FillRight == nil {
 			return nil, nil
 		}
 		left, right, dst, err := adjustBinaryOpTags(bfa.be, left, right)
@@ -183,6 +189,16 @@ func newBinaryOpFunc(bf func(left, right float64, isBool bool) float64) binaryOp
 			logger.Panicf("BUG: len(left) must match len(right) and len(dst); got %d vs %d vs %d", len(left), len(right), len(dst))
 		}
 		isBool := bfa.be.Bool
+		fillLeft := bfa.be.FillLeft
+		fillRight := bfa.be.FillRight
+		// A NaN produced by a vector comparison denotes a filtered-out sample rather
+		// than an explicitly present NaN value. Drop it when this result is used as
+		// the right-hand side of another comparison.
+		// Note that filtered-out samples surviving as NaN inside other wrappers such as
+		// transform functions or arithmetic operations aren't detected, since such NaN
+		// is indistinguishable from an explicitly present NaN value there.
+		// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10018.
+		dropNaNRight := isCmpOp && isVectorComparisonExpr(bfa.be.Right)
 		for i, tsLeft := range left {
 			leftValues := tsLeft.Values
 			rightValues := right[i].Values
@@ -193,6 +209,24 @@ func newBinaryOpFunc(bf func(left, right float64, isBool bool) float64) binaryOp
 			}
 			for j, a := range leftValues {
 				b := rightValues[j]
+				leftIsNaN := math.IsNaN(a)
+				rightIsNaN := math.IsNaN(b)
+				// Both sides are NaN.
+				if leftIsNaN && rightIsNaN {
+					dstValues[j] = bf(a, b, isBool)
+					continue
+				}
+				if dropNaNRight && rightIsNaN && fillRight == nil {
+					dstValues[j] = nan
+					continue
+				}
+				// Apply the fill value when either the left or right side is NaN, but not both.
+				if leftIsNaN && fillLeft != nil {
+					a = fillLeft.N
+				}
+				if rightIsNaN && fillRight != nil {
+					b = fillRight.N
+				}
 				dstValues[j] = bf(a, b, isBool)
 			}
 		}
@@ -202,6 +236,38 @@ func newBinaryOpFunc(bf func(left, right float64, isBool bool) float64) binaryOp
 	}
 }
 
+// isVectorComparisonExpr returns whether e is a comparison operation
+// with at least one non-scalar operand.
+func isVectorComparisonExpr(e metricsql.Expr) bool {
+	if re, ok := e.(*metricsql.RollupExpr); ok && re.Window == nil {
+		// Unwrap `(...) offset <d>` and `(...) @ <t>`, which do not change
+		// the shape of the result. Subqueries are left as is, since rollup
+		// functions skip NaN values on their own.
+		e = re.Expr
+	}
+	be, ok := e.(*metricsql.BinaryOpExpr)
+	if !ok || !metricsql.IsBinaryOpCmp(be.Op) {
+		return false
+	}
+	return !isScalarLikeExpr(be.Left) || !isScalarLikeExpr(be.Right)
+}
+
+// isScalarLikeExpr returns whether e always evaluates to a scalar value.
+func isScalarLikeExpr(e metricsql.Expr) bool {
+	switch e := e.(type) {
+	case *metricsql.NumberExpr, *metricsql.DurationExpr:
+		return true
+	case *metricsql.BinaryOpExpr:
+		return isScalarLikeExpr(e.Left) && isScalarLikeExpr(e.Right)
+	case *metricsql.FuncExpr:
+		switch strings.ToLower(e.Name) {
+		case "now", "pi", "scalar", "start", "end", "step", "time", "timezone_offset":
+			return true
+		}
+	}
+	return false
+}
+
 func adjustBinaryOpTags(be *metricsql.BinaryOpExpr, left, right []*timeseries) ([]*timeseries, []*timeseries, []*timeseries, error) {
 	if len(be.GroupModifier.Op) == 0 && len(be.JoinModifier.Op) == 0 {
 		if isScalar(left) {
@@ -226,7 +292,7 @@ func adjustBinaryOpTags(be *metricsql.BinaryOpExpr, left, right []*timeseries) (
 		}
 	}
 
-	// Slow path: `vector op vector` or `a op {on|ignoring} {group_left|group_right} b`
+	// Slow path: `vector op vector` or `a op {on|ignoring} {group_left|group_right} {fill|fill_left|fill_right} b`
 	var rvsLeft, rvsRight []*timeseries
 	mLeft, mRight := createTimeseriesMapByTagSet(be, left, right)
 	joinOp := strings.ToLower(be.JoinModifier.Op)
@@ -239,10 +305,27 @@ func adjustBinaryOpTags(be *metricsql.BinaryOpExpr, left, right []*timeseries) (
 		// Add __name__ to groupTags if metric name must be preserved.
 		groupTags = append(groupTags[:len(groupTags):len(groupTags)], "__name__")
 	}
+	// Add missing keys from mRight to mLeft when fill_left()/fill() modifier is used
+	if be.FillLeft != nil {
+		for k := range mRight {
+			if _, ok := mLeft[k]; !ok {
+				mLeft[k] = nil
+			}
+		}
+	}
 	for k, tssLeft := range mLeft {
 		tssRight := mRight[k]
+		if len(tssLeft) == 0 {
+			if be.FillLeft == nil {
+				logger.Panicf("BUG: unexpected empty tssLeft for key %q when FillLeft is nil", k)
+			}
+			tssLeft = []*timeseries{newFillTimeseries(be, tssRight[0])}
+		}
 		if len(tssRight) == 0 {
-			continue
+			if be.FillRight == nil {
+				continue
+			}
+			tssRight = []*timeseries{newFillTimeseries(be, tssLeft[0])}
 		}
 		switch joinOp {
 		case "group_left":
@@ -287,6 +370,28 @@ func adjustBinaryOpTags(be *metricsql.BinaryOpExpr, left, right []*timeseries) (
 	return rvsLeft, rvsRight, dst, nil
 }
 
+// newFillTimeseries returns a time series filled with NaN values for the fill_left()/fill_right()/fill() modifiers.
+// These NaN values will be replaced later with the fill value if needed.
+func newFillTimeseries(be *metricsql.BinaryOpExpr, src *timeseries) *timeseries {
+	var ts timeseries
+	ts.CopyFromShallowTimestamps(src)
+	if !be.KeepMetricNames {
+		ts.MetricName.ResetMetricGroup()
+	}
+	groupTags := be.GroupModifier.Args
+	switch strings.ToLower(be.GroupModifier.Op) {
+	case "on":
+		ts.MetricName.RemoveTagsOn(groupTags)
+	default:
+		ts.MetricName.RemoveTagsIgnoring(groupTags)
+	}
+	values := ts.Values
+	for i := range values {
+		values[i] = math.NaN()
+	}
+	return &ts
+}
+
 func ensureSingleTimeseries(side string, be *metricsql.BinaryOpExpr, tss []*timeseries) error {
 	if len(tss) == 0 {
 		logger.Panicf("BUG: tss must contain at least one value")

app/vmselect/promql/eval.go

@@ -132,6 +132,9 @@ type EvalConfig struct {
 	// Whether the response can be cached.
 	MayCache bool
 
+	// Whether repeated cacheable binary op subexpressions can be optimized.
+	OptimizeRepeatedBinaryOpSubexprs bool
+
 	// LookbackDelta is analog to `-query.lookback-delta` from Prometheus.
 	LookbackDelta int64
 
@@ -171,6 +174,7 @@ func copyEvalConfig(src *EvalConfig) *EvalConfig {
 	ec.MaxPointsPerSeries = src.MaxPointsPerSeries
 	ec.Deadline = src.Deadline
 	ec.MayCache = src.MayCache
+	ec.OptimizeRepeatedBinaryOpSubexprs = src.OptimizeRepeatedBinaryOpSubexprs
 	ec.LookbackDelta = src.LookbackDelta
 	ec.RoundDigits = src.RoundDigits
 	ec.EnforcedTagFilterss = src.EnforcedTagFilterss
@@ -420,18 +424,7 @@ func evalBinaryOp(qt *querytracer.Tracer, ec *EvalConfig, be *metricsql.BinaryOp
 	if bf == nil {
 		return nil, fmt.Errorf(`unknown binary op %q`, be.Op)
 	}
-	var err error
-	var tssLeft, tssRight []*timeseries
-	switch strings.ToLower(be.Op) {
-	case "and", "if":
-		// Fetch right-side series at first, since it usually contains
-		// lower number of time series for `and` and `if` operator.
-		// This should produce more specific label filters for the left side of the query.
-		// This, in turn, should reduce the time to select series for the left side of the query.
-		tssRight, tssLeft, err = execBinaryOpArgs(qt, ec, be.Right, be.Left, be)
-	default:
-		tssLeft, tssRight, err = execBinaryOpArgs(qt, ec, be.Left, be.Right, be)
-	}
+	tssLeft, tssRight, err := execBinaryOpArgs(qt, ec, be)
 	if err != nil {
 		return nil, fmt.Errorf("cannot execute %q: %w", be.AppendString(nil), err)
 	}
@@ -447,6 +440,29 @@ func evalBinaryOp(qt *querytracer.Tracer, ec *EvalConfig, be *metricsql.BinaryOp
 	return rv, nil
 }
 
+// binaryOpEvalOrder might change the order of evaluation of the left and right sides of a binary operation,
+// when there is chance to push down common label filters from exprFirst to exprSecond in the following executions.
+func binaryOpEvalOrder(be *metricsql.BinaryOpExpr) (exprFirst, exprSecond metricsql.Expr) {
+	exprFirst, exprSecond = be.Left, be.Right
+	switch strings.ToLower(be.Op) {
+	case "and", "if":
+		// For `and` and `if`, fetch the right-side series first, since it usually contains
+		// fewer time series and yields more specific filters for the left side.
+		exprFirst, exprSecond = be.Right, be.Left
+	}
+	if be.FillLeft != nil && be.FillRight == nil {
+		// For `fill_left(<value>)`, the unmatched series can only come from the right side, so evaluate it first.
+		exprFirst, exprSecond = be.Right, be.Left
+	}
+	return exprFirst, exprSecond
+}
+
+// canPushdownCommonFilters decides if common label filters can be pushed down from one side of a binary operation to the other.
+//
+// Common filters cannot be pushed down when:
+// - the operator is `or` or `default`;
+// - either side is an aggregation function without explicit grouping;
+// - fill(<value>) modifier is used.
 func canPushdownCommonFilters(be *metricsql.BinaryOpExpr) bool {
 	switch strings.ToLower(be.Op) {
 	case "or", "default":
@@ -455,6 +471,10 @@ func canPushdownCommonFilters(be *metricsql.BinaryOpExpr) bool {
 	if isAggrFuncWithoutGrouping(be.Left) || isAggrFuncWithoutGrouping(be.Right) {
 		return false
 	}
+	// Filters cannot be propagated when fill(<value>) modifier is used.
+	if be.FillLeft != nil && be.FillRight != nil {
+		return false
+	}
 	return true
 }
 
@@ -466,82 +486,185 @@ func isAggrFuncWithoutGrouping(e metricsql.Expr) bool {
 	return len(afe.Modifier.Args) == 0
 }
 
-func execBinaryOpArgs(qt *querytracer.Tracer, ec *EvalConfig, exprFirst, exprSecond metricsql.Expr, be *metricsql.BinaryOpExpr) ([]*timeseries, []*timeseries, error) {
-	if !canPushdownCommonFilters(be) {
-		// Execute exprFirst and exprSecond in parallel, since it is impossible to pushdown common filters
-		// from exprFirst to exprSecond.
-		// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2886
-		qt = qt.NewChild("execute left and right sides of %q in parallel", be.Op)
-		defer qt.Done()
-		var wg sync.WaitGroup
-
-		var tssFirst []*timeseries
-		var errFirst error
-		qtFirst := qt.NewChild("expr1")
-		wg.Go(func() {
-			tssFirst, errFirst = evalExpr(qtFirst, ec, exprFirst)
-			qtFirst.Done()
-		})
-
-		var tssSecond []*timeseries
-		var errSecond error
-		qtSecond := qt.NewChild("expr2")
-		wg.Go(func() {
-			tssSecond, errSecond = evalExpr(qtSecond, ec, exprSecond)
-			qtSecond.Done()
-		})
-
-		wg.Wait()
-		if errFirst != nil {
-			return nil, nil, errFirst
+func execBinaryOpArgs(qt *querytracer.Tracer, ec *EvalConfig, be *metricsql.BinaryOpExpr) ([]*timeseries, []*timeseries, error) {
+	exprFirst, exprSecond := binaryOpEvalOrder(be)
+	firstIsLeft := exprFirst == be.Left
+	sortResult := func(tssFirst, tssSecond []*timeseries) ([]*timeseries, []*timeseries, error) {
+		if firstIsLeft {
+			return tssFirst, tssSecond, nil
 		}
-		if errSecond != nil {
-			return nil, nil, errSecond
+		return tssSecond, tssFirst, nil
+	}
+	if canPushdownCommonFilters(be) {
+		// Execute binary operation in the following way:
+		//
+		// 1) execute the exprFirst
+		// 2) get common label filters for series returned at step 1
+		// 3) push down the found common label filters to exprSecond. This filters out unneeded series
+		//    during exprSecond execution instead of spending compute resources on extracting and processing these series
+		//    before they are dropped later when matching time series according to https://prometheus.io/docs/prometheus/latest/querying/operators/#vector-matching
+		// 4) execute the exprSecond with possible additional filters found at step 3
+		//
+		// Typical use cases:
+		// - Kubernetes-related: show pod creation time with the node name:
+		//
+		//     kube_pod_created{namespace="prod"} * on (uid) group_left(node) kube_pod_info
+		//
+		//   Without the optimization `kube_pod_info` would select and spend compute resources
+		//   for more time series than needed. The selected time series would be dropped later
+		//   when matching time series on the right and left sides of binary operand.
+		//
+		// - Generic alerting queries, which rely on `info` metrics.
+		//   See https://grafana.com/blog/2021/08/04/how-to-use-promql-joins-for-more-effective-queries-of-prometheus-metrics-at-scale/
+		//
+		// - Queries, which get additional labels from `info` metrics.
+		//   See https://www.robustperception.io/exposing-the-software-version-to-prometheus
+		tssFirst, err := evalExpr(qt, ec, exprFirst)
+		if err != nil {
+			return nil, nil, err
+		}
+		if len(tssFirst) == 0 && !strings.EqualFold(be.Op, "or") {
+			// Fast path: there is no sense in executing the exprSecond when exprFirst returns an empty result,
+			// since the "exprFirst op exprSecond" would return an empty result in any case.
+			// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3349
+			return nil, nil, nil
+		}
+		lfs := getCommonLabelFilters(tssFirst)
+		lfs = metricsql.TrimFiltersByGroupModifier(lfs, be)
+		exprSecond = metricsql.PushdownBinaryOpFilters(exprSecond, lfs)
+		tssSecond, err := evalExpr(qt, ec, exprSecond)
+		if err != nil {
+			return nil, nil, err
+		}
+		return sortResult(tssFirst, tssSecond)
+	}
+
+	// Execute exprFirst and exprSecond sequentially if there are cacheable repeated subexpressions
+	// in exprFirst and exprSecond.
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10575
+	if shouldOptimizeRepeatedBinaryOpSubexprs(ec, exprFirst, exprSecond) {
+		qt = qt.NewChild("execute left and right sides of %q sequentially because repeated cacheable subexpression was found", be.Op)
+		defer qt.Done()
+
+		qtFirst := qt.NewChild("expr1")
+		tssFirst, err := evalExpr(qtFirst, ec, exprFirst)
+		qtFirst.Done()
+		if err != nil {
+			return nil, nil, err
+		}
+
+		qtSecond := qt.NewChild("expr2")
+		tssSecond, err := evalExpr(qtSecond, ec, exprSecond)
+		qtSecond.Done()
+		if err != nil {
+			return nil, nil, err
 		}
 		return tssFirst, tssSecond, nil
 	}
 
-	// Execute binary operation in the following way:
-	//
-	// 1) execute the exprFirst
-	// 2) get common label filters for series returned at step 1
-	// 3) push down the found common label filters to exprSecond. This filters out unneeded series
-	//    during exprSecond execution instead of spending compute resources on extracting and processing these series
-	//    before they are dropped later when matching time series according to https://prometheus.io/docs/prometheus/latest/querying/operators/#vector-matching
-	// 4) execute the exprSecond with possible additional filters found at step 3
-	//
-	// Typical use cases:
-	// - Kubernetes-related: show pod creation time with the node name:
-	//
-	//     kube_pod_created{namespace="prod"} * on (uid) group_left(node) kube_pod_info
-	//
-	//   Without the optimization `kube_pod_info` would select and spend compute resources
-	//   for more time series than needed. The selected time series would be dropped later
-	//   when matching time series on the right and left sides of binary operand.
-	//
-	// - Generic alerting queries, which rely on `info` metrics.
-	//   See https://grafana.com/blog/2021/08/04/how-to-use-promql-joins-for-more-effective-queries-of-prometheus-metrics-at-scale/
-	//
-	// - Queries, which get additional labels from `info` metrics.
-	//   See https://www.robustperception.io/exposing-the-software-version-to-prometheus
-	tssFirst, err := evalExpr(qt, ec, exprFirst)
-	if err != nil {
-		return nil, nil, err
+	// Execute exprFirst and exprSecond in parallel, since it is impossible to pushdown common filters
+	// from exprFirst to exprSecond.
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2886
+	qt = qt.NewChild("execute left and right sides of %q in parallel", be.Op)
+	defer qt.Done()
+	var wg sync.WaitGroup
+
+	var tssFirst []*timeseries
+	var errFirst error
+	qtFirst := qt.NewChild("expr1")
+	wg.Go(func() {
+		tssFirst, errFirst = evalExpr(qtFirst, ec, exprFirst)
+		qtFirst.Done()
+	})
+
+	var tssSecond []*timeseries
+	var errSecond error
+	qtSecond := qt.NewChild("expr2")
+	wg.Go(func() {
+		tssSecond, errSecond = evalExpr(qtSecond, ec, exprSecond)
+		qtSecond.Done()
+	})
+
+	wg.Wait()
+	if errFirst != nil {
+		return nil, nil, errFirst
 	}
-	if len(tssFirst) == 0 && !strings.EqualFold(be.Op, "or") {
-		// Fast path: there is no sense in executing the exprSecond when exprFirst returns an empty result,
-		// since the "exprFirst op exprSecond" would return an empty result in any case.
-		// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3349
-		return nil, nil, nil
+	if errSecond != nil {
+		return nil, nil, errSecond
 	}
-	lfs := getCommonLabelFilters(tssFirst)
-	lfs = metricsql.TrimFiltersByGroupModifier(lfs, be)
-	exprSecond = metricsql.PushdownBinaryOpFilters(exprSecond, lfs)
-	tssSecond, err := evalExpr(qt, ec, exprSecond)
-	if err != nil {
-		return nil, nil, err
+	return sortResult(tssFirst, tssSecond)
+}
+
+func shouldOptimizeRepeatedBinaryOpSubexprs(ec *EvalConfig, exprFirst, exprSecond metricsql.Expr) bool {
+	if !ec.OptimizeRepeatedBinaryOpSubexprs {
+		return false
 	}
-	return tssFirst, tssSecond, nil
+	if ec.Start == ec.End {
+		return false
+	}
+	if !ec.mayCache() {
+		return false
+	}
+
+	candidatesFirst := make(map[string]struct{}, 1)
+	var b []byte
+	visitOptimizedAggrs(exprFirst, func(ae *metricsql.AggrFuncExpr) {
+		if hasUnseededVolatileFunc(ae) {
+			return
+		}
+		b = ae.AppendString(b[:0])
+		candidatesFirst[string(b)] = struct{}{}
+	})
+	if len(candidatesFirst) == 0 {
+		return false
+	}
+
+	repeated := false
+	visitOptimizedAggrs(exprSecond, func(ae *metricsql.AggrFuncExpr) {
+		if repeated {
+			return
+		}
+		b = ae.AppendString(b[:0])
+		_, repeated = candidatesFirst[string(b)]
+	})
+	return repeated
+}
+
+func visitOptimizedAggrs(e metricsql.Expr, f func(ae *metricsql.AggrFuncExpr)) {
+	metricsql.VisitAll(e, func(expr metricsql.Expr) {
+		ae, ok := expr.(*metricsql.AggrFuncExpr)
+		if !ok {
+			return
+		}
+		if getIncrementalAggrFuncCallbacks(ae.Name) == nil {
+			return
+		}
+		fe, _ := tryGetArgRollupFuncWithMetricExpr(ae)
+		if fe == nil {
+			return
+		}
+		f(ae)
+	})
+}
+
+func hasUnseededVolatileFunc(e metricsql.Expr) bool {
+	found := false
+	metricsql.VisitAll(e, func(expr metricsql.Expr) {
+		if found {
+			return
+		}
+		fe, ok := expr.(*metricsql.FuncExpr)
+		if !ok {
+			return
+		}
+		switch strings.ToLower(fe.Name) {
+		case "now":
+			found = true
+		case "rand", "rand_normal", "rand_exponential":
+			found = len(fe.Args) == 0
+		}
+	})
+	return found
 }
 
 func getCommonLabelFilters(tss []*timeseries) []metricsql.LabelFilter {

app/vmselect/promql/eval_test.go

@@ -170,3 +170,87 @@ func TestGetSumInstantValues(t *testing.T) {
 		[]*timeseries{ts("foo", 100, 1)},
 	)
 }
+
+func TestShouldOptimizeRepeatedBinaryOpSubexprsGate(t *testing.T) {
+	e, err := metricsql.Parse(`count(count(vm_requests_total) by (action,addr,cluster,endpoint)) by (action,addr,cluster) / count(count(vm_requests_total) by (action,addr,cluster,endpoint))`)
+	if err != nil {
+		t.Fatalf("unexpected error in metricsql.Parse(): %s", err)
+	}
+	be, ok := e.(*metricsql.BinaryOpExpr)
+	if !ok {
+		t.Fatalf("unexpected expr type; got %T; want *metricsql.BinaryOpExpr", e)
+	}
+
+	f := func(name string, ec *EvalConfig, resultExpected bool) {
+		t.Helper()
+		result := shouldOptimizeRepeatedBinaryOpSubexprs(ec, be.Left, be.Right)
+		if result != resultExpected {
+			t.Fatalf("unexpected result for %q; got %v; want %v", name, result, resultExpected)
+		}
+	}
+
+	f("disabled optimization", &EvalConfig{
+		Start: 1000,
+		End:   2000,
+		Step:  1000,
+	}, false)
+	f("disabled cache", &EvalConfig{
+		Start:                            1000,
+		End:                              2000,
+		Step:                             1000,
+		OptimizeRepeatedBinaryOpSubexprs: true,
+	}, false)
+	f("instant query", &EvalConfig{
+		Start:                            1000,
+		End:                              1000,
+		Step:                             1000,
+		MayCache:                         true,
+		OptimizeRepeatedBinaryOpSubexprs: true,
+	}, false)
+	f("repeated cacheable aggregate subexpression", &EvalConfig{
+		Start:                            1000,
+		End:                              2000,
+		Step:                             1000,
+		MayCache:                         true,
+		OptimizeRepeatedBinaryOpSubexprs: true,
+	}, true)
+	f("unaligned range query", &EvalConfig{
+		Start:                            1001,
+		End:                              2000,
+		Step:                             1000,
+		MayCache:                         true,
+		OptimizeRepeatedBinaryOpSubexprs: true,
+	}, false)
+}
+
+func TestShouldOptimizeRepeatedBinaryOpSubexprsExpressions(t *testing.T) {
+	f := func(name, q string, resultExpected bool) {
+		t.Helper()
+		e, err := metricsql.Parse(q)
+		if err != nil {
+			t.Fatalf("unexpected error in metricsql.Parse(%q) for %q: %s", q, name, err)
+		}
+		be, ok := e.(*metricsql.BinaryOpExpr)
+		if !ok {
+			t.Fatalf("unexpected expr type for %q; got %T; want *metricsql.BinaryOpExpr", name, e)
+		}
+		ec := &EvalConfig{Start: 1000, End: 2000, Step: 1000, MayCache: true, OptimizeRepeatedBinaryOpSubexprs: true}
+		result := shouldOptimizeRepeatedBinaryOpSubexprs(ec, be.Left, be.Right)
+		if result != resultExpected {
+			t.Fatalf("unexpected result for %q; got %v; want %v; query: %q", name, result, resultExpected, q)
+		}
+	}
+
+	f("original issue query", `count(count(vm_requests_total) by (action,addr,cluster,endpoint)) by (action,addr,cluster) / count(count(vm_requests_total) by (action,addr,cluster,endpoint))`, true)
+	f("right side contains repeated count aggregate", `count(foo) by (job) / (count(foo) by (job) + 1)`, true)
+	f("same sum aggregate", `sum(rate(foo[5m])) by (job) / sum(rate(foo[5m])) by (job)`, true)
+	f("same inner rollup but different aggregates", `sum(rate(foo[5m])) by (job) / count(rate(foo[5m])) by (job)`, false)
+	f("different count aggregates", `count(foo) by (job) / count(bar) by (job)`, false)
+	f("bare metric selector", `foo / foo`, false)
+	f("bare rollup function", `rate(a[5m]) / rate(a[5m])`, false)
+	f("now at modifier", `sum(rate(foo[5m] @ now())) by (job) / sum(rate(foo[5m] @ now())) by (job)`, false)
+	f("unseeded rand at modifier", `sum(rate(foo[5m] @ rand())) by (job) / sum(rate(foo[5m] @ rand())) by (job)`, false)
+	f("unseeded rand_normal at modifier", `sum(rate(foo[5m] @ rand_normal())) by (job) / sum(rate(foo[5m] @ rand_normal())) by (job)`, false)
+	f("unseeded rand_exponential at modifier", `sum(rate(foo[5m] @ rand_exponential())) by (job) / sum(rate(foo[5m] @ rand_exponential())) by (job)`, false)
+	f("seeded rand at modifier", `sum(rate(foo[5m] @ rand(1))) by (job) / sum(rate(foo[5m] @ rand(1))) by (job)`, true)
+}

app/vmselect/promql/exec_test.go

@@ -2994,6 +2994,136 @@ func TestExecSuccess(t *testing.T) {
 		resultExpected := []netstorage.Result{}
 		f(q, resultExpected)
 	})
+	t.Run(`compare_to_nan_left_vector_right_scalar`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_set(time(), "foo", "bar") != NaN`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{1000, 1200, 1400, 1600, 1800, 2000},
+			Timestamps: timestampsExpected,
+		}
+		r.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("bar"),
+		}}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_non_nan_scalar_right`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_set(time(), "foo", "bar") != 1200`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{1000, nan, 1400, 1600, 1800, 2000},
+			Timestamps: timestampsExpected,
+		}
+		r.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("bar"),
+		}}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_nan_vector_right`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_set(time(), "foo", "bar") != label_set(NaN, "foo", "bar")`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{1000, 1200, 1400, 1600, 1800, 2000},
+			Timestamps: timestampsExpected,
+		}
+		r.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("bar"),
+		}}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_nan_scalar_comparison_right`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_set(time(), "foo", "bar") != (1 > 2)`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{1000, 1200, 1400, 1600, 1800, 2000},
+			Timestamps: timestampsExpected,
+		}
+		r.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("bar"),
+		}}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_empty_vector_right`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_set(time(), "foo", "bar") != (label_set(time(), "foo", "bar") > 100000)`
+		resultExpected := []netstorage.Result{}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_empty_vector_right_offset`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_set(time(), "foo", "bar") != ((label_set(time(), "foo", "bar") > 100000) offset 0s)`
+		resultExpected := []netstorage.Result{}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_empty_vector_left`, func(t *testing.T) {
+		// A missing sample on the left side results in NaN for every comparison
+		// operation, so no special handling is needed there.
+		t.Parallel()
+		q := `(label_set(time(), "foo", "bar") > 100000) != label_set(time(), "foo", "bar")`
+		resultExpected := []netstorage.Result{}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_empty_series_right_bool`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_set(time(), "foo", "bar") == bool (label_set(time(), "foo", "bar") > 100000)`
+		resultExpected := []netstorage.Result{}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_partially_empty_series_right`, func(t *testing.T) {
+		// Prometheus drops the timestamps filtered out by the comparison on the right.
+		// See https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11100#issuecomment-4933952390.
+		t.Parallel()
+		q := `label_set(time(), "foo", "bar") != (label_set(time(), "foo", "bar") * 2 > 2800)`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{nan, nan, nan, 1600, 1800, 2000},
+			Timestamps: timestampsExpected,
+		}
+		r.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("bar"),
+		}}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_empty_unlabeled_vector_right`, func(t *testing.T) {
+		t.Parallel()
+		q := `sum(label_set(time(), "foo", "bar")) != (sum(label_set(time(), "foo", "bar")) > 100000)`
+		resultExpected := []netstorage.Result{}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_empty_series_right_with_fill_left`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_set(time(), "foo", "bar") != fill_left(0) (label_set(time(), "foo", "bar") > 100000)`
+		resultExpected := []netstorage.Result{}
+		f(q, resultExpected)
+	})
+	t.Run(`compare_to_empty_series_right_with_fill_right`, func(t *testing.T) {
+		t.Parallel()
+		q := `label_set(time(), "foo", "bar") != fill_right(0) (label_set(time(), "foo", "bar") > 100000)`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{1000, 1200, 1400, 1600, 1800, 2000},
+			Timestamps: timestampsExpected,
+		}
+		r.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("bar"),
+		}}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
 	t.Run(`-1 < 2`, func(t *testing.T) {
 		t.Parallel()
 		q := `-1 < 2`
@@ -4006,6 +4136,275 @@ func TestExecSuccess(t *testing.T) {
 		resultExpected := []netstorage.Result{r1, r2}
 		f(q, resultExpected)
 	})
+	t.Run(`vector + vector fill()`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(1, "foo", "common")
+			or label_set(2, "foo", "left_only")
+		) + fill(0) (
+			label_set(3, "foo", "common")
+			or label_set(4, "foo", "right_only")
+		), "foo")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("common"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{2, 2, 2, 2, 2, 2},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("left_only"),
+		}}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("right_only"),
+		}}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
+	t.Run(`vector + vector fill() both sides NaN case`, func(t *testing.T) {
+		t.Parallel()
+		q := `(
+			label_set(time() <= 1200, "foo", "common")
+		) + fill(10) (
+			label_set(time() >= 1600, "foo", "common")
+		)`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{1010, 1210, nan, 1610, 1810, 2010},
+			Timestamps: timestampsExpected,
+		}
+		r.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("common"),
+		}}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`vector + vector fill_left() fill_right()`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(1, "foo", "common")
+			or label_set(2, "foo", "left_only")
+		) + fill_left(10) fill_right(20) (
+			label_set(3, "foo", "common")
+			or label_set(4, "foo", "right_only")
+		), "foo")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("common"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{22, 22, 22, 22, 22, 22},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("left_only"),
+		}}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{14, 14, 14, 14, 14, 14},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("right_only"),
+		}}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
+	t.Run(`vector + vector fill_right() only`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(1, "foo", "common")
+			or label_set(2, "foo", "left_only")
+		) + fill_right(20) (
+			label_set(3, "foo", "common")
+			or label_set(4, "foo", "right_only")
+		), "foo")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("common"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{22, 22, 22, 22, 22, 22},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("left_only"),
+		}}
+		resultExpected := []netstorage.Result{r1, r2}
+		f(q, resultExpected)
+	})
+	t.Run(`vector + vector on() fill()`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(1, "foo", "common", "extra", "l")
+			or label_set(2, "foo", "left_only", "extra", "l")
+		) + on(foo) fill(0) (
+			label_set(3, "foo", "common", "extra", "r")
+			or label_set(4, "foo", "right_only", "extra", "r")
+		), "foo")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("common"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{2, 2, 2, 2, 2, 2},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("left_only"),
+		}}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("right_only"),
+		}}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
+	t.Run(`vector + vector on() group_left() fill_right()`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(1, "method", "get", "code", "500")
+			or label_set(2, "method", "get", "code", "404")
+			or label_set(3, "method", "put", "code", "501")
+		) + on(method) group_left() fill_right(0) (
+			label_set(10, "method", "get")
+		), "method", "code")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{12, 12, 12, 12, 12, 12},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("code"),
+				Value: []byte("404"),
+			},
+			{
+				Key:   []byte("method"),
+				Value: []byte("get"),
+			},
+		}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{11, 11, 11, 11, 11, 11},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("code"),
+				Value: []byte("500"),
+			},
+			{
+				Key:   []byte("method"),
+				Value: []byte("get"),
+			},
+		}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{3, 3, 3, 3, 3, 3},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("code"),
+				Value: []byte("501"),
+			},
+			{
+				Key:   []byte("method"),
+				Value: []byte("put"),
+			},
+		}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
+	t.Run(`vector / vector ignoring() fill()`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(6, "method", "get", "code", "500")
+			or label_set(1, "method", "put", "code", "500")
+		) / ignoring(code) fill(0) (
+			label_set(12, "method", "get")
+			or label_set(5, "method", "post")
+			or label_set(10, "method", "put")
+		), "method")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{0.5, 0.5, 0.5, 0.5, 0.5, 0.5},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("method"),
+				Value: []byte("get"),
+			},
+		}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{0, 0, 0, 0, 0, 0},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("method"),
+				Value: []byte("post"),
+			},
+		}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{0.1, 0.1, 0.1, 0.1, 0.1, 0.1},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("method"),
+				Value: []byte("put"),
+			},
+		}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
 	t.Run(`histogram_quantile(scalar)`, func(t *testing.T) {
 		t.Parallel()
 		q := `histogram_quantile(0.6, time())`
@@ -4833,13 +5232,137 @@ func TestExecSuccess(t *testing.T) {
 		resultExpected := []netstorage.Result{}
 		f(q, resultExpected)
 	})
-	t.Run(`buckets_limit(zero)`, func(t *testing.T) {
+	// buckets that are consecutively empty at left and right ends will not be preserved.
+	t.Run(`buckets_limit(trim_zero_preserve_empty_when_limit_not_reached)`, func(t *testing.T) {
 		t.Parallel()
-		q := `buckets_limit(0, (
-			alias(label_set(100, "le", "inf", "x", "y"), "metric"),
-			alias(label_set(50, "le", "120", "x", "y"), "metric"),
-		))`
-		resultExpected := []netstorage.Result{}
+		q := `sort(buckets_limit(3, (
+			alias(label_set(36, "le", "+Inf"), "metric"),
+			alias(label_set(36, "le", "25"), "metric"),
+			alias(label_set(36, "le", "21"), "metric"),
+			alias(label_set(36, "le", "19"), "metric"),
+			alias(label_set(36, "le", "18"), "metric"),
+			alias(label_set(36, "le", "17"), "metric"),
+			alias(label_set(36, "le", "16"), "metric"),
+			alias(label_set(27, "le", "12"), "metric"),
+			alias(label_set(14, "le", "9"), "metric"),
+			alias(label_set(0, "le", "6"), "metric"),
+			alias(label_set(0, "le", "1"), "metric"),
+			)))`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{14, 14, 14, 14, 14, 14},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.MetricGroup = []byte("metric")
+		r1.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("le"),
+				Value: []byte("9"),
+			},
+		}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{27, 27, 27, 27, 27, 27},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.MetricGroup = []byte("metric")
+		r2.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("le"),
+				Value: []byte("12"),
+			},
+		}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{36, 36, 36, 36, 36, 36},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.MetricGroup = []byte("metric")
+		r3.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("le"),
+				Value: []byte("16"),
+			},
+		}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
+
+	// the number of non-empty bucket doesn't reach the given "limit", so some empty buckets will be preserved, and left buckets are preferred to be kept.
+	t.Run(`buckets_limit(trim_zero)`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort(buckets_limit(5, (
+			alias(label_set(36, "le", "18"), "metric"),
+			alias(label_set(36, "le", "17"), "metric"),
+			alias(label_set(36, "le", "16"), "metric"),
+			alias(label_set(27, "le", "12"), "metric"),
+			alias(label_set(14, "le", "9"), "metric"),
+			alias(label_set(0, "le", "6"), "metric"),
+			alias(label_set(0, "le", "1"), "metric"),
+			)))`
+
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{0, 0, 0, 0, 0, 0},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.MetricGroup = []byte("metric")
+		r1.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("le"),
+				Value: []byte("1"),
+			},
+		}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{0, 0, 0, 0, 0, 0},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.MetricGroup = []byte("metric")
+		r2.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("le"),
+				Value: []byte("6"),
+			},
+		}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{14, 14, 14, 14, 14, 14},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.MetricGroup = []byte("metric")
+		r3.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("le"),
+				Value: []byte("9"),
+			},
+		}
+		r4 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{27, 27, 27, 27, 27, 27},
+			Timestamps: timestampsExpected,
+		}
+		r4.MetricName.MetricGroup = []byte("metric")
+		r4.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("le"),
+				Value: []byte("12"),
+			},
+		}
+		r5 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{36, 36, 36, 36, 36, 36},
+			Timestamps: timestampsExpected,
+		}
+		r5.MetricName.MetricGroup = []byte("metric")
+		r5.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("le"),
+				Value: []byte("16"),
+			},
+		}
+
+		resultExpected := []netstorage.Result{r1, r2, r3, r4, r5}
 		f(q, resultExpected)
 	})
 	t.Run(`buckets_limit(unused)`, func(t *testing.T) {
@@ -6228,50 +6751,6 @@ func TestExecSuccess(t *testing.T) {
 		resultExpected := []netstorage.Result{r1, r2, r3, r4, r5, r6, r7}
 		f(q, resultExpected)
 	})
-	t.Run(`sum(histogram_over_time) by (vmrange)`, func(t *testing.T) {
-		t.Parallel()
-		q := `sort_by_label(
-			buckets_limit(
-				3,
-				sum(histogram_over_time(alias(label_set(rand(0)*1.3+1.1, "foo", "bar"), "xxx")[200s:5s])) by (vmrange)
-			), "le"
-		)`
-		r1 := netstorage.Result{
-			MetricName: metricNameExpected,
-			Values:     []float64{40, 40, 40, 40, 40, 40},
-			Timestamps: timestampsExpected,
-		}
-		r1.MetricName.Tags = []storage.Tag{
-			{
-				Key:   []byte("le"),
-				Value: []byte("+Inf"),
-			},
-		}
-		r2 := netstorage.Result{
-			MetricName: metricNameExpected,
-			Values:     []float64{0, 0, 0, 0, 0, 0},
-			Timestamps: timestampsExpected,
-		}
-		r2.MetricName.Tags = []storage.Tag{
-			{
-				Key:   []byte("le"),
-				Value: []byte("1.000e+00"),
-			},
-		}
-		r3 := netstorage.Result{
-			MetricName: metricNameExpected,
-			Values:     []float64{40, 40, 40, 40, 40, 40},
-			Timestamps: timestampsExpected,
-		}
-		r3.MetricName.Tags = []storage.Tag{
-			{
-				Key:   []byte("le"),
-				Value: []byte("2.448e+00"),
-			},
-		}
-		resultExpected := []netstorage.Result{r1, r2, r3}
-		f(q, resultExpected)
-	})
 	t.Run(`sum(histogram_over_time)`, func(t *testing.T) {
 		t.Parallel()
 		q := `sum(histogram_over_time(alias(label_set(rand(0)*1.3+1.1, "foo", "bar"), "xxx")[200s:5s]))`

app/vmselect/promql/rollup.go

@@ -1055,7 +1055,7 @@ func newRollupHoltWinters(args []any) (rollupFunc, error) {
 			return nan
 		}
 
-		// See https://en.wikipedia.org/wiki/Exponential_smoothing#Double_exponential_smoothing .
+		// See https://en.wikipedia.org/wiki/Exponential_smoothing#Double_exponential_smoothing_%28Holt_linear%29 .
 		// TODO: determine whether this shit really works.
 		s0 := rfa.prevValue
 		if math.IsNaN(s0) {

app/vmselect/promql/timeseries.go

@@ -2,6 +2,7 @@ package promql
 
 import (
 	"fmt"
+	"math"
 	"sort"
 	"strconv"
 	"sync"
@@ -288,6 +289,9 @@ func marshalMetricTagsSorted(dst []byte, mn *storage.MetricName) []byte {
 }
 
 func marshalBytesFast(dst []byte, s []byte) []byte {
+	if len(s) > math.MaxUint16 {
+		logger.Panicf("BUG: s len %d cannot exceed %d", len(s), math.MaxUint16)
+	}
 	dst = encoding.MarshalUint16(dst, uint16(len(s)))
 	dst = append(dst, s...)
 	return dst

app/vmselect/promql/transform.go

@@ -393,7 +393,7 @@ func transformBucketsLimit(tfa *transformFuncArg) ([]*timeseries, error) {
 		return nil, err
 	}
 	if limit <= 0 {
-		return nil, nil
+		return nil, fmt.Errorf("limit must be greater than 0; got %d", limit)
 	}
 	if limit < 3 {
 		// Preserve the first and the last bucket for better accuracy for min and max values.
@@ -461,6 +461,23 @@ func transformBucketsLimit(tfa *transformFuncArg) ([]*timeseries, error) {
 				prevValue = value
 			}
 		}
+
+		// Remove buckets that are consecutively empty at left and right ends to obtain more accurate max and min values.
+		// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10417.
+		epsilon := 1e-9
+		isEmptyBucket := func(hits float64) bool {
+			return !math.IsNaN(hits) && math.Abs(hits) < epsilon
+		}
+		l := 0
+		r := len(leGroup) - 1
+		for r-l+1 > limit && isEmptyBucket(leGroup[r].hits) {
+			r--
+		}
+		for r-l+1 > limit && isEmptyBucket(leGroup[l].hits) {
+			l++
+		}
+		leGroup = leGroup[l : r+1]
+
 		for len(leGroup) > limit {
 			// Preserve the first and the last bucket for better accuracy for min and max values
 			xxMinIdx := 1
@@ -1121,29 +1138,29 @@ func groupLeTimeseries(tss []*timeseries) map[string][]leTimeseries {
 
 func fixBrokenBuckets(i int, xss []leTimeseries) {
 	// Buckets are already sorted by le, so their values must be in ascending order,
-	// since the next bucket includes all the previous buckets.
-	// If the next bucket has lower value than the current bucket,
-	// then the next bucket must be substituted with the current bucket value.
+	// since the upper bucket includes all the lower buckets.
+	// If the upper bucket has lower value than the current bucket,
+	// then the upper bucket must be substituted with the current bucket value.
 	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4580#issuecomment-2186659102
 	if len(xss) < 2 {
 		return
 	}
 
-	vNext := xss[0].ts.Values[i]
+	vPrev := xss[0].ts.Values[i]
 	// Set the lowest bucket to 0 if its value is NaN, so it can be properly
 	// compared with upper buckets in the loop below.
-	if math.IsNaN(vNext) {
-		vNext = 0
-		xss[0].ts.Values[i] = vNext
+	if math.IsNaN(vPrev) {
+		vPrev = 0
+		xss[0].ts.Values[i] = vPrev
 	}
 	// Substitute upper bucket values with lower bucket values if the upper values are NaN
-	// or are bigger than the lower bucket values.
+	// or are smaller than the lower bucket values.
 	for j := 1; j < len(xss); j++ {
 		v := xss[j].ts.Values[i]
-		if math.IsNaN(v) || vNext > v {
-			xss[j].ts.Values[i] = vNext
+		if math.IsNaN(v) || vPrev > v {
+			xss[j].ts.Values[i] = vPrev
 		} else {
-			vNext = v
+			vPrev = v
 		}
 	}
 }

app/vmselect/vmui/assets/MetricsQL-CQAtdgI6.md

@@ -91,9 +91,9 @@ The list of MetricsQL features on top of PromQL:
   Labels from the `on()` list aren't copied.
 * [Aggregate functions](#aggregate-functions) accept arbitrary number of args.
   For example, `avg(q1, q2, q3)` would return the average values for every point across time series returned by `q1`, `q2` and `q3`.
-* [@ modifier](https://prometheus.io/docs/prometheus/latest/querying/basics/#modifier) can be put anywhere in the query.
+* [@ modifier](https://prometheus.io/docs/prometheus/latest/querying/basics/#-modifier) can be put anywhere in the query.
   For example, `sum(foo) @ end()` calculates `sum(foo)` at the `end` timestamp of the selected time range `[start ... end]`.
-* Arbitrary subexpression can be used as [@ modifier](https://prometheus.io/docs/prometheus/latest/querying/basics/#modifier).
+* Arbitrary subexpression can be used as [@ modifier](https://prometheus.io/docs/prometheus/latest/querying/basics/#-modifier).
   For example, `foo @ (end() - 1h)` calculates `foo` at the `end - 1 hour` timestamp on the selected time range `[start ... end]`.
 * [offset](https://prometheus.io/docs/prometheus/latest/querying/basics/#offset-modifier), lookbehind window in square brackets
   and `step` value for [subquery](#subqueries) may refer to the current step aka `$__interval` value from Grafana with `[Ni]` syntax.
@@ -482,7 +482,7 @@ See also [hoeffding_bound_lower](#hoeffding_bound_lower).
 #### holt_winters
 
 `holt_winters(series_selector[d], sf, tf)` is a [rollup function](#rollup-functions), which calculates Holt-Winters value
-(aka [double exponential smoothing](https://en.wikipedia.org/wiki/Exponential_smoothing#Double_exponential_smoothing)) for [raw samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples)
+(aka [double exponential smoothing](https://en.wikipedia.org/wiki/Exponential_smoothing#Double_exponential_smoothing_%28Holt_linear%29)) for [raw samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples)
 over the given lookbehind window `d` using the given smoothing factor `sf` and the given trend factor `tf`.
 Both `sf` and `tf` must be in the range `[0...1]`.
 
@@ -1154,7 +1154,7 @@ See also [asin](#asin) and [cos](#cos).
 #### acosh
 
 `acosh(q)` is a [transform function](#transform-functions), which returns
-[inverse hyperbolic cosine](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Inverse_hyperbolic_cosine) for every point of every time series returned by `q`.
+[inverse hyperbolic cosine](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Definitions_in_terms_of_logarithms) for every point of every time series returned by `q`.
 
 Metric names are stripped from the resulting series. Add [keep_metric_names](#keep_metric_names) modifier in order to keep metric names.
 
@@ -1176,7 +1176,7 @@ See also [acos](#acos) and [sin](#sin).
 #### asinh
 
 `asinh(q)` is a [transform function](#transform-functions), which returns
-[inverse hyperbolic sine](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Inverse_hyperbolic_sine) for every point of every time series returned by `q`.
+[inverse hyperbolic sine](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Definitions_in_terms_of_logarithms) for every point of every time series returned by `q`.
 
 Metric names are stripped from the resulting series. Add [keep_metric_names](#keep_metric_names) modifier in order to keep metric names.
 
@@ -1198,7 +1198,7 @@ See also [tan](#tan).
 #### atanh
 
 `atanh(q)` is a [transform function](#transform-functions), which returns
-[inverse hyperbolic tangent](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Inverse_hyperbolic_tangent) for every point of every time series returned by `q`.
+[inverse hyperbolic tangent](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Definitions_in_terms_of_logarithms) for every point of every time series returned by `q`.
 
 Metric names are stripped from the resulting series. Add [keep_metric_names](#keep_metric_names) modifier in order to keep metric names.
 
@@ -1229,8 +1229,7 @@ Metric names are stripped from the resulting series. Add [keep_metric_names](#ke
 `buckets_limit(limit, buckets)` is a [transform function](#transform-functions), which limits the number
 of [histogram buckets](https://valyala.medium.com/improving-histogram-usability-for-prometheus-and-grafana-bc7e5df0e350) to the given `limit`. 
 
-The result will preserve the first and the last bucket to improve accuracy for min and max values. 
-So, if the `limit` is greater than 0 and less than 3, the function will still return 3 buckets: the first bucket, the last bucket, and a selected bucket.
+The given `limit` should be greater than `0`. If it is less than `3`, it will be automatically raised to `3` to preserve the first and last buckets for better accuracy of min and max values.
 
 See also [prometheus_buckets](#prometheus_buckets) and [histogram_quantile](#histogram_quantile).
 

app/vmselect/vmui/assets/rolldown-runtime-CNC7AqOf.js

@@ -0,0 +1 @@
+var e=Object.create,t=Object.defineProperty,n=Object.getOwnPropertyDescriptor,r=Object.getOwnPropertyNames,i=Object.getPrototypeOf,a=Object.prototype.hasOwnProperty,o=(e,t,n)=>()=>{if(n)throw n[0];try{return e&&(t=e(e=0)),t}catch(e){throw n=[e],e}},s=(e,t)=>()=>(t||(e((t={exports:{}}).exports,t),e=null),t.exports),c=(e,n)=>{let r={};for(var i in e)t(r,i,{get:e[i],enumerable:!0});return n||t(r,Symbol.toStringTag,{value:`Module`}),r},l=(e,i,o,s)=>{if(i&&typeof i==`object`||typeof i==`function`)for(var c=r(i),l=0,u=c.length,d;l<u;l++)d=c[l],!a.call(e,d)&&d!==o&&t(e,d,{get:(e=>i[e]).bind(null,d),enumerable:!(s=n(i,d))||s.enumerable});return e},u=(n,r,a)=>(a=n==null?{}:e(i(n)),l(r||!n||!n.__esModule?t(a,`default`,{value:n,enumerable:!0}):a,n)),d=e=>a.call(e,`module.exports`)?e[`module.exports`]:l(t({},`__esModule`,{value:!0}),e);export{u as a,d as i,o as n,c as r,s as t};

app/vmselect/vmui/assets/rolldown-runtime-Cyuzqnbw.js

@@ -1 +0,0 @@
-var e=Object.create,t=Object.defineProperty,n=Object.getOwnPropertyDescriptor,r=Object.getOwnPropertyNames,i=Object.getPrototypeOf,a=Object.prototype.hasOwnProperty,o=(e,t)=>()=>(e&&(t=e(e=0)),t),s=(e,t)=>()=>(t||(e((t={exports:{}}).exports,t),e=null),t.exports),c=(e,n)=>{let r={};for(var i in e)t(r,i,{get:e[i],enumerable:!0});return n||t(r,Symbol.toStringTag,{value:`Module`}),r},l=(e,i,o,s)=>{if(i&&typeof i==`object`||typeof i==`function`)for(var c=r(i),l=0,u=c.length,d;l<u;l++)d=c[l],!a.call(e,d)&&d!==o&&t(e,d,{get:(e=>i[e]).bind(null,d),enumerable:!(s=n(i,d))||s.enumerable});return e},u=(n,r,a)=>(a=n==null?{}:e(i(n)),l(r||!n||!n.__esModule?t(a,`default`,{value:n,enumerable:!0}):a,n)),d=e=>a.call(e,`module.exports`)?e[`module.exports`]:l(t({},`__esModule`,{value:!0}),e);export{u as a,d as i,o as n,c as r,s as t};

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-CusQvJzs.js"></script>
-  <link rel="modulepreload" crossorigin href="./assets/rolldown-runtime-Cyuzqnbw.js">
-  <link rel="modulepreload" crossorigin href="./assets/vendor-B83wxFqK.js">
+  <script type="module" crossorigin src="./assets/index-D5egN2id.js"></script>
+  <link rel="modulepreload" crossorigin href="./assets/rolldown-runtime-CNC7AqOf.js">
+  <link rel="modulepreload" crossorigin href="./assets/vendor-DwJYpOdw.js">
   <link rel="stylesheet" crossorigin href="./assets/vendor-CnsZ1jie.css">
-  <link rel="stylesheet" crossorigin href="./assets/index-BBUnmLOr.css">
+  <link rel="stylesheet" crossorigin href="./assets/index-BJqoElx2.css">
 </head>
 <body>
 <noscript>You need to enable JavaScript to run this app.</noscript>

app/vmstorage/main.go

@@ -30,6 +30,17 @@ var (
 		"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")
+	maxBackfillAge = flagutil.NewRetentionDuration("maxBackfillAge", "0", "The maximum allowed age for the ingested samples with historical timestamps. "+
+		"Samples with timestamps older than now-maxBackfillAge are rejected during data ingestion. "+
+		"By default, or when set to 0, -maxBackfillAge equals to -retentionPeriod, e.g. it is unlimited within the configured retention. "+
+		"This can be useful for limiting ingestion of historical samples, for example, when older data has been moved to another storage tier. "+
+		"See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#retention")
+	vmselectAddr = flag.String("vmselectAddr", "", "TCP address to listen for incoming connections from vmselect. "+
+		"When set, the node will be able to accept cluster-native vmselect RPC requests as if it were vmstorage. "+
+		"The tenant ID assigned to this node's data is controlled by -accountID and -projectID flags. "+
+		"See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#multi-tenancy")
+	vmselectDisableRPCCompression = flag.Bool("rpc.disableCompression", false, "Whether to disable compression of the data sent from vmstorage to vmselect. "+
+		"This reduces CPU usage at the cost of higher network bandwidth usage")
 	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.*")
@@ -108,7 +119,7 @@ func DataPath() string {
 }
 
 // Init initializes vmstorage.
-func Init(vmselectMaxConcurrentRequests int, resetCacheIfNeeded func(mrs []storage.MetricRow)) {
+func Init(vmselectMaxConcurrentRequests int, vmselectMaxQueueDuration time.Duration, resetCacheIfNeeded func(mrs []storage.MetricRow)) {
 	storage.SetDedupInterval(*minScrapeInterval)
 	storage.SetDataFlushInterval(*inmemoryDataFlushInterval)
 	storage.LegacySetRetentionTimezoneOffset(*retentionTimezoneOffset)
@@ -143,6 +154,7 @@ func Init(vmselectMaxConcurrentRequests int, resetCacheIfNeeded func(mrs []stora
 	opts := storage.OpenOptions{
 		Retention:                   retentionPeriod.Duration(),
 		FutureRetention:             futureRetention.Duration(),
+		MaxBackfillAge:              maxBackfillAge.Duration(),
 		DenyQueriesOutsideRetention: *denyQueriesOutsideRetention,
 		MaxHourlySeries:             getMaxHourlySeries(),
 		MaxDailySeries:              getMaxDailySeries(),
@@ -169,6 +181,21 @@ func Init(vmselectMaxConcurrentRequests int, resetCacheIfNeeded func(mrs []stora
 	storageMetrics.RegisterMetricsWriter(vmStorage.writeStorageMetrics)
 	metrics.RegisterSet(storageMetrics)
 
+	if *vmselectAddr != "" {
+		var err error
+		limits := vmselectapi.Limits{
+			MaxConcurrentRequests:         vmselectMaxConcurrentRequests,
+			MaxConcurrentRequestsFlagName: "search.maxConcurrentRequests",
+			MaxQueueDuration:              vmselectMaxQueueDuration,
+			MaxQueueDurationFlagName:      "search.maxQueueDuration",
+		}
+		api := newVMStorageWithTenantID(vmStorage)
+		vmselectSrv, err = vmselectapi.NewServer(*vmselectAddr, api, limits, *vmselectDisableRPCCompression)
+		if err != nil {
+			logger.Fatalf("cannot create a server with -vmselectAddr=%s: %s", *vmselectAddr, err)
+		}
+	}
+
 	VMInsertAPI = vmStorage
 	VMSelectAPI = vmStorage
 	GetSearch = vmStorage.GetSearch
@@ -191,6 +218,8 @@ var (
 
 	// TODO(@rtm0): Remove this dependency from vmalert-tool unit tests.
 	DebugFlush func()
+
+	vmselectSrv *vmselectapi.Server
 )
 
 // Stop stops the vmstorage
@@ -201,6 +230,10 @@ func Stop() {
 
 	logger.Infof("gracefully closing the storage at %s", *storageDataPath)
 	startTime := time.Now()
+
+	if vmselectSrv != nil {
+		vmselectSrv.MustStop()
+	}
 	vmStorage.Stop()
 	logger.Infof("successfully closed the storage in %.3f seconds", time.Since(startTime).Seconds())
 
@@ -443,8 +476,10 @@ func (vms *VMStorage) writeStorageMetrics(w io.Writer) {
 	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="storage/inmemory"}`, tm.InmemorySizeBytes)
 	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="storage/small"}`, tm.SmallSizeBytes)
 	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="storage/big"}`, tm.BigSizeBytes)
+	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="storage/metaindex"}`, tm.MetaindexSizeBytes)
 	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="indexdb/inmemory"}`, idbm.InmemorySizeBytes)
 	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="indexdb/file"}`, idbm.FileSizeBytes)
+	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="indexdb/metaindex"}`, idbm.MetaindexSizeBytes)
 
 	metrics.WriteCounterUint64(w, `vm_rows_received_by_storage_total`, m.RowsReceivedTotal)
 	metrics.WriteCounterUint64(w, `vm_rows_added_to_storage_total`, m.RowsAddedTotal)

app/vmstorage/vmstorage.go

@@ -164,6 +164,10 @@ func (vms *VMStorage) IsReadOnly() bool {
 }
 
 func (vms *VMStorage) InitSearch(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (vmselectapi.BlockIterator, error) {
+	return vms.initSearch(qt, sq, marshalDefault, deadline)
+}
+
+func (vms *VMStorage) initSearch(qt *querytracer.Tracer, sq *storage.SearchQuery, marshal marshalFunc, deadline uint64) (vmselectapi.BlockIterator, error) {
 	vms.wg.Add(1)
 
 	tr := sq.GetTimeRange()
@@ -178,6 +182,7 @@ func (vms *VMStorage) InitSearch(qt *querytracer.Tracer, sq *storage.SearchQuery
 		return nil, fmt.Errorf("missing tag filters")
 	}
 	bi := getBlockIterator()
+	bi.marshal = marshal
 	bi.wgDone = vms.wg.Done
 	bi.sr.Init(qt, vms.s, tfss, tr, maxMetrics, deadline)
 	if err := bi.sr.Error(); err != nil {
@@ -198,11 +203,19 @@ func (vms *VMStorage) getMaxMetrics(searchQueryLimit int) int {
 	return searchQueryLimit
 }
 
+type marshalFunc func(dst []byte, src *storage.MetricBlock) []byte
+
+// marshalDefault is the default implementation of the MetricBlock marshaling.
+func marshalDefault(dst []byte, src *storage.MetricBlock) []byte {
+	return src.Marshal(dst)
+}
+
 // blockIterator implements vmselectapi.BlockIterator
 type blockIterator struct {
-	sr     storage.Search
-	mb     storage.MetricBlock
-	wgDone func()
+	sr      storage.Search
+	mb      storage.MetricBlock
+	marshal marshalFunc
+	wgDone  func()
 }
 
 var blockIteratorsPool sync.Pool
@@ -228,10 +241,10 @@ func (bi *blockIterator) NextBlock(dst []byte) ([]byte, bool) {
 	if !bi.sr.NextMetricBlock() {
 		return dst, false
 	}
-	mb := bi.mb
+	mb := &bi.mb
 	mb.MetricName = bi.sr.MetricBlockRef.MetricName
 	bi.sr.MetricBlockRef.BlockRef.MustReadBlock(&mb.Block)
-	dst = mb.Marshal(dst[:0])
+	dst = bi.marshal(dst[:0], mb)
 	return dst, true
 }
 

app/vmstorage/vmstorage_with_tenant_id.go

@@ -0,0 +1,282 @@
+package vmstorage
+
+import (
+	"flag"
+	"fmt"
+	"math"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
+	"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/vmselectapi"
+)
+
+var (
+	accountID = flag.Uint64("accountID", 0, "The accountID of the stored data")
+	projectID = flag.Uint64("projectID", 0, "The projectID of the stored data")
+)
+
+func newVMStorageWithTenantID(vms *VMStorage) *VMStorageWithTenantID {
+	if *accountID > math.MaxUint32 {
+		logger.Fatalf("-accountID must be in the range [0, %d], got %d", uint32(math.MaxUint32), *accountID)
+	}
+	if *projectID > math.MaxUint32 {
+		logger.Fatalf("-projectID must be in the range [0, %d], got %d", uint32(math.MaxUint32), *projectID)
+	}
+	return &VMStorageWithTenantID{
+		vms:       vms,
+		accountID: uint32(*accountID),
+		projectID: uint32(*projectID),
+	}
+}
+
+// VMStorageWithTenantID is a thin wrapper around VMStorage type that overrides
+// its methods to properly serve requests coming from a vmselect (require
+// tenantID).
+//
+// A new instance of this type should be created using
+// newVMStorageWithTenantID(). The created instance does not require closing.
+// The instance also does not take ownership of vms and it is the responsibility
+// of the caller to close vms.
+type VMStorageWithTenantID struct {
+	vms *VMStorage
+
+	accountID uint32
+	projectID uint32
+}
+
+// InitSearch initializes a storage search for a request initiated by a
+// vmselect.
+//
+// The search is initialized only if the search query is either multitenant or
+// its accountID and projectID match -accountID and -projectID flag values.
+// Otherwise, the method returns an interator that will return no data.
+//
+// The method also overrides the data format of the data returned by the
+// iterator by prepending accountID and projectID bytes to the metric name and
+// the data block (a format used in vmcluster).
+func (vmst *VMStorageWithTenantID) InitSearch(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (vmselectapi.BlockIterator, error) {
+	if !vmst.hasValidTenantID(sq) {
+		return emptyBI, nil
+	}
+	return vmst.vms.initSearch(qt, sq, vmst.marshalMetricBlock, deadline)
+}
+
+var emptyBI = &emptyBlockIterator{}
+
+// emptyBlockIterator is an implementation of vmselectapi.BlockIterator that
+// always returns no data.
+type emptyBlockIterator struct{}
+
+func (*emptyBlockIterator) MustClose() {}
+
+func (*emptyBlockIterator) NextBlock(dst []byte) ([]byte, bool) {
+	return dst, false
+}
+
+func (*emptyBlockIterator) Error() error {
+	return nil
+}
+
+// marshalMetricBlock serializes a metric block in the format expected by
+// vmselect.
+//
+// vmselect expects metric names and data blocks to have the tenantID but
+// vmsingle does not have it. Therefore the tenantID needs to be included to
+// every metric name and block.
+func (vmst *VMStorageWithTenantID) marshalMetricBlock(dst []byte, src *storage.MetricBlock) []byte {
+	// Marshal metric name:
+	// 1. Marshal metric name length + accountID length + projectID length (in
+	//    bytes).
+	// 2. append accountID and projectID bytes
+	// 3. Finally append metric name bytes
+	dst = encoding.MarshalVarUint64(dst, uint64(len(src.MetricName))+8)
+	dst = encoding.MarshalUint32(dst, vmst.accountID)
+	dst = encoding.MarshalUint32(dst, vmst.projectID)
+	dst = append(dst, src.MetricName...)
+
+	// Marshal data block.
+	dst = encoding.MarshalUint32(dst, vmst.accountID)
+	dst = encoding.MarshalUint32(dst, vmst.projectID)
+	dst = storage.MarshalBlock(dst, &src.Block)
+
+	return dst
+}
+
+// SearchMetricNames searches the storage for metric names that match the query.
+//
+// If the query is not multitenant or the query accountID and projectID do not
+// match the -accoutID and -projectID flag values, the method will return an
+// empty result.
+//
+// Found metric names are prepended with accountID and projectID bytes (a format
+// used in vmcluster).
+func (vmst *VMStorageWithTenantID) SearchMetricNames(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) ([]string, error) {
+	if !vmst.hasValidTenantID(sq) {
+		return nil, nil
+	}
+
+	metricNames, err := vmst.vms.SearchMetricNames(qt, sq, deadline)
+	if err != nil {
+		return nil, err
+	}
+
+	// vmselect expects metric names to have the tenantID but vmsingle does not
+	// have it. Therefore the tenantID needs to be prepended to every metric
+	// name.
+	dst := make([]byte, 0, 8)
+	dst = encoding.MarshalUint32(dst, vmst.accountID)
+	dst = encoding.MarshalUint32(dst, vmst.projectID)
+	tenantID := string(dst)
+
+	for i, metricName := range metricNames {
+		metricNames[i] = tenantID + metricName
+	}
+	return metricNames, nil
+}
+
+// LabelValues searches the storage for label values that match the query and
+// correspond to a label whose name is `labelName`. The returned result
+// will contain not more than `maxLabelValues`.
+//
+// If the query is not multitenant or the query accountID and projectID do not
+// match the -accoutID and -projectID flag values, the method will return an
+// empty result.
+func (vmst *VMStorageWithTenantID) LabelValues(qt *querytracer.Tracer, sq *storage.SearchQuery, labelName string, maxLabelValues int, deadline uint64) ([]string, error) {
+	if !vmst.hasValidTenantID(sq) {
+		return nil, nil
+	}
+	return vmst.vms.LabelValues(qt, sq, labelName, maxLabelValues, deadline)
+}
+
+// TagValueSuffixes searches the storage for Graphite tag value suffixes. The
+// returned result will contain not more than `maxSuffixes`.
+//
+// If the query is not multitenant or the query accountID and projectID do not
+// match the -accoutID and -projectID flag values, the method will return an
+// empty result.
+func (vmst *VMStorageWithTenantID) TagValueSuffixes(qt *querytracer.Tracer, accountID, projectID uint32, tr storage.TimeRange, tagKey, tagValuePrefix string, delimiter byte, maxSuffixes int, deadline uint64) ([]string, error) {
+	if !vmst.isValidTenantID(accountID, projectID) {
+		return nil, nil
+	}
+	return vmst.vms.TagValueSuffixes(qt, accountID, projectID, tr, tagKey, tagValuePrefix, delimiter, maxSuffixes, deadline)
+}
+
+// LabelNames searches the storage for label names that match the query.
+// The returned result will contain not more than `maxLabelNames`.
+//
+// If the query is not multitenant or the query accountID and projectID do not
+// match the -accoutID and -projectID flag values, the method will return an
+// empty result.
+func (vmst *VMStorageWithTenantID) LabelNames(qt *querytracer.Tracer, sq *storage.SearchQuery, maxLabelNames int, deadline uint64) ([]string, error) {
+	if !vmst.hasValidTenantID(sq) {
+		return nil, nil
+	}
+	return vmst.vms.LabelNames(qt, sq, maxLabelNames, deadline)
+}
+
+// SeriesCount returns the total number of metrics stored in the database.
+//
+// The method may return inflated numbers. How inflated the count depends
+// on the churn rate and the retention period. For example, if a metric lasts
+// for 2 months, it will be counted twice.
+//
+// The method also counts the deleted metrics.
+//
+// If the query is not multitenant or the query accountID and projectID do not
+// match the -accoutID and -projectID flag values, the method will return 0.
+func (vmst *VMStorageWithTenantID) SeriesCount(qt *querytracer.Tracer, accountID, projectID uint32, deadline uint64) (uint64, error) {
+	if !vmst.isValidTenantID(accountID, projectID) {
+		return 0, nil
+	}
+	return vmst.vms.SeriesCount(qt, accountID, projectID, deadline)
+}
+
+// Tenants returns just one tenant consisting of the -accountID and -projectID
+// flag values.
+func (vmst *VMStorageWithTenantID) Tenants(qt *querytracer.Tracer, tr storage.TimeRange, deadline uint64) ([]string, error) {
+	tenantID := fmt.Sprintf("%d:%d", vmst.accountID, vmst.projectID)
+	return []string{tenantID}, nil
+}
+
+// TSDBStatus retrieves the status for metrics that match to the search query.
+//
+// If the query is not multitenant or the query accountID and projectID do not
+// match the -accoutID and -projectID flag values, the method will return empty
+// status.
+func (vmst *VMStorageWithTenantID) TSDBStatus(qt *querytracer.Tracer, sq *storage.SearchQuery, focusLabel string, topN int, deadline uint64) (*storage.TSDBStatus, error) {
+	if !vmst.hasValidTenantID(sq) {
+		return &storage.TSDBStatus{}, nil
+	}
+	return vmst.vms.TSDBStatus(qt, sq, focusLabel, topN, deadline)
+}
+
+// DeleteSeries marks as deleted metrics that match the search query.
+// The method returns the number of deleted metrics.
+//
+// If the query is not multitenant or the query accountID and projectID do not
+// match the -accoutID and -projectID flag values, no metrics will be deleted
+// and the method will return 0.
+func (vmst *VMStorageWithTenantID) DeleteSeries(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (int, error) {
+	if !vmst.hasValidTenantID(sq) {
+		return 0, nil
+	}
+	return vmst.vms.DeleteSeries(qt, sq, deadline)
+}
+
+// RegisterMetricNames registers metric names in the index, the sample values
+// and timestamps are ignored.
+func (vmst *VMStorageWithTenantID) RegisterMetricNames(qt *querytracer.Tracer, mrs []storage.MetricRow, deadline uint64) error {
+	return vmst.vms.RegisterMetricNames(qt, mrs, deadline)
+}
+
+// GetMetricNamesUsageStats retrieves the usage stats for metrics whose name
+// matches the pattern.
+//
+// If the request is not multitenant or the request accountID and projectID do
+// not match the -accoutID and -projectID flag values, no metrics will be
+// deleted and the method will return 0.
+func (vmst *VMStorageWithTenantID) GetMetricNamesUsageStats(qt *querytracer.Tracer, tt *storage.TenantToken, limit, le int, matchPattern string, deadline uint64) (metricnamestats.StatsResult, error) {
+	if !vmst.isValidTenantToken(tt) {
+		return metricnamestats.StatsResult{}, nil
+	}
+	return vmst.vms.GetMetricNamesUsageStats(qt, tt, limit, le, matchPattern, deadline)
+}
+
+// ResetMetricNamesUsageStats resets the metric name usage stats.
+func (vmst *VMStorageWithTenantID) ResetMetricNamesUsageStats(qt *querytracer.Tracer, deadline uint64) error {
+	return vmst.vms.ResetMetricNamesUsageStats(qt, deadline)
+}
+
+// GetMetadataRecords retrieves the metadata for the metricName.
+//
+// If the request is not multitenant or the request accountID and projectID do
+// not match the -accoutID and -projectID flag values, no metrics will be
+// deleted and the method will return 0.
+func (vmst *VMStorageWithTenantID) GetMetadataRecords(qt *querytracer.Tracer, tt *storage.TenantToken, limit int, metricName string, deadline uint64) ([]*metricsmetadata.Row, error) {
+	if !vmst.isValidTenantToken(tt) {
+		return nil, nil
+	}
+	return vmst.vms.GetMetadataRecords(qt, tt, limit, metricName, deadline)
+}
+
+// hasValidTenantID returns true if the search query is either multitenant or
+// its accountID and projectID match -accountID and -projectID flag values.
+func (vmst *VMStorageWithTenantID) hasValidTenantID(sq *storage.SearchQuery) bool {
+	return sq.IsMultiTenant || vmst.isValidTenantID(sq.AccountID, sq.ProjectID)
+}
+
+// isValidTenantToken returns true if the TenantToken is either multitenant or
+// its accountID and projectID match -accountID and -projectID flag values.
+func (vmst *VMStorageWithTenantID) isValidTenantToken(tt *storage.TenantToken) bool {
+	return tt == nil || vmst.isValidTenantID(tt.AccountID, tt.ProjectID)
+}
+
+// isValidTenantID returns true if the accountID and projectID match -accountID
+// and -projectID flag values.
+func (vmst *VMStorageWithTenantID) isValidTenantID(accountID, projectID uint32) bool {
+	return accountID == vmst.accountID && projectID == vmst.projectID
+}

app/vmui/Dockerfile-web

@@ -1,4 +1,4 @@
-FROM golang:1.26.4 AS build-web-stage
+FROM golang:1.26.5 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:3.23.4
+FROM alpine:3.24.1
 USER root
 
 COPY --from=build-web-stage /build/web-amd64 /app/web

app/vmui/packages/vmui/package.json

@@ -9,7 +9,7 @@
     "start": "vite",
     "start:playground": "cross-env PLAYGROUND=true npm run start",
     "build": "vite build",
-    "lint": "eslint --output-file vmui-lint-report.json --format json 'src/**/*.{ts,tsx}'",
+    "lint": "eslint --format stylish 'src/**/*.{ts,tsx}'",
     "lint:local": "eslint --ext .ts,.tsx -f stylish src",
     "lint:fix": "eslint 'src/**/*.{ts,tsx}' --fix",
     "copy-metricsql-docs": "cp ../../../../docs/MetricsQL.md src/assets/MetricsQL.md || true",
@@ -23,40 +23,40 @@
     "classnames": "^2.5.1",
     "dayjs": "^1.11.21",
     "lodash.debounce": "^4.0.8",
-    "marked": "^18.0.5",
-    "preact": "^10.29.2",
-    "qs": "^6.15.2",
+    "marked": "^18.0.6",
+    "preact": "^10.29.7",
+    "qs": "^6.15.3",
     "react-input-mask": "^2.0.4",
-    "react-router-dom": "^7.17.0",
+    "react-router-dom": "^7.18.1",
     "uplot": "^1.6.32",
-    "vite": "^8.0.16",
+    "vite": "^8.1.4",
     "web-vitals": "^5.3.0"
   },
   "devDependencies": {
-    "@eslint/eslintrc": "^3.3.5",
+    "@eslint/eslintrc": "^3.3.6",
     "@eslint/js": "^9.39.2",
     "@preact/preset-vite": "^2.10.5",
     "@testing-library/jest-dom": "^6.9.1",
     "@testing-library/preact": "^3.2.4",
     "@types/lodash.debounce": "^4.0.9",
-    "@types/node": "^25.9.2",
+    "@types/node": "^26.1.1",
     "@types/qs": "^6.15.1",
     "@types/react": "^19.2.17",
     "@types/react-input-mask": "^3.0.6",
     "@types/react-router-dom": "^5.3.3",
-    "@typescript-eslint/eslint-plugin": "^8.61.0",
-    "@typescript-eslint/parser": "^8.61.0",
+    "@typescript-eslint/eslint-plugin": "^8.64.0",
+    "@typescript-eslint/parser": "^8.64.0",
     "cross-env": "^10.1.0",
     "eslint": "^9.39.2",
     "eslint-plugin-react": "^7.37.5",
     "eslint-plugin-unused-imports": "^4.4.1",
-    "globals": "^17.6.0",
-    "http-proxy-middleware": "^4.1.0",
+    "globals": "^17.7.0",
+    "http-proxy-middleware": "^4.2.0",
     "jsdom": "^29.1.1",
-    "postcss": "^8.5.15",
+    "postcss": "^8.5.19",
     "sass-embedded": "^1.100.0",
     "typescript": "^6.0.3",
-    "vitest": "^4.1.8"
+    "vitest": "^4.1.10"
   },
   "browserslist": {
     "production": [

app/vmui/packages/vmui/src/assets/MetricsQL.md

@@ -91,9 +91,9 @@ The list of MetricsQL features on top of PromQL:
   Labels from the `on()` list aren't copied.
 * [Aggregate functions](#aggregate-functions) accept arbitrary number of args.
   For example, `avg(q1, q2, q3)` would return the average values for every point across time series returned by `q1`, `q2` and `q3`.
-* [@ modifier](https://prometheus.io/docs/prometheus/latest/querying/basics/#modifier) can be put anywhere in the query.
+* [@ modifier](https://prometheus.io/docs/prometheus/latest/querying/basics/#-modifier) can be put anywhere in the query.
   For example, `sum(foo) @ end()` calculates `sum(foo)` at the `end` timestamp of the selected time range `[start ... end]`.
-* Arbitrary subexpression can be used as [@ modifier](https://prometheus.io/docs/prometheus/latest/querying/basics/#modifier).
+* Arbitrary subexpression can be used as [@ modifier](https://prometheus.io/docs/prometheus/latest/querying/basics/#-modifier).
   For example, `foo @ (end() - 1h)` calculates `foo` at the `end - 1 hour` timestamp on the selected time range `[start ... end]`.
 * [offset](https://prometheus.io/docs/prometheus/latest/querying/basics/#offset-modifier), lookbehind window in square brackets
   and `step` value for [subquery](#subqueries) may refer to the current step aka `$__interval` value from Grafana with `[Ni]` syntax.
@@ -482,7 +482,7 @@ See also [hoeffding_bound_lower](#hoeffding_bound_lower).
 #### holt_winters
 
 `holt_winters(series_selector[d], sf, tf)` is a [rollup function](#rollup-functions), which calculates Holt-Winters value
-(aka [double exponential smoothing](https://en.wikipedia.org/wiki/Exponential_smoothing#Double_exponential_smoothing)) for [raw samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples)
+(aka [double exponential smoothing](https://en.wikipedia.org/wiki/Exponential_smoothing#Double_exponential_smoothing_%28Holt_linear%29)) for [raw samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples)
 over the given lookbehind window `d` using the given smoothing factor `sf` and the given trend factor `tf`.
 Both `sf` and `tf` must be in the range `[0...1]`.
 
@@ -1154,7 +1154,7 @@ See also [asin](#asin) and [cos](#cos).
 #### acosh
 
 `acosh(q)` is a [transform function](#transform-functions), which returns
-[inverse hyperbolic cosine](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Inverse_hyperbolic_cosine) for every point of every time series returned by `q`.
+[inverse hyperbolic cosine](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Definitions_in_terms_of_logarithms) for every point of every time series returned by `q`.
 
 Metric names are stripped from the resulting series. Add [keep_metric_names](#keep_metric_names) modifier in order to keep metric names.
 
@@ -1176,7 +1176,7 @@ See also [acos](#acos) and [sin](#sin).
 #### asinh
 
 `asinh(q)` is a [transform function](#transform-functions), which returns
-[inverse hyperbolic sine](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Inverse_hyperbolic_sine) for every point of every time series returned by `q`.
+[inverse hyperbolic sine](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Definitions_in_terms_of_logarithms) for every point of every time series returned by `q`.
 
 Metric names are stripped from the resulting series. Add [keep_metric_names](#keep_metric_names) modifier in order to keep metric names.
 
@@ -1198,7 +1198,7 @@ See also [tan](#tan).
 #### atanh
 
 `atanh(q)` is a [transform function](#transform-functions), which returns
-[inverse hyperbolic tangent](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Inverse_hyperbolic_tangent) for every point of every time series returned by `q`.
+[inverse hyperbolic tangent](https://en.wikipedia.org/wiki/Inverse_hyperbolic_functions#Definitions_in_terms_of_logarithms) for every point of every time series returned by `q`.
 
 Metric names are stripped from the resulting series. Add [keep_metric_names](#keep_metric_names) modifier in order to keep metric names.
 
@@ -1229,8 +1229,7 @@ Metric names are stripped from the resulting series. Add [keep_metric_names](#ke
 `buckets_limit(limit, buckets)` is a [transform function](#transform-functions), which limits the number
 of [histogram buckets](https://valyala.medium.com/improving-histogram-usability-for-prometheus-and-grafana-bc7e5df0e350) to the given `limit`. 
 
-The result will preserve the first and the last bucket to improve accuracy for min and max values. 
-So, if the `limit` is greater than 0 and less than 3, the function will still return 3 buckets: the first bucket, the last bucket, and a selected bucket.
+The given `limit` should be greater than `0`. If it is less than `3`, it will be automatically raised to `3` to preserve the first and last buckets for better accuracy of min and max values.
 
 See also [prometheus_buckets](#prometheus_buckets) and [histogram_quantile](#histogram_quantile).
 

app/vmui/packages/vmui/src/components/Configurators/StepConfigurator/StepConfigurator.tsx

@@ -183,7 +183,7 @@ const StepConfigurator: FC = () => {
           <div className="vm-step-control-popper-info">
             <p>
               <code>step</code> - the <Hyperlink
-                href="https://prometheus.io/docs/prometheus/latest/querying/basics/#time-durations"
+                href="https://prometheus.io/docs/prometheus/latest/querying/basics/#float-literals-and-time-durations"
                 text="interval"
               /> between datapoints, which must be returned from the range query.
               The <code>query</code> is executed

app/vmui/packages/vmui/src/components/Configurators/TimeRangeSettings/ExecutionControls/ExecutionControls.tsx

@@ -1,4 +1,4 @@
-import { FC, useEffect, useRef, useState } from "preact/compat";
+import { FC, useEffect, useRef } from "preact/compat";
 import { useTimeDispatch } from "../../../../state/time/TimeStateContext";
 import { getAppModeEnable } from "../../../../utils/app-mode";
 import Button from "../../../Main/Button/Button";
@@ -9,27 +9,38 @@ import classNames from "classnames";
 import Tooltip from "../../../Main/Tooltip/Tooltip";
 import useDeviceDetect from "../../../../hooks/useDeviceDetect";
 import useBoolean from "../../../../hooks/useBoolean";
+import { getMillisecondsFromDuration } from "../../../../utils/time";
+import { useSearchParams } from "react-router";
 
-interface AutoRefreshOption {
-  seconds: number
-  title: string
-}
-
-const delayOptions: AutoRefreshOption[] = [
-  { seconds: 0, title: "Off" },
-  { seconds: 1, title: "1s" },
-  { seconds: 2, title: "2s" },
-  { seconds: 5, title: "5s" },
-  { seconds: 10, title: "10s" },
-  { seconds: 30, title: "30s" },
-  { seconds: 60, title: "1m" },
-  { seconds: 300, title: "5m" },
-  { seconds: 900, title: "15m" },
-  { seconds: 1800, title: "30m" },
-  { seconds: 3600, title: "1h" },
-  { seconds: 7200, title: "2h" }
+const delayOptions = [
+  "Off",
+  "1s",
+  "2s",
+  "5s",
+  "10s",
+  "30s",
+  "1m",
+  "5m",
+  "15m",
+  "30m",
+  "1h",
+  "2h"
 ];
 
+const DEFAULT_OPTION = delayOptions[0];
+
+const MIN_REFRESH_MS = 1000;
+const MAX_REFRESH_MS = getMillisecondsFromDuration(delayOptions[delayOptions.length - 1]);
+const REFRESH_URL_PARAM = "refresh";
+
+const durationToMs = (dur: string | null) => {
+  return dur ? getMillisecondsFromDuration(dur) : 0;
+};
+
+const isValidDelay = (ms: number) => {
+  return ms >= MIN_REFRESH_MS && ms <= MAX_REFRESH_MS;
+};
+
 interface ExecutionControlsProps {
   tooltip: string;
   useAutorefresh?: boolean;
@@ -38,12 +49,14 @@ interface ExecutionControlsProps {
 
 export const ExecutionControls: FC<ExecutionControlsProps> = ({ tooltip, useAutorefresh, closeModal }) => {
   const { isMobile } = useDeviceDetect();
+  const [searchParams, setSearchParams] = useSearchParams();
 
   const dispatch = useTimeDispatch();
   const appModeEnable = getAppModeEnable();
-  const [autoRefresh, setAutoRefresh] = useState(false);
 
-  const [selectedDelay, setSelectedDelay] = useState<AutoRefreshOption>(delayOptions[0]);
+  const rawDelay = searchParams.get(REFRESH_URL_PARAM);
+  const msDelay = durationToMs(rawDelay);
+  const selectedDelay = isValidDelay(msDelay) ? rawDelay : DEFAULT_OPTION;
 
   const {
     value: openOptions,
@@ -52,11 +65,20 @@ export const ExecutionControls: FC<ExecutionControlsProps> = ({ tooltip, useAuto
   } = useBoolean(false);
   const optionsButtonRef = useRef<HTMLDivElement>(null);
 
-  const handleChange = (d: AutoRefreshOption) => {
-    if ((autoRefresh && !d.seconds) || (!autoRefresh && d.seconds)) {
-      setAutoRefresh(prev => !prev);
-    }
-    setSelectedDelay(d);
+  const handleChange = (dur: string) => () => {
+    setSearchParams(prev => {
+      const nextParams = new URLSearchParams(prev);
+      const ms = durationToMs(dur);
+
+      if (ms) {
+        nextParams.set(REFRESH_URL_PARAM, `${dur}`);
+      } else {
+        nextParams.delete(REFRESH_URL_PARAM);
+      }
+
+      return nextParams;
+    });
+
     handleCloseOptions();
   };
 
@@ -68,23 +90,19 @@ export const ExecutionControls: FC<ExecutionControlsProps> = ({ tooltip, useAuto
   };
 
   useEffect(() => {
-    const delay = selectedDelay.seconds;
+    const ms = durationToMs(selectedDelay);
     let timer: number;
-    if (autoRefresh) {
+
+    if (useAutorefresh && isValidDelay(ms)) {
       timer = setInterval(() => {
         dispatch({ type: "RUN_QUERY" });
-      }, delay * 1000) as unknown as number;
-    } else {
-      setSelectedDelay(delayOptions[0]);
+      }, ms) as unknown as number;
     }
-    return () => {
-      timer && clearInterval(timer);
-    };
-  }, [selectedDelay, autoRefresh]);
 
-  const createHandlerChange = (d: AutoRefreshOption) => () => {
-    handleChange(d);
-  };
+    return () => {
+      clearInterval(timer);
+    };
+  }, [selectedDelay, useAutorefresh]);
 
   return (
     <>
@@ -106,7 +124,7 @@ export const ExecutionControls: FC<ExecutionControlsProps> = ({ tooltip, useAuto
                 <span className="vm-mobile-option__icon"><RestartIcon/></span>
                 <div className="vm-mobile-option-text">
                   <span className="vm-mobile-option-text__label">Auto-refresh</span>
-                  <span className="vm-mobile-option-text__value">{selectedDelay.title}</span>
+                  <span className="vm-mobile-option-text__value">{selectedDelay}</span>
                 </div>
                 <span className="vm-mobile-option__arrow"><ArrowDownIcon/></span>
               </div>
@@ -139,7 +157,7 @@ export const ExecutionControls: FC<ExecutionControlsProps> = ({ tooltip, useAuto
                       )}
                       onClick={toggleOpenOptions}
                     >
-                      {selectedDelay.title}
+                      {selectedDelay}
                     </Button>
                   </div>
                 </Tooltip>
@@ -187,12 +205,12 @@ export const ExecutionControls: FC<ExecutionControlsProps> = ({ tooltip, useAuto
                 className={classNames({
                   "vm-list-item": true,
                   "vm-list-item_mobile": isMobile,
-                  "vm-list-item_active": d.seconds === selectedDelay.seconds
+                  "vm-list-item_active": d === selectedDelay
                 })}
-                key={d.seconds}
-                onClick={createHandlerChange(d)}
+                key={d}
+                onClick={handleChange(d)}
               >
-                {d.title}
+                {d}
               </div>
             ))}
           </div>

app/vmui/packages/vmui/src/components/ExploreAlerts/BaseAlert/index.tsx

@@ -98,7 +98,7 @@ const BaseAlert = ({ item, group }: BaseAlertProps) => {
       {!!Object.keys(item.annotations || {}).length && (
         <>
           <span className="vm-alerts-title">Annotations</span>
-          <table>
+          <table className="vm-annotations-table">
             <colgroup>
               <col className="vm-col-md"/>
               <col/>

app/vmui/packages/vmui/src/components/ExploreAlerts/BaseRule/index.tsx

@@ -121,7 +121,7 @@ const BaseRule = ({ item, group }: BaseRuleProps) => {
       {!!Object.keys(item?.annotations || {}).length && (
         <>
           <span className="vm-alerts-title">Annotations</span>
-          <table>
+          <table className="vm-annotations-table">
             <colgroup>
               <col className="vm-col-md"/>
               <col/>

app/vmui/packages/vmui/src/layouts/Header/HeaderNav/HeaderNav.tsx

@@ -17,10 +17,12 @@ interface HeaderNavProps {
 const HeaderNav: FC<HeaderNavProps> = ({ color, background, direction }) => {
   const { pathname } = useLocation();
   const [activeMenu, setActiveMenu] = useState(pathname);
+  const [openMenu, setOpenMenu] = useState<string | null>(null);
   const menu = useNavigationMenu();
 
   useEffect(() => {
     setActiveMenu(pathname);
+    setOpenMenu(null);
   }, [pathname]);
 
   return (
@@ -41,6 +43,8 @@ const HeaderNav: FC<HeaderNavProps> = ({ color, background, direction }) => {
               color={color}
               background={background}
               direction={direction}
+              openMenu={openMenu}
+              setOpenMenu={setOpenMenu}
             />
           )
           : (

app/vmui/packages/vmui/src/layouts/Header/HeaderNav/NavSubItem.tsx

@@ -1,11 +1,8 @@
-import { FC, useRef, useState } from "preact/compat";
-import { useLocation } from "react-router-dom";
+import { FC, useRef, useState, Dispatch, SetStateAction } from "preact/compat";
 import classNames from "classnames";
 import { ArrowDropDownIcon } from "../../../components/Main/Icons";
 import Popper from "../../../components/Main/Popper/Popper";
 import NavItem from "./NavItem";
-import { useEffect } from "react";
-import useBoolean from "../../../hooks/useBoolean";
 import { NavigationItem, NavigationItemType } from "../../../router/navigation";
 
 interface NavItemProps {
@@ -15,6 +12,8 @@ interface NavItemProps {
   color?: string
   background?: string
   direction?: "row" | "column"
+  openMenu: string | null,
+  setOpenMenu: Dispatch<SetStateAction<string | null>>,
 }
 
 const NavSubItem: FC<NavItemProps> = ({
@@ -23,21 +22,18 @@ const NavSubItem: FC<NavItemProps> = ({
   color,
   background,
   submenu,
-  direction = "row"
+  direction = "row",
+  openMenu,
+  setOpenMenu,
 }) => {
-  const { pathname } = useLocation();
-
   const [menuTimeout, setMenuTimeout] = useState<NodeJS.Timeout | null>(null);
   const buttonRef = useRef<HTMLDivElement>(null);
 
-  const {
-    value: openSubmenu,
-    setFalse: handleCloseSubmenu,
-    setTrue: setOpenSubmenu,
-  } = useBoolean(false);
+  const openSubmenu = openMenu === label;
+  const handleCloseSubmenu = () => setOpenMenu(prev => (prev === label ? null : prev));
 
   const handleOpenSubmenu = () => {
-    if (direction === "row" || !openSubmenu) setOpenSubmenu();
+    if (direction === "row" || !openSubmenu) setOpenMenu(label);
     if (direction === "column" && openSubmenu) handleCloseSubmenu();
     if (direction === "row" && menuTimeout) clearTimeout(menuTimeout);
   };
@@ -52,10 +48,6 @@ const NavSubItem: FC<NavItemProps> = ({
     if (menuTimeout) clearTimeout(menuTimeout);
   };
 
-  useEffect(() => {
-    handleCloseSubmenu();
-  }, [pathname]);
-
   return (
     <div
       className={classNames({

app/vmui/packages/vmui/src/pages/CardinalityPanel/CardinalityTotals/CardinalityTotals.tsx

@@ -32,7 +32,7 @@ const CardinalityTotals: FC<CardinalityTotalsProps> = ({
   const match = searchParams.get("match");
   const focusLabel = searchParams.get("focusLabel");
   const isMetric = /__name__/.test(match || "");
-
+  const showMetricNameStats = !(match || focusLabel);
   const progress = totalSeries / totalSeriesAll * 100;
   const diff = totalSeries - totalSeriesPrev;
   const dynamic = Math.abs(diff) / totalSeriesPrev * 100;
@@ -56,7 +56,7 @@ const CardinalityTotals: FC<CardinalityTotalsProps> = ({
     },
   ].filter(t => t.display);
 
-  if (!totals.length) {
+  if (!totals.length && !showMetricNameStats) {
     return null;
   }
 
@@ -75,7 +75,7 @@ const CardinalityTotals: FC<CardinalityTotalsProps> = ({
           <h4 className="vm-cardinality-totals-card__title">
             {info && (
               <Tooltip title={<p className="vm-cardinality-totals-card__tooltip">{info}</p>}>
-                <div className="vm-cardinality-totals-card__info-icon"><InfoOutlinedIcon/></div>
+                <div className="vm-cardinality-totals-card__info-icon"><InfoOutlinedIcon /></div>
               </Tooltip>
             )}
             {title}
@@ -99,7 +99,10 @@ const CardinalityTotals: FC<CardinalityTotalsProps> = ({
           )}
         </div>
       ))}
-      <CardinalityMetricNameStats metricNameStats={metricNameStats}/>
+      {
+        showMetricNameStats &&
+        <CardinalityMetricNameStats metricNameStats={metricNameStats} />
+      }
     </div>
   );
 };

app/vmui/packages/vmui/src/pages/ExploreAlerts/style.scss

@@ -62,6 +62,13 @@
     padding-right: 40px;
   }
 
+  .vm-annotations-table {
+    tbody > tr > td {
+      vertical-align: top;
+      white-space: pre-line;
+    }
+  }
+
   display: flex;
   flex-direction: column;
   align-items: stretch;

app/vmui/packages/vmui/src/pages/Relabel/index.tsx

@@ -72,7 +72,7 @@ const Relabel: FC = () => {
         <div className="vm-relabeling-header-configs">
           <TextField
             type="textarea"
-            label="Relabel configs"
+            label="Config"
             value={config}
             autofocus
             onChange={handleChangeConfig}
@@ -82,8 +82,9 @@ const Relabel: FC = () => {
         <div className="vm-relabeling-header__labels">
           <TextField
             type="textarea"
-            label="Labels"
+            label="A time series"
             value={labels}
+            placeholder="up{job=&quot;job_name&quot;,instance=&quot;host:port&quot;}"
             onChange={handleChangeLabels}
             onEnter={handleRunQuery}
           />
@@ -97,22 +98,22 @@ const Relabel: FC = () => {
             <Button
               variant="text"
               color="gray"
-              startIcon={<InfoIcon/>}
+              startIcon={<WikiIcon/>}
             >
               Relabeling cookbook
             </Button>
           </a>
           <a
             target="_blank"
-            href="https://docs.victoriametrics.com/victoriametrics/relabeling/"
+            href="https://docs.victoriametrics.com/victoriametrics/relabeling/#relabeling-stages"
             rel="help noreferrer"
           >
             <Button
               variant="text"
               color="gray"
-              startIcon={<WikiIcon/>}
+              startIcon={<InfoIcon/>}
             >
-              Documentation
+              Relabeling Stages
             </Button>
           </a>
           <Button

app/vmui/packages/vmui/src/utils/time.test.ts

@@ -1,6 +1,6 @@
 import { describe, it, expect } from "vitest";
 import dayjs from "dayjs";
-import { getNanoTimestamp, parseSupportedDuration } from "./time";
+import { getMillisecondsFromDuration, getNanoTimestamp, parseSupportedDuration } from "./time";
 
 describe("Time utils", () => {
   describe("getNanoTimestamp", () => {
@@ -82,4 +82,19 @@ describe("Time utils", () => {
       expect(parseSupportedDuration("   ")).toBeUndefined();
     });
   });
+
+  describe("getMillisecondsFromDuration", () => {
+    it("should convert valid durations to milliseconds", () => {
+      expect(getMillisecondsFromDuration("1s")).toBe(1000);
+      expect(getMillisecondsFromDuration("1.5s")).toBe(1500);
+      expect(getMillisecondsFromDuration("1m30s")).toBe(90000);
+      expect(getMillisecondsFromDuration("1h 30m")).toBe(5400000);
+      expect(getMillisecondsFromDuration("500ms")).toBe(500);
+    });
+
+    it("should return zero when duration cannot be parsed", () => {
+      expect(getMillisecondsFromDuration("garbage")).toBe(0);
+      expect(getMillisecondsFromDuration("")).toBe(0);
+    });
+  });
 });

app/vmui/packages/vmui/src/utils/time.ts

@@ -16,7 +16,7 @@ const supportedValuesOf = Intl.supportedValuesOf;
 export const supportedTimezones = supportedValuesOf ? supportedValuesOf("timeZone") as string[] : timezones;
 
 // The list of supported units could be the following -
-// https://prometheus.io/docs/prometheus/latest/querying/basics/#time-durations
+// https://prometheus.io/docs/prometheus/latest/querying/basics/#float-literals-and-time-durations
 export const supportedDurations = [
   { long: "years", short: "y", possible: "year" },
   { long: "weeks", short: "w", possible: "week" },
@@ -100,6 +100,10 @@ export const getSecondsFromDuration = (dur: string) => {
   return dayjs.duration(durObject).asSeconds();
 };
 
+export const getMillisecondsFromDuration = (dur: string): number => {
+  return getSecondsFromDuration(dur) * 1000;
+};
+
 const instantQueryViews = [DisplayType.table, DisplayType.code];
 export const getStepFromDuration = (dur: number, histogram?: boolean, displayType?: DisplayType): string => {
   if (displayType && instantQueryViews.includes(displayType)) return roundStep(dur);
@@ -276,4 +280,3 @@ export const getNanoTimestamp = (dateStr: string): bigint => {
   // Return the full timestamp in nanoseconds as a BigInt
   return BigInt(baseMs) * 1000000n + BigInt(extraNano);
 };
-

apptest/README.md

@@ -4,7 +4,7 @@ The `apptest` package contains the integration tests for the VictoriaMetrics
 applications (such as vmstorage, vminsert, and vmselect).
 
 An integration test aims at verifying the behavior of an application as a whole,
-as apposed to a unit test that verifies the behavior of a building block of an
+as opposed to a unit test that verifies the behavior of a building block of an
 application.
 
 To achieve that an integration test starts an application in a separate process
@@ -19,10 +19,10 @@ work together as a system.
 The package provides a collection of helpers to start applications and make
 queries to them:
 
--   `app.go` - contains the generic code for staring an application and should
+-   `app.go` - contains the generic code for starting an application and should
     not be used by integration tests directly.
 -   `{vmstorage,vminsert,etc}.go` - build on top of `app.go` and provide the
-    code for staring a specific application.
+    code for starting a specific application.
 -   `client.go` - provides helper functions for sending HTTP requests to
     applications.
 
@@ -36,7 +36,7 @@ the application binary files to be built and put into the `bin` directory. The
 build rule used for running integration tests, `make apptest`,
 accounts for that, it builds all application binaries before running the tests.
 But if you want to run the tests without `make`, i.e. by executing
-`go test ./app/apptest`, you will need to build the binaries first (for example,
+`go test ./apptest/tests`, you will need to build the binaries first (for example,
 by executing `make all`).
 
 Not all binaries can be built from `master` branch, cluster binaries can be built

apptest/testdata.go

@@ -0,0 +1,358 @@
+package apptest
+
+import (
+	"fmt"
+	"slices"
+)
+
+type TestData struct {
+	Samples              []string
+	Step                 int64
+	WantSeries           []map[string]string
+	WantLabels           []string
+	WantLabelValues      []string
+	WantQueryResults     []*QueryResult
+	WantMetadata         map[string][]MetadataEntry
+	WantMetricNamesStats []MetricNamesStatsRecord
+}
+
+func GenerateTestData(prefix string, numMetrics, start, end int64) TestData {
+	d := TestData{
+		Samples:              []string{},
+		Step:                 (end - start) / numMetrics,
+		WantSeries:           make([]map[string]string, numMetrics),
+		WantLabels:           make([]string, numMetrics),
+		WantLabelValues:      make([]string, numMetrics),
+		WantQueryResults:     make([]*QueryResult, numMetrics),
+		WantMetadata:         make(map[string][]MetadataEntry),
+		WantMetricNamesStats: make([]MetricNamesStatsRecord, numMetrics),
+	}
+	for i := range numMetrics {
+		metricName := fmt.Sprintf("%s_%04d", prefix, i)
+		metricHelp := fmt.Sprintf("# HELP %s some help message", metricName)
+		metricType := fmt.Sprintf("# TYPE %s gauge", metricName)
+		labelName := fmt.Sprintf("label_%04d", i)
+		labelValue := fmt.Sprintf("value_%04d", i)
+		value := i
+		timestamp := start + i*d.Step
+		sample := fmt.Sprintf(`%s{%s="value", label="%s"} %d %d`, metricName, labelName, labelValue, value, timestamp)
+
+		d.Samples = append(d.Samples, metricHelp, metricType, sample)
+		d.WantSeries[i] = map[string]string{
+			"__name__": metricName,
+			labelName:  "value",
+			"label":    labelValue,
+		}
+		d.WantLabels[i] = labelName
+		d.WantLabelValues[i] = labelValue
+		d.WantQueryResults[i] = &QueryResult{
+			Metric: map[string]string{
+				"__name__": metricName,
+				labelName:  "value",
+				"label":    labelValue,
+			},
+			Samples: []*Sample{{Timestamp: timestamp, Value: float64(value)}},
+		}
+		d.WantMetadata[metricName] = []MetadataEntry{{Help: "some help message", Type: "gauge"}}
+		d.WantMetricNamesStats[i].MetricName = metricName
+	}
+	d.WantLabels = append(d.WantLabels, "__name__", "label")
+	slices.Sort(d.WantLabels)
+	return d
+}
+
+// AssertSeries retrieves metric names from the storage and compares the result
+// with the expected one.
+func AssertSeries(tc *TestCase, app PrometheusQuerier, metricNameRE, tenantID string, start, end int64, want []map[string]string) {
+	tc.T().Helper()
+
+	query := fmt.Sprintf(`{__name__=~"%s"}`, metricNameRE)
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /prometheus/api/v1/series response",
+		Got: func() any {
+			tc.T().Helper()
+			return app.PrometheusAPIV1Series(tc.T(), query, QueryOpts{
+				Tenant: tenantID,
+				Start:  fmt.Sprintf("%d", start),
+				End:    fmt.Sprintf("%d", end),
+			}).Sort()
+		},
+		Want: &PrometheusAPIV1SeriesResponse{
+			Status: "success",
+			Data:   want,
+		},
+		Retries: 1000,
+		FailNow: true,
+	})
+}
+
+// AssertSeriesCount retrieves series count and compares it with expected one.
+func AssertSeriesCount(tc *TestCase, app PrometheusQuerier, tenantID string, start, end int64, want uint64) {
+	tc.T().Helper()
+
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /prometheus/api/v1/series/count response",
+		Got: func() any {
+			tc.T().Helper()
+			return app.PrometheusAPIV1SeriesCount(tc.T(), QueryOpts{
+				Tenant: tenantID,
+				Start:  fmt.Sprintf("%d", start),
+				End:    fmt.Sprintf("%d", end),
+			})
+		},
+		Want: &PrometheusAPIV1SeriesCountResponse{
+			Status: "success",
+			Data:   []uint64{want},
+		},
+		FailNow: true,
+	})
+}
+
+// AssertLabels retrieves label names from the storage and compares the result
+// with the expected one.
+func AssertLabels(tc *TestCase, app PrometheusQuerier, metricNameRE, tenantID string, start, end int64, want []string) {
+	tc.T().Helper()
+
+	query := fmt.Sprintf(`{__name__=~"%s"}`, metricNameRE)
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /prometheus/api/v1/labels response",
+		Got: func() any {
+			tc.T().Helper()
+			res := app.PrometheusAPIV1Labels(tc.T(), query, QueryOpts{
+				Tenant: tenantID,
+				Start:  fmt.Sprintf("%d", start),
+				End:    fmt.Sprintf("%d", end),
+			})
+			slices.Sort(res.Data)
+			return res
+		},
+		Want: &PrometheusAPIV1LabelsResponse{
+			Status: "success",
+			Data:   want,
+		},
+		FailNow: true,
+	})
+}
+
+// AssertLabelValues retrieves values for the label whose name is labelName for
+// the series whose name mathes metricNameRE, compares the result with the
+// expected one.
+func AssertLabelValues(tc *TestCase, app PrometheusQuerier, metricNameRE, labelName, tenantID string, start, end int64, want []string) {
+	tc.T().Helper()
+
+	query := fmt.Sprintf(`{__name__=~"%s"}`, metricNameRE)
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /prometheus/api/v1/labels/.../values response",
+		Got: func() any {
+			tc.T().Helper()
+			res := app.PrometheusAPIV1LabelValues(tc.T(), labelName, query, QueryOpts{
+				Tenant: tenantID,
+				Start:  fmt.Sprintf("%d", start),
+				End:    fmt.Sprintf("%d", end),
+			})
+			slices.Sort(res.Data)
+			return res
+		},
+		Want: &PrometheusAPIV1LabelValuesResponse{
+			Status: "success",
+			Data:   want,
+		},
+		FailNow: true,
+	})
+}
+
+// AssertQueryResults sends a data query to storage and compares the query
+// result with the expected one.
+func AssertQueryResults(tc *TestCase, app PrometheusQuerier, metricNameRE, tenantID string, start, end, step int64, want []*QueryResult) {
+	tc.T().Helper()
+
+	query := fmt.Sprintf(`{__name__=~"%s"}`, metricNameRE)
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /prometheus/api/v1/query_range response",
+		Got: func() any {
+			tc.T().Helper()
+			return app.PrometheusAPIV1QueryRange(tc.T(), query, QueryOpts{
+				Tenant:      tenantID,
+				Start:       fmt.Sprintf("%d", start),
+				End:         fmt.Sprintf("%d", end),
+				Step:        fmt.Sprintf("%dms", step),
+				MaxLookback: fmt.Sprintf("%dms", step-1),
+				NoCache:     "1",
+			})
+		},
+		Want: &PrometheusAPIV1QueryResponse{
+			Status: "success",
+			Data: &QueryData{
+				ResultType: "matrix",
+				Result:     want,
+			},
+		},
+		FailNow: true,
+	})
+}
+
+func AssertMetadata(tc *TestCase, app PrometheusQuerier, metricName, tenantID string, want map[string][]MetadataEntry) {
+	tc.T().Helper()
+
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /prometheus/api/v1/metadata response",
+		Got: func() any {
+			tc.T().Helper()
+			return app.PrometheusAPIV1Metadata(tc.T(), metricName, 0, QueryOpts{
+				Tenant: tenantID,
+			})
+		},
+		Want: &PrometheusAPIV1Metadata{
+			Status: "success",
+			Data:   want,
+		},
+		FailNow: true,
+	})
+}
+
+func AssertMetricNamesStats(tc *TestCase, app PrometheusQuerier, metricNameRE, tenantID string, want []MetricNamesStatsRecord) {
+	tc.T().Helper()
+
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /prometheus/api/v1/status/metric_names_stats response",
+		Got: func() any {
+			tc.T().Helper()
+			return app.PrometheusAPIV1StatusMetricNamesStats(tc.T(), "", "", metricNameRE, QueryOpts{
+				Tenant: tenantID,
+			})
+		},
+		Want: MetricNamesStatsResponse{
+			Records: want,
+		},
+		FailNow: true,
+	})
+
+}
+
+// GraphiteTestData holds the data samples in Graphite Pickle format, distance
+// between samples in milliseconds and expected responses for various Graphite
+// API endpoints.
+type GraphiteTestData struct {
+	Samples             []string
+	Step                int64
+	WantMetricsIndex    []string
+	WantMetricsFind     []GraphiteMetric
+	WantMetricsExpand   []string
+	WantRenderedTargets []GraphiteRenderedTarget
+}
+
+// GenerateGraphiteTestData generates Graphite test data.
+func GenerateGraphiteTestData(prefix string, numMetrics, start, end int64) GraphiteTestData {
+	d := GraphiteTestData{
+		Samples:             make([]string, numMetrics),
+		Step:                (end - start) / numMetrics,
+		WantMetricsIndex:    make([]string, numMetrics),
+		WantMetricsFind:     make([]GraphiteMetric, numMetrics),
+		WantMetricsExpand:   make([]string, numMetrics),
+		WantRenderedTargets: make([]GraphiteRenderedTarget, numMetrics),
+	}
+
+	datapoints := make([][2]float64, numMetrics)
+	for i := range numMetrics {
+		timestamp := (start + i*d.Step) / 1000
+		datapoints[i][1] = float64(timestamp)
+	}
+
+	for i := range numMetrics {
+		suffix := fmt.Sprintf("%04d", i)
+		metricName := fmt.Sprintf("%s.%s", prefix, suffix)
+		value := i
+		timestamp := (start + i*d.Step) / 1000
+		sample := fmt.Sprintf(`%s %d %d`, metricName, value, timestamp)
+
+		d.Samples[i] = sample
+		d.WantMetricsIndex[i] = metricName
+		d.WantMetricsFind[i].Id = metricName
+		d.WantMetricsFind[i].Text = suffix
+		d.WantMetricsFind[i].Leaf = 1
+		d.WantMetricsExpand[i] = metricName
+		d.WantRenderedTargets[i].Target = metricName
+		d.WantRenderedTargets[i].Datapoints = slices.Clone(datapoints)
+		d.WantRenderedTargets[i].Datapoints[i][0] = float64(value)
+	}
+	return d
+}
+
+// AssertGraphiteMetricsIndex retrieves all metrics by sending a request to
+// /graphite/metrics/index.json and compares the result with the expected one.
+func AssertGraphiteMetricsIndex(tc *TestCase, app PrometheusQuerier, tenantID string, want []string) {
+	tc.T().Helper()
+
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /graphite/metrics/index.json response",
+		Got: func() any {
+			tc.T().Helper()
+			return app.GraphiteMetricsIndex(tc.T(), QueryOpts{
+				Tenant: tenantID,
+			})
+		},
+		Want:    want,
+		Retries: 30,
+		FailNow: true,
+	})
+
+}
+
+// AssertGraphiteMetricsFind finds metric names by sending a request to
+// /graphite/metrics/find and compares the result with the expected one.
+func AssertGraphiteMetricsFind(tc *TestCase, app PrometheusQuerier, query, tenantID string, want []GraphiteMetric) {
+	tc.T().Helper()
+
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /graphite/metrics/find response",
+		Got: func() any {
+			tc.T().Helper()
+			return app.GraphiteMetricsFind(tc.T(), query, QueryOpts{
+				Tenant: tenantID,
+			})
+		},
+		Want:    want,
+		FailNow: true,
+	})
+
+}
+
+// AssertGraphiteMetricsFind expands metric names by sending a request to
+// /graphite/metrics/expand and compares the result with the expected one.
+func AssertGraphiteMetricsExpand(tc *TestCase, app PrometheusQuerier, query, tenantID string, want []string) {
+	tc.T().Helper()
+
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /graphite/metrics/expand response",
+		Got: func() any {
+			tc.T().Helper()
+			return app.GraphiteMetricsExpand(tc.T(), query, QueryOpts{
+				Tenant: tenantID,
+			})
+		},
+		Want:    want,
+		FailNow: true,
+	})
+
+}
+
+// AssertGraphiteRender retieves metric raw data by sending a request to
+// /graphite/render and compares the result with the expected one.
+func AssertGraphiteRender(tc *TestCase, app PrometheusQuerier, target, tenantID string, from, until, step int64, want []GraphiteRenderedTarget) {
+	tc.T().Helper()
+
+	tc.Assert(&AssertOptions{
+		Msg: "unexpected /graphite/render response",
+		Got: func() any {
+			tc.T().Helper()
+			return app.GraphiteRender(tc.T(), target, QueryOpts{
+				Tenant:      tenantID,
+				From:        fmt.Sprintf("%d", from/1000),
+				Until:       fmt.Sprintf("%d", until/1000),
+				StorageStep: fmt.Sprintf("%dms", step),
+			})
+		},
+		Want:    want,
+		FailNow: true,
+	})
+}

apptest/tests/max_backfill_age_test.go

@@ -0,0 +1,222 @@
+package tests
+
+import (
+	"fmt"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/apptest"
+)
+
+func TestSingleMaxBackfillAge(t *testing.T) {
+	tc := apptest.NewTestCase(t)
+	defer tc.Stop()
+
+	opts := maxBackfillAgeOpts{
+		start: func(retentionPeriod, maxBackfillAge string) apptest.PrometheusWriteQuerier {
+			return tc.MustStartVmsingle("vmsingle", []string{
+				"-storageDataPath=" + filepath.Join(tc.Dir(), "vmsingle"),
+				"-retentionPeriod=" + retentionPeriod,
+				"-maxBackfillAge=" + maxBackfillAge,
+			})
+		},
+		stop: func() {
+			tc.StopApp("vmsingle")
+		},
+	}
+
+	testMaxBackfillAge(tc, opts)
+}
+
+func TestClusterMaxBackfillAge(t *testing.T) {
+	tc := apptest.NewTestCase(t)
+	defer tc.Stop()
+
+	opts := maxBackfillAgeOpts{
+		start: func(retentionPeriod, maxBackfillAge string) apptest.PrometheusWriteQuerier {
+			return tc.MustStartCluster(&apptest.ClusterOptions{
+				Vmstorage1Instance: "vmstorage1",
+				Vmstorage1Flags: []string{
+					"-storageDataPath=" + filepath.Join(tc.Dir(), "vmstorage1"),
+					"-retentionPeriod=" + retentionPeriod,
+					"-maxBackfillAge=" + maxBackfillAge,
+				},
+				Vmstorage2Instance: "vmstorage2",
+				Vmstorage2Flags: []string{
+					"-storageDataPath=" + filepath.Join(tc.Dir(), "vmstorage2"),
+					"-retentionPeriod=" + retentionPeriod,
+					"-maxBackfillAge=" + maxBackfillAge,
+				},
+				VminsertInstance: "vminsert",
+				VminsertFlags:    []string{},
+				VmselectInstance: "vmselect",
+				VmselectFlags:    []string{},
+			})
+		},
+		stop: func() {
+			tc.StopApp("vminsert")
+			tc.StopApp("vmselect")
+			tc.StopApp("vmstorage1")
+			tc.StopApp("vmstorage2")
+		},
+	}
+
+	testMaxBackfillAge(tc, opts)
+}
+
+type maxBackfillAgeOpts struct {
+	start func(retentionPeriod, maxBackfillAge string) apptest.PrometheusWriteQuerier
+	stop  func()
+}
+
+func testMaxBackfillAge(tc *apptest.TestCase, opts maxBackfillAgeOpts) {
+	t := tc.T()
+
+	assertSeries := func(app apptest.PrometheusQuerier, prefix string, start, end int64, want []map[string]string) {
+		t.Helper()
+
+		query := fmt.Sprintf(`{__name__=~"metric_%s.*"}`, prefix)
+		tc.Assert(&apptest.AssertOptions{
+			Msg: "unexpected /api/v1/series response",
+			Got: func() any {
+				return app.PrometheusAPIV1Series(t, query, apptest.QueryOpts{
+					Start: fmt.Sprintf("%d", start),
+					End:   fmt.Sprintf("%d", end),
+				}).Sort()
+			},
+			Want: &apptest.PrometheusAPIV1SeriesResponse{
+				Status: "success",
+				Data:   want,
+			},
+			FailNow: true,
+		})
+	}
+
+	assertQueryResults := func(app apptest.PrometheusQuerier, prefix string, start, end, step int64, want []*apptest.QueryResult) {
+		t.Helper()
+
+		query := fmt.Sprintf(`{__name__=~"metric_%s.*"}`, prefix)
+		tc.Assert(&apptest.AssertOptions{
+			Msg: "unexpected /api/v1/query_range response",
+			Got: func() any {
+				return app.PrometheusAPIV1QueryRange(t, query, apptest.QueryOpts{
+					Start:       fmt.Sprintf("%d", start),
+					End:         fmt.Sprintf("%d", end),
+					Step:        fmt.Sprintf("%dms", step),
+					MaxLookback: fmt.Sprintf("%dms", step-1),
+					NoCache:     "1",
+				})
+			},
+			Want: &apptest.PrometheusAPIV1QueryResponse{
+				Status: "success",
+				Data: &apptest.QueryData{
+					ResultType: "matrix",
+					Result:     want,
+				},
+			},
+			FailNow: true,
+		})
+	}
+
+	const numMetrics = 1000
+	now := time.Now().UTC()
+	var start, end, step int64
+	emptySeries := []map[string]string{}
+	emptyQueryResults := []*apptest.QueryResult{}
+
+	// Start sut with the same -retentionPeriod and -maxBackfillAge.
+	sut := opts.start("1y", "1y")
+
+	// Verify that samples older than the retention period are rejected.
+	start = now.Add(-365 * 24 * time.Hour).Add(-time.Hour).UnixMilli()
+	end = now.Add(-365 * 24 * time.Hour).UnixMilli()
+	step = (end - start) / numMetrics
+	outsideRetention := genMaxBackfillAgeData("outside_retention", numMetrics, start, step)
+	sut.PrometheusAPIV1ImportPrometheus(t, outsideRetention.samples, apptest.QueryOpts{})
+	sut.ForceFlush(t)
+	assertSeries(sut, "outside_retention", start, end, emptySeries)
+	assertQueryResults(sut, "outside_retention", start, end, step, emptyQueryResults)
+
+	// Verify that samples within the retention period are accepted and
+	// searcheable.
+	start = now.Add(-365 * 24 * time.Hour).Add(time.Hour).UnixMilli()
+	end = now.Add(-365 * 24 * time.Hour).Add(2 * time.Hour).UnixMilli()
+	step = (end - start) / numMetrics
+	insideRetention := genMaxBackfillAgeData("inside_retention", numMetrics, start, step)
+	sut.PrometheusAPIV1ImportPrometheus(t, insideRetention.samples, apptest.QueryOpts{})
+	sut.ForceFlush(t)
+	assertSeries(sut, "inside_retention", start, end, insideRetention.wantSeries)
+	assertQueryResults(sut, "inside_retention", start, end, step, insideRetention.wantQueryResults)
+
+	// Restart sut with -maxBackfillAge shorter than the -retentionPeriod.
+	opts.stop()
+	sut = opts.start("1y", "6M")
+
+	// Verify that new samples older than max backfill age but still within the
+	// retention period are rejected but existing samples are still searcheable.
+	start = now.Add(-365 * 24 * time.Hour).Add(time.Hour).UnixMilli()
+	end = now.Add(-365 * 24 * time.Hour).Add(2 * time.Hour).UnixMilli()
+	step = (end - start) / numMetrics
+	insideRetention2 := genMaxBackfillAgeData("inside_retention2", numMetrics, start, step)
+	sut.PrometheusAPIV1ImportPrometheus(t, insideRetention2.samples, apptest.QueryOpts{})
+	sut.ForceFlush(t)
+	assertSeries(sut, "inside_retention2", start, end, emptySeries)
+	assertQueryResults(sut, "inside_retention2", start, end, step, emptyQueryResults)
+	assertSeries(sut, "inside_retention", start, end, insideRetention.wantSeries)
+	assertQueryResults(sut, "inside_retention", start, end, step, insideRetention.wantQueryResults)
+
+	// Verify that the metrics that are outside the backfill window can still
+	// be deleted.
+	sut.PrometheusAPIV1AdminTSDBDeleteSeries(t, `{__name__=~".*inside_retention.*"}`, apptest.QueryOpts{})
+	sut.ForceFlush(t)
+	assertSeries(sut, "inside_retention", start, end, emptySeries)
+	assertQueryResults(sut, "inside_retention", start, end, step, emptyQueryResults)
+
+	// Verify that the samples that are within the backfill window are accepted
+	// and searchable.
+	start = now.Add(-180 * 24 * time.Hour).UnixMilli()
+	end = now.Add(-180 * 24 * time.Hour).Add(1 * time.Hour).UnixMilli()
+	step = (end - start) / numMetrics
+	insideMaxBackfillAge := genMaxBackfillAgeData("inside_max_backfill_age", numMetrics, start, step)
+	sut.PrometheusAPIV1ImportPrometheus(t, insideMaxBackfillAge.samples, apptest.QueryOpts{})
+	sut.ForceFlush(t)
+	assertSeries(sut, "inside_max_backfill_age", start, end, insideMaxBackfillAge.wantSeries)
+	assertQueryResults(sut, "inside_max_backfill_age", start, end, step, insideMaxBackfillAge.wantQueryResults)
+
+	opts.stop()
+}
+
+type maxBackfillAgeData struct {
+	samples          []string
+	wantSeries       []map[string]string
+	wantQueryResults []*apptest.QueryResult
+}
+
+func genMaxBackfillAgeData(prefix string, numMetrics, start, step int64) maxBackfillAgeData {
+	samples := make([]string, numMetrics)
+	wantSeries := make([]map[string]string, numMetrics)
+	wantQueryResults := make([]*apptest.QueryResult, numMetrics)
+	for i := range numMetrics {
+		metricName := fmt.Sprintf("metric_%s_%04d", prefix, i)
+		labelName := fmt.Sprintf("label_%s_%04d", prefix, i)
+		labelValue := fmt.Sprintf("value_%s_%04d", prefix, i)
+		value := i
+		timestamp := start + i*step
+		samples[i] = fmt.Sprintf(`%s{%s="value", label="%s"} %d %d`, metricName, labelName, labelValue, value, timestamp)
+		wantSeries[i] = map[string]string{
+			"__name__": metricName,
+			labelName:  "value",
+			"label":    labelValue,
+		}
+		wantQueryResults[i] = &apptest.QueryResult{
+			Metric: map[string]string{
+				"__name__": metricName,
+				labelName:  "value",
+				"label":    labelValue,
+			},
+			Samples: []*apptest.Sample{{Timestamp: timestamp, Value: float64(value)}},
+		}
+	}
+	return maxBackfillAgeData{samples, wantSeries, wantQueryResults}
+}

apptest/tests/metricsmetadata_test.go

@@ -2,6 +2,7 @@ package tests
 
 import (
 	"fmt"
+	"math"
 	"testing"
 
 	"github.com/google/go-cmp/cmp"
@@ -25,7 +26,11 @@ func TestSingleMetricsMetadata(t *testing.T) {
 	if len(resp.Data) != 0 {
 		t.Fatalf("unexpected resp Records: %d, want: %d", len(resp.Data), 0)
 	}
-
+	generateValueExceedLimit := func(prefix string) string {
+		buf := make([]byte, math.MaxUint16+len(prefix))
+		copy(buf, prefix)
+		return string(buf)
+	}
 	const ingestTimestamp = 1707123456700
 	prometheusTextDataSet := []string{
 		`# HELP metric_name_1 some help message`,
@@ -40,6 +45,12 @@ func TestSingleMetricsMetadata(t *testing.T) {
 		`# TYPE metric_name_3 gauge`,
 		`metric_name_3{label="baz"} 30`,
 	}
+	prometheusTextDataSet = append(prometheusTextDataSet,
+		`# HELP metric_name_4 `+generateValueExceedLimit("large help"),
+		`# TYPE metric_name_4 gauge`,
+		`metric_name_4{label="baz"} 30`,
+	)
+
 	prometheusRemoteWriteDataSet := prompb.WriteRequest{
 		Timeseries: []prompb.TimeSeries{
 			{Labels: []prompb.Label{{Name: "__name__", Value: "metric_name_4"}}, Samples: []prompb.Sample{{Value: 40, Timestamp: ingestTimestamp}}},
@@ -52,6 +63,9 @@ func TestSingleMetricsMetadata(t *testing.T) {
 			{MetricFamilyName: "metric_name_5", Help: "some help message", Type: prompb.MetricTypeSummary},
 			{MetricFamilyName: "metric_name_6", Help: "some help message", Type: prompb.MetricTypeStateset},
 			{MetricFamilyName: `metric_name_7_!@"_suffix`, Help: "some help message", Type: prompb.MetricTypeStateset},
+			{MetricFamilyName: "metric_name_8", Help: generateValueExceedLimit("large_help"), Type: prompb.MetricTypeStateset},
+			{MetricFamilyName: "metric_name_9", Help: "some help message", Type: prompb.MetricTypeStateset, Unit: generateValueExceedLimit("large_unit")},
+			{MetricFamilyName: generateValueExceedLimit("metric_name_10"), Help: "some help message", Type: prompb.MetricTypeStateset},
 		},
 	}
 
@@ -137,6 +151,11 @@ func TestClusterMetricsMetadata(t *testing.T) {
 	if len(resp.Data) != 0 {
 		t.Fatalf("unexpected resp Records: %d, want: %d", len(resp.Data), 0)
 	}
+	generateValueExceedLimit := func(prefix string) string {
+		buf := make([]byte, math.MaxUint16+len(prefix))
+		copy(buf, prefix)
+		return string(buf)
+	}
 
 	const ingestTimestamp = 1707123456700
 	prometheusTextDataSet := []string{
@@ -152,6 +171,11 @@ func TestClusterMetricsMetadata(t *testing.T) {
 		`# TYPE metric_name_3 gauge`,
 		`metric_name_3{label="baz"} 30`,
 	}
+	prometheusTextDataSet = append(prometheusTextDataSet,
+		`# HELP metric_name_4 `+generateValueExceedLimit("large help"),
+		`# TYPE metric_name_4 gauge`,
+		`metric_name_4{label="baz"} 30`,
+	)
 	prometheusRemoteWriteDataSet := prompb.WriteRequest{
 		Timeseries: []prompb.TimeSeries{
 			{Labels: []prompb.Label{{Name: "__name__", Value: "metric_name_4"}}, Samples: []prompb.Sample{{Value: 40, Timestamp: ingestTimestamp}}},
@@ -164,6 +188,9 @@ func TestClusterMetricsMetadata(t *testing.T) {
 			{MetricFamilyName: "metric_name_5", Help: "some help message", Type: prompb.MetricTypeSummary},
 			{MetricFamilyName: "metric_name_6", Help: "some help message", Type: prompb.MetricTypeStateset},
 			{MetricFamilyName: `metric_name_7_!@"_suffix`, Help: "some help message", Type: prompb.MetricTypeStateset},
+			{MetricFamilyName: "metric_name_8", Help: generateValueExceedLimit("large_help"), Type: prompb.MetricTypeStateset},
+			{MetricFamilyName: "metric_name_9", Help: "some help message", Type: prompb.MetricTypeStateset, Unit: generateValueExceedLimit("large_unit")},
+			{MetricFamilyName: generateValueExceedLimit("metric_name_10"), Help: "some help message", Type: prompb.MetricTypeStateset},
 		},
 	}
 

apptest/tests/vmagent_remotewrite_test.go

@@ -517,10 +517,15 @@ func TestClusterVMAgentForwardMetricsMetadata(t *testing.T) {
 		"-remoteWrite.tmpDataPath=" + tc.Dir() + "/vmagent",
 		fmt.Sprintf(`-remoteWrite.url=http://%s/insert/multitenant/prometheus/api/v1/write`, sut.Vminsert.HTTPAddr()),
 	})
-
+	generateValueExceedLimit := func(prefix string) string {
+		buf := make([]byte, math.MaxUint16+len(prefix))
+		copy(buf, prefix)
+		return string(buf)
+	}
 	prometheusRemoteWriteDataSet := prompb.WriteRequest{
 		Metadata: []prompb.MetricMetadata{
 			{MetricFamilyName: "metric_name_4", Help: "some help message", Type: prompb.MetricTypeSummary, AccountID: 100},
+			{MetricFamilyName: "metric_name_8", Help: generateValueExceedLimit("large_help"), Type: prompb.MetricTypeStateset, AccountID: 100},
 		},
 	}
 	vmagent.PrometheusAPIV1Write(t, prometheusRemoteWriteDataSet, apptest.QueryOpts{Tenant: "multitenant"})

apptest/tests/vmsingle_vmselect_rpc_test.go

@@ -0,0 +1,216 @@
+package tests
+
+import (
+	"fmt"
+	"path/filepath"
+	"slices"
+	"strconv"
+	"testing"
+	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/apptest"
+	"github.com/google/go-cmp/cmp"
+)
+
+func TestMixedPrometheusQueries(t *testing.T) {
+	tc := apptest.NewTestCase(t)
+	defer tc.Stop()
+
+	const (
+		accountID1 = 12
+		projectID1 = 34
+		accountID2 = 56
+		projectID2 = 78
+		numMetrics = 10
+	)
+	tenantID1 := fmt.Sprintf("%d:%d", accountID1, projectID1)
+	tenantID2 := fmt.Sprintf("%d:%d", accountID2, projectID2)
+	start := time.Date(2026, 1, 1, 0, 0, 0, 0, time.UTC).UnixMilli()
+	end := time.Date(2026, 1, 2, 0, 0, 0, 0, time.UTC).UnixMilli()
+	data := apptest.GenerateTestData("metric", numMetrics, start, end)
+	emptySeries := []map[string]string{}
+	emptyLabels := []string{}
+	emptyLabelValues := []string{}
+	emptyQueryResults := []*apptest.QueryResult{}
+	emptyMetadata := map[string][]apptest.MetadataEntry{}
+	emptyMetricNamesStats := []apptest.MetricNamesStatsRecord{}
+
+	vmsingle := tc.MustStartVmsingle("vmsingle", []string{
+		"-storageDataPath=" + filepath.Join(tc.Dir(), "vmsingle"),
+		"-retentionPeriod=100y",
+		fmt.Sprintf("-accountID=%d", accountID1),
+		fmt.Sprintf("-projectID=%d", projectID1),
+	})
+	vmselect := tc.MustStartVmselect("vmselect", []string{
+		"-storageNode=" + vmsingle.VmselectAddr(),
+	})
+
+	vmsingle.PrometheusAPIV1ImportPrometheus(tc.T(), data.Samples, apptest.QueryOpts{})
+	vmsingle.ForceFlush(t)
+
+	// Ensure vmsingle returns data.
+	apptest.AssertSeries(tc, vmsingle, "metric.*", "", start, end, data.WantSeries)
+	apptest.AssertSeriesCount(tc, vmsingle, "", start, end, numMetrics)
+	apptest.AssertLabels(tc, vmsingle, "metric.*", "", start, end, data.WantLabels)
+	apptest.AssertLabelValues(tc, vmsingle, "metric.*", "label", "", start, end, data.WantLabelValues)
+	apptest.AssertQueryResults(tc, vmsingle, "metric.*", "", start, end, data.Step, data.WantQueryResults)
+	apptest.AssertMetadata(tc, vmsingle, "", "", data.WantMetadata)
+	for i := range data.WantMetricNamesStats {
+		data.WantMetricNamesStats[i].QueryRequestsCount = 1
+	}
+	apptest.AssertMetricNamesStats(tc, vmsingle, "", "", data.WantMetricNamesStats)
+
+	// Check that current vmsingle tenant (configured via flags) is tenant1.
+	gotAdminTenantsResponse := vmselect.APIV1AdminTenants(t, apptest.QueryOpts{})
+	wantAdminTenantsResponse := &apptest.AdminTenantsResponse{
+		Status: "success",
+		Data:   []string{tenantID1},
+	}
+	if diff := cmp.Diff(wantAdminTenantsResponse, gotAdminTenantsResponse); diff != "" {
+		t.Fatalf("unexpected tenants (-want, +got):\n%s", diff)
+	}
+
+	// Ensure vmselect returns data for tenant1.
+	apptest.AssertSeries(tc, vmselect, "metric.*", tenantID1, start, end, data.WantSeries)
+	apptest.AssertSeriesCount(tc, vmselect, tenantID1, start, end, numMetrics)
+	apptest.AssertLabels(tc, vmselect, "metric.*", tenantID1, start, end, data.WantLabels)
+	apptest.AssertLabelValues(tc, vmselect, "metric.*", "label", tenantID1, start, end, data.WantLabelValues)
+	apptest.AssertQueryResults(tc, vmselect, "metric.*", tenantID1, start, end, data.Step, data.WantQueryResults)
+	apptest.AssertMetadata(tc, vmselect, "", tenantID1, data.WantMetadata)
+	for i := range data.WantMetricNamesStats {
+		data.WantMetricNamesStats[i].QueryRequestsCount = 2
+	}
+	apptest.AssertMetricNamesStats(tc, vmselect, "", tenantID1, data.WantMetricNamesStats)
+
+	// Ensure vmselect does not return any data for tenant2.
+	apptest.AssertSeries(tc, vmselect, "metric.*", tenantID2, start, end, emptySeries)
+	apptest.AssertSeriesCount(tc, vmselect, tenantID2, start, end, 0)
+	apptest.AssertLabels(tc, vmselect, "metric.*", tenantID2, start, end, emptyLabels)
+	apptest.AssertLabelValues(tc, vmselect, "metric.*", "label", tenantID2, start, end, emptyLabelValues)
+	apptest.AssertQueryResults(tc, vmselect, "metric.*", tenantID2, start, end, data.Step, emptyQueryResults)
+	apptest.AssertMetadata(tc, vmselect, "", tenantID2, emptyMetadata)
+	apptest.AssertMetricNamesStats(tc, vmselect, "", tenantID2, emptyMetricNamesStats)
+
+	// Ensure vmselect returns data for multitenant.
+	for _, v := range data.WantSeries {
+		v["vm_account_id"] = strconv.Itoa(accountID1)
+		v["vm_project_id"] = strconv.Itoa(projectID1)
+	}
+	apptest.AssertSeries(tc, vmselect, "metric.*", "multitenant", start, end, data.WantSeries)
+	data.WantLabels = append(data.WantLabels, "vm_account_id", "vm_project_id")
+	apptest.AssertLabels(tc, vmselect, "metric.*", "multitenant", start, end, data.WantLabels)
+	apptest.AssertLabelValues(tc, vmselect, "metric.*", "label", "multitenant", start, end, data.WantLabelValues)
+	for _, v := range data.WantQueryResults {
+		v.Metric["vm_account_id"] = strconv.Itoa(accountID1)
+		v.Metric["vm_project_id"] = strconv.Itoa(projectID1)
+	}
+	apptest.AssertQueryResults(tc, vmselect, "metric.*", "multitenant", start, end, data.Step, data.WantQueryResults)
+	apptest.AssertMetadata(tc, vmselect, "", "multitenant", data.WantMetadata)
+	for i := range data.WantMetricNamesStats {
+		data.WantMetricNamesStats[i].QueryRequestsCount = 3
+	}
+	apptest.AssertMetricNamesStats(tc, vmselect, "", "multitenant", data.WantMetricNamesStats)
+}
+
+func TestMixedDeleteSeries(t *testing.T) {
+	tc := apptest.NewTestCase(t)
+	defer tc.Stop()
+
+	const (
+		accountID1 = 12
+		projectID1 = 34
+		accountID2 = 56
+		projectID2 = 78
+		numMetrics = 10
+	)
+	tenantID1 := fmt.Sprintf("%d:%d", accountID1, projectID1)
+	tenantID2 := fmt.Sprintf("%d:%d", accountID2, projectID2)
+	start := time.Date(2026, 1, 1, 0, 0, 0, 0, time.UTC).UnixMilli()
+	end := time.Date(2026, 1, 2, 0, 0, 0, 0, time.UTC).UnixMilli()
+	data1 := apptest.GenerateTestData("metric1", numMetrics, start, end)
+	data2 := apptest.GenerateTestData("metric2", numMetrics, start, end)
+	emptySeries := []map[string]string{}
+
+	vmsingle := tc.MustStartVmsingle("vmsingle", []string{
+		"-storageDataPath=" + filepath.Join(tc.Dir(), "vmsingle"),
+		"-retentionPeriod=100y",
+		fmt.Sprintf("-accountID=%d", accountID1),
+		fmt.Sprintf("-projectID=%d", projectID1),
+	})
+	vmselect := tc.MustStartVmselect("vmselect", []string{
+		"-storageNode=" + vmsingle.VmselectAddr(),
+	})
+
+	vmsingle.PrometheusAPIV1ImportPrometheus(tc.T(), data1.Samples, apptest.QueryOpts{})
+	vmsingle.PrometheusAPIV1ImportPrometheus(tc.T(), data2.Samples, apptest.QueryOpts{})
+	vmsingle.ForceFlush(t)
+
+	wantSeries12 := slices.Concat(data1.WantSeries, data2.WantSeries)
+	apptest.AssertSeries(tc, vmsingle, "metric.*", "", start, end, wantSeries12)
+
+	vmselect.PrometheusAPIV1AdminTSDBDeleteSeries(tc.T(), `{__name__=~"metric1.*"}`, apptest.QueryOpts{
+		Tenant: tenantID1,
+	})
+	apptest.AssertSeries(tc, vmsingle, "metric.*", "", start, end, data2.WantSeries)
+	vmselect.PrometheusAPIV1AdminTSDBDeleteSeries(tc.T(), `{__name__=~"metric2.*"}`, apptest.QueryOpts{
+		Tenant: tenantID2,
+	})
+	apptest.AssertSeries(tc, vmsingle, "metric.*", "", start, end, data2.WantSeries)
+	vmselect.PrometheusAPIV1AdminTSDBDeleteSeries(tc.T(), `{__name__=~"metric2.*"}`, apptest.QueryOpts{
+		Tenant: "multitenant",
+	})
+	apptest.AssertSeries(tc, vmsingle, "metric.*", "", start, end, emptySeries)
+}
+
+func TestMixedGraphiteQueries(t *testing.T) {
+	tc := apptest.NewTestCase(t)
+	defer tc.Stop()
+
+	const (
+		accountID1 = 12
+		projectID1 = 34
+		accountID2 = 56
+		projectID2 = 78
+		numMetrics = 10
+	)
+	tenantID1 := fmt.Sprintf("%d:%d", accountID1, projectID1)
+	tenantID2 := fmt.Sprintf("%d:%d", accountID2, projectID2)
+	start := time.Date(2026, 1, 1, 0, 0, 0, 0, time.UTC).UnixMilli()
+	end := time.Date(2026, 1, 2, 0, 0, 0, 0, time.UTC).UnixMilli()
+	data := apptest.GenerateGraphiteTestData("metric", numMetrics, start, end)
+	emptyMetricsIndex := []string{}
+	emptyMetricsFind := []apptest.GraphiteMetric{}
+	emptyMetricsExpand := []string{}
+	emptyRenderedTargets := []apptest.GraphiteRenderedTarget{}
+
+	vmsingle := tc.MustStartVmsingle("vmsingle", []string{
+		"-storageDataPath=" + filepath.Join(tc.Dir(), "vmsingle"),
+		"-retentionPeriod=100y",
+		fmt.Sprintf("-accountID=%d", accountID1),
+		fmt.Sprintf("-projectID=%d", projectID1),
+	})
+	vmselect := tc.MustStartVmselect("vmselect", []string{
+		"-storageNode=" + vmsingle.VmselectAddr(),
+	})
+
+	vmsingle.GraphiteWrite(tc.T(), data.Samples, apptest.QueryOpts{})
+	vmsingle.ForceFlush(t)
+
+	// Ensure vmsingle returns data.
+	apptest.AssertGraphiteMetricsIndex(tc, vmsingle, "", data.WantMetricsIndex)
+	apptest.AssertGraphiteMetricsFind(tc, vmsingle, "metric.*", "", data.WantMetricsFind)
+	apptest.AssertGraphiteMetricsExpand(tc, vmsingle, "metric.*", "", data.WantMetricsExpand)
+	apptest.AssertGraphiteRender(tc, vmsingle, "metric.*", "", start, end, data.Step, data.WantRenderedTargets)
+
+	// Ensure vmselect returns data for tenant1.
+	apptest.AssertGraphiteMetricsIndex(tc, vmselect, tenantID1, data.WantMetricsIndex)
+	apptest.AssertGraphiteMetricsFind(tc, vmselect, "metric.*", tenantID1, data.WantMetricsFind)
+	apptest.AssertGraphiteMetricsExpand(tc, vmselect, "metric.*", tenantID1, data.WantMetricsExpand)
+	apptest.AssertGraphiteRender(tc, vmselect, "metric.*", tenantID1, start, end, data.Step, data.WantRenderedTargets)
+
+	// Ensure vmselect does not return any data for tenant2.
+	apptest.AssertGraphiteMetricsIndex(tc, vmselect, tenantID2, emptyMetricsIndex)
+	apptest.AssertGraphiteMetricsFind(tc, vmselect, "metric.*", tenantID2, emptyMetricsFind)
+	apptest.AssertGraphiteMetricsExpand(tc, vmselect, "metric.*", tenantID2, emptyMetricsExpand)
+	apptest.AssertGraphiteRender(tc, vmselect, "metric.*", tenantID2, start, end, data.Step, emptyRenderedTargets)
+}

apptest/vmsingle.go

@@ -25,12 +25,14 @@ func StartVmsingle(instance string, flags []string, cli *Client, output io.Write
 			"-httpListenAddr":     "127.0.0.1:0",
 			"-graphiteListenAddr": "127.0.0.1:0",
 			"-opentsdbListenAddr": "127.0.0.1:0",
+			"-vmselectAddr":       "127.0.0.1:0",
 		},
 		extractREs: []*regexp.Regexp{
 			storageDataPathRE,
 			httpListenAddrRE,
 			graphiteListenAddrRE,
 			openTSDBListenAddrRE,
+			vmselectAddrRE,
 		},
 		output: output,
 	})
@@ -43,6 +45,7 @@ func StartVmsingle(instance string, flags []string, cli *Client, output io.Write
 		httpListenAddr:     stderrExtracts[1],
 		graphiteListenAddr: stderrExtracts[2],
 		openTSDBListenAddr: stderrExtracts[3],
+		vmselectAddr:       stderrExtracts[4],
 	}), nil
 }
 
@@ -51,6 +54,7 @@ type vmsingleRuntimeValues struct {
 	httpListenAddr     string
 	graphiteListenAddr string
 	openTSDBListenAddr string
+	vmselectAddr       string
 }
 
 func newVmsingle(app *app, cli *Client, rt vmsingleRuntimeValues) *Vmsingle {
@@ -85,6 +89,7 @@ func newVmsingle(app *app, cli *Client, rt vmsingleRuntimeValues) *Vmsingle {
 		},
 		storageDataPath: rt.storageDataPath,
 		httpListenAddr:  rt.httpListenAddr,
+		vmselectAddr:    rt.vmselectAddr,
 	}
 }
 
@@ -99,6 +104,7 @@ type Vmsingle struct {
 
 	storageDataPath string
 	httpListenAddr  string
+	vmselectAddr    string
 }
 
 // HTTPAddr returns the address at which the vminsert process is
@@ -107,6 +113,12 @@ func (app *Vmsingle) HTTPAddr() string {
 	return app.httpListenAddr
 }
 
+// VmselectAddr returns the address at which the vmsingle process is listening
+// for vmselect connections.
+func (app *Vmsingle) VmselectAddr() string {
+	return app.vmselectAddr
+}
+
 // String returns the string representation of the vmsingle app state.
 func (app *Vmsingle) String() string {
 	return fmt.Sprintf("{app: %s storageDataPath: %q httpListenAddr: %q}", []any{

dashboards/query-stats.json

@@ -1,4 +1,59 @@
 {
+  "__inputs": [
+    {
+      "name": "DS_VICTORIALOGS",
+      "label": "VictoriaLogs",
+      "description": "",
+      "type": "datasource",
+      "pluginId": "victoriametrics-logs-datasource",
+      "pluginName": "VictoriaLogs"
+    }
+  ],
+  "__elements": {},
+  "__requires": [
+    {
+      "type": "grafana",
+      "id": "grafana",
+      "name": "Grafana",
+      "version": "12.4.3"
+    },
+    {
+      "type": "panel",
+      "id": "logs",
+      "name": "Logs",
+      "version": ""
+    },
+    {
+      "type": "panel",
+      "id": "stat",
+      "name": "Stat",
+      "version": ""
+    },
+    {
+      "type": "panel",
+      "id": "table",
+      "name": "Table",
+      "version": ""
+    },
+    {
+      "type": "panel",
+      "id": "text",
+      "name": "Text",
+      "version": ""
+    },
+    {
+      "type": "panel",
+      "id": "timeseries",
+      "name": "Time series",
+      "version": ""
+    },
+    {
+      "type": "datasource",
+      "id": "victoriametrics-logs-datasource",
+      "name": "VictoriaLogs",
+      "version": "0.29.0"
+    }
+  ],
   "annotations": {
     "list": [
       {
@@ -18,7 +73,6 @@
   "editable": true,
   "fiscalYearStartMonth": 0,
   "graphTooltip": 0,
-  "id": 861,
   "links": [
     {
       "icon": "doc",
@@ -78,7 +132,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               }
             ]
           }
@@ -109,7 +164,7 @@
         "textMode": "auto",
         "wideLayout": true
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -117,7 +172,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| type:=\"instant\"\n| count()",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| type:=\"instant\"\n| count()",
           "queryType": "stats",
           "refId": "A"
         }
@@ -140,7 +195,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               }
             ]
           }
@@ -171,7 +227,7 @@
         "textMode": "auto",
         "wideLayout": true
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -179,7 +235,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash) \n| type:=\"range\"\n| count()",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash \n| type:=\"range\"\n| count()",
           "queryType": "stats",
           "refId": "A"
         }
@@ -202,7 +258,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               }
             ]
           }
@@ -233,7 +290,7 @@
         "textMode": "auto",
         "wideLayout": true
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -241,7 +298,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash) \n| series_fetched:=0\n| count()",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash \n| series_fetched:=0\n| count()",
           "queryType": "stats",
           "refId": "A"
         }
@@ -265,7 +322,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               }
             ]
           },
@@ -297,7 +355,7 @@
         "textMode": "auto",
         "wideLayout": true
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -305,7 +363,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash) \n| stats min(start_ms)\n",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash \n| stats min(start_ms)\n",
           "queryType": "stats",
           "refId": "A"
         }
@@ -327,10 +385,6 @@
       "type": "row"
     },
     {
-      "fieldConfig": {
-        "defaults": {},
-        "overrides": []
-      },
       "gridPos": {
         "h": 2,
         "w": 24,
@@ -347,8 +401,7 @@
         "content": "To filter by specific query copy its hash from the table and put it into `query_hash` filter on the top. To disable filtering enter `*`.",
         "mode": "markdown"
       },
-      "pluginVersion": "11.6.0",
-      "title": "",
+      "pluginVersion": "12.4.3",
       "transparent": true,
       "type": "text"
     },
@@ -367,6 +420,9 @@
             "cellOptions": {
               "type": "auto"
             },
+            "footer": {
+              "reducers": []
+            },
             "inspect": false
           },
           "mappings": [],
@@ -374,7 +430,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -423,18 +480,6 @@
                 "value": 204
               }
             ]
-          },
-          {
-            "matcher": {
-              "id": "byName",
-              "options": "duration_max"
-            },
-            "properties": [
-              {
-                "id": "custom.width",
-                "value": 122
-              }
-            ]
           }
         ]
       },
@@ -447,18 +492,10 @@
       "id": 4,
       "options": {
         "cellHeight": "sm",
-        "footer": {
-          "countRows": false,
-          "fields": "",
-          "reducer": [
-            "sum"
-          ],
-          "show": false
-        },
         "showHeader": true,
         "sortBy": []
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -466,7 +503,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats by(tenant,query,query_hash) max(execution_duration_ms) duration_max \n| sort by(duration_max) desc | limit $top",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats by(tenant,query,query_hash) max(execution_duration_ms) duration_max \n| sort by(duration_max) desc | limit $top",
           "queryType": "instant",
           "refId": "A"
         }
@@ -478,7 +515,7 @@
           "options": {
             "delimiter": ",",
             "replace": true,
-            "source": "Line"
+            "source": "labels"
           }
         },
         {
@@ -533,6 +570,7 @@
               "type": "linear"
             },
             "showPoints": "auto",
+            "showValues": false,
             "spanNulls": false,
             "stacking": {
               "group": "A",
@@ -547,7 +585,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -580,7 +619,7 @@
           "sort": "none"
         }
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -588,7 +627,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats max(execution_duration_ms) execution_duration_max",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats max(execution_duration_ms) execution_duration_max",
           "queryType": "statsRange",
           "refId": "A"
         }
@@ -611,6 +650,9 @@
             "cellOptions": {
               "type": "auto"
             },
+            "footer": {
+              "reducers": []
+            },
             "inspect": false
           },
           "mappings": [],
@@ -618,7 +660,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -680,18 +723,10 @@
       "interval": "1m",
       "options": {
         "cellHeight": "sm",
-        "footer": {
-          "countRows": false,
-          "fields": "",
-          "reducer": [
-            "sum"
-          ],
-          "show": false
-        },
         "showHeader": true,
         "sortBy": []
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -699,7 +734,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats by(tenant,query,query_hash) max(series_fetched) series_fetched_max\n| sort by(series_fetched_max) desc | limit $top",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats by(tenant,query,query_hash) max(series_fetched) series_fetched_max\n| sort by(series_fetched_max) desc | limit $top",
           "queryType": "instant",
           "refId": "A"
         }
@@ -711,7 +746,7 @@
           "options": {
             "delimiter": ",",
             "replace": true,
-            "source": "Line"
+            "source": "labels"
           }
         },
         {
@@ -766,6 +801,7 @@
               "type": "linear"
             },
             "showPoints": "auto",
+            "showValues": false,
             "spanNulls": false,
             "stacking": {
               "group": "A",
@@ -780,7 +816,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -813,7 +850,7 @@
           "sort": "none"
         }
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -821,7 +858,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats max(series_fetched) series_fetched_max",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats max(series_fetched) series_fetched_max",
           "queryType": "statsRange",
           "refId": "A"
         }
@@ -844,6 +881,9 @@
             "cellOptions": {
               "type": "auto"
             },
+            "footer": {
+              "reducers": []
+            },
             "inspect": false
           },
           "mappings": [],
@@ -851,7 +891,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -912,18 +953,10 @@
       "id": 5,
       "options": {
         "cellHeight": "sm",
-        "footer": {
-          "countRows": false,
-          "fields": "",
-          "reducer": [
-            "sum"
-          ],
-          "show": false
-        },
         "showHeader": true,
         "sortBy": []
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -931,7 +964,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats by(tenant,query,query_hash) max(samples_fetched) samples_fetched_max\n| sort by(samples_fetched_max) desc | limit $top",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats by(tenant,query,query_hash) max(samples_fetched) samples_fetched_max\n| sort by(samples_fetched_max) desc | limit $top",
           "queryType": "instant",
           "refId": "A"
         }
@@ -942,8 +975,10 @@
           "id": "extractFields",
           "options": {
             "delimiter": ",",
+            "format": "json",
+            "keepTime": false,
             "replace": true,
-            "source": "Line"
+            "source": "labels"
           }
         },
         {
@@ -998,6 +1033,7 @@
               "type": "linear"
             },
             "showPoints": "auto",
+            "showValues": false,
             "spanNulls": false,
             "stacking": {
               "group": "A",
@@ -1012,7 +1048,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -1045,7 +1082,7 @@
           "sort": "none"
         }
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -1053,7 +1090,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats max(samples_fetched) samples_fetched_max",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats max(samples_fetched) samples_fetched_max",
           "queryType": "statsRange",
           "refId": "A"
         }
@@ -1076,6 +1113,9 @@
             "cellOptions": {
               "type": "auto"
             },
+            "footer": {
+              "reducers": []
+            },
             "inspect": false
           },
           "mappings": [],
@@ -1083,7 +1123,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -1144,18 +1185,10 @@
       "id": 11,
       "options": {
         "cellHeight": "sm",
-        "footer": {
-          "countRows": false,
-          "fields": "",
-          "reducer": [
-            "sum"
-          ],
-          "show": false
-        },
         "showHeader": true,
         "sortBy": []
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -1163,7 +1196,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats by(tenant,query,query_hash) max(bytes) bytes_fetched_max \n| sort by(bytes_fetched_max) desc | limit $top",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats by(tenant,query,query_hash) max(bytes) bytes_fetched_max \n| sort by(bytes_fetched_max) desc | limit $top",
           "queryType": "instant",
           "refId": "A"
         }
@@ -1175,7 +1208,7 @@
           "options": {
             "delimiter": ",",
             "replace": true,
-            "source": "Line"
+            "source": "labels"
           }
         },
         {
@@ -1230,6 +1263,7 @@
               "type": "linear"
             },
             "showPoints": "auto",
+            "showValues": false,
             "spanNulls": false,
             "stacking": {
               "group": "A",
@@ -1244,7 +1278,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -1277,7 +1312,7 @@
           "sort": "none"
         }
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -1285,7 +1320,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats max(bytes) bytes_fetched",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats max(bytes) bytes_fetched",
           "queryType": "statsRange",
           "refId": "A"
         }
@@ -1308,6 +1343,9 @@
             "cellOptions": {
               "type": "auto"
             },
+            "footer": {
+              "reducers": []
+            },
             "inspect": false
           },
           "mappings": [],
@@ -1315,7 +1353,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -1388,18 +1427,10 @@
       "id": 13,
       "options": {
         "cellHeight": "sm",
-        "footer": {
-          "countRows": false,
-          "fields": "",
-          "reducer": [
-            "sum"
-          ],
-          "show": false
-        },
         "showHeader": true,
         "sortBy": []
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -1407,7 +1438,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats by(tenant,query,query_hash) max(memory_estimated_bytes) memory_estimated_max\n| sort by(memory_estimated_max) desc | limit $top",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats by(tenant,query,query_hash) max(memory_estimated_bytes) memory_estimated_max\n| sort by(memory_estimated_max) desc | limit $top",
           "queryType": "instant",
           "refId": "A"
         }
@@ -1419,7 +1450,7 @@
           "options": {
             "delimiter": ",",
             "replace": true,
-            "source": "Line"
+            "source": "labels"
           }
         },
         {
@@ -1474,6 +1505,7 @@
               "type": "linear"
             },
             "showPoints": "auto",
+            "showValues": false,
             "spanNulls": false,
             "stacking": {
               "group": "A",
@@ -1488,7 +1520,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -1521,7 +1554,7 @@
           "sort": "none"
         }
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -1529,7 +1562,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats max(memory_estimated_bytes) memory_estimated_bytes",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats max(memory_estimated_bytes) memory_estimated_bytes",
           "queryType": "statsRange",
           "refId": "A"
         }
@@ -1552,6 +1585,9 @@
             "cellOptions": {
               "type": "auto"
             },
+            "footer": {
+              "reducers": []
+            },
             "inspect": false
           },
           "mappings": [],
@@ -1559,7 +1595,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -1620,18 +1657,10 @@
       "id": 18,
       "options": {
         "cellHeight": "sm",
-        "footer": {
-          "countRows": false,
-          "fields": "",
-          "reducer": [
-            "sum"
-          ],
-          "show": false
-        },
         "showHeader": true,
         "sortBy": []
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -1639,7 +1668,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:in($query_hash)\n| stats by(tenant,query,query_hash) max(range_ms) range_max \n| sort by(range_max) desc | limit $top",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| tenant:in($tenant)\n| query_hash:$query_hash\n| stats by(tenant,query,query_hash) max(range_ms) range_max \n| sort by(range_max) desc | limit $top",
           "queryType": "instant",
           "refId": "A"
         }
@@ -1651,7 +1680,7 @@
           "options": {
             "delimiter": ",",
             "replace": true,
-            "source": "Line"
+            "source": "labels"
           }
         },
         {
@@ -1706,6 +1735,7 @@
               "type": "linear"
             },
             "showPoints": "auto",
+            "showValues": false,
             "spanNulls": false,
             "stacking": {
               "group": "A",
@@ -1720,7 +1750,8 @@
             "mode": "absolute",
             "steps": [
               {
-                "color": "green"
+                "color": "green",
+                "value": 0
               },
               {
                 "color": "red",
@@ -1753,7 +1784,7 @@
           "sort": "none"
         }
       },
-      "pluginVersion": "11.6.0",
+      "pluginVersion": "12.4.3",
       "targets": [
         {
           "datasource": {
@@ -1761,7 +1792,7 @@
             "uid": "${ds}"
           },
           "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| query_hash:in($query_hash)\n| stats max(range_ms) range_max",
+          "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| query_hash:$query_hash\n| stats max(range_ms) range_max",
           "queryType": "statsRange",
           "refId": "A"
         }
@@ -1770,7 +1801,7 @@
       "type": "timeseries"
     },
     {
-      "collapsed": false,
+      "collapsed": true,
       "gridPos": {
         "h": 1,
         "w": 24,
@@ -1778,52 +1809,59 @@
         "y": 56
       },
       "id": 12,
-      "panels": [],
-      "title": "Query log",
-      "type": "row"
-    },
-    {
-      "datasource": {
-        "type": "victoriametrics-logs-datasource",
-        "uid": "${ds}"
-      },
-      "fieldConfig": {
-        "defaults": {},
-        "overrides": []
-      },
-      "gridPos": {
-        "h": 14,
-        "w": 24,
-        "x": 0,
-        "y": 57
-      },
-      "id": 6,
-      "options": {
-        "dedupStrategy": "none",
-        "enableInfiniteScrolling": false,
-        "enableLogDetails": true,
-        "prettifyLogMessage": false,
-        "showCommonLabels": false,
-        "showLabels": false,
-        "showTime": false,
-        "sortOrder": "Descending",
-        "wrapLogMessage": false
-      },
-      "pluginVersion": "11.6.0",
-      "targets": [
+      "panels": [
         {
           "datasource": {
             "type": "victoriametrics-logs-datasource",
             "uid": "${ds}"
           },
-          "editorMode": "code",
-          "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| query_hash:in($query_hash)\n| limit 200",
-          "queryType": "instant",
-          "refId": "A"
+          "fieldConfig": {
+            "defaults": {},
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 14,
+            "w": 24,
+            "x": 0,
+            "y": 57
+          },
+          "id": 6,
+          "options": {
+            "dedupStrategy": "none",
+            "detailsMode": "sidebar",
+            "enableInfiniteScrolling": false,
+            "enableLogDetails": true,
+            "fontSize": "small",
+            "prettifyLogMessage": false,
+            "showCommonLabels": false,
+            "showControls": true,
+            "showLabels": false,
+            "showTime": false,
+            "sortOrder": "Descending",
+            "syntaxHighlighting": false,
+            "unwrappedColumns": false,
+            "wrapLogMessage": true
+          },
+          "pluginVersion": "12.4.3",
+          "targets": [
+            {
+              "datasource": {
+                "type": "victoriametrics-logs-datasource",
+                "uid": "${ds}"
+              },
+              "direction": "desc",
+              "editorMode": "code",
+              "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| query_hash:$query_hash\n| limit 200",
+              "queryType": "instant",
+              "refId": "A"
+            }
+          ],
+          "title": "Raw logs",
+          "type": "logs"
         }
       ],
-      "title": "Raw logs",
-      "type": "logs"
+      "title": "Query log",
+      "type": "row"
     },
     {
       "collapsed": true,
@@ -1831,7 +1869,7 @@
         "h": 1,
         "w": 24,
         "x": 0,
-        "y": 71
+        "y": 57
       },
       "id": 17,
       "panels": [
@@ -1848,7 +1886,7 @@
             "h": 14,
             "w": 24,
             "x": 0,
-            "y": 70
+            "y": 58
           },
           "id": 15,
           "options": {
@@ -1857,12 +1895,14 @@
             "enableLogDetails": true,
             "prettifyLogMessage": false,
             "showCommonLabels": false,
+            "showControls": false,
             "showLabels": false,
             "showTime": false,
             "sortOrder": "Descending",
+            "unwrappedColumns": false,
             "wrapLogMessage": false
           },
-          "pluginVersion": "11.6.0",
+          "pluginVersion": "12.4.3",
           "targets": [
             {
               "datasource": {
@@ -1870,7 +1910,7 @@
                 "uid": "${ds}"
               },
               "editorMode": "code",
-              "expr": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| series_fetched:=0\n| query_hash:in($query_hash)",
+              "expr": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats \n| series_fetched:=0\n| query_hash:$query_hash",
               "queryType": "instant",
               "refId": "A"
             }
@@ -1885,7 +1925,7 @@
   ],
   "preload": false,
   "refresh": "",
-  "schemaVersion": 41,
+  "schemaVersion": 42,
   "tags": [
     "victoriametrics",
     "victorialogs"
@@ -1894,8 +1934,9 @@
     "list": [
       {
         "current": {
-          "text": "VictoriaLogs",
-          "value": "PD775F2863313E6C7"
+          "text": "",
+          "value": "${ds}",
+          "selected": true
         },
         "name": "ds",
         "options": [],
@@ -1933,21 +1974,17 @@
           }
         ],
         "query": "5,10,15,20",
-        "type": "custom"
+        "type": "custom",
+        "valuesFormat": "csv"
       },
       {
         "allValue": "*",
-        "current": {
-          "text": "All",
-          "value": [
-            "$__all"
-          ]
-        },
+        "current": {},
         "datasource": {
           "type": "victoriametrics-logs-datasource",
           "uid": "${ds}"
         },
-        "definition": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats  | fields tenant",
+        "definition": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats  | fields tenant",
         "includeAll": true,
         "multi": true,
         "name": "tenant",
@@ -1955,12 +1992,13 @@
         "query": {
           "field": "tenant",
           "limit": 25,
-          "query": "\"\\tvm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats  | fields tenant",
+          "query": "\"vm_slow_query_stats\" | extract 'vm_slow_query_stats <vm_slow_query_stats>' | unpack_logfmt from vm_slow_query_stats  | fields tenant",
           "refId": "VictoriaLogsVariableQueryEditor-VariableQuery",
           "type": "fieldValue"
         },
-        "refresh": 1,
+        "refresh": 2,
         "regex": "",
+        "regexApplyTo": "value",
         "type": "query"
       },
       {
@@ -2000,5 +2038,6 @@
   "timezone": "browser",
   "title": "Query Stats (cluster)",
   "uid": "feg3od1zt1fy8e",
-  "version": 1
-}
+  "version": 1,
+  "weekStart": ""
+}

dashboards/victoriametrics-cluster.json

@@ -2083,7 +2083,7 @@
             "type": "prometheus",
             "uid": "$ds"
           },
-          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSeу major page faults rate panel in Troubleshooting section if this metric continued to be high.",
+          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSee major page faults rate panel in Troubleshooting section if this metric continued to be high.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -5136,6 +5136,110 @@
           ],
           "title": "Major page faults rate  ($instance)",
           "type": "timeseries"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "$ds"
+          },
+          "description": "Average duration of fsync system calls. Spikes or high latency indicates that storage I/O cannot keep up with the write rate, which may slow down data ingestion.\n\nIf this value is elevated:\n- Check disk-related metrics.\n- If using cloud storage, check IOPS and throughput limits.\n- If using NFS, check network performance, server resource usage, and any errors on the NFS server side.\n- Check for known bugs in the filesystem or kernel version being used.",
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "custom": {
+                "axisBorderShow": false,
+                "axisCenteredZero": false,
+                "axisColorMode": "text",
+                "axisLabel": "",
+                "axisPlacement": "auto",
+                "axisSoftMin": 0,
+                "barAlignment": 0,
+                "barWidthFactor": 0.6,
+                "drawStyle": "line",
+                "fillOpacity": 0,
+                "gradientMode": "none",
+                "hideFrom": {
+                  "legend": false,
+                  "tooltip": false,
+                  "viz": false
+                },
+                "insertNulls": false,
+                "lineInterpolation": "linear",
+                "lineWidth": 1,
+                "pointSize": 5,
+                "scaleDistribution": {
+                  "type": "linear"
+                },
+                "showPoints": "never",
+                "showValues": false,
+                "spanNulls": false,
+                "stacking": {
+                  "group": "A",
+                  "mode": "none"
+                },
+                "thresholdsStyle": {
+                  "mode": "off"
+                }
+              },
+              "links": [],
+              "mappings": [],
+              "thresholds": {
+                "mode": "absolute",
+                "steps": [
+                  {
+                    "color": "green",
+                    "value": 0
+                  }
+                ]
+              },
+              "unit": "s"
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 0,
+            "y": 305
+          },
+          "id": 230,
+          "options": {
+            "legend": {
+              "calcs": [
+                "lastNotNull",
+                "max"
+              ],
+              "displayMode": "table",
+              "placement": "bottom",
+              "showLegend": true,
+              "sortBy": "Last *",
+              "sortDesc": true
+            },
+            "tooltip": {
+              "hideZeros": true,
+              "mode": "multi",
+              "sort": "desc"
+            }
+          },
+          "pluginVersion": "12.2.0",
+          "targets": [
+            {
+              "datasource": {
+                "type": "prometheus",
+                "uid": "$ds"
+              },
+              "editorMode": "code",
+              "expr": "rate(vm_filestream_fsync_duration_seconds_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval]) / rate(vm_filestream_fsync_calls_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])",
+              "format": "time_series",
+              "instant": false,
+              "legendFormat": "{{instance}}",
+              "refId": "A"
+            }
+          ],
+          "title": "Fsync avg duration ($instance)",
+          "type": "timeseries"
         }
       ],
       "title": "Troubleshooting",

dashboards/victoriametrics.json

@@ -896,7 +896,7 @@
             "uid": "$ds"
           },
           "editorMode": "code",
-          "expr": "sum(min_over_time(vm_app_version{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job)",
+          "expr": "sum(min_over_time(up{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job)",
           "format": "time_series",
           "instant": false,
           "legendFormat": "{{job}}",
@@ -2388,7 +2388,7 @@
             "type": "prometheus",
             "uid": "$ds"
           },
-          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSeу major page faults rate panel in Troubleshooting section if this metric continued to be high.",
+          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSee major page faults rate panel in Troubleshooting section if this metric continued to be high.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -5181,6 +5181,110 @@
           ],
           "title": "Major page faults rate",
           "type": "timeseries"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "$ds"
+          },
+          "description": "Average duration of fsync system calls. Spikes or high latency indicates that storage I/O cannot keep up with the write rate, which may slow down data ingestion.\n\nIf this value is elevated:\n- Check disk-related metrics.\n- If using cloud storage, check IOPS and throughput limits.\n- If using NFS, check network performance, server resource usage, and any errors on the NFS server side.\n- Check for known bugs in the filesystem or kernel version being used.",
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "custom": {
+                "axisBorderShow": false,
+                "axisCenteredZero": false,
+                "axisColorMode": "text",
+                "axisLabel": "",
+                "axisPlacement": "auto",
+                "axisSoftMin": 0,
+                "barAlignment": 0,
+                "barWidthFactor": 0.6,
+                "drawStyle": "line",
+                "fillOpacity": 0,
+                "gradientMode": "none",
+                "hideFrom": {
+                  "legend": false,
+                  "tooltip": false,
+                  "viz": false
+                },
+                "insertNulls": false,
+                "lineInterpolation": "linear",
+                "lineWidth": 1,
+                "pointSize": 5,
+                "scaleDistribution": {
+                  "type": "linear"
+                },
+                "showPoints": "never",
+                "showValues": false,
+                "spanNulls": false,
+                "stacking": {
+                  "group": "A",
+                  "mode": "none"
+                },
+                "thresholdsStyle": {
+                  "mode": "off"
+                }
+              },
+              "links": [],
+              "mappings": [],
+              "thresholds": {
+                "mode": "absolute",
+                "steps": [
+                  {
+                    "color": "green",
+                    "value": 0
+                  }
+                ]
+              },
+              "unit": "s"
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 81
+          },
+          "id": 159,
+          "options": {
+            "legend": {
+              "calcs": [
+                "lastNotNull",
+                "max"
+              ],
+              "displayMode": "table",
+              "placement": "bottom",
+              "showLegend": true,
+              "sortBy": "Last *",
+              "sortDesc": true
+            },
+            "tooltip": {
+              "hideZeros": true,
+              "mode": "multi",
+              "sort": "desc"
+            }
+          },
+          "pluginVersion": "12.2.0",
+          "targets": [
+            {
+              "datasource": {
+                "type": "prometheus",
+                "uid": "$ds"
+              },
+              "editorMode": "code",
+              "expr": "rate(vm_filestream_fsync_duration_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]) / rate(vm_filestream_fsync_calls_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])",
+              "format": "time_series",
+              "instant": false,
+              "legendFormat": "{{instance}}",
+              "refId": "A"
+            }
+          ],
+          "title": "Fsync avg duration ($instance)",
+          "type": "timeseries"
         }
       ],
       "title": "Troubleshooting",
@@ -8742,4 +8846,4 @@
   "uid": "wNf0q_kZk",
   "version": 1,
   "weekStart": ""
-}
+}

dashboards/vm/victoriametrics-cluster.json

@@ -2084,7 +2084,7 @@
             "type": "victoriametrics-metrics-datasource",
             "uid": "$ds"
           },
-          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSeу major page faults rate panel in Troubleshooting section if this metric continued to be high.",
+          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSee major page faults rate panel in Troubleshooting section if this metric continued to be high.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -5137,6 +5137,110 @@
           ],
           "title": "Major page faults rate  ($instance)",
           "type": "timeseries"
+        },
+        {
+          "datasource": {
+            "type": "victoriametrics-metrics-datasource",
+            "uid": "$ds"
+          },
+          "description": "Average duration of fsync system calls. Spikes or high latency indicates that storage I/O cannot keep up with the write rate, which may slow down data ingestion.\n\nIf this value is elevated:\n- Check disk-related metrics.\n- If using cloud storage, check IOPS and throughput limits.\n- If using NFS, check network performance, server resource usage, and any errors on the NFS server side.\n- Check for known bugs in the filesystem or kernel version being used.",
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "custom": {
+                "axisBorderShow": false,
+                "axisCenteredZero": false,
+                "axisColorMode": "text",
+                "axisLabel": "",
+                "axisPlacement": "auto",
+                "axisSoftMin": 0,
+                "barAlignment": 0,
+                "barWidthFactor": 0.6,
+                "drawStyle": "line",
+                "fillOpacity": 0,
+                "gradientMode": "none",
+                "hideFrom": {
+                  "legend": false,
+                  "tooltip": false,
+                  "viz": false
+                },
+                "insertNulls": false,
+                "lineInterpolation": "linear",
+                "lineWidth": 1,
+                "pointSize": 5,
+                "scaleDistribution": {
+                  "type": "linear"
+                },
+                "showPoints": "never",
+                "showValues": false,
+                "spanNulls": false,
+                "stacking": {
+                  "group": "A",
+                  "mode": "none"
+                },
+                "thresholdsStyle": {
+                  "mode": "off"
+                }
+              },
+              "links": [],
+              "mappings": [],
+              "thresholds": {
+                "mode": "absolute",
+                "steps": [
+                  {
+                    "color": "green",
+                    "value": 0
+                  }
+                ]
+              },
+              "unit": "s"
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 0,
+            "y": 305
+          },
+          "id": 230,
+          "options": {
+            "legend": {
+              "calcs": [
+                "lastNotNull",
+                "max"
+              ],
+              "displayMode": "table",
+              "placement": "bottom",
+              "showLegend": true,
+              "sortBy": "Last *",
+              "sortDesc": true
+            },
+            "tooltip": {
+              "hideZeros": true,
+              "mode": "multi",
+              "sort": "desc"
+            }
+          },
+          "pluginVersion": "12.2.0",
+          "targets": [
+            {
+              "datasource": {
+                "type": "victoriametrics-metrics-datasource",
+                "uid": "$ds"
+              },
+              "editorMode": "code",
+              "expr": "rate(vm_filestream_fsync_duration_seconds_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval]) / rate(vm_filestream_fsync_calls_total{job=~\"$job_storage\", instance=~\"$instance\"}[$__rate_interval])",
+              "format": "time_series",
+              "instant": false,
+              "legendFormat": "{{instance}}",
+              "refId": "A"
+            }
+          ],
+          "title": "Fsync avg duration ($instance)",
+          "type": "timeseries"
         }
       ],
       "title": "Troubleshooting",

dashboards/vm/victoriametrics.json

@@ -897,7 +897,7 @@
             "uid": "$ds"
           },
           "editorMode": "code",
-          "expr": "sum(min_over_time(vm_app_version{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job)",
+          "expr": "sum(min_over_time(up{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job)",
           "format": "time_series",
           "instant": false,
           "legendFormat": "{{job}}",
@@ -2389,7 +2389,7 @@
             "type": "victoriametrics-metrics-datasource",
             "uid": "$ds"
           },
-          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSeу major page faults rate panel in Troubleshooting section if this metric continued to be high.",
+          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSee major page faults rate panel in Troubleshooting section if this metric continued to be high.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -5182,6 +5182,110 @@
           ],
           "title": "Major page faults rate",
           "type": "timeseries"
+        },
+        {
+          "datasource": {
+            "type": "victoriametrics-metrics-datasource",
+            "uid": "$ds"
+          },
+          "description": "Average duration of fsync system calls. Spikes or high latency indicates that storage I/O cannot keep up with the write rate, which may slow down data ingestion.\n\nIf this value is elevated:\n- Check disk-related metrics.\n- If using cloud storage, check IOPS and throughput limits.\n- If using NFS, check network performance, server resource usage, and any errors on the NFS server side.\n- Check for known bugs in the filesystem or kernel version being used.",
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "custom": {
+                "axisBorderShow": false,
+                "axisCenteredZero": false,
+                "axisColorMode": "text",
+                "axisLabel": "",
+                "axisPlacement": "auto",
+                "axisSoftMin": 0,
+                "barAlignment": 0,
+                "barWidthFactor": 0.6,
+                "drawStyle": "line",
+                "fillOpacity": 0,
+                "gradientMode": "none",
+                "hideFrom": {
+                  "legend": false,
+                  "tooltip": false,
+                  "viz": false
+                },
+                "insertNulls": false,
+                "lineInterpolation": "linear",
+                "lineWidth": 1,
+                "pointSize": 5,
+                "scaleDistribution": {
+                  "type": "linear"
+                },
+                "showPoints": "never",
+                "showValues": false,
+                "spanNulls": false,
+                "stacking": {
+                  "group": "A",
+                  "mode": "none"
+                },
+                "thresholdsStyle": {
+                  "mode": "off"
+                }
+              },
+              "links": [],
+              "mappings": [],
+              "thresholds": {
+                "mode": "absolute",
+                "steps": [
+                  {
+                    "color": "green",
+                    "value": 0
+                  }
+                ]
+              },
+              "unit": "s"
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 81
+          },
+          "id": 159,
+          "options": {
+            "legend": {
+              "calcs": [
+                "lastNotNull",
+                "max"
+              ],
+              "displayMode": "table",
+              "placement": "bottom",
+              "showLegend": true,
+              "sortBy": "Last *",
+              "sortDesc": true
+            },
+            "tooltip": {
+              "hideZeros": true,
+              "mode": "multi",
+              "sort": "desc"
+            }
+          },
+          "pluginVersion": "12.2.0",
+          "targets": [
+            {
+              "datasource": {
+                "type": "victoriametrics-metrics-datasource",
+                "uid": "$ds"
+              },
+              "editorMode": "code",
+              "expr": "rate(vm_filestream_fsync_duration_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]) / rate(vm_filestream_fsync_calls_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])",
+              "format": "time_series",
+              "instant": false,
+              "legendFormat": "{{instance}}",
+              "refId": "A"
+            }
+          ],
+          "title": "Fsync avg duration ($instance)",
+          "type": "timeseries"
         }
       ],
       "title": "Troubleshooting",
@@ -8743,4 +8847,4 @@
   "uid": "wNf0q_kZk_vm",
   "version": 1,
   "weekStart": ""
-}
+}

dashboards/vm/vmagent.json

@@ -892,7 +892,7 @@
             "uid": "$ds"
           },
           "editorMode": "code",
-          "expr": "sum(up{job=~\"$job\", instance=~\"$instance\"}) by (job)",
+          "expr": "sum(min_over_time(up{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job)",
           "format": "time_series",
           "instant": false,
           "interval": "",
@@ -2165,7 +2165,7 @@
             "type": "victoriametrics-metrics-datasource",
             "uid": "$ds"
           },
-          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSeу major page faults rate panel in Troubleshooting section if this metric continued to be high.",
+          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSee major page faults rate panel in Troubleshooting section if this metric continued to be high.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -4581,6 +4581,110 @@
           ],
           "title": "Rows ignored for last 1h ($instance)",
           "type": "timeseries"
+        },
+        {
+          "datasource": {
+            "type": "victoriametrics-metrics-datasource",
+            "uid": "$ds"
+          },
+          "description": "Average duration of fsync system calls. Spikes or high latency indicates that storage I/O cannot keep up with the write rate, which may slow down data ingestion.\n\nIf this value is elevated:\n- Check disk-related metrics.\n- If using cloud storage, check IOPS and throughput limits.\n- If using NFS, check network performance, server resource usage, and any errors on the NFS server side.\n- Check for known bugs in the filesystem or kernel version being used.",
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "custom": {
+                "axisBorderShow": false,
+                "axisCenteredZero": false,
+                "axisColorMode": "text",
+                "axisLabel": "",
+                "axisPlacement": "auto",
+                "axisSoftMin": 0,
+                "barAlignment": 0,
+                "barWidthFactor": 0.6,
+                "drawStyle": "line",
+                "fillOpacity": 0,
+                "gradientMode": "none",
+                "hideFrom": {
+                  "legend": false,
+                  "tooltip": false,
+                  "viz": false
+                },
+                "insertNulls": false,
+                "lineInterpolation": "linear",
+                "lineWidth": 1,
+                "pointSize": 5,
+                "scaleDistribution": {
+                  "type": "linear"
+                },
+                "showPoints": "never",
+                "showValues": false,
+                "spanNulls": false,
+                "stacking": {
+                  "group": "A",
+                  "mode": "none"
+                },
+                "thresholdsStyle": {
+                  "mode": "off"
+                }
+              },
+              "links": [],
+              "mappings": [],
+              "thresholds": {
+                "mode": "absolute",
+                "steps": [
+                  {
+                    "color": "green",
+                    "value": 0
+                  }
+                ]
+              },
+              "unit": "s"
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 189
+          },
+          "id": 171,
+          "options": {
+            "legend": {
+              "calcs": [
+                "lastNotNull",
+                "max"
+              ],
+              "displayMode": "table",
+              "placement": "bottom",
+              "showLegend": true,
+              "sortBy": "Last *",
+              "sortDesc": true
+            },
+            "tooltip": {
+              "hideZeros": true,
+              "mode": "multi",
+              "sort": "desc"
+            }
+          },
+          "pluginVersion": "12.2.0",
+          "targets": [
+            {
+              "datasource": {
+                "type": "victoriametrics-metrics-datasource",
+                "uid": "$ds"
+              },
+              "editorMode": "code",
+              "expr": "rate(vm_filestream_fsync_duration_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]) / rate(vm_filestream_fsync_calls_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])",
+              "format": "time_series",
+              "instant": false,
+              "legendFormat": "{{instance}}",
+              "refId": "A"
+            }
+          ],
+          "title": "Fsync avg duration ($instance)",
+          "type": "timeseries"
         }
       ],
       "title": "Troubleshooting",

dashboards/vm/vmalert.json

@@ -1840,7 +1840,7 @@
             "type": "victoriametrics-metrics-datasource",
             "uid": "$ds"
           },
-          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSeу major page faults rate panel in Troubleshooting section if this metric continued to be high.",
+          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSee major page faults rate panel in Troubleshooting section if this metric continued to be high.",
           "fieldConfig": {
             "defaults": {
               "color": {

dashboards/vmagent.json

@@ -891,7 +891,7 @@
             "uid": "$ds"
           },
           "editorMode": "code",
-          "expr": "sum(up{job=~\"$job\", instance=~\"$instance\"}) by (job)",
+          "expr": "sum(min_over_time(up{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by (job)",
           "format": "time_series",
           "instant": false,
           "interval": "",
@@ -2164,7 +2164,7 @@
             "type": "prometheus",
             "uid": "$ds"
           },
-          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSeу major page faults rate panel in Troubleshooting section if this metric continued to be high.",
+          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSee major page faults rate panel in Troubleshooting section if this metric continued to be high.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -4580,6 +4580,110 @@
           ],
           "title": "Rows ignored for last 1h ($instance)",
           "type": "timeseries"
+        },
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "$ds"
+          },
+          "description": "Average duration of fsync system calls. Spikes or high latency indicates that storage I/O cannot keep up with the write rate, which may slow down data ingestion.\n\nIf this value is elevated:\n- Check disk-related metrics.\n- If using cloud storage, check IOPS and throughput limits.\n- If using NFS, check network performance, server resource usage, and any errors on the NFS server side.\n- Check for known bugs in the filesystem or kernel version being used.",
+          "fieldConfig": {
+            "defaults": {
+              "color": {
+                "mode": "palette-classic"
+              },
+              "custom": {
+                "axisBorderShow": false,
+                "axisCenteredZero": false,
+                "axisColorMode": "text",
+                "axisLabel": "",
+                "axisPlacement": "auto",
+                "axisSoftMin": 0,
+                "barAlignment": 0,
+                "barWidthFactor": 0.6,
+                "drawStyle": "line",
+                "fillOpacity": 0,
+                "gradientMode": "none",
+                "hideFrom": {
+                  "legend": false,
+                  "tooltip": false,
+                  "viz": false
+                },
+                "insertNulls": false,
+                "lineInterpolation": "linear",
+                "lineWidth": 1,
+                "pointSize": 5,
+                "scaleDistribution": {
+                  "type": "linear"
+                },
+                "showPoints": "never",
+                "showValues": false,
+                "spanNulls": false,
+                "stacking": {
+                  "group": "A",
+                  "mode": "none"
+                },
+                "thresholdsStyle": {
+                  "mode": "off"
+                }
+              },
+              "links": [],
+              "mappings": [],
+              "thresholds": {
+                "mode": "absolute",
+                "steps": [
+                  {
+                    "color": "green",
+                    "value": 0
+                  }
+                ]
+              },
+              "unit": "s"
+            },
+            "overrides": []
+          },
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 189
+          },
+          "id": 171,
+          "options": {
+            "legend": {
+              "calcs": [
+                "lastNotNull",
+                "max"
+              ],
+              "displayMode": "table",
+              "placement": "bottom",
+              "showLegend": true,
+              "sortBy": "Last *",
+              "sortDesc": true
+            },
+            "tooltip": {
+              "hideZeros": true,
+              "mode": "multi",
+              "sort": "desc"
+            }
+          },
+          "pluginVersion": "12.2.0",
+          "targets": [
+            {
+              "datasource": {
+                "type": "prometheus",
+                "uid": "$ds"
+              },
+              "editorMode": "code",
+              "expr": "rate(vm_filestream_fsync_duration_seconds_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval]) / rate(vm_filestream_fsync_calls_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])",
+              "format": "time_series",
+              "instant": false,
+              "legendFormat": "{{instance}}",
+              "refId": "A"
+            }
+          ],
+          "title": "Fsync avg duration ($instance)",
+          "type": "timeseries"
         }
       ],
       "title": "Troubleshooting",

dashboards/vmalert.json

@@ -1839,7 +1839,7 @@
             "type": "prometheus",
             "uid": "$ds"
           },
-          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSeу major page faults rate panel in Troubleshooting section if this metric continued to be high.",
+          "description": "Shows memory pressure based on [Pressure Stall Information](https://docs.kernel.org/accounting/psi.html).\n\n**Lower is better.**\n\nPressure is measured as amount of time within 1sec time window the process was:\n- waiting: at least one thread was blocked on memory.\n- stalled: every thread was blocked on memory (severe pressure).\n\nElevated memory pressure can slowdown the process performance by utilizing more disk IO. Consider increasing amount of available RAM limit or decreasing the load on the process.\n\nSee major page faults rate panel in Troubleshooting section if this metric continued to be high.",
           "fieldConfig": {
             "defaults": {
               "color": {

deployment/docker/Makefile

@@ -3,11 +3,11 @@
 DOCKER_REGISTRIES ?= docker.io quay.io
 DOCKER_NAMESPACE ?= victoriametrics
 
-ROOT_IMAGE ?= alpine:3.23.4
+ROOT_IMAGE ?= alpine:3.24.1
 ROOT_IMAGE_SCRATCH ?= scratch
-CERTS_IMAGE := alpine:3.23.4
+CERTS_IMAGE := alpine:3.24.1
 
-GO_BUILDER_IMAGE := golang:1.26.4
+GO_BUILDER_IMAGE := golang:1.26.5
 
 BUILDER_IMAGE := local/builder:2.0.0-$(shell echo $(GO_BUILDER_IMAGE) | tr :/ __)-1
 BASE_IMAGE := local/base:1.1.4-$(shell echo $(ROOT_IMAGE) | tr :/ __)-$(shell echo $(CERTS_IMAGE) | tr :/ __)

deployment/docker/README.md

@@ -163,6 +163,8 @@ The list of alerting rules is the following:
   alerting rules related to [vmauth](https://docs.victoriametrics.com/victoriametrics/vmauth/) component;
 * [alerts-vmanomaly.yml](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/deployment/docker/rules/alerts-vmanomaly.yml):
   alerting rules related to [VictoriaMetrics Anomaly Detection](https://docs.victoriametrics.com/anomaly-detection/);
+* [alerts-vmbackupmanager.yml](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/deployment/docker/rules/alerts-vmbackupmanager.yml):
+  alerting rules related to [vmbackupmanager](https://docs.victoriametrics.com/victoriametrics/vmbackupmanager/) component;
 
 Please, also see [how to monitor VictoriaMetrics installations](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#monitoring).
 ## Troubleshooting

deployment/docker/compose-vm-cluster.yml

@@ -3,7 +3,7 @@ services:
   #  It scrapes targets defined in --promscrape.config
   #  And forward them to --remoteWrite.url
   vmagent:
-    image: victoriametrics/vmagent:v1.145.0
+    image: victoriametrics/vmagent:v1.148.0
     depends_on:
       - "vmauth"
     ports:
@@ -42,14 +42,14 @@ services:
   # vmstorage shards. Each shard receives 1/N of all metrics sent to vminserts,
   # where N is number of vmstorages (2 in this case).
   vmstorage-1:
-    image: victoriametrics/vmstorage:v1.145.0-cluster
+    image: victoriametrics/vmstorage:v1.148.0-cluster
     volumes:
       - strgdata-1:/storage
     command:
       - "--storageDataPath=/storage"
     restart: always
   vmstorage-2:
-    image: victoriametrics/vmstorage:v1.145.0-cluster
+    image: victoriametrics/vmstorage:v1.148.0-cluster
     volumes:
       - strgdata-2:/storage
     command:
@@ -59,7 +59,7 @@ services:
   # vminsert is ingestion frontend. It receives metrics pushed by vmagent,
   # pre-process them and distributes across configured vmstorage shards.
   vminsert-1:
-    image: victoriametrics/vminsert:v1.145.0-cluster
+    image: victoriametrics/vminsert:v1.148.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -68,7 +68,7 @@ services:
       - "--storageNode=vmstorage-2:8400"
     restart: always
   vminsert-2:
-    image: victoriametrics/vminsert:v1.145.0-cluster
+    image: victoriametrics/vminsert:v1.148.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -80,7 +80,7 @@ services:
   # vmselect is a query fronted. It serves read queries in MetricsQL or PromQL.
   # vmselect collects results from configured `--storageNode` shards.
   vmselect-1:
-    image: victoriametrics/vmselect:v1.145.0-cluster
+    image: victoriametrics/vmselect:v1.148.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -90,7 +90,7 @@ services:
       - "--vmalert.proxyURL=http://vmalert:8880"
     restart: always
   vmselect-2:
-    image: victoriametrics/vmselect:v1.145.0-cluster
+    image: victoriametrics/vmselect:v1.148.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -105,7 +105,7 @@ services:
   # read requests from Grafana, vmui, vmalert among vmselects.
   # It can be used as an authentication proxy.
   vmauth:
-    image: victoriametrics/vmauth:v1.145.0
+    image: victoriametrics/vmauth:v1.148.0
     depends_on:
       - "vmselect-1"
       - "vmselect-2"
@@ -119,7 +119,7 @@ services:
 
   # vmalert executes alerting and recording rules
   vmalert:
-    image: victoriametrics/vmalert:v1.145.0
+    image: victoriametrics/vmalert:v1.148.0
     depends_on:
       - "vmauth"
     ports:
@@ -129,6 +129,7 @@ services:
       - ./rules/alerts-health.yml:/etc/alerts/alerts-health.yml
       - ./rules/alerts-vmagent.yml:/etc/alerts/alerts-vmagent.yml
       - ./rules/alerts-vmalert.yml:/etc/alerts/alerts-vmalert.yml
+      - ./rules/alerts-vmbackupmanager.yml:/etc/alerts/alerts-vmbackupmanager.yml
     command:
       - "--datasource.url=http://vmauth:8427/select/0/prometheus"
       - "--datasource.basicAuth.username=foo"

deployment/docker/compose-vm-single.yml

@@ -3,7 +3,7 @@ services:
   #  It scrapes targets defined in --promscrape.config
   #  And forward them to --remoteWrite.url
   vmagent:
-    image: victoriametrics/vmagent:v1.145.0
+    image: victoriametrics/vmagent:v1.148.0
     depends_on:
       - "victoriametrics"
     ports:
@@ -18,7 +18,7 @@ services:
   # VictoriaMetrics instance, a single process responsible for
   # storing metrics and serve read requests.
   victoriametrics:
-    image: victoriametrics/victoria-metrics:v1.145.0
+    image: victoriametrics/victoria-metrics:v1.148.0
     ports:
       - 8428:8428
       - 8089:8089
@@ -59,7 +59,7 @@ services:
 
   # vmalert executes alerting and recording rules
   vmalert:
-    image: victoriametrics/vmalert:v1.145.0
+    image: victoriametrics/vmalert:v1.148.0
     depends_on:
       - "victoriametrics"
       - "alertmanager"
@@ -70,6 +70,7 @@ services:
       - ./rules/alerts-health.yml:/etc/alerts/alerts-health.yml
       - ./rules/alerts-vmagent.yml:/etc/alerts/alerts-vmagent.yml
       - ./rules/alerts-vmalert.yml:/etc/alerts/alerts-vmalert.yml
+      - ./rules/alerts-vmbackupmanager.yml:/etc/alerts/alerts-vmbackupmanager.yml
     command:
       - "--datasource.url=http://victoriametrics:8428/"
       - "--remoteRead.url=http://victoriametrics:8428/"

deployment/docker/rules/alerts-vmagent.yml

@@ -10,7 +10,7 @@ groups:
     concurrency: 2
     rules:
       - alert: PersistentQueueIsDroppingData
-        expr: sum(increase(vm_persistentqueue_bytes_dropped_total[5m])) without (path) > 0
+        expr: sum(increase(vm_persistentqueue_bytes_dropped_total[5m])) without (path,name) > 0
         for: 10m
         labels:
           severity: critical
@@ -201,4 +201,4 @@ groups:
         annotations:
           summary: "Persistent Queue (url {{ $labels.url }}) of {{ $labels.instance }} (job:{{ $labels.job }}) will run out of space in 4 hours."
           description: "RemoteWrite destination ({{ $labels.url }}) is unavailable or unable to receive data in a timely manner, so the persistent queue size is growing. 
-            Once the available space is exhausted, some samples will be discarded and cause incident. Please check the health of remoteWrite destination ({{ $labels.url }})."
+            Once the available space is exhausted, some samples will be discarded and cause incident. Please check the health of remoteWrite destination ({{ $labels.url }})."

deployment/docker/rules/alerts-vmalert.yml

@@ -120,3 +120,39 @@ groups:
           summary: "vmalert instance {{ $labels.instance }} is failing to send notifications to Alertmanager"
           description: "vmalert instance {{ $labels.instance }} is failing to send alert notifications to \"{{ $labels.addr }}\".
             Check vmalert's logs for detailed error message."
+
+      - alert: AlertingRuleResultsApproachingLimit
+        expr: |
+          (
+            vmalert_alerting_rules_last_evaluation_samples
+              > on(group,file) group_left()
+            (vmalert_group_rule_results_limit * 0.9)
+          )
+            and on(group,file)
+          (vmalert_group_rule_results_limit > 0)
+        for: 5m
+        labels:
+          severity: warning
+        annotations:
+          summary: "Alerting rule {{ $labels.alertname }} in group {{ $labels.group }} is approaching the configured results limit"
+          description: "Alerting rule \"{{ $labels.alertname }}\" from group \"{{ $labels.group }}\" in file \"{{ $labels.file }}\" produced {{ $value }} samples in last evaluation, which approaches the configured results limit.
+            If the produced results exceed the limit, the rule will be marked with an error and all its results will be discarded. 
+            Try increasing the results limit for the group or reducing the number of series produced by the rule. See https://docs.victoriametrics.com/victoriametrics/vmalert/#groups."
+
+      - alert: RecordingRuleResultsApproachingLimit
+        expr: |
+          (
+            vmalert_recording_rules_last_evaluation_samples
+              > on(group,file) group_left()
+            (vmalert_group_rule_results_limit * 0.9)
+          )
+            and on(group,file)
+          (vmalert_group_rule_results_limit > 0)
+        for: 5m
+        labels:
+          severity: warning
+        annotations:
+          summary: "Recording rule {{ $labels.recording }} in group {{ $labels.group }} is approaching the configured results limit"
+          description: "Recording rule \"{{ $labels.recording }}\" from group \"{{ $labels.group }}\" in file \"{{ $labels.file }}\" produced {{ $value }} samples in last evaluation, which approaches the configured results limit.
+            If the produced results exceed the limit, the rule will be marked with an error and all its results will be discarded. 
+            Try increasing the results limit for the group or reducing the number of series produced by the rule. See https://docs.victoriametrics.com/victoriametrics/vmalert/#groups."