docs/vmanomaly: v1.30.0

The change applies only to the master and must not be ported to the
cluster version

Follow-up on
https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10926

.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@8aad20d150bbac5944a9f9d289da16a4b0d87c1e  # v4.36.2
         with:
           languages: go
 
       - name: Autobuild
-        uses: github/codeql-action/autobuild@e46ed2cbd01164d986452f91f178727624ae40d7  # v4.35.3
+        uses: github/codeql-action/autobuild@8aad20d150bbac5944a9f9d289da16a4b0d87c1e  # v4.36.2
 
       - name: Perform CodeQL Analysis
-        uses: github/codeql-action/analyze@e46ed2cbd01164d986452f91f178727624ae40d7  # v4.35.3
+        uses: github/codeql-action/analyze@8aad20d150bbac5944a9f9d289da16a4b0d87c1e  # v4.36.2
         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/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

@@ -571,6 +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.
@@ -1056,7 +1064,7 @@ func (rwctx *remoteWriteCtx) TryPushTimeSeries(tss []prompb.TimeSeries, forceDro
 
 	copyTimeSeriesIfNeeded := func() {
 		if v == nil {
-			v := tssPool.Get().(*[]prompb.TimeSeries)
+			v = tssPool.Get().(*[]prompb.TimeSeries)
 			tss = append(*v, tss...)
 		}
 	}

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/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/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/main.go

@@ -175,11 +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
 	}
@@ -192,23 +193,24 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 		if tkn == nil {
 			logger.Panicf("BUG: unexpected nil jwt token for user %q", ui.name())
 		}
-		if !tkn.HasVMAccessClaim() && ui.JWT.DefaultVMAccessClaim == nil {
-			http.Error(w, "Unauthorized", http.StatusUnauthorized)
+		defer putToken(tkn)
+		// 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
 		}
-		defer putToken(tkn)
-		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{
@@ -384,7 +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,
 			}
 		}
 

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

@@ -785,7 +785,26 @@ statusCode=401
 Unauthorized`
 	f(simpleCfgStr, request, responseExpected)
 
-	// token without vm_access claim is accepted when it
+	// 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 = `

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/promql/binary_op.go

@@ -172,7 +172,13 @@ func newBinaryOpFunc(bf func(left, right float64, isBool bool) float64) binaryOp
 			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 +189,8 @@ 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
 		for i, tsLeft := range left {
 			leftValues := tsLeft.Values
 			rightValues := right[i].Values
@@ -193,6 +201,19 @@ 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)
+				// apply the fill value when either the left or right side is NaN, but not both.
+				if leftIsNaN && rightIsNaN {
+					dstValues[j] = bf(a, b, isBool)
+					continue
+				}
+				if leftIsNaN && fillLeft != nil {
+					a = fillLeft.N
+				}
+				if rightIsNaN && fillRight != nil {
+					b = fillRight.N
+				}
 				dstValues[j] = bf(a, b, isBool)
 			}
 		}
@@ -226,7 +247,7 @@ 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 +260,27 @@ 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 +325,28 @@ 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

@@ -424,18 +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)
 	}
@@ -451,6 +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":
@@ -459,6 +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
 }
 
@@ -470,42 +486,63 @@ 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) {
-	canPushdown := canPushdownCommonFilters(be)
-	if !canPushdown && !shouldOptimizeRepeatedBinaryOpSubexprs(ec, exprFirst, exprSecond) {
-		// 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 tssFirst, tssSecond, nil
+		return tssSecond, tssFirst, nil
 	}
-	if !canPushdown {
+	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()
 
@@ -525,47 +562,37 @@ func execBinaryOpArgs(qt *querytracer.Tracer, ec *EvalConfig, exprFirst, exprSec
 		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 tssFirst, tssSecond, nil
+	return sortResult(tssFirst, tssSecond)
 }
 
 func shouldOptimizeRepeatedBinaryOpSubexprs(ec *EvalConfig, exprFirst, exprSecond metricsql.Expr) bool {
@@ -1687,10 +1714,6 @@ func assertInstantValues(tss []*timeseries) {
 
 var memoryIntensiveQueries = metrics.NewCounter(`vm_memory_intensive_queries_total`)
 
-var _ = metrics.NewGauge(`vm_max_memory_per_query`, func() float64 {
-	return float64(maxMemoryPerQuery.N)
-})
-
 func evalRollupFuncWithMetricExpr(qt *querytracer.Tracer, ec *EvalConfig, funcName string, rf rollupFunc,
 	expr metricsql.Expr, me *metricsql.MetricExpr, iafc *incrementalAggrFuncContext, windowExpr *metricsql.DurationExpr,
 ) ([]*timeseries, error) {

app/vmselect/promql/exec_test.go

@@ -4006,6 +4006,275 @@ 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 +5102,137 @@ 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 +6621,6 @@ 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

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/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/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" },

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.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}}",

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}}",

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": "",

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": "",

deployment/docker/Makefile

@@ -7,7 +7,7 @@ ROOT_IMAGE ?= alpine:3.24.1
 ROOT_IMAGE_SCRATCH ?= scratch
 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/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:v1.146.0
+    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:v1.146.0-cluster
+    image: victoriametrics/vmstorage:v1.148.0-cluster
     volumes:
       - strgdata-1:/storage
     command:
       - "--storageDataPath=/storage"
     restart: always
   vmstorage-2:
-    image: victoriametrics/vmstorage:v1.146.0-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:v1.146.0-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:v1.146.0-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:v1.146.0-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:v1.146.0-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:v1.146.0
+    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:v1.146.0
+    image: victoriametrics/vmalert:v1.148.0
     depends_on:
       - "vmauth"
     ports:

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:v1.146.0
+    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:v1.146.0
+    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:v1.146.0
+    image: victoriametrics/vmalert:v1.148.0
     depends_on:
       - "victoriametrics"
       - "alertmanager"

deployment/docker/rules/alerts-cluster.yml

@@ -223,16 +223,4 @@ groups:
             Unexpected TSID misses for \"{{ $labels.job }}\" ({{ $labels.instance }}) for the last 15 minutes.
             If this happens after unclean shutdown of VictoriaMetrics process (via \"kill -9\", OOM or power off),
             then this is OK - the alert must go away in a few minutes after the restart.
-            Otherwise this may point to the corruption of index data.
-
-      - alert: VMSelectConcurrentQueriesExceedMemoryLimit
-        expr: (vm_max_memory_per_query * on(job, instance) vm_concurrent_select_capacity) > on(job, instance) vm_available_memory_bytes
-        for: 5m
-        labels:
-          severity: warning
-        annotations:
-          summary: "vmselect ({{ $labels.instance }}) concurrent query memory may exceed pod limit"
-          description: "Current concurrent queries ({{ $value | humanize1024 }} combined max memory) exceed
-            the available memory on instance {{ $labels.instance }}.
-            This may result in OOM kills. Consider reducing -maxConcurrentRequests,
-            lowering -maxMemoryPerQuery, or scaling up pod memory limits."
+            Otherwise this may point to the corruption of index data.

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."

deployment/docker/rules/alerts-vmanomaly.yml

@@ -30,6 +30,35 @@ groups:
           summary: "Service {{ $labels.job }} is down on {{ $labels.instance }}"
           description: "{{ $labels.instance }} of job {{ $labels.job }} has been down for more than 5m"
 
+      - alert: SchedulerWorkerDown
+        expr: >
+          (vmanomaly_scheduler_alive{job=~".*vmanomaly.*"} == 0)
+          and on (job, instance, scheduler_alias, preset)
+          (sum(vmanomaly_scheduler_restarts_total{job=~".*vmanomaly.*"})
+          by (job, instance, scheduler_alias, preset) >= 0)
+        for: 2m
+        labels:
+          severity: critical
+        annotations:
+          summary: "Scheduler {{ $labels.scheduler_alias }} is down in {{ $labels.job }} ({{ $labels.instance }})"
+          description: |
+            Periodic scheduler {{ $labels.scheduler_alias }} from preset {{ $labels.preset }} has remained down for more than 2 minutes
+            after entering automatic recovery. The restart policy may be exhausted.
+            Check vmanomaly logs and the vmanomaly_scheduler_restarts_total metric for restart failures.
+
+      - alert: FrequentSchedulerRestarts
+        expr: >
+          sum(increase(vmanomaly_scheduler_restarts_total{job=~".*vmanomaly.*"}[15m]))
+          by (job, instance, scheduler_alias, preset) > 2
+        labels:
+          severity: warning
+        annotations:
+          summary: "Scheduler {{ $labels.scheduler_alias }} is restarting frequently in {{ $labels.job }} ({{ $labels.instance }})"
+          description: |
+            Scheduler {{ $labels.scheduler_alias }} from preset {{ $labels.preset }} had more than two restart attempts
+            during the last 15 minutes. The service may be self-healing, but repeated restarts indicate an unstable
+            scheduler, configuration, datasource, or runtime dependency. Check vmanomaly logs for the original failure.
+
       # default value of 900 Should be changed to the scrape_interval for pull metrics. For push metrics this should be the lowest fit_every or infer_every in your vmanomaly config.
       - alert: NoSelfMonitoringMetrics
         expr: >

deployment/docker/rules/alerts-vmauth.yml

@@ -56,3 +56,20 @@ groups:
           summary: "Too many errors served for user {{ $labels.username }} (instance {{ $labels.instance }})"
           description: "Requests from user {{ $labels.username }} are receiving errors.
             Please check the vmauth logs to verify that the configuration is correct and clients are sending valid requests."
+      - alert: InvalidAuthTokenRequestErrors
+        expr: sum(increase(vmauth_http_request_errors_total{reason="invalid_auth_token"}[5m])) without (instance, reason) > 0
+        for: 15m
+        labels:
+          severity: warning
+        annotations:
+          dashboard: "{{ $externalURL }}/d/nbuo5Mr4k?viewPanel=16&var-job={{ $labels.job }}"
+          summary: "vmauth {{ $labels.job }} is receiving many requests with invalid auth tokens"
+          description: |
+            vmauth {{ $labels.job }} received {{ $value }} requests with invalid auth tokens in the last 5 minutes.
+            This may indicate:
+              - credentials have been updated on vmauth but not on clients
+              - client misconfiguration or use of an expired token
+              - a brute-force attack.
+
+            Check vmauth metrics for longevity and scale of the issue. 
+            Check access log for detailed information: https://docs.victoriametrics.com/victoriametrics/vmauth/#access-log

deployment/docker/vmanomaly/vmanomaly-integration/compose.yml

@@ -1,6 +1,6 @@
 services:
   vmagent:
-    image: victoriametrics/vmagent:v1.146.0
+    image: victoriametrics/vmagent:v1.148.0
     depends_on:
       - "victoriametrics"
     ports:
@@ -14,7 +14,7 @@ services:
     restart: always
 
   victoriametrics:
-    image: victoriametrics/victoria-metrics:v1.146.0
+    image: victoriametrics/victoria-metrics:v1.148.0
     ports:
       - 8428:8428
     volumes:
@@ -40,7 +40,7 @@ services:
     restart: always
 
   vmalert:
-    image: victoriametrics/vmalert:v1.146.0
+    image: victoriametrics/vmalert:v1.148.0
     depends_on:
       - "victoriametrics"
     ports:
@@ -59,7 +59,7 @@ services:
       - '--external.alert.source=explore?orgId=1&left=["now-1h","now","VictoriaMetrics",{"expr": },{"mode":"Metrics"},{"ui":[true,true,true,"none"]}]'
     restart: always
   vmanomaly:
-    image: victoriametrics/vmanomaly:v1.29.7
+    image: victoriametrics/vmanomaly:v1.30.0
     depends_on:
       - "victoriametrics"
     ports:

deployment/docker/vmanomaly/vmanomaly-integration/vmanomaly_config.yml

@@ -1,16 +1,18 @@
 schedulers:
   periodic:
     infer_every: "1m"
-    fit_every: "1h"
-    fit_window: "2d" # 2d-14d based on the presence of weekly seasonality in your data
+    fit_every: "100w" # the online model keeps learning during inference
+    fit_window: "2w"
 
 models:
-  prophet:
-    class: "prophet"
-    args:
-      interval_width: 0.98
-      weekly_seasonality: False  # comment it if your data has weekly seasonality
-      yearly_seasonality: False
+  temporal_envelope:
+    class: "temporal_envelope"
+    queries: "all"
+    schedulers: "all"
+    seasonalities: ["hod_smooth", "dow_smooth"]
+    alpha: 0.005
+    loss_reactivity: 5
+    iqr_threshold: 2
 
 reader:
   datasource_url: "http://victoriametrics:8428/"
@@ -26,4 +28,4 @@ writer:
 monitoring:
   pull: # Enable /metrics endpoint.
     addr: "0.0.0.0"
-    port: 8490
+    port: 8490

docs/ai-tools/README.md

@@ -69,18 +69,21 @@ See more details at [VictoriaMetrics/mcp-victoriatraces](https://github.com/Vict
 REST API and documentation for AI-assisted anomaly detection, model management, and observability insights.
 
 Capabilities include:
-- Health Monitoring: Check `vmanomaly` server health and build information
-- Model Management: List, validate, and configure anomaly detection models (like `zscore_online`, `prophet`, and more)
-- Configuration Generation: Generate complete `vmanomaly` YAML configurations
-- Alert Rule Generation: Generate [`vmalert`](https://docs.victoriametrics.com/victoriametrics/vmalert/) [alerting rules](https://docs.victoriametrics.com/victoriametrics/vmalert/#alerting-rules) based on [anomaly score metrics](https://docs.victoriametrics.com/anomaly-detection/faq/#what-is-anomaly-score) to simplify alerting setup
-- Documentation Search: Full-text search across embedded `vmanomaly` documentation with fuzzy matching
+- Check `vmanomaly` health, build information, compatibility, and self-monitoring metrics
+- Inspect model schemas and validate model or complete service configurations
+- Profile sampled query results for trends, seasonalities, changepoints, gaps, and intermittent behavior
+- Run asynchronous autotune tasks and turn their results into data-driven model recommendations
+- Generate complete `vmanomaly` YAML configurations and [`vmalert`](https://docs.victoriametrics.com/victoriametrics/vmalert/) [alerting rules](https://docs.victoriametrics.com/victoriametrics/vmalert/#alerting-rules)
+- Search embedded `vmanomaly` documentation with fuzzy matching
 
 See more details at [VictoriaMetrics/mcp-vmanomaly](https://github.com/VictoriaMetrics/mcp-vmanomaly).
+# UI Copilot
+vmanomaly UI includes an [AI Copilot](https://docs.victoriametrics.com/anomaly-detection/ui/#ai-assistance) that can assist users with anomaly detection tasks, model configuration, and troubleshooting, changing the UI state based on user queries and providing actionable suggestions through automated data profiling and validation. The AI Copilot is powered by respective [MCP Server](#vmanomaly-mcp-server) and [Agent Skills](#agent-skills), enabling it to understand the context of the user's actions and provide relevant guidance.
 
 # Agent Skills
 
-[Agent skills](https://github.com/VictoriaMetrics/skills) help AI agents and automation tools understand, operate, 
-and troubleshoot VictoriaMetrics observability components, including metrics, logs, and traces.
+[Agent skills](https://github.com/VictoriaMetrics/skills) help AI agents and automation tools understand, operate,
+and troubleshoot VictoriaMetrics observability components, including metrics, logs, traces, and [`vmanomaly`](https://docs.victoriametrics.com/anomaly detection/).
 
 These skills provide predefined workflows and capabilities such as:
 * Query metrics, logs, traces and alerts
@@ -89,6 +92,9 @@ These skills provide predefined workflows and capabilities such as:
 * Cardinality optimization 
 * Unused metric detection
 * Stream aggregation configuration
+* Build validated `vmanomaly` configurations from measured time-series characteristics (e.g., seasonality, changepoints, trends)
+* Query and operate the `vmanomaly` API
+* Review existing anomaly detection configurations against real data and identify false-positive or model-data fit issues
 
 To install the available skills for AI agents, run:
 ```sh

docs/ai-tools/_index.md

@@ -9,6 +9,7 @@ tags:
   - metrics
   - logs
   - traces
+  - anomaly detection
   - AI
   - AI integration
   - agent

docs/anomaly-detection/CHANGELOG.md

@@ -14,6 +14,29 @@ aliases:
 ---
 Please find the changelog for VictoriaMetrics Anomaly Detection below.
 
+## v1.30.0
+Released: 23-07-2026
+
+- FEATURE: Added univariate and multivariate online [Temporal Envelope](https://docs.victoriametrics.com/anomaly-detection/components/models/#temporal-envelope) models for complex time-series profiles. They combine robust causal trend, compact calendar and learned holiday behavior, changepoint adaptation, residual prediction intervals, future-horizon forecasts, and optional cross-series dependency detection in bounded state.
+
+- FEATURE: Added asynchronous [`/api/v1/autotune/tasks`](https://docs.victoriametrics.com/anomaly-detection/components/models/#shared-asynchronous-autotune-workflow) endpoints and improved unsupervised tuning with model-capability-aware objectives, optional causal exact validation for online models, constrained alert-volume selection, fitted-state complexity tie-breaking, and frozen model-specific parameters.
+
+- FEATURE: Added `/api/v1/timeseries/characteristics` for bounded analysis of trend, calendar seasonality, changepoints, gaps, and intermittent or spiky behavior across sampled query results. This improves automated model selection and search-space suggestions in both agentic workflows and [AI Copilot](https://docs.victoriametrics.com/anomaly-detection/ui/#ai-assistance)-backed [UI](https://docs.victoriametrics.com/anomaly-detection/ui/) experiments.
+
+- UI: Updated [vmanomaly UI](https://docs.victoriametrics.com/anomaly-detection/ui/) from [v1.7.2](https://docs.victoriametrics.com/anomaly-detection/ui/#v172) to [v1.8.0](https://docs.victoriametrics.com/anomaly-detection/ui/#v180). Notable mentions are model and query configuration, query prettification, fullscreen charts, out-of-date results alert and improved [AI Copilot](https://docs.victoriametrics.com/anomaly-detection/ui/#ai-assistance) suggestion and cancellation behavior.
+
+- FEATURE: Made AI-assisted configuration more reliable with reusable [vmanomaly workflow skills](https://github.com/VictoriaMetrics/skills), data-aware model suggestions, protocol-safe cancellation and synchronized query, model, and anomaly-setting updates in [UI](https://docs.victoriametrics.com/anomaly-detection/ui/).
+
+- IMPROVEMENT: Fitted history can be retained as a stronger prior through `history_strength` argument, requiring a reduced number of observations (say 2 weeks for weekly seasonality) to be effective if the history window is a good indicator of "normal" data patterns. Models supported: `mad_online`, `zscore_online`, `quantile_online`.
+
+- IMPROVEMENT: Improved bounded datasource operation with optional ad-hoc series limits, stale-series lookback caps, and propagation of configured per-query sampling periods to models and autotune trials.
+
+- IMPROVEMENT: Added `reader.fetch_timeout` and `reader.processing_timeout` for [`VmReader`](https://docs.victoriametrics.com/anomaly-detection/components/reader/#vm-reader) and [`VLogsReader`](https://docs.victoriametrics.com/anomaly-detection/components/reader/#victorialogs-reader), allowing datasource reads and post-fetch processing to be tuned *independently* while preserving `reader.timeout` as the backward-compatible default for both phases.
+
+- BUGFIX: Fixed [backtesting runs](https://docs.victoriametrics.com/anomaly-detection/components/scheduler/#backtesting-scheduler) across multiple fit cycles and for auto-tuned online wrappers, preventing duplicate or missing predictions while retaining causal model updates and restoration of compatible legacy auto-tuned state.
+
+- BUGFIX: Prevented service shutdown after a [`PeriodicScheduler`](https://docs.victoriametrics.com/anomaly-detection/components/scheduler/#periodic-scheduler) worker dies by adding bounded restart attempts with backoff and exposing restart health through [startup metrics](https://docs.victoriametrics.com/anomaly-detection/components/monitoring/#startup-metrics).
+
 ## v1.29.7
 Released: 2026-06-25
 
@@ -283,7 +306,7 @@ Released: 2025-06-13
 ## v1.23.2
 Released: 2025-06-09
 
-- IMPROVEMENT: Increased convergence speed for [OnlineZScoreModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-z-score), [ZScoreModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#z-score), [MADModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#mad), and [OnlineMADModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-mad) models. Now it works better for tight optimization budgets (n_trials < 10, timeout < 1s)
+- IMPROVEMENT: Increased convergence speed for [OnlineZScoreModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-z-score), [ZScoreModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#z-score), [MADModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#mad-median-absolute-deviation), and [OnlineMADModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-mad) models. Now it works better for tight optimization budgets (n_trials < 10, timeout < 1s)
 
 - BUGFIX: Now mean and variance of [OnlineZScoreModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-z-score) with exponential `decay` < 1 [arg](https://docs.victoriametrics.com/anomaly-detection/components/models/#decay) are properly calculated for unbiased predictions.
 
@@ -538,7 +561,7 @@ Released: 2024-10-01
 
 > A bug was discovered in this release that causes the service to crash. Please use the patch [v1.16.1](#v1161) to resolve this issue.
 
-- FEATURE: Introduced data dumps to a host filesystem for [VmReader](https://docs.victoriametrics.com/anomaly-detection/#vm-reader).  Resource-intensive setups (multiple queries returning many metrics, bigger `fit_window` arg) will have RAM consumption reduced during fit calls.
+- FEATURE: Introduced data dumps to a host filesystem for [VmReader](https://docs.victoriametrics.com/anomaly-detection/components/reader/#vm-reader).  Resource-intensive setups (multiple queries returning many metrics, bigger `fit_window` arg) will have RAM consumption reduced during fit calls.
 - IMPROVEMENT: Added a `groupby` argument for logical grouping in [multivariate models](https://docs.victoriametrics.com/anomaly-detection/components/models/#multivariate-models). When specified, a separate multivariate model is trained for each unique combination of label values in the `groupby` columns. For example, to perform multivariate anomaly detection on metrics at the machine level without cross-entity interference, you can use `groupby: [host]` or `groupby: [instance]`, ensuring one model per entity being trained (e.g., per host). Please find more details [here](https://docs.victoriametrics.com/anomaly-detection/components/models/#group-by).
 - IMPROVEMENT: Improved performance of [VmReader](https://docs.victoriametrics.com/anomaly-detection/components/reader/#vm-reader) on multicore instances for reading and data processing.
 - IMPROVEMENT: Introduced new CLI argument aliases to enhance compatibility with [Helm charts](https://github.com/VictoriaMetrics/helm-charts/blob/master/charts/victoria-metrics-anomaly/README.md) (i.e. using secrets) and better align with [VictoriaMetrics flags](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#list-of-command-line-flags):
@@ -685,7 +708,7 @@ Released: 2024-02-15
 
 ## v1.9.2
 Released: 2024-01-29
-- BUGFIX: now multivariate models (like [`IsolationForestMultivariateModel`](https://docs.victoriametrics.com/anomaly-detection/components/models/#isolation-foresthttpsenwikipediaorgwikiisolation_forest-multivariate)) are properly handled throughout fit/infer phases.
+- BUGFIX: now multivariate models (like [`IsolationForestMultivariateModel`](https://docs.victoriametrics.com/anomaly-detection/components/models/#isolation-forest-multivariate)) are properly handled throughout fit/infer phases.
 
 
 ## v1.9.1

docs/anomaly-detection/FAQ.md

@@ -143,9 +143,15 @@ For information on migrating between different versions of `vmanomaly`, please r
 
 > {{% available_from "v1.28.3" anomaly %}} Try our [MCP Server](https://github.com/VictoriaMetrics/mcp-vmanomaly) to get AI-assisted recommendations on selecting the best model and its configuration for your use case. See [installation guide](https://github.com/VictoriaMetrics/mcp-vmanomaly#installation) for more details.
 
-Selecting the best model for `vmanomaly` depends on the data's nature and the [types of anomalies](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-2/#categories-of-anomalies) to detect. For instance, [Z-score](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-z-score) is suitable for data without trends or seasonality, while more complex patterns might require models like [Prophet](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet).
+Selecting the best model for `vmanomaly` depends on the data's nature and the [types of anomalies](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-2/#categories-of-anomalies) to detect:
 
-Also, there is an option to auto-tune the most important (hyper)parameters of selected model class {{% available_from "v1.12.0" anomaly %}}, find [the details here](https://docs.victoriametrics.com/anomaly-detection/components/models/#autotuned).
+- Use [Online MAD](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-mad) for simple, mostly stationary data with no-to-slow trend, when robustness to outliers is important.
+- Use [Online Z-score](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-z-score) for simple, light-tailed data where standard-deviation units are meaningful.
+- Use [Temporal Envelope](https://docs.victoriametrics.com/anomaly-detection/components/models/#temporal-envelope) {{% available_from "v1.30.0" anomaly %}} for complex data with trends, calendar patterns, holidays, or persistent shifts. It is the preferred *online* alternative to Prophet (which will be deprecated in the future releases).
+- Use multivariate Temporal Envelope when normal relationships between aligned metrics matter. This should replace [Isolation Forest](https://docs.victoriametrics.com/anomaly-detection/components/models/#isolation-forest-multivariate) used in previous versions of `vmanomaly`, which will be deprecated in future releases.
+- Use [Prophet](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet) when Prophet-specific decomposition outputs, or offline batch behavior are required. Consider using Temporal Envelope instead, as it is more efficient and provides better results in most cases.
+
+There is also an option to auto-tune the most important parameters of a selected model class {{% available_from "v1.12.0" anomaly %}}. {{% available_from "v1.30.0" anomaly %}} The asynchronous autotune API can first profile a bounded sample through `/api/v1/timeseries/characteristics`, then tune a shared concrete configuration through `/api/v1/autotune/tasks`. See the [autotune workflow](https://docs.victoriametrics.com/anomaly-detection/components/models/#shared-asynchronous-autotune-workflow).
 
 Please refer to [respective blogpost on anomaly types and alerting heuristics](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-2/) for more details.
 
@@ -298,7 +304,7 @@ Configuration above will produce N intervals of full length (`fit_window`=14d +
 
 ## Forecasting
 
-`vmanomaly` can generate future forecasts (e.g. using [ProphetModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet) {{% available_from "v1.25.3" anomaly %}}), which is helpful for capacity planning, resource allocation, or trend analysis when the underlying data is complex and exceeds what inline MetricsQL queries, including [predict_linear](https://docs.victoriametrics.com/victoriametrics/metricsql/#predict_linear), can handle.
+`vmanomaly` can generate future forecasts using [Temporal Envelope](https://docs.victoriametrics.com/anomaly-detection/components/models/#temporal-envelope) {{% available_from "v1.30.0" anomaly %}} or [ProphetModel](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet) {{% available_from "v1.25.3" anomaly %}}. This is helpful for capacity planning, resource allocation, or trend analysis when the underlying data is complex and exceeds what inline MetricsQL queries, including [predict_linear](https://docs.victoriametrics.com/victoriametrics/metricsql/#predict_linear), can handle.
 
 > However, please note that this mode should be used with care, as the model will produce `yhat_{h}` (and probably `yhat_lower_{h}`, and `yhat_upper_{h}`) time series **for each timeseries returned by input queries and for each forecasting horizon specified in `forecast_at` argument, which can lead to a significant increase in the number of active timeseries in VictoriaMetrics TSDB**.
 
@@ -309,12 +315,12 @@ Here's an example of how to produce forecasts using `vmanomaly` and combine it w
 schedulers:
   periodic_5m:  # this scheduler will be used to produce anomaly scores each 5 minutes using "regular" simple model
     class: 'periodic'
-    fit_every: '30d'
+    fit_every: '100w'
     fit_window: '3d'
     infer_every: '5m'
   periodic_forecast:  # this scheduler will be used to produce forecasts each 24h using "daily" model
     class: 'periodic'
-    fit_every: '7d'
+    fit_every: '1000w'
     fit_window: '730d' # to fit the model on 2 years of data to account for seasonality and holidays
     infer_every: '24h'
 # https://docs.victoriametrics.com/anomaly-detection/components/reader/#vm-reader
@@ -356,15 +362,17 @@ models:
     quantiles: [0.25, 0.5, 0.75]  # to produce median and upper quartiles
     iqr_threshold: 2.0
 
-  prophet_1d:
-    class: 'prophet'
+  envelope_1d:
+    class: 'temporal_envelope'
     queries: ['disk_usage_perc_1d']
     schedulers: ['periodic_forecast']
     clip_predictions: True
     detection_direction: 'above_expected'  # as we are interested in spikes in capacity planning
     forecast_at: ['3d', '7d']  # this will produce forecasts for 3 and 7 days ahead
     provide_series: ['yhat', 'yhat_upper']  # to write forecasts back to VictoriaMetrics, omitting `yhat_lower` as it is not needed in this example
-    # other model params, yearly_seasonality may stay
+    # other model params
+    seasonalities:
+
 
     # https://facebook.github.io/prophet/docs/quick_start#python-api
     args:
@@ -423,7 +431,7 @@ services:
   # ...
   vmanomaly:
     container_name: vmanomaly
-    image: victoriametrics/vmanomaly:v1.29.7
+    image: victoriametrics/vmanomaly:v1.30.0
     # ...
     restart: always
     volumes:
@@ -460,6 +468,8 @@ With the introduction of [online models](https://docs.victoriametrics.com/anomal
 - **Optimized resource utilization**: By spreading the computational load over time and reducing peak demands, online models make more efficient use of resources and inducing less data transfer from VictoriaMetrics TSDB, improving overall system performance.
 - **Faster convergence**: Online models can adapt {{% available_from "v1.23.0" anomaly %}} to changes in data patterns more quickly, which is particularly beneficial in dynamic environments where data characteristics may shift frequently. See `decay` argument description [here](https://docs.victoriametrics.com/anomaly-detection/components/models/#decay).
 
+{{% available_from "v1.30.0" anomaly %}} Online models expose a **common** `min_n_samples_seen` warmup: they continue learning, but emit `anomaly_score: 0` until enough observations have been seen. Online MAD, Z-score, and Seasonal Quantile also support `history_strength`, which retains fitted history as a stronger prior while inference continues to adapt the state.
+
 > {{% available_from "v1.24.0" anomaly %}} Online models are best used in conjunction with [stateful mode](https://docs.victoriametrics.com/anomaly-detection/components/settings/#state-restoration) to preserve the model state across service restarts. This allows the model to continue adapting to new data without losing previously learned patterns, thus avoiding the need for a full `fit` stage to start working again. {{% available_from "v1.28.1" anomaly %}} Additionally, setting [retention policies](https://docs.victoriametrics.com/anomaly-detection/components/settings/#retention) helps manage disk space or RAM used by periodical cleanup of old model instances.
 
 Here's an example of how we can switch from (offline) [Z-score model](https://docs.victoriametrics.com/anomaly-detection/components/models/#z-score) to [Online Z-score model](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-z-score):
@@ -641,7 +651,7 @@ options:
 Here’s an example of using the config splitter to divide configurations based on the `extra_filters` argument from the reader section:
 
 ```sh
-docker pull victoriametrics/vmanomaly:v1.29.7 && docker image tag victoriametrics/vmanomaly:v1.29.7 vmanomaly
+docker pull victoriametrics/vmanomaly:v1.30.0 && docker image tag victoriametrics/vmanomaly:v1.30.0 vmanomaly
 ```
 
 ```sh

docs/anomaly-detection/Migration.md

@@ -45,7 +45,7 @@ There are 2 types of compatibility to consider when migrating in stateful mode:
 
 | Group start | Group end | Compatibility | Notes |
 |---------|--------- |------------|-------|
-| [v1.29.1](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1291) | [v1.29.7](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1297) | Fully Compatible | - |
+| [v1.29.1](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1291) | [v1.30.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1300) | Fully Compatible | v1.30.0 adds new [Temporal Envelope](https://docs.victoriametrics.com/anomaly-detection/components/models/#temporal-envelope) model state without changing the compatibility of existing model and data artifacts. |
 | [v1.28.7](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1287) | [v1.29.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1290) | Partially compatible* | Dumped models of class [prophet](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet) and [seasonal quantile](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-seasonal-quantile) have problems with loading to [v1.29.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1290) due to dropped `pytz` library. **Upgrading directly from v1.28.7 to [v1.29.1](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1291) with a fix is suggested** |
 | [v1.26.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1262) | [v1.28.7](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1287) | Fully Compatible | [v1.28.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1280) introduced [rolling](https://docs.victoriametrics.com/anomaly-detection/components/models/#rolling-models) model class drop in favor of [online](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-models) models (`rolling_quantile` and `std` models), however, it does not impact compatibility, as artifacts were not produced by default for rolling models. Also, offline `mad` and `zscore` models are redirecting to their respective online counterparts since [v1.28.4](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1284). |
 | [v1.25.3](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1253) | [v1.26.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1270) | Partially Compatible* | [v1.25.3](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1253) introduced `forecast_at` argument for base [univariate](https://docs.victoriametrics.com/anomaly-detection/components/models/#univariate-models) and `Prophet` [models](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet), however, itself remains backward-reversible from newer states like [v1.26.2](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1262), [v1.27.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1270). (All models except `isolation_forest_multivariate` class will be dropped) |

docs/anomaly-detection/QuickStart.md

@@ -132,7 +132,7 @@ Below are the steps to get `vmanomaly` up and running inside a Docker container:
 1. Pull Docker image:
 
 ```sh
-docker pull victoriametrics/vmanomaly:v1.29.7
+docker pull victoriametrics/vmanomaly:v1.30.0
 ```
 
 2. Create the license file with your license key.
@@ -152,7 +152,7 @@ docker run -it \
     -v ./license:/license \
     -v ./config.yaml:/config.yaml \
     -p 8490:8490 \
-    victoriametrics/vmanomaly:v1.29.7 \
+    victoriametrics/vmanomaly:v1.30.0 \
     /config.yaml \
     --licenseFile=/license \
     --loggerLevel=INFO \
@@ -169,7 +169,7 @@ docker run -it \
     -e VMANOMALY_DATA_DUMPS_DIR=/tmp/vmanomaly/data \
     -e VMANOMALY_MODEL_DUMPS_DIR=/tmp/vmanomaly/models \
     -p 8490:8490 \
-    victoriametrics/vmanomaly:v1.29.7 \
+    victoriametrics/vmanomaly:v1.30.0 \
     /config.yaml \
     --licenseFile=/license \
     --loggerLevel=INFO \
@@ -182,7 +182,7 @@ services:
   # ...
   vmanomaly:
     container_name: vmanomaly
-    image: victoriametrics/vmanomaly:v1.29.7
+    image: victoriametrics/vmanomaly:v1.30.0
     # ...
     restart: always
     volumes:
@@ -245,7 +245,7 @@ Before deploying, check the correctness of your configuration validate config fi
 
 ### Example
 
-Here is an example of a config file that will run the [Prophet](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet) model on `vm_cache_entries` metric, with periodic scheduler that runs inference every minute and fits the model every day. The model will be trained on the last 2 weeks of data each time it is (re)fitted. The model will produce `anomaly_score`, `yhat`, `yhat_lower`, and `yhat_upper` [series](https://docs.victoriametrics.com/anomaly-detection/components/models/#vmanomaly-output) for debugging purposes. The model will be timezone-aware and will use cyclical encoding for the hour of the day and day of the week seasonality.
+Here is an example of a config file that runs the online [Temporal Envelope](https://docs.victoriametrics.com/anomaly-detection/components/models/#temporal-envelope) model on a CPU metric. The scheduler runs inference every five minutes and performs a full refit only every 100 weeks; between refits the model updates causally from each inference batch. The initial fit uses four weeks of data. The model produces `anomaly_score`, `yhat`, `yhat_lower`, and `yhat_upper` [series](https://docs.victoriametrics.com/anomaly-detection/components/models/#vmanomaly-output) for debugging, and its hour-of-day and day-of-week profiles follow the query timezone and daylight-saving-time changes.
 
 ```yaml
 settings:
@@ -260,38 +260,30 @@ settings:
     # scheduler: INFO
     # reader: INFO
     # writer: INFO
-    model.prophet: WARNING
+    model.online.temporal_envelope: WARNING
 
 schedulers:
-  1d_5m:
+  100w_5m:
     # https://docs.victoriametrics.com/anomaly-detection/components/scheduler/#periodic-scheduler
     class: 'periodic'
     infer_every: '5m'
     scatter_infer_jobs: true
-    fit_every: '1d'
+    # Temporal Envelope learns online between full refits.
+    fit_every: '100w'
     fit_window: '4w'
 
 models:
-  # https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet
-  prophet_model:
-    class: 'prophet'
+  # https://docs.victoriametrics.com/anomaly-detection/components/models/#temporal-envelope
+  temporal_envelope_model:
+    class: 'temporal_envelope'
+    queries: ['cpu_user']
+    schedulers: ['100w_5m']
     provide_series: ['anomaly_score', 'yhat', 'yhat_lower', 'yhat_upper']  # for debugging
-    tz_aware: True  # set to True if your data is timezone-aware, to deal with DST changes correctly
-    tz_use_cyclical_encoding: True
-    tz_seasonalities: # intra-day + intra-week seasonality
-      - name: 'hod'  # intra-day seasonality, hour of the day
-        fourier_order: 4  # keep it 3-8 based on intraday pattern complexity
-        prior_scale: 10
-      - name: 'dow'  # intra-week seasonality, time of the week
-        fourier_order: 2  # keep it 2-4, as dependencies are learned separately for each weekday
-    compression:  # available since v1.28.1
-      window: "30m"               # downsample 5m data into 30m intervals before fitting
-      agg_method: "mean"          # use mean aggregation within each window
-      adjust_boundaries: true     # adjust confidence intervals after downsampling
-    # inner model args (key-value pairs) accepted by
-    # https://facebook.github.io/prophet/docs/quick_start#python-api
-    args:
-      interval_width: 0.98  # see https://facebook.github.io/prophet/docs/uncertainty_intervals
+    seasonalities: ['hod_smooth', 'dow_smooth']
+    alpha: 0.005  # trend reactivity; try 0.0025-0.02
+    loss_reactivity: 5  # try 1-5; lower values reduce the influence of spikes
+    iqr_threshold: 2  # try 1-4 to adjust data-driven interval width
+    min_n_samples_seen: 16
 
 reader:
   class: 'vm'  # use VictoriaMetrics as a data source
@@ -299,6 +291,7 @@ reader:
   datasource_url: "https://play.victoriametrics.com/" # [YOUR_DATASOURCE_URL]
   tenant_id: '0:0'
   sampling_period: "5m"
+  tz: 'UTC'  # set the IANA timezone that defines local calendar patterns, e.g. 'America/New_York'
   series_processing_batch_size: 8  # number of time series to process together while preparing data for fit or infer stages
   queries:
     # define your queries with MetricsQL - https://docs.victoriametrics.com/victoriametrics/metricsql/

docs/anomaly-detection/Self-monitoring.md

@@ -135,6 +135,8 @@ These alerting rules complements the [dashboard](#grafana-dashboard) to monitor
 `vmanomaly-health` alerting group:
 - **`TooManyRestarts`**: Triggers if an instance restarts more than twice within 15 minutes, suggesting the process might be crashlooping and needs investigation.
 - **`ServiceDown`**: Alerts if an instance is down for more than 5 minutes, indicating a service outage.
+- **`SchedulerWorkerDown`**: {{% available_from "v1.30.0" anomaly %}} Alerts when an individual periodic scheduler remains down for more than 2 minutes after automatic recovery attempts. This catches partial service failures that a process-level `ServiceDown` alert cannot detect.
+- **`FrequentSchedulerRestarts`**: {{% available_from "v1.30.0" anomaly %}} Warns when an individual scheduler has more than two restart attempts within 15 minutes. A single successful self-heal does not alert, while repeated attempts indicate a flapping scheduler or dependency.
 - **`ProcessNearFDLimits`**: Alerts when the number of available file descriptors falls below 100, which could lead to severe degradation if the limit is exhausted.
 - **`TooHighCPUUsage`**: Alerts when CPU usage exceeds 90% for a continuous 5-minute period, indicating possible resource exhaustion and the need to adjust resource allocation or load.
 - **`TooHighMemoryUsage`**: Alerts when RAM usage exceeds 85% for a continuous 5-minute period and the need to adjust resource allocation or load.

docs/anomaly-detection/UI.md

@@ -193,9 +193,10 @@ The best applications of this mode are:
 
 ### What you can do with Copilot
 
-- **Ask questions** about any model (e.g. [Prophet](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet) or [Z-score](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-z-score) - parameters, trade-offs, when to use each)
+- **Ask questions** about any model (e.g. [Temporal Envelope](https://docs.victoriametrics.com/anomaly-detection/components/models/#temporal-envelope), [Prophet](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet), or [Z-score](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-z-score) - parameters, trade-offs, when to use each)
 - **Improve detection quality** - describe what's wrong ("too many false positives", "missing spikes") and Copilot reads the config, searches the docs, and proposes a validated configuration change to fix the issue.
 - **Get config suggestions inline** - suggestions appear as interactive cards with an explanation and a YAML diff; click **Apply** to write the change directly to your current settings, or **Decline** to keep the conversation going.
+- {{% available_from "v1.30.0" anomaly %}} **Profile and tune the real query** - with [mcp-vmanomaly](#mcp-tools-server) connected, Copilot can inspect bounded time-series characteristics, recommend an online model, start an asynchronous autotune task, and apply its validated query and model suggestions.
 
 ### How it works
 
@@ -212,7 +213,7 @@ AI Assistant is disabled by default; enable it with `VMANOMALY_COPILOT_ENABLED=t
 Supported providers and model formats:
 
 - **Anthropic** - set `ANTHROPIC_API_KEY`; model format: `anthropic:<model>`
-  - Examples: `claude-haiku-4-5`, `claude-sonnet-4-6`; see [full list](https://platform.claude.com/docs/en/about-claude/models/overview#latest-models-comparison)
+  - Examples: `claude-haiku-4-5`, `claude-sonnet-5`; see [full list](https://platform.claude.com/docs/en/about-claude/models/overview#latest-models-comparison)
 - **OpenAI** - set `OPENAI_API_KEY`; model format: `openai:<model>` or `openai-responses:<model>`
   - Examples: `gpt-5-mini`, `gpt-5.2`; see [full list](https://platform.openai.com/docs/models)
   - {{% available_from "v1.29.1" anomaly %}} OpenAI-compatible non-OpenAI providers are supported through `OPENAI_BASE_URL` + `OPENAI_API_KEY`
@@ -315,7 +316,7 @@ docker run -it --rm \
   -e VMANOMALY_MCP_SERVER_URL=http://mcp-vmanomaly:8081/mcp \
   -p 8080:8080 \
   -p 8490:8490 \
-  victoriametrics/vmanomaly:v1.29.7 \
+  victoriametrics/vmanomaly:v1.30.0 \
   vmanomaly_config.yaml
 ```
 
@@ -553,7 +554,7 @@ preset: ui
 # other optional server/settings parameters, e.g. port, max_concurrent_tasks, n_workers, logger_levels, etc.
 ```
 
-using one of the [deployment methods](https://docs.victoriametrics.com/anomaly-detection/quickstart/#how-to-install-and-run-vmanomaly) in a [QuickStart guide](https://docs.victoriametrics.com/anomaly-detection/quickstart/#quickstart), e.g. via Docker.
+using one of the [deployment methods](https://docs.victoriametrics.com/anomaly-detection/quickstart/#how-to-install-and-run-vmanomaly) in a [QuickStart guide](https://docs.victoriametrics.com/anomaly-detection/quickstart/), e.g. via Docker.
 
 Retrieve the UI at `http://<vmanomaly-host>:<port>` (e.g. at `http://localhost:8490` if running locally with default port) and start exploring anomaly detection models and their configurations interactively.
 
@@ -640,6 +641,29 @@ If the **results** look good and the **model configuration should be deployed in
 
 ## Changelog
 
+### v1.8.0
+Released: 23-07-2026
+
+vmanomaly version: [v1.30.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1300)
+
+- FEATURE: UX Improvements, refreshed interface with clearer model and query forms, inline configuration previews, and fullscreen chart controls.
+
+- IMPROVEMENT: Boosted [AI Copilot](#ai-assistance) stability and suggestions quality, aligned with MCP/skills toolset, and proper handling of canceled or incomplete tool calls.
+
+- Server (production-running) models and queries can be accessed and selected from the UI, with a new "Queries" button, while model selection now includes a drop-down for server-configured scheduled models in model wizard.
+
+- IMPROVEMENT: Added query prettification and stable value formatting, including platform-aware keyboard-shortcut hints.
+
+- IMPROVEMENT: Results now visibly switch to an out-of-date state after the query, time range, or model configuration changes. The action updates to rerun detection and remains visible in both expanded and collapsed model views.
+
+- IMPROVEMENT: Chart range navigation commits one query on interaction completion instead of issuing many intermediate requests.
+
+- IMPROVEMENT: Compact consecutive identical [AI Copilot](#ai-assistance) tool calls, keep query/model/anomaly suggestions synchronized, and recover cleanly from canceled or incomplete tool calls.
+
+- BUGFIX: Fixed exact UI backtesting across multiple fit cycles and for auto-tuned online wrappers, preventing *duplicate or missing predictions* while retaining causal model updates and restoration of compatible legacy auto-tuned state.
+
+- BUGFIX: Kept automatic trailing-slash redirects for configured path prefixes and `/vmui` relative to the public origin, preventing internal backend hostnames from leaking through reverse proxies such as `vmauth`.
+
 ### v1.7.2
 Released: 2026-06-25
 

docs/anomaly-detection/components/models.md

@@ -551,7 +551,6 @@ Main differences between offline and online:
 **Limitations**:
 
 - Until the online model sees enough data (especially if the data shows strong seasonality), its predictions might be unstable, producing more [false positives](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#false-positive) (`anomaly_score > 1`) or making [false negative predictions, skipping real anomalies](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#false-negative).
-- Not all models (e.g., complex ones like [Prophet](#prophet)) have a direct online alternative, thus their applicability can be somewhat limited.
 
 Each of the ([built-in](#built-in-models) or [custom](#custom-model-guide)) online models (like [`OnlineZscoreModel`](#online-z-score)) shares the following common parameters and properties:
 - `n_samples_seen_` (int) - this model *property* refers to the number of datapoints the model was trained on and increases from 0 (before the first `fit`) with each consecutive `infer` call.
@@ -568,23 +567,24 @@ Every other model that isn't [online](#online-models). Offline models are comple
 ### Overview
 Built-in models support 2 groups of arguments:
 
-- **`vmanomaly`-specific** arguments - please refer to *Parameters specific for vmanomaly* and *Default model parameters* subsections for each of the models below.
+- **Model-specific arguments** - please refer to the collapsed *Model-specific arguments* section for each model below.
 - Arguments to **inner model** (say, [Facebook's Prophet](https://facebook.github.io/prophet/docs/quick_start#python-api)), passed inside `args` argument as key-value pairs, that will be directly given to the model during initialization to allow granular control. Optional.
 
 > For users who may not be familiar with Python data types such as `list[dict]`, a [dictionary](https://www.w3schools.com/python/python_dictionaries.asp) in Python is a data structure that stores data values in key-value pairs. This structure allows for efficient data retrieval and management.
 
 
 **Models**:
-* [AutoTuned](#autotuned) - designed to take the cognitive load off the user, allowing any of built-in models below to be re-tuned for best hyperparameters on data seen during each `fit` phase of the algorithm. Tradeoff is between increased computational time and optimized results / simpler maintenance.
-* [Prophet](#prophet) - the most versatile one for production usage, especially for complex data ([trends](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#trend), [change points](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-2/#novelties), [multi-seasonality](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#seasonality))
-* [Online Z-score](#online-z-score) - useful for initial testing and for simpler data ([de-trended](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#trend) data without strict [seasonality](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#seasonality) and with anomalies of similar magnitude as your "normal" data)
-* [MAD](#online-mad) - similarly to [Z-score](#online-z-score), is effective for **identifying outliers in relatively consistent data**. Useful for detecting sudden, stark deviations from the median, being less prone to outlier's magnitude than z-score.
-* [Rolling Quantile](#rolling-quantile) - best for **data with evolving patterns**, as it adapts to changes over a rolling window.
-* [Online Seasonal Quantile](#online-seasonal-quantile) - best used on **[de-trended](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#trend) data with strong (possibly multiple) [seasonalities](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#seasonality)**. Can act as a (slightly less powerful) [online](#online-models) replacement to [`ProphetModel`](#prophet).
-* [Seasonal Trend Decomposition](#seasonal-trend-decomposition) - similarly to Holt-Winters, is best for **data with pronounced [seasonal](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#seasonality) and [trend](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#trend) components**
-* [Isolation forest (Multivariate)](#isolation-forest-multivariate) - useful for **metrics data interaction** (several queries/metrics -> single anomaly score) and **efficient in detecting anomalies in high-dimensional datasets**
-* [Holt-Winters](#holt-winters) - well-suited for **data with moderate complexity**, exhibiting distinct [trends](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#trend) and/or [single seasonal pattern](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#seasonality).
-* [Custom model](#custom-model-guide) - benefit from your own models and expertise to better support your **unique use case**.
+- [AutoTuned](#autotuned) - designed to take the cognitive load off the user, allowing any of built-in models below to be re-tuned for best hyperparameters on data seen during each `fit` phase of the algorithm. Tradeoff is between increased computational time and optimized results / simpler maintenance.
+- [Temporal Envelope](#temporal-envelope) - the preferred **online model for complex operational data** with trends, changepoints, multiple calendar patterns, holidays, capable of [forecasting](https://docs.victoriametrics.com/anomaly-detection/components/faq/#forecasting). Its multivariate form also learns cross-series relationships.
+- [Prophet](#prophet) - an offline forecasting alternative when Prophet-specific decomposition outputs are required. Favor `Temporal Envelope` for online adaptation and multivariate support.
+- [Online Z-score](#online-z-score) - useful for initial testing and for simpler data ([de-trended](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#trend) data without strict [seasonality](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#seasonality) and with anomalies of similar magnitude as your "normal" data)
+- [MAD](#online-mad) - similarly to [Z-score](#online-z-score), is effective for **identifying outliers in relatively consistent data**. Useful for detecting sudden, stark deviations from the median, being less prone to outlier's magnitude than z-score.
+- [Rolling Quantile](#rolling-quantile) - best for **data with evolving patterns**, as it adapts to changes over a rolling window.
+- [Online Seasonal Quantile](#online-seasonal-quantile) - best used on **[de-trended](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#trend) data with strong (possibly multiple) [seasonalities](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#seasonality)**. Can act as a (slightly less powerful) [online](#online-models) replacement to [`ProphetModel`](#prophet).
+- [Seasonal Trend Decomposition](#seasonal-trend-decomposition) - similarly to Holt-Winters, is best for **data with pronounced [seasonal](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#seasonality) and [trend](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#trend) components**
+- [Isolation forest (Multivariate)](#isolation-forest-multivariate) - an offline alternative for **metrics data interaction** (several queries/metrics -> single anomaly score) and high-dimensional feature-space outliers. Prefer multivariate Temporal Envelope when temporal profiles and *online* adaptation matter.
+- [Holt-Winters](#holt-winters) - well-suited for **data with moderate complexity**, exhibiting distinct [trends](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#trend) and/or [single seasonal pattern](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#seasonality).
+- [Custom model](#custom-model-guide) - benefit from your own models and expertise to better support your **unique use case**.
 
 
 ### AutoTuned
@@ -613,11 +613,13 @@ Tuning [hyperparameters](https://en.wikipedia.org/wiki/Hyperparameter_(machine_l
 > ```
 > will produce **one model per each timeseries** returned by `your_query`, with **the same** hyperparameters, such as `z_threshold`, but different parameters, such as mean, std, etc.
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
 
-* `class` (string) - model class name `"model.auto.AutoTunedModel"` (or `auto` with class alias support{{% available_from "v1.13.0" anomaly %}})
-* `tuned_class_name` (string) - [Built-in model class](#built-in-models) to wrap, i.e. `zscore_online`
-* `optimization_params` (dict) - Optimization parameters for *unsupervised* model tuning. Control percentage of found anomalies, as well as a tradeoff between time spent and the accuracy. The higher `timeout` and `n_trials` are, the better model configuration can be found for `tuned_class_name`, but the longer it takes and vice versa. Set `n_jobs` to `-1` to use all the CPUs available, it makes sense if only you have a big dataset to train on during `fit` calls, otherwise overhead isn't worth it.
+{{% collapse name="Model-specific arguments" %}}
+
+- `class` (string) - model class name `"model.auto.AutoTunedModel"` (or `auto` with class alias support{{% available_from "v1.13.0" anomaly %}})
+- `tuned_class_name` (string) - [Built-in model class](#built-in-models) to wrap, i.e. `zscore_online`
+- `optimization_params` (dict) - Optimization parameters for *unsupervised* model tuning. Control percentage of found anomalies, as well as a tradeoff between time spent and the accuracy. The higher `timeout` and `n_trials` are, the better model configuration can be found for `tuned_class_name`, but the longer it takes and vice versa. Set `n_jobs` to `-1` to use all the CPUs available, it makes sense if only you have a big dataset to train on during `fit` calls, otherwise overhead isn't worth it.
   - `anomaly_percentage` (float) - Expected percentage of anomalies that can be seen in training data, from `[0, 0.5)` interval (i.e. 0.01 means it's expected ~ 1% of anomalies to be present in training data). This is a *required* parameter.
   - `optimized_business_params` (list[string]) - {{% available_from "v1.15.0" anomaly %}} this argument allows particular [business-specific parameters](#common-args) such as [`detection_direction`](https://docs.victoriametrics.com/anomaly-detection/components/models/#detection-direction) or [`min_dev_from_expected`](https://docs.victoriametrics.com/anomaly-detection/components/models/#minimal-deviation-from-expected) to remain **unchanged during optimizations, retaining their initial values**. I.e. setting `optimized_business_params` to  `['detection_direction']` will allow to optimize only `detection_direction` business-specific arg, while `min_dev_from_expected` will retain its default value of (e.g. [1, 2] if set to that value in model config). By default and if not set, will be equal to `[]` (empty list), meaning no business params will be optimized. **A recommended option is to leave it empty** as this feature is still experimental and may lead to unexpected results.
   - `seed` (int) - Random seed for reproducibility and deterministic nature of underlying optimizations.
@@ -626,9 +628,13 @@ Tuning [hyperparameters](https://en.wikipedia.org/wiki/Hyperparameter_(machine_l
   - `train_val_ratio` (float) - {{% available_from "v1.25.1" anomaly %}} the ratio of training to validation data size when constructing folds, e.g. setting it to 2 will result in 2/3 of the data being used for training and 1/3 for validation in each of the splits. Defaults to 3 (3/4 of the data for training and 1/4 for validation). Isn't used when `validation_scheme` is set to `leaky`.
   - `n_trials` (int) - How many trials to sample from hyperparameter search space. The higher, the longer it takes but the better the results can be. Defaults to 128.
   - `timeout` (float) - How many seconds in total can be spent on each model to tune hyperparameters. The higher, the longer it takes, allowing to test more trials out of defined `n_trials`, but the better the results can be.
+  - `exact` (boolean) - {{% available_from "v1.30.0" anomaly %}} evaluate online models causally during validation: predict from the current state, then update it with newly observed data. Set this to `true` when production uses exact online inference. Defaults to `false` for backward compatibility.
+  - `optimize_complexity` (boolean) - {{% available_from "v1.30.0" anomaly %}} prefer a smaller fitted model state when validation quality is effectively tied. Defaults to `true`.
+  - `frozen_params` (dict) - {{% available_from "v1.30.0" anomaly %}} model-specific parameters that **must remain fixed** while the remaining search space is tuned. Nested dictionaries are merged recursively. Use this for non-optimizable context such as configured holidays or multivariate `groupby`; `class` and `class_name` cannot be frozen.
 
-![vmanomaly-autotune-schema](autotune.webp)
+{{% /collapse %}}
 
+{{% collapse name="Configuration example" %}}
 
 ```yaml
 # ...
@@ -647,15 +653,153 @@ models:
       timeout: 10  # how many seconds to spend on optimization for each trained model during `fit` phase call
       n_jobs: 1  # how many jobs in parallel to launch. Consider making it > 1 only if you have fit window containing > 10000 datapoints for each series
       optimized_business_params: []  # business-specific params to include in optimization, if not set - defaults to empty list, meaning no business params will be optimized, which is a recommended option as business arguments are better set by stakeholders rather than algorithms
+      exact: true  # causally validate online models the same way they run in production
+      optimize_complexity: true  # prefer smaller fitted state when quality is tied
+      frozen_params:
+        history_strength: 2
   # ...
 ```
 
+{{% /collapse %}}
+
+</div>
+
+![vmanomaly-autotune-schema](autotune.webp)
+
+#### Shared asynchronous autotune workflow
+
+{{% available_from "v1.30.0" anomaly %}} Agents, the UI, and external automation can tune one shared configuration across a bounded sample of query results without adding an `auto` wrapper to the production configuration:
+
+1. Inspect the query with `GET /api/v1/timeseries/characteristics` to identify trend, calendar seasonality, changepoints, gaps, and intermittent behavior.
+2. Start a task with `POST /api/v1/autotune/tasks`. Provide the actual query, the candidate model class, the same query `step` used in production, and a bounded `limit`.
+3. Poll `GET /api/v1/autotune/tasks/{task_id}` until `status` is `done`; cancel unnecessary work with `DELETE` on the same path.
+4. Validate and deploy `result_data.data.modelConfig`, which is a concrete configuration for the selected model class.
+
+> [!TIP]
+> Use [skills](https://docs.victoriametrics.com/ai-tools/#agent-skills) where abovementioned workflow is automated. Also, [AI Copilot](https://docs.victoriametrics.com/anomaly-detection/ui/#ai-assistant) can generate a tuned model configuration to interactively backtest in UI, based on the query data characteristics and user's anomaly expectations.
+
+Example request for an online model:
+
+```json
+{
+  "query": "sum(rate(http_requests_total[5m])) by (service)",
+  "tuned_class_name": "temporal_envelope",
+  "anomaly_percentage": 0.01,
+  "step": "5m",
+  "limit": 100,
+  "use_profile_hints": true,
+  "optimization_params": {
+    "exact": true,
+    "n_splits": 3,
+    "n_trials": 64,
+    "timeout": 60,
+    "optimize_complexity": true
+  },
+  "frozen_params": {
+    "holidays": {"countries": ["US"], "group": true}
+  }
+}
+```
+
+The requested anomaly percentage is treated as an alert-volume constraint rather than a target that must be reached in every validation fold. Model-specific search ranges may be narrowed using the sampled characteristics, while `frozen_params` preserves operator-supplied context. For online models, `exact: true` usually gives the most representative choice when production inference is causal.
+
 > There are some expected limitations of Autotune mode:
 > - It can't be made on your [custom model](#custom-model-guide).
 > - It can't be applied to itself (like `tuned_class_name: 'model.auto.AutoTunedModel'`)
-> - `AutoTunedModel` can't be used on [rolling models](https://docs.victoriametrics.com/anomaly-detection/components/models/#rolling-models) like [`RollingQuantile`](https://docs.victoriametrics.com/anomaly-detection/components/models/#rolling-quantile) in combination with [on-disk model storage mode](https://docs.victoriametrics.com/anomaly-detection/faq/#on-disk-mode), as the rolling models exists only during `infer` calls and aren't persisted neither in RAM, nor on disk.
 
 
+### Temporal Envelope
+
+{{% available_from "v1.30.0" anomaly %}} Temporal Envelope is the preferred online model for complex operational and business metrics. It learns an evolving expected range from robust trend, calendar and holiday patterns, persistent level shifts, uncertainty, and optional future forecasts. The model adapts during inference while limiting the lasting influence of short-lived spikes.
+
+> `TemporalEnvelopeModel` is [univariate](#univariate-models) and [online](#online-models). `TemporalEnvelopeMultivariateModel` also learns normal cross-series relationships as a [multivariate](#multivariate-models) model.
+
+Use it for:
+
+- infrastructure, application, or business metrics with trends and daily, weekly, monthly, or holiday behavior;
+- deployments, traffic migrations, and capacity changes that create persistent shifts, including short-horizon forecasts through `forecast_at`;
+- aligned related metrics where each channel keeps its own temporal pattern while their joint behavior contributes to one anomaly score.
+
+For simple profiles without strong trend or seasonality, prefer [Online MAD](#online-mad) or [Online Z-score](#online-z-score). [Prophet](#prophet) and [Isolation Forest](#isolation-forest-multivariate) remain offline alternatives when their distinct capabilities are required or validation favors them.
+
+<div class="model-details">
+
+{{% collapse name="Model-specific arguments" %}}
+
+- `class` (string) - `temporal_envelope` or the full class name `model.online.TemporalEnvelopeModel`.
+- `quantiles` (list[float], default `[0.25, 0.75]`) - two ordered probabilities used as the base lower and upper residual boundaries. Values must be within `[0, 1]`.
+- `iqr_threshold` (float, default `2.0`) - primary data-derived interval-width control. Higher values widen the ordinary residual envelope and produce fewer detections. Values from `1` to `4` are a practical experimentation range.
+- `alpha` (float, default `0.005`) - **Trend Reactivity**. Higher values follow persistent level and trend changes faster; lower values produce steadier expectations. Values from `0.0025` to `0.02` are a practical experimentation range.
+- `loss_reactivity` (float, default `5.0`) - controls how readily new deviations influence the model. Higher values adapt more readily; lower values reduce the influence of spikes. Values from `1` to `5` are a good starting range.
+- `min_n_samples_seen` (integer, default `16`) - observations each model or group must see before anomaly scores are trusted. State and forecasts continue learning during warmup, while anomaly scores remain zero. Approximate warmup duration is this value multiplied by the query step.
+- `changepoint_window` (integer, default `16`) - consecutive same-direction residuals required before treating a deviation as a persistent level shift. Use a value longer than ordinary anomaly bursts; approximate confirmation time is this value multiplied by the query step.
+- `seasonalities` (list[string], default `['hod_smooth', 'dow_smooth']`) - recurring calendar patterns to model. Use an empty list when calendar patterns are known to be irrelevant. Supported presets are:
+  - hour of day: `hod_smooth`, `hod_spiky`, `hod_plateau`;
+  - day of week: `dow_smooth`, `dow_spiky`, `dow_plateau`;
+  - weekday versus weekend: `weekpart_plateau`;
+  - month of year: `month_smooth`, `month_plateau`.
+- `holidays` (object, default `{}`) - known country holidays and special events. Supported keys are `countries` (country codes or names), `special_events` (recurring `MM-DD` or one-off `YYYY-MM-DD` dates), and `group` (whether configured events share one learned response). Holiday response shape is learned from data rather than configured publicly.
+- `forecast_at` (list[string], default `[]`) - positive future offsets such as `['1h', '1d']`. Each offset adds point and, when requested through `provide_series`, lower and upper forecast series.
+
+Preset suffixes describe expected profile shape: `smooth` represents gradual recurring curves, `spiky` represents narrow phase peaks, and `plateau` represents sustained calendar levels. Choose only profiles supported by the data. Calendar and holiday features use civil time from the configured query timezone, so hour/day profiles remain aligned across daylight-saving-time transitions.
+
+Temporal Envelope also supports the [common model arguments](#common-args), including `queries`, `schedulers`, `provide_series`, `detection_direction`, `scale`, `clip_predictions`, `min_dev_from_expected`, and `min_rel_dev_from_expected`. Input `data_range` and query timezone are configured on the [reader](https://docs.victoriametrics.com/anomaly-detection/components/reader/#config-parameters).
+
+The multivariate variant uses `class: temporal_envelope_multivariate` or `model.online.TemporalEnvelopeMultivariateModel` and adds:
+
+- `dependency_rank` (integer, default `8`) - compact rank used to represent cross-channel dependencies. Set it to `0` to disable dependency features.
+- `score_aggregation` (`max`, `mean`, or `l2`; default `l2`) - combines per-channel boundary deviations into one anomaly score. `l2` uses their root-mean-square magnitude.
+- `groupby` (list[string], optional) - metric label names used to fit a separate multivariate model for each unique label-value combination.
+- `max_channels` (integer, default `1000`) - hard safety limit for input time series in one model instance.
+- `recommended_max_channels` (integer, default `100`) - advisory limit; larger inputs up to `max_channels` are accepted with a warning because high-dimensional estimates may be less stable and more expensive.
+- `random_state` (integer, default `42`) - seed used for reproducible dependency features.
+
+The multivariate model emits only the joint `anomaly_score` by default to keep output cardinality bounded. Explicitly request channel-level `y`, `yhat`, `yhat_lower`, and `yhat_upper` families through `provide_series` only when those diagnostics are needed and the downstream cardinality is acceptable.
+
+{{% /collapse %}}
+
+{{% collapse name="Configuration examples" %}}
+
+Univariate monitoring with calendar and holiday behavior:
+
+```yaml
+models:
+  operations_envelope:
+    class: temporal_envelope
+    queries: [request_rate, latency]
+    seasonalities: [hod_smooth, dow_smooth, weekpart_plateau]
+    alpha: 0.005
+    loss_reactivity: 5
+    iqr_threshold: 2
+    changepoint_window: 16
+    holidays:
+      countries: [US]
+      special_events: ['12-24']
+      group: true
+    forecast_at: [1h, 1d]
+    provide_series: [anomaly_score, yhat, yhat_lower, yhat_upper]
+```
+
+Multivariate monitoring with one dependency model per cluster:
+
+```yaml
+models:
+  service_dependency_envelope:
+    class: temporal_envelope_multivariate
+    queries: [request_rate, error_rate, latency]
+    groupby: [cluster]
+    dependency_rank: 8
+    score_aggregation: l2
+    seasonalities: [hod_smooth, dow_smooth]
+    provide_series: [anomaly_score]
+```
+
+For independent per-series detection, use `temporal_envelope`. Use `temporal_envelope_multivariate` only when the relationship between aligned series is itself meaningful; each channel still learns its own trend and calendar patterns.
+
+{{% /collapse %}}
+
+</div>
+
 ### [Prophet](https://facebook.github.io/prophet/)
 `vmanomaly` uses the Facebook Prophet implementation for time series forecasting, with detailed usage provided in the [Prophet library documentation](https://facebook.github.io/prophet/docs/quick_start#python-api). All original Prophet parameters are supported and can be directly passed to the model via `args` argument.
 
@@ -663,7 +807,9 @@ models:
 
 > {{% available_from "v1.25.3" anomaly %}} Producing forecasts for future timestamps is now supported. To enable this, set the `forecast_at` argument to a list of relative future offsets (e.g., `['1h', '1d']`). The model will then generate forecasts for these future timestamps, which can be useful for planning and resource allocation. Output series are affected by [provide_series](#provide-series) argument, which need to include at least `yhat` for point-wise forecasts (and `yhat_lower` or/and `yhat_upper` for respective confidence intervals). See the example below for more details.
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
+
+{{% collapse name="Model-specific arguments" %}}
 
 - `class` (string) - model class name `"model.prophet.ProphetModel"` (or `prophet` with class alias support{{% available_from "v1.13.0" anomaly %}})
 - `seasonalities` (list[dict], optional): Additional seasonal components to include in Prophet. See Prophet’s [`add_seasonality()`](https://facebook.github.io/prophet/docs/seasonality,_holiday_effects,_and_regressors#modeling-holidays-and-special-events:~:text=modeling%20the%20cycle-,Specifying,-Custom%20Seasonalities) documentation for details.
@@ -681,19 +827,9 @@ models:
   - `agg_method` (str, optional, default="mean"): Aggregation function to apply within each window. Supported values: "mean", "median".
   - `adjust_boundaries` (bool, optional, default=true): Whether to adjust confidence interval boundaries after downsampling. If true, `yhat_lower` and `yhat_upper` will be adjusted based on the aggregated vs original data variability.
 
-> Apart from standard [`vmanomaly` output](#vmanomaly-output), Prophet model can provide additional metrics.
+{{% /collapse %}}
 
-**Additional output metrics produced by FB Prophet**
-Depending on chosen `seasonality` parameter FB Prophet can return additional metrics such as:
-- `trend`, `trend_lower`, `trend_upper`
-- `additive_terms`, `additive_terms_lower`, `additive_terms_upper`,
-- `multiplicative_terms`, `multiplicative_terms_lower`, `multiplicative_terms_upper`,
-- `daily`, `daily_lower`, `daily_upper`,
-- `hourly`, `hourly_lower`, `hourly_upper`,
-- `holidays`, `holidays_lower`, `holidays_upper`,
-- and a number of columns for each holiday if `holidays` param is set
-
-*Config Example*
+{{% collapse name="Configuration examples" %}}
 
 Timezone-unaware example:
 
@@ -766,6 +902,21 @@ models:
       country_holidays: 'US'
 ```
 
+{{% /collapse %}}
+
+</div>
+
+> Apart from standard [`vmanomaly` output](#vmanomaly-output), Prophet model can provide additional metrics.
+
+**Additional output metrics produced by FB Prophet**
+Depending on chosen `seasonality` parameter FB Prophet can return additional metrics such as:
+- `trend`, `trend_lower`, `trend_upper`
+- `additive_terms`, `additive_terms_lower`, `additive_terms_upper`,
+- `multiplicative_terms`, `multiplicative_terms_lower`, `multiplicative_terms_upper`,
+- `daily`, `daily_lower`, `daily_upper`,
+- `hourly`, `hourly_lower`, `hourly_upper`,
+- `holidays`, `holidays_lower`, `holidays_upper`,
+- and a number of columns for each holiday if `holidays` param is set
 
 Resulting metrics of the model are described [here](#vmanomaly-output)
 
@@ -775,13 +926,18 @@ Resulting metrics of the model are described [here](#vmanomaly-output)
 
 Online version of existing [Z-score](#z-score) implementation with the same exact behavior and implications {{% available_from "v1.15.0" anomaly %}}.
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
 
-* `class` (string) - model class name `"model.online.OnlineZscoreModel"` (or `zscore_online`with class alias support{{% available_from "v1.13.0" anomaly %}})
-* `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculation boundaries and anomaly score. Defaults to `2.5`.
-* `min_n_samples_seen` (int, optional) - the minimum number of samples to be seen (`n_samples_seen_` property) before computing the anomaly score. Otherwise, the **anomaly score will be 0**, as there is not enough data to trust the model's predictions. Defaults to 16.
+{{% collapse name="Model-specific arguments" %}}
 
-*Config Example*
+- `class` (string) - model class name `"model.online.OnlineZscoreModel"` (or `zscore_online`with class alias support{{% available_from "v1.13.0" anomaly %}})
+- `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculation boundaries and anomaly score. Defaults to `2.5`.
+- `min_n_samples_seen` (int, optional) - the minimum number of samples to be seen (`n_samples_seen_` property) before computing the anomaly score. Otherwise, the **anomaly score will be 0**, as there is not enough data to trust the model's predictions. Defaults to 16.
+- `history_strength` (float, optional) - {{% available_from "v1.30.0" anomaly %}} strength of the initial history learned by `fit`. Values above `1` keep fitted mean and variance unchanged initially but reduce the leverage of subsequent updates. Defaults to `1`.
+
+{{% /collapse %}}
+
+{{% collapse name="Configuration example" %}}
 
 ```yaml
 models:
@@ -789,6 +945,7 @@ models:
     class: "zscore_online"  # or 'model.online.OnlineZscoreModel'
     z_threshold: 3.5
     min_n_samples_seen: 128  # i.e. calculate it as full seasonality / data freq
+    history_strength: 2  # retain fitted history as a stronger prior
     provide_series: ['anomaly_score', 'yhat']  # common arg example
     # Common arguments for built-in model, if not set, default to
     # See https://docs.victoriametrics.com/anomaly-detection/components/models/#common-args
@@ -803,6 +960,10 @@ models:
     # anomaly_score_outside_data_range: 1.01  # auto anomaly score (1.01) if `y` (real value) is outside of data_range, if set
 ```
 
+{{% /collapse %}}
+
+</div>
+
 Resulting metrics of the model are described [here](#vmanomaly-output).
 
 
@@ -812,14 +973,19 @@ Resulting metrics of the model are described [here](#vmanomaly-output).
 
 The MAD model is a robust method for anomaly detection that is *less sensitive* to outliers in data compared to standard deviation-based models. It considers a point as an anomaly if the absolute deviation from the median is significantly large. This is the online approximate version, based on [t-digests](https://www.sciencedirect.com/science/article/pii/S2665963820300403) for online quantile estimation{{% available_from "v1.15.0" anomaly %}}.
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
 
-* `class` (string) - model class name `"model.online.OnlineMADModel"` (or `mad_online` with class alias support{{% available_from "v1.13.0" anomaly %}})
-* `threshold` (float, optional) - The threshold multiplier for the MAD to determine anomalies. Defaults to `2.5`. Higher values will identify fewer points as anomalies.
-* `min_n_samples_seen` (int, optional) - the minimum number of samples to be seen (`n_samples_seen_` property) before computing the anomaly score. Otherwise, the **anomaly score will be 0**, as there is not enough data to trust the model's predictions. Defaults to 16.
-* `compression` (int, optional) - the compression parameter for  underlying [t-digest](https://www.sciencedirect.com/science/article/pii/S2665963820300403). Higher values mean higher accuracy but higher memory usage. By default 100.
+{{% collapse name="Model-specific arguments" %}}
 
-*Config Example*
+- `class` (string) - model class name `"model.online.OnlineMADModel"` (or `mad_online` with class alias support{{% available_from "v1.13.0" anomaly %}})
+- `threshold` (float, optional) - The threshold multiplier for the MAD to determine anomalies. Defaults to `2.5`. Higher values will identify fewer points as anomalies.
+- `min_n_samples_seen` (int, optional) - the minimum number of samples to be seen (`n_samples_seen_` property) before computing the anomaly score. Otherwise, the **anomaly score will be 0**, as there is not enough data to trust the model's predictions. Defaults to 16.
+- `history_strength` (float, optional) - {{% available_from "v1.30.0" anomaly %}} strength of the initial history learned by `fit`. Values above `1` preserve fitted quantiles initially but reduce the leverage of subsequent updates. Defaults to `1`.
+- `compression` (int, optional) - the compression parameter for  underlying [t-digest](https://www.sciencedirect.com/science/article/pii/S2665963820300403). Higher values mean higher accuracy but higher memory usage. By default 100.
+
+{{% /collapse %}}
+
+{{% collapse name="Configuration example" %}}
 
 
 ```yaml
@@ -828,6 +994,7 @@ models:
     class: "mad_online"  # or 'model.online.OnlineMADModel'
     threshold: 2.5
     min_n_samples_seen: 128  # i.e. calculate it as full seasonality / data freq
+    history_strength: 2  # retain fitted history as a stronger prior
     compression: 100  # higher values mean higher accuracy but higher memory usage
     provide_series: ['anomaly_score', 'yhat']  # common arg example
     # Common arguments for built-in model, if not set, default to
@@ -843,6 +1010,10 @@ models:
     # anomaly_score_outside_data_range: 1.01  # auto anomaly score (1.01) if `y` (real value) is outside of data_range, if set
 ```
 
+{{% /collapse %}}
+
+</div>
+
 Resulting metrics of the model are described [here](#vmanomaly-output).
 
 
@@ -852,13 +1023,18 @@ Resulting metrics of the model are described [here](#vmanomaly-output).
 
 This model is best used on **data with short evolving patterns** (i.e. 10-100 datapoints of particular frequency), as it adapts to changes over a rolling window.
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
 
-* `class` (string) - model class name `"model.rolling_quantile.RollingQuantileModel"` (or `rolling_quantile` with class alias support{{% available_from "v1.13.0" anomaly %}})
-* `quantile` (float) - quantile value, from 0.5 to 1.0. This constraint is implied by 2-sided confidence interval.
-* `window_steps` (integer) - size of the moving window. (see 'sampling_period')
+{{% collapse name="Model-specific arguments" %}}
 
-*Config Example*
+- `class` (string) - model class name `"model.rolling_quantile.RollingQuantileModel"` (or `rolling_quantile` with class alias support {{% available_from "v1.13.0" anomaly %}})
+- `quantile` (float) - quantile value, from 0.5 to 1.0. This constraint is implied by 2-sided confidence interval.
+- `window_steps` (integer) - size of the moving window. (see 'sampling_period')
+- `iqr_threshold` (float, optional) - {{% available_from "v1.25.0" anomaly %}} The [interquartile range (IQR)](https://en.wikipedia.org/wiki/Interquartile_range) multiplier to increase the width of the prediction intervals. Defaults to 0 (no adjustment) for backward compatibility. If set > 0, the model will add half IQR * `iqr_threshold` to `yhat_lower` and  `yhat_upper`. This is useful for data with high variance or outliers, as it helps to avoid false positives in anomaly detection.
+
+{{% /collapse %}}
+
+{{% collapse name="Configuration example" %}}
 
 ```yaml
 models:
@@ -866,6 +1042,7 @@ models:
     class: "rolling_quantile"
     quantile: 0.9
     window_steps: 96
+    iqr_threshold: 1
     # Common arguments for built-in model, if not set, default to
     # See https://docs.victoriametrics.com/anomaly-detection/components/models/#common-args
     #
@@ -879,6 +1056,10 @@ models:
     # anomaly_score_outside_data_range: 1.01  # auto anomaly score (1.01) if `y` (real value) is outside of data_range, if set
 ```
 
+{{% /collapse %}}
+
+</div>
+
 Resulting metrics of the model are described [here](#vmanomaly-output).
 
 
@@ -892,21 +1073,26 @@ Best used on **[de-trended](https://victoriametrics.com/blog/victoriametrics-ano
 
 It uses the `quantiles` triplet to calculate `yhat_lower`, `yhat`, and `yhat_upper` [output](#vmanomaly-output), respectively, for each of the `min_subseasons` sub-intervals contained in `seasonal_interval`. For example, with '4d' + '2h' seasonality patterns (multiple), it will hold and update 24*4 / 2 = 48 consecutive estimates (each 2 hours long).
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
 
-* `class` (string) - model class name `"model.online.OnlineQuantileModel"` (or `quantile_online` with class alias support{{% available_from "v1.13.0" anomaly %}})
-* `quantiles` (list[float], optional) - The quantiles to estimate. `yhat_lower`, `yhat`, `yhat_upper` are the quantile order. By default (0.01, 0.5, 0.99).
-* `iqr_threshold` (float, optional) - {{% available_from "v1.25.0" anomaly %}} The [interquartile range (IQR)](https://en.wikipedia.org/wiki/Interquartile_range) multiplier to increase the width of the prediction intervals. Defaults to 0 (no adjustment) for backward compatibility. If set > 0, the model will add IQR * `iqr_threshold` to `yhat_lower` and  `yhat_upper` (respecting `min_subseason` seasonal buckets). This is useful for data with high variance or outliers, as it helps to avoid false positives in anomaly detection. Best used with **robust** `quantiles` set to (0.25, 0.5, 0.75) or similar.
-* `seasonal_interval` (string, optional) - the interval for the seasonal adjustment. If not set, the model will equal to a simple online quantile model. By default not set.
-* `min_subseason` (str, optional) - the minimum interval to estimate quantiles for. By default not set. Note that the minimum interval should be a multiple of the seasonal interval, i.e. if seasonal_interval='2h', then min_subseason='15m' is valid, but '37m' is not.
-* `use_transform` (bool, optional) - whether to internally apply a `log1p(abs(x)) * sign(x)` transformation to the data to stabilize internal quantile estimation. Does not affect the scale of produced output (i.e. `yhat`) By default False.
-* `global_smoothing` (float, optional) - the smoothing parameter for the global quantiles. i.e. the output is a weighted average of the global and seasonal quantiles (if `seasonal_interval` and `min_subseason` args are set). Should be from `[0, 1]` interval, where 0 means no smoothing and 1 means using only global quantile values.
-* `scale` (float, optional) - Is used to adjust the margins between `yhat` and [`yhat_lower`, `yhat_upper`]. New margin = `|yhat_* - yhat_lower| * scale`. Defaults to 1 (no scaling is applied). See `scale`[common arg](https://docs.victoriametrics.com/anomaly-detection/components/models/#scale) section for detailed instructions and 2-sided option.
-* `season_starts_from` (str, optional) - the start date for the seasonal adjustment, as a reference point to start counting the intervals. By default '1970-01-01'.
-* `min_n_samples_seen` (int, optional) - the minimum number of samples to be seen (`n_samples_seen_` property) before computing the anomaly score. Otherwise, the **anomaly score will be 0**, as there is not enough data to trust the model's predictions. Defaults to 16.
-* `compression` (int, optional) - the compression parameter for the underlying [t-digests](https://www.sciencedirect.com/science/article/pii/S2665963820300403). Higher values mean higher accuracy but higher memory usage. By default 100.
+{{% collapse name="Model-specific arguments" %}}
 
-*Config Example*
+- `class` (string) - model class name `"model.online.OnlineQuantileModel"` (or `quantile_online` with class alias support{{% available_from "v1.13.0" anomaly %}})
+- `quantiles` (list[float], optional) - The quantiles to estimate. `yhat_lower`, `yhat`, `yhat_upper` are the quantile order. By default (0.01, 0.5, 0.99).
+- `iqr_threshold` (float, optional) - {{% available_from "v1.25.0" anomaly %}} The [interquartile range (IQR)](https://en.wikipedia.org/wiki/Interquartile_range) multiplier to increase the width of the prediction intervals. Defaults to 0 (no adjustment) for backward compatibility. If set > 0, the model will add IQR * `iqr_threshold` to `yhat_lower` and  `yhat_upper` (respecting `min_subseason` seasonal buckets). This is useful for data with high variance or outliers, as it helps to avoid false positives in anomaly detection. Best used with **robust** `quantiles` set to (0.25, 0.5, 0.75) or similar.
+- `seasonal_interval` (string, optional) - the interval for the seasonal adjustment. If not set, the model will equal to a simple online quantile model. By default not set.
+- `min_subseason` (str, optional) - the minimum interval to estimate quantiles for. By default not set. Note that the minimum interval should be a multiple of the seasonal interval, i.e. if seasonal_interval='2h', then min_subseason='15m' is valid, but '37m' is not.
+- `use_transform` (bool, optional) - whether to internally apply a `log1p(abs(x)) * sign(x)` transformation to the data to stabilize internal quantile estimation. Does not affect the scale of produced output (i.e. `yhat`) By default False.
+- `global_smoothing` (float, optional) - the smoothing parameter for the global quantiles. i.e. the output is a weighted average of the global and seasonal quantiles (if `seasonal_interval` and `min_subseason` args are set). Should be from `[0, 1]` interval, where 0 means no smoothing and 1 means using only global quantile values.
+- `scale` (float, optional) - Is used to adjust the margins between `yhat` and [`yhat_lower`, `yhat_upper`]. New margin = `|yhat_* - yhat_lower| * scale`. Defaults to 1 (no scaling is applied). See `scale`[common arg](https://docs.victoriametrics.com/anomaly-detection/components/models/#scale) section for detailed instructions and 2-sided option.
+- `season_starts_from` (str, optional) - the start date for the seasonal adjustment, as a reference point to start counting the intervals. By default '1970-01-01'.
+- `min_n_samples_seen` (int, optional) - the minimum number of samples to be seen (`n_samples_seen_` property) before computing the anomaly score. Otherwise, the **anomaly score will be 0**, as there is not enough data to trust the model's predictions. Defaults to 16.
+- `history_strength` (float, optional) - {{% available_from "v1.30.0" anomaly %}} strength of the initial history learned by `fit`. Values above `1` preserve fitted quantiles initially but reduce the leverage of subsequent updates. Defaults to `1`.
+- `compression` (int, optional) - the compression parameter for the underlying [t-digests](https://www.sciencedirect.com/science/article/pii/S2665963820300403). Higher values mean higher accuracy but higher memory usage. By default 100.
+
+{{% /collapse %}}
+
+{{% collapse name="Configuration example" %}}
 
 Suppose we have a data with strong intra-day (hourly) and intra-week (daily) seasonality, data granularity is '5m' with up to 5% expected outliers present in data. Then you can apply similar config:
 
@@ -919,6 +1105,7 @@ models:
     seasonal_interval: '7d'  # longest seasonality (week, day) = week, starting from `season_starts_from`
     min_subseason: '1h'  # smallest seasonality (week, day, hour) = hour, will have its own quantile estimates
     min_n_samples_seen: 288 # 1440 / 5 - at least 1 full day, ideal = 1440 / 5 * 7 - one full week (seasonal_interval)
+    history_strength: 2  # retain fitted history as a stronger prior
     scale: 1.1  # to compensate lowered quantile boundaries with wider intervals
     season_starts_from: '2024-01-01'  # interval calculation starting point, especially for uncommon seasonalities like '36h' or '12d'
     compression: 100  # higher values mean higher accuracy but higher memory usage
@@ -936,6 +1123,10 @@ models:
     # anomaly_score_outside_data_range: 1.01  # auto anomaly score (1.01) if `y` (real value) is outside of data_range, if set
 ```
 
+{{% /collapse %}}
+
+</div>
+
 Resulting metrics of the model are described [here](#vmanomaly-output).
 
 
@@ -945,14 +1136,17 @@ Resulting metrics of the model are described [here](#vmanomaly-output).
 
 Here we use Seasonal Decompose implementation from `statsmodels` [library](https://www.statsmodels.org/dev/generated/statsmodels.tsa.seasonal.seasonal_decompose). Parameters from this library can be passed to the model. Some parameters are specifically predefined in `vmanomaly` and can't be changed by user (`model`='additive', `two_sided`=False).
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
 
-* `class` (string) - model class name `"model.std.StdModel"` (or `std` with class alias support{{% available_from "v1.13.0" anomaly %}})
-* `period` (integer) -  Number of datapoints in one season.
-* `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculating boundaries to define anomaly score. Defaults to `2.5`.
+{{% collapse name="Model-specific arguments" %}}
 
+- `class` (string) - model class name `"model.std.StdModel"` (or `std` with class alias support{{% available_from "v1.13.0" anomaly %}})
+- `period` (integer) -  Number of datapoints in one season.
+- `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculating boundaries to define anomaly score. Defaults to `2.5`.
 
-*Config Example*
+{{% /collapse %}}
+
+{{% collapse name="Configuration example" %}}
 
 
 ```yaml
@@ -973,13 +1167,16 @@ models:
     # anomaly_score_outside_data_range: 1.01  # auto anomaly score (1.01) if `y` (real value) is outside of data_range, if set
 ```
 
+{{% /collapse %}}
+
+</div>
 
 Resulting metrics of the model are described [here](#vmanomaly-output).
 
 **Additional output metrics produced by Seasonal Trend Decomposition model**
-* `resid` - The residual component of the data series.
-* `trend` - The trend component of the data series.
-* `seasonal` - The seasonal component of the data series.
+- `resid` - The residual component of the data series.
+- `trend` - The trend component of the data series.
+- `seasonal` - The seasonal component of the data series.
 
 
 ### [Isolation forest](https://en.wikipedia.org/wiki/Isolation_forest) (Multivariate)
@@ -994,13 +1191,15 @@ Detects anomalies using binary trees. The algorithm has a linear time complexity
 
 Here we use Isolation Forest implementation from `scikit-learn` [library](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.IsolationForest). All parameters from this library can be passed to the model.
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
 
-* `class` (string) - model class name `"model.isolation_forest.IsolationForestMultivariateModel"` (or `isolation_forest_multivariate` with class alias support {{% available_from "v1.13.0" anomaly %}})
+{{% collapse name="Model-specific arguments" %}}
 
-* `contamination` (float or string, optional) - The amount of contamination of the data set, i.e. the proportion of outliers in the data set. Used when fitting to define the threshold on the scores of the samples. Default value - "auto". Should be either `"auto"` or be in the range (0.0, 0.5]. {{% available_from "v1.29.5" anomaly %}} Numeric strings, such as `"0.01"`, are accepted, while invalid non-finite values, such as `nan`, `inf`, and `-inf`, are rejected during config validation.
+- `class` (string) - model class name `"model.isolation_forest.IsolationForestMultivariateModel"` (or `isolation_forest_multivariate` with class alias support {{% available_from "v1.13.0" anomaly %}})
 
-* `seasonal_features` (list of string) - List of seasonality to encode through [cyclical encoding](https://towardsdatascience.com/cyclical-features-encoding-its-about-time-ce23581845ca), i.e. `dow` (day of week). **Introduced in [1.12.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1120)**. 
+- `contamination` (float or string, optional) - The amount of contamination of the data set, i.e. the proportion of outliers in the data set. Used when fitting to define the threshold on the scores of the samples. Default value - "auto". Should be either `"auto"` or be in the range (0.0, 0.5]. {{% available_from "v1.29.5" anomaly %}} Numeric strings, such as `"0.01"`, are accepted, while invalid non-finite values, such as `nan`, `inf`, and `-inf`, are rejected during config validation.
+
+- `seasonal_features` (list of string) - List of seasonality to encode through [cyclical encoding](https://towardsdatascience.com/cyclical-features-encoding-its-about-time-ce23581845ca), i.e. `dow` (day of week). **Introduced in [1.12.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1120)**.
   - Empty by default for backward compatibility.
   - Example: `seasonal_features: ['dow', 'hod']`.
   - Supported seasonalities:
@@ -1009,9 +1208,11 @@ Here we use Isolation Forest implementation from `scikit-learn` [library](https:
     - "dow" - day of week (1-7)
     - "month" - month of year (1-12)
 
-* `args` (dict, optional) - Inner model args (key-value pairs). See accepted params in [model documentation](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.IsolationForest). Defaults to empty (not provided). Example:  {"random_state": 42, "n_estimators": 100}
+- `args` (dict, optional) - Inner model args (key-value pairs). See accepted params in [model documentation](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.IsolationForest). Defaults to empty (not provided). Example:  {"random_state": 42, "n_estimators": 100}
 
-*Config Example*
+{{% /collapse %}}
+
+{{% collapse name="Configuration example" %}}
 
 
 ```yaml
@@ -1035,6 +1236,10 @@ models:
     # anomaly_score_outside_data_range: 1.01  # auto anomaly score (1.01) if `y` (real value) is outside of data_range, if set
 ```
 
+{{% /collapse %}}
+
+</div>
+
 Resulting metrics of the model are described [here](#vmanomaly-output).
 
 ### [Holt-Winters](https://en.wikipedia.org/wiki/Exponential_smoothing)
@@ -1043,29 +1248,33 @@ Resulting metrics of the model are described [here](#vmanomaly-output).
 
 Here we use Holt-Winters Exponential Smoothing implementation from `statsmodels` [library](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing). All parameters from this library can be passed to the model.
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
 
-* `class` (string) - model class name `"model.holtwinters.HoltWinters"` (or `holtwinters` with class alias support{{% available_from "v1.13.0" anomaly %}})
+{{% collapse name="Model-specific arguments" %}}
 
-* `frequency` (string) - Must be set equal to sampling_period. Model needs to know expected data-points frequency (e.g. '10m'). If omitted, frequency is guessed during fitting as **the median of intervals between fitting data timestamps**. During inference, if incoming data doesn't have the same frequency, then it will be interpolated.  E.g. data comes at 15 seconds resolution, and our resample_freq is '1m'. Then fitting data will be downsampled to '1m' and internal model is trained at '1m' intervals. So, during inference, prediction data would be produced at '1m' intervals, but interpolated to "15s" to match with expected output, as output data must have the same timestamps. As accepted by pandas.Timedelta (e.g. '5m').
+- `class` (string) - model class name `"model.holtwinters.HoltWinters"` (or `holtwinters` with class alias support{{% available_from "v1.13.0" anomaly %}})
 
-* `seasonality` (string, optional) - As accepted by pandas.Timedelta.
+- `frequency` (string) - Must be set equal to sampling_period. Model needs to know expected data-points frequency (e.g. '10m'). If omitted, frequency is guessed during fitting as **the median of intervals between fitting data timestamps**. During inference, if incoming data doesn't have the same frequency, then it will be interpolated.  E.g. data comes at 15 seconds resolution, and our resample_freq is '1m'. Then fitting data will be downsampled to '1m' and internal model is trained at '1m' intervals. So, during inference, prediction data would be produced at '1m' intervals, but interpolated to "15s" to match with expected output, as output data must have the same timestamps. As accepted by pandas.Timedelta (e.g. '5m').
 
-* If `seasonal_periods` is not specified, it is calculated as `seasonality` / `frequency`
+- `seasonality` (string, optional) - As accepted by pandas.Timedelta.
+
+- If `seasonal_periods` is not specified, it is calculated as `seasonality` / `frequency`
 Used to compute "seasonal_periods" param for the model (e.g. '1D' or '1W').
 
-* `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculating boundaries to define anomaly score. Defaults to 2.5.
+- `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculating boundaries to define anomaly score. Defaults to 2.5.
 
 
 *Default model parameters*:
 
-* If [parameter](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing#statsmodels.tsa.holtwinters.ExponentialSmoothing-parameters) `seasonal` is not specified, default value will be `add`.
+- If [parameter](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing#statsmodels.tsa.holtwinters.ExponentialSmoothing-parameters) `seasonal` is not specified, default value will be `add`.
 
-* If [parameter](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing#statsmodels.tsa.holtwinters.ExponentialSmoothing-parameters) `initialization_method` is not specified, default value will be `estimated`.
+- If [parameter](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing#statsmodels.tsa.holtwinters.ExponentialSmoothing-parameters) `initialization_method` is not specified, default value will be `estimated`.
 
-* `args` (dict, optional) - Inner model args (key-value pairs). See accepted params in [model documentation](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing#statsmodels.tsa.holtwinters.ExponentialSmoothing-parameters). Defaults to empty (not provided). Example:  {"seasonal": "add", "initialization_method": "estimated"}
+- `args` (dict, optional) - Inner model args (key-value pairs). See accepted params in [model documentation](https://www.statsmodels.org/dev/generated/statsmodels.tsa.holtwinters.ExponentialSmoothing#statsmodels.tsa.holtwinters.ExponentialSmoothing-parameters). Defaults to empty (not provided). Example:  {"seasonal": "add", "initialization_method": "estimated"}
 
-*Config Example*
+{{% /collapse %}}
+
+{{% collapse name="Configuration example" %}}
 
 ```yaml
 models:
@@ -1090,6 +1299,10 @@ models:
     # anomaly_score_outside_data_range: 1.01  # auto anomaly score (1.01) if `y` (real value) is outside of data_range, if set
 ```
 
+{{% /collapse %}}
+
+</div>
+
 Resulting metrics of the model are described [here](#vmanomaly-output).
 
 ## vmanomaly output
@@ -1137,7 +1350,7 @@ Here in this guide, we will
 
 We'll create `custom_model.py` file with `CustomModel` class that will inherit from `vmanomaly`'s `Model` base class.
 In the `CustomModel` class, the following methods are required: - `__init__`, `fit`, `infer`, `serialize` and `deserialize`:
-* `__init__` method should initiate parameters for the model.
+- `__init__` method should initiate parameters for the model.
 
   if your model relies on configs that have `arg` [key-value pair argument, like Prophet](#prophet), do not forget to use Python's `**kwargs` in method's signature and to explicitly call
 
@@ -1145,10 +1358,10 @@ In the `CustomModel` class, the following methods are required: - `__init__`, `f
   super().__init__(**kwargs)
   ``` 
   to initialize the base class each model derives from
-* `fit` method should contain the model training process.
-* `infer` should return Pandas.DataFrame object with model's inferences.
-* `serialize` method that saves the model on disk.
-* `deserialize` load the saved model from disk.
+- `fit` method should contain the model training process.
+- `infer` should return Pandas.DataFrame object with model's inferences.
+- `serialize` method that saves the model on disk.
+- `deserialize` load the saved model from disk.
 
 For the sake of simplicity, the model in this example will return one of two values of `anomaly_score` - 0 or 1 depending on input parameter `percentage`.
 
@@ -1265,7 +1478,7 @@ monitoring:
 Let's pull the docker image for `vmanomaly`:
 
 ```sh
-docker pull victoriametrics/vmanomaly:v1.29.7
+docker pull victoriametrics/vmanomaly:v1.30.0
 ```
 
 Now we can run the docker container putting as volumes both config and model file:
@@ -1279,7 +1492,7 @@ docker run -it \
 -v $(PWD)/license:/license \
 -v $(PWD)/custom_model.py:/vmanomaly/model/custom.py \
 -v $(PWD)/custom.yaml:/config.yaml \
-victoriametrics/vmanomaly:v1.29.7 /config.yaml \
+victoriametrics/vmanomaly:v1.30.0 /config.yaml \
 --licenseFile=/license
 --watch
 ```
@@ -1348,12 +1561,16 @@ Produced model instances are **stored in-memory** between consecutive re-fit cal
 
 Model is useful for initial testing and for simpler data ([de-trended](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#trend) data without strict [seasonality](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#seasonality) and with anomalies of similar magnitude as your "normal" data).
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
 
-* `class` (string) - model class name `"model.zscore.ZscoreModel"` (or `zscore` with class alias support{{% available_from "v1.13.0" anomaly %}})
-* `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculation boundaries and anomaly score. Defaults to `2.5`.
+{{% collapse name="Model-specific arguments" %}}
 
-*Config Example*
+- `class` (string) - model class name `"model.zscore.ZscoreModel"` (or `zscore` with class alias support{{% available_from "v1.13.0" anomaly %}})
+- `z_threshold` (float, optional) - [standard score](https://en.wikipedia.org/wiki/Standard_score) for calculation boundaries and anomaly score. Defaults to `2.5`.
+
+{{% /collapse %}}
+
+{{% collapse name="Configuration example" %}}
 
 ```yaml
 models:
@@ -1373,6 +1590,10 @@ models:
     # anomaly_score_outside_data_range: 1.01  # auto anomaly score (1.01) if `y` (real value) is outside of data_range, if set
 ```
 
+{{% /collapse %}}
+
+</div>
+
 Resulting metrics of the model are described [here](#vmanomaly-output).
 
 
@@ -1382,12 +1603,16 @@ Resulting metrics of the model are described [here](#vmanomaly-output).
 
 The MAD model is a robust method for anomaly detection that is *less sensitive* to outliers in data compared to standard deviation-based models. It considers a point as an anomaly if the absolute deviation from the median is significantly large.
 
-*Parameters specific for vmanomaly*:
+<div class="model-details">
 
-* `class` (string) - model class name `"model.mad.MADModel"` (or `mad` with class alias support{{% available_from "v1.13.0" anomaly %}})
-* `threshold` (float, optional) - The threshold multiplier for the MAD to determine anomalies. Defaults to `2.5`. Higher values will identify fewer points as anomalies.
+{{% collapse name="Model-specific arguments" %}}
 
-*Config Example*
+- `class` (string) - model class name `"model.mad.MADModel"` (or `mad` with class alias support{{% available_from "v1.13.0" anomaly %}})
+- `threshold` (float, optional) - The threshold multiplier for the MAD to determine anomalies. Defaults to `2.5`. Higher values will identify fewer points as anomalies.
+
+{{% /collapse %}}
+
+{{% collapse name="Configuration example" %}}
 
 
 ```yaml
@@ -1408,4 +1633,8 @@ models:
     # anomaly_score_outside_data_range: 1.01  # auto anomaly score (1.01) if `y` (real value) is outside of data_range, if set
 ```
 
+{{% /collapse %}}
+
+</div>
+
 Resulting metrics of the model are described [here](#vmanomaly-output).

docs/anomaly-detection/components/monitoring.md

@@ -352,6 +352,20 @@ For detailed guidance on configuring mTLS parameters such as `verify_tls`, `tls_
             <td>Gauge</td>
             <td>Timestamp of the last successful config [hot-reload](https://docs.victoriametrics.com/anomaly-detection/components/#hot-reload) in seconds since epoch {{% available_from "v1.25.1" anomaly %}}</td>
         </tr>
+        <tr>
+            <td>
+<span style="white-space: nowrap;">`vmanomaly_scheduler_alive`</span>
+            </td>
+            <td>Gauge</td>
+            <td>Whether the scheduler worker thread identified by `scheduler_alias` and `preset` is alive (`1`) or not (`0`) {{% available_from "v1.30.0" anomaly %}}.</td>
+        </tr>
+        <tr>
+            <td>
+<span style="white-space: nowrap;">`vmanomaly_scheduler_restarts_total`</span>
+            </td>
+            <td>Counter</td>
+            <td>Number of bounded scheduler restart attempts {{% available_from "v1.30.0" anomaly %}}, labeled by `scheduler_alias`, `preset`, and `status` (`success` or `failure`).</td>
+        </tr>
     </tbody>
 </table>
 

docs/anomaly-detection/components/reader.md

@@ -263,7 +263,33 @@ BasicAuth password. If set, it will be used to authenticate the request.
 `30s`
             </td>
             <td>
-Timeout for the requests, passed as a string
+Backward-compatible timeout used for both datasource fetches and post-fetch processing when `fetch_timeout` or `processing_timeout` are not set.
+            </td>
+        </tr>
+        <tr>
+            <td>
+
+<span style="white-space: nowrap;">`fetch_timeout`</span>
+            </td>
+            <td>
+
+Not set (`timeout` fallback)
+            </td>
+            <td>
+Optional timeout {{% available_from "v1.30.0" anomaly %}} for each datasource read request. Use values such as `5s`, `30s`, or `1m`.
+            </td>
+        </tr>
+        <tr>
+            <td>
+
+<span style="white-space: nowrap;">`processing_timeout`</span>
+            </td>
+            <td>
+
+Not set (`timeout` fallback)
+            </td>
+            <td>
+Optional timeout {{% available_from "v1.30.0" anomaly %}} for post-fetch processing that prepares returned data for fit or inference. High-cardinality queries may need a larger processing timeout than their datasource fetch timeout.
             </td>
         </tr>
         <tr>
@@ -362,6 +388,19 @@ If True, then query will be performed from the last seen timestamp for a given s
         <tr>
             <td>
 
+<span style="white-space: nowrap;">`query_last_seen_max_lookback`</span>
+            </td>
+            <td>
+
+`None`
+            </td>
+            <td>
+Optional hard cap {{% available_from "v1.30.0" anomaly %}} for how far `query_from_last_seen_timestamp` may move a query start into the past to recover skipped inference intervals. When configured below the query step, the effective cap is raised to one step. Examples: `5m`, `1h`.
+            </td>
+        </tr>
+        <tr>
+            <td>
+
 <span style="white-space: nowrap;">`latency_offset`</span>
             </td>
             <td>
@@ -395,7 +434,7 @@ Optional arg{{% available_from "v1.17.0" anomaly %}} overrides how `search.maxPo
 `UTC`
             </td>
             <td>
-Optional argument{{% available_from "v1.18.0" anomaly %}} specifies the [IANA](https://nodatime.org/TimeZones) timezone to account for local shifts, like [DST](https://en.wikipedia.org/wiki/Daylight_saving_time), in models sensitive to seasonal patterns (e.g., [`ProphetModel`](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet) or [`OnlineQuantileModel`](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-seasonal-quantile)). Defaults to `UTC` if not set and can be overridden on a [per-query basis](#per-query-parameters).
+Optional argument {{% available_from "v1.18.0" anomaly %}} specifies the [IANA](https://nodatime.org/TimeZones) timezone to account for local shifts, like [DST](https://en.wikipedia.org/wiki/Daylight_saving_time), in models sensitive to seasonal patterns (e.g., [`TemporalEnvelopeModel`](https://docs.victoriametrics.com/anomaly-detection/components/models/#temporal-envelope), [`ProphetModel`](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet), or [`OnlineQuantileModel`](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-seasonal-quantile)). Defaults to `UTC` if not set and can be overridden on a [per-query basis](#per-query-parameters).
             </td>
         </tr>
         <tr>
@@ -461,6 +500,9 @@ reader:
       # tenant_id: '1:0'  # if set, overrides reader-level tenant_id
       # offset: '-15s'  # if set, overrides reader-level offset
   sampling_period: '1m'
+  timeout: '30s'  # backward-compatible default for both phases
+  fetch_timeout: '30s'  # timeout for each datasource request, overrides `timeout` if set
+  processing_timeout: '1m'  # timeout for preparing fetched series for fit/infer, overrides `timeout` if set
   query_from_last_seen_timestamp: True  # false by default
   latency_offset: '1ms'
   series_processing_batch_size: 8
@@ -800,7 +842,33 @@ Frequency of the points returned. Will be converted to `/select/stats_query_rang
 `30s`
             </td>
             <td>
-(Optional) Specifies the maximum duration to wait for a query to complete before timing out. Can be set on a [per-query basis](#per-query-parameters-1) to override the reader-level setting.
+(Optional) Backward-compatible timeout used for both datasource fetches and post-fetch processing when `fetch_timeout` or `processing_timeout` are not set.
+            </td>
+        </tr>
+        <tr>
+            <td>
+
+<span style="white-space: nowrap;">`fetch_timeout`</span>
+            </td>
+            <td>
+
+Not set (`timeout` fallback)
+            </td>
+            <td>
+Optional timeout {{% available_from "v1.30.0" anomaly %}} for each VictoriaLogs datasource read request. Use values such as `5s`, `30s`, or `1m`.
+            </td>
+        </tr>
+        <tr>
+            <td>
+
+<span style="white-space: nowrap;">`processing_timeout`</span>
+            </td>
+            <td>
+
+Not set (`timeout` fallback)
+            </td>
+            <td>
+Optional timeout {{% available_from "v1.30.0" anomaly %}} for post-fetch processing that prepares returned data for fit or inference. High-cardinality results may need a larger processing timeout than their datasource fetch timeout.
             </td>
         </tr>
         <tr>
@@ -906,6 +974,19 @@ If a path to a CA bundle file (like `ca.crt`), it will verify the certificate us
 Optional argument {{% available_from "v1.29.7" anomaly %}}, allows specifying the number of time series to process together while preparing data for fit or infer stages. Defaults to `8`. Suggested values are 4-16 for high-cardinality queries.
             </td>
         </tr>
+        <tr>
+            <td>
+
+<span style="white-space: nowrap;">`query_last_seen_max_lookback`</span>
+            </td>
+            <td>
+
+`None`
+            </td>
+            <td>
+Optional hard cap {{% available_from "v1.30.0" anomaly %}} for how far last-seen recovery may move a query start into the past. Examples: `5m`, `1h`.
+            </td>
+        </tr>
     </tbody>
 </table>
 
@@ -927,7 +1008,9 @@ reader:
   series_processing_batch_size: 8
   data_range: [0, 'inf']  # reader-level
   offset: '0s'  # reader-level
-  timeout: '30s'
+  timeout: '30s'  # backward-compatible default for both phases
+  fetch_timeout: '30s'  # timeout for each datasource request, overrides `timeout` if set
+  processing_timeout: '1m'  # timeout for preparing fetched series for fit/infer, overrides `timeout` if set
   queries:
     # one query returning 1 result fields (avg_duration), it will have __name__ label (series name) as `duration_30m__avg`
     duration_avg_30m:

docs/anomaly-detection/components/scheduler.md

@@ -72,6 +72,8 @@ options={`"scheduler.periodic.PeriodicScheduler"`, `"scheduler.oneoff.OneoffSche
 
 > If `start_from` [parameter](#parameters-1) is used, it's suggested to also set `restore_state: true` in the [Settings section](https://docs.victoriametrics.com/anomaly-detection/components/settings/#state-restoration) of a config, so that the scheduler can restore its state from the previous run **if terminated or restarted in between scheduled runs** and continue producing anomaly scores without interruptions, otherwise the service will be idle until future `start_from` time is reached. E.g. if `start_from` is set to `20:00` and the service is started and then terminated and restarted at `20:30`, it will not produce any anomaly scores until the next day's `20:00` is reached (+23:30 of being idle), which introduces inconvenience for the users.
 
+> {{% available_from "v1.30.0" anomaly %}} If a periodic scheduler worker exits unexpectedly, the service attempts bounded restarts with exponential backoff instead of shutting down unrelated schedulers. Monitor [`vmanomaly_scheduler_alive`](https://docs.victoriametrics.com/anomaly-detection/components/monitoring/#startup-metrics) and `vmanomaly_scheduler_restarts_total` to alert on persistent failures.
+
 ### Parameters
 
 For periodic scheduler parameters are defined as differences in times, expressed in difference units, e.g. days, hours, minutes, seconds. Time granularity is defined by the last characters of a string. Examples: `"50s"` (seconds), `"4m"` (minutes), `"3h"` (hours), `"2d"` (days), `"1w"` (weeks).

docs/anomaly-detection/components/server.md

@@ -32,7 +32,7 @@ Server component of VictoriaMetrics Anomaly Detection (`vmanomaly`) is responsib
 ### Example Configuration
 
 > [!TIP]
-> If [hot-reloading](https://docs.victoriametrics.com/anomaly-detection/components/scheduler/#hot-reloading) is enabled in vmanomaly service, the server will automatically pick up changes made to the configuration file without requiring a restart.
+> If [hot-reloading](https://docs.victoriametrics.com/anomaly-detection/components/#hot-reload) is enabled in vmanomaly service, the server will automatically pick up changes made to the configuration file without requiring a restart.
 
 ```yaml
 server:
@@ -63,4 +63,16 @@ reader:
 
 After starting the `vmanomaly` server with the above configuration, UI can be accessed at `<vmanomaly-host>:8490/vmanomaly/vmui/` (e.g. `http://localhost:8490/vmanomaly/vmui/`).
 
-Rest API endpoints (e.g. `/metrics`) can be accessed at `<vmanomaly-host>:8490/vmanomaly/metrics` (e.g. `http://localhost:8490/vmanomaly/metrics`).
+Rest API endpoints (e.g. `/metrics`) can be accessed at `<vmanomaly-host>:8490/vmanomaly/metrics` (e.g. `http://localhost:8490/vmanomaly/metrics`).
+
+### Time-series analysis and autotune API
+
+{{% available_from "v1.30.0" anomaly %}} The server exposes bounded endpoints for UI, MCP, and automation workflows:
+
+- `GET /api/v1/timeseries/characteristics` samples the supplied query and summarizes trend, calendar seasonality, changepoints, gaps, and intermittent or spiky behavior. Use `limit` (default 100) to cap sampled series and pass the production `step` and timezone.
+- `POST /api/v1/autotune/tasks` starts asynchronous shared-model tuning. The request contains the query, candidate `tuned_class_name`, expected `anomaly_percentage`, data-source settings, and optimization parameters.
+- `GET /api/v1/autotune/tasks/{task_id}` returns progress and the concrete suggested `modelConfig` when complete.
+- `DELETE /api/v1/autotune/tasks/{task_id}` cancels pending work cooperatively.
+
+> [!TIP]
+> For a complete request and recommended workflow, see [Shared asynchronous autotune workflow](https://docs.victoriametrics.com/anomaly-detection/components/models/#shared-asynchronous-autotune-workflow). OpenAPI schemas for the running version are available at `/docs` endpoint of a running `vmanomaly` instance.

docs/anomaly-detection/guides/guide-vmanomaly-vmalert/README.md

@@ -10,9 +10,9 @@ sitemap:
 
 - To use *vmanomaly*, part of the enterprise package, a license key is required. Obtain your key [here](https://victoriametrics.com/products/enterprise/trial/) for this tutorial or for enterprise use.
 - In the tutorial, we'll be using the following VictoriaMetrics components:
-  -  [VictoriaMetrics Single-Node](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) (v1.146.0)
-  -  [vmalert](https://docs.victoriametrics.com/victoriametrics/vmalert/) (v1.146.0)
-  -  [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/) (v1.146.0)
+  -  [VictoriaMetrics Single-Node](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) (v1.148.0)
+  -  [vmalert](https://docs.victoriametrics.com/victoriametrics/vmalert/) (v1.148.0)
+  -  [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/) (v1.148.0)
 - [Grafana](https://grafana.com/) (v12.2.0)
 - [Docker](https://docs.docker.com/get-docker/) and [Docker Compose](https://docs.docker.com/compose/)
 - [Node exporter](https://github.com/prometheus/node_exporter#node-exporter) (v1.9.1) and [Alertmanager](https://prometheus.io/docs/alerting/latest/alertmanager/) (v0.28.1)
@@ -323,7 +323,7 @@ Let's wrap it all up together into the `docker-compose.yml` file.
 services:
   vmagent:
     container_name: vmagent
-    image: victoriametrics/vmagent:v1.146.0
+    image: victoriametrics/vmagent:v1.148.0
     depends_on:
       - "victoriametrics"
     ports:
@@ -340,7 +340,7 @@ services:
 
   victoriametrics:
     container_name: victoriametrics
-    image: victoriametrics/victoria-metrics:v1.146.0
+    image: victoriametrics/victoria-metrics:v1.148.0
     ports:
       - 8428:8428
     volumes:
@@ -373,7 +373,7 @@ services:
 
   vmalert:
     container_name: vmalert
-    image: victoriametrics/vmalert:v1.146.0
+    image: victoriametrics/vmalert:v1.148.0
     depends_on:
       - "victoriametrics"
     ports:
@@ -395,7 +395,7 @@ services:
     restart: always
   vmanomaly:
     container_name: vmanomaly
-    image: victoriametrics/vmanomaly:v1.29.7
+    image: victoriametrics/vmanomaly:v1.30.0
     depends_on:
       - "victoriametrics"
     ports:

docs/guides/connecting-vm-components-to-cloud-storage/README.md

@@ -0,0 +1,426 @@
+Several VictoriaMetrics components can connect to cloud storage to read or write object data.
+
+The following table shows the supported types of storage for each component:
+
+| Component | AWS S3 and S3-compatible | Google Cloud Storage | Azure Blob Storage |
+|-----------|----|----------------------|--------------------|
+| [vmbackup](https://docs.victoriametrics.com/victoriametrics/vmbackup/) | ✅ | ✅ | ✅ |
+| [vmrestore](https://docs.victoriametrics.com/victoriametrics/vmrestore/) | ✅ | ✅ | ✅ |
+| [vmbackupmanager](https://docs.victoriametrics.com/victoriametrics/vmbackupmanager/) | ✅ | ✅ | ✅ |
+| [vmalert](https://docs.victoriametrics.com/victoriametrics/vmalert/) |  ✅ | ✅ | ❌ |
+
+All these components use the same underlying libraries, so the authentication setup is largely the same. The main difference is in the command-line flags: 
+
+- vmalert uses `-s3.*` prefixed flags (e.g., `-s3.credsFilePath`)
+- backup and restore tools use unprefixed flags (e.g., `-credsFilePath`)
+
+See the [component reference](https://docs.victoriametrics.com/guides/connecting-vm-components-to-cloud-storage/#per-component-flag-reference) for details.
+
+## Obtaining credentials
+
+You need to supply credentials so the component can connect to the cloud storage. The setup differs by provider; the sections below cover AWS S3, S3‑compatible systems, GCS, and Azure Blob Storage.
+
+### AWS S3
+
+1. In AWS, [create an IAM user](https://docs.aws.amazon.com/IAM/latest/UserGuide/id_users_create.html) or role with the minimum permissions for the target bucket (see table below).
+1. [Create an access key](https://docs.aws.amazon.com/IAM/latest/UserGuide/id_credentials_access-keys.html) for that IAM identity.
+1. Copy the **Access key ID** and **Secret access key** values. You will use them in the credentials file or environment variables.
+
+| Component       | S3 usage                                                    | Minimum S3 permissions |
+|-----------------|-------------------------------------------------------------|------------------------|
+| vmalert (Enterprise) | Reads alerting/recording rules from bucket.    | `s3:GetObject`, `s3:ListBucket` on the rules bucket/prefix.|
+| vmrestore       | Restores backups from S3 buckets                            | `s3:GetObject`, `s3:ListBucket` on the backup bucket/prefix.|
+| vmbackup        | Uploads backups to S3 and may delete old backup objects during maintenance.| `s3:PutObject`, `s3:GetObject`, `s3:ListBucket`, `s3:DeleteObject` on the backup bucket/prefix.|
+| vmbackupmanager (Enterprise) | Automates backups using vmbackup behavior.                  | Same as vmbackup. |
+
+
+### S3-compatible storage (MinIO, Ceph)
+
+Generate access keys using your storage system's admin interface or CLI. The credentials follow the same format as AWS S3.
+
+### Google Cloud Storage
+
+1. Open the Google Cloud Console and go to **IAM & Admin > Service Accounts**.
+1. Click **Create service account**, enter a name, and assign a Storage role (for example, Storage Object Admin).
+1. Open the service account, go to **Keys**, then click **Add key > Create new key**.
+1. Choose **JSON** as the key type and click **Create**.
+1. Store the JSON file on the machine running the VictoriaMetrics component.
+
+### Azure Blob Storage
+
+> Azure does not use credential files.
+
+1. Log in to the Azure Portal.
+1. Search for and select **Storage accounts**.
+1. Click on your specific storage account name. (This is your `AZURE_STORAGE_ACCOUNT_NAME`).
+1. In the left menu under **Security + networking**, click **Access keys**.
+1. Copy the key value from either **key1** or **key2** (this is your `AZURE_STORAGE_ACCOUNT_KEY`).
+1. Define the access keys as environment variables.
+    ```sh
+    export AZURE_STORAGE_ACCOUNT_NAME=mystorageaccount
+    export AZURE_STORAGE_ACCOUNT_KEY=myaccountkey
+    ```
+
+## Authenticating with the cloud provider
+
+Provide the credentials as a file or with environment variables, along with the path to the cloud storage bucket. The syntax for the bucket name depends on the cloud provider:
+
+- `s3://`: for AWS S3 and self-hosted S3-compatible storage (MinIO, Ceph)
+- `gs://`: Google Cloud Storage
+- `azblob://`: Azure Blob Storage
+
+### vmbackup and vmrestore
+
+The following example backups to an AWS S3 bucket using a credentials file:
+
+```sh
+vmbackup \
+  -storageDataPath=/data \
+  -snapshot.createURL=http://localhost:8428/snapshot/create \
+  -dst=s3://victoriametrics-backup/backup01 \
+  -credsFilePath=/etc/credentials
+```
+
+In order to restore from the same backup from AWS S3:
+
+```sh
+vmrestore \
+  -src=s3://victoriametrics-backup/backup01 \
+  -storageDataPath=/data \
+  -credsFilePath=/etc/credentials
+
+```
+
+> To use non-AWS S3 buckets such as MinIO or Ceph, you must [supply the `-customS3Endpoint` argument](https://docs.victoriametrics.com/guides/connecting-vm-components-to-cloud-storage/#s3-compatible).
+
+Alternatively, you can set the access keys as environment variables instead of using a credential file:
+
+```sh
+export AWS_ACCESS_KEY_ID=YOUR_AWS_ACCESS_KEY
+export AWS_SECRET_ACCESS_KEY=YOUR_SECRET_AWS_ACCESS_KEY
+
+vmbackup \
+  -storageDataPath=/data \
+  -snapshot.createURL=http://localhost:8428/snapshot/create \
+  -dst=s3://victoriametrics-backup/backup01
+
+vmrestore \
+  -src=s3://victoriametrics-backup/backup01 \
+  -storageDataPath=/data
+```
+
+
+Backups on Google Cloud Storage use the `gs://` prefix in the destination:
+
+```sh
+vmbackup \
+  -storageDataPath=/data \
+  -snapshot.createURL=http://localhost:8428/snapshot/create \
+  -dst=gs://victoriametrics-backup/backup01 \
+  -credsFilePath=/etc/credentials
+```
+
+You can restore this backup with:
+
+```sh
+vmrestore \
+  -src=gs://victoriametrics-backup/backup01 \
+  -storageDataPath=/data \
+  -credsFilePath=/etc/credentials
+```
+
+On Google Cloud, you can define the path to the JSON credential file with `GOOGLE_APPLICATION_CREDENTIALS`. For example:
+
+```sh
+export GOOGLE_APPLICATION_CREDENTIALS=/etc/credentials
+
+vmbackup \
+  -storageDataPath=/data \
+  -snapshot.createURL=http://localhost:8428/snapshot/create \
+  -dst=gs://victoriametrics-backup/backup01
+
+vmrestore \
+  -src=gs://victoriametrics-backup/backup01 \
+  -storageDataPath=/data
+```
+
+For Azure Blob Storage, use the `azblob://` prefix and rely on environment variables instead of `-credsFilePath`.
+
+```sh
+export AZURE_STORAGE_ACCOUNT_NAME=myaccount
+export AZURE_STORAGE_ACCOUNT_KEY=mykey
+
+vmbackup \
+  -storageDataPath=/data \
+  -snapshot.createURL=http://localhost:8428/snapshot/create \
+  -dst=azblob://victoriametrics-backup/backup01
+
+vmrestore \
+  -src=azblob://victoriametrics-backup/backup01 \
+  -storageDataPath=/data
+```
+
+### vmbackupmanager 
+
+> vmbackupmanager only works in the [Enterprise](https://docs.victoriametrics.com/victoriametrics/enterprise/) edition.
+
+To manage backups with vmbackupmanager on AWS S3, add the credentials with the `-credsFilePath` flag:
+
+```sh
+vmbackupmanager \
+  -dst=s3://vmstorage-data/backups \
+  -credsFilePath=/etc/credentials \
+  -storageDataPath=/vmstorage-data \
+  -snapshot.createURL=http://vmstorage:8482/snapshot/create \
+  -licenseFile=/etc/vm-license
+```
+
+Or define the access keys using environment variables:
+
+```sh
+export AWS_ACCESS_KEY_ID=YOUR_AWS_ACCESS_KEY
+export AWS_SECRET_ACCESS_KEY=YOUR_SECRET_AWS_ACCESS_KEY
+
+vmbackupmanager \
+  -dst=s3://vmstorage-data/backups \
+  -storageDataPath=/vmstorage-data \
+  -snapshot.createURL=http://vmstorage:8482/snapshot/create \
+  -licenseFile=/etc/vm-license
+```
+
+> To use non-AWS S3 buckets, you must [supply the `-customS3Endpoint` argument](#s3-compatible).
+
+Automated backups on Google Cloud Storage take the following form:
+
+```sh
+vmbackupmanager \
+  -dst=gs://vmstorage-data/backups \
+  -credsFilePath=/etc/credentials \
+  -storageDataPath=/vmstorage-data \
+  -snapshot.createURL=http://vmstorage:8482/snapshot/create \
+  -licenseFile=/etc/vm-license
+```
+
+As with vmbackup and vmrestore, you can also define the path to the JSON credential file with `GOOGLE_APPLICATION_CREDENTIALS`:
+
+```sh
+export GOOGLE_APPLICATION_CREDENTIALS=/etc/credentials
+
+vmbackupmanager \
+  -dst=gs://vmstorage-data/backups \
+  -storageDataPath=/vmstorage-data \
+  -snapshot.createURL=http://vmstorage:8482/snapshot/create \
+  -licenseFile=/etc/vm-license
+```
+
+vmbackupmanager can also use Azure Blob Storage by defining environment variables:
+
+```sh
+export AZURE_STORAGE_ACCOUNT_NAME=mystorageaccount
+export AZURE_STORAGE_ACCOUNT_KEY=myaccountkey
+
+vmbackupmanager \
+  -dst=azblob://vmstorage-data/backups \
+  -storageDataPath=/vmstorage-data \
+  -snapshot.createURL=http://vmstorage:8482/snapshot/create \
+  -licenseFile=/etc/vm-license
+```
+
+### vmalert
+
+> - vmalert cloud storage command line flags are prefixed with `-s3.` for S3 buckets *and* Google Cloud Storage.
+> - Cloud storage only works in the [Enterprise](https://docs.victoriametrics.com/victoriametrics/enterprise/) edition.
+
+Read alerting rules from an S3 bucket. The `-rule` flag accepts a prefix, so it matches all files starting with `alerts_` in the `rules` folder:
+
+```sh
+vmalert \
+  -rule=s3://my-alert-bucket/rules/alerts_ \
+  -s3.credsFilePath=/etc/vmalert/aws-credentials \
+  -datasource.url=http://vmselect:8481/select/0/prometheus \
+  -notifier.url=http://alertmanager:9093 \
+  -licenseFile=/etc/vm-license
+```
+
+Instead of a credential file, you can supply the access keys using environment variables:
+
+```sh
+export AWS_ACCESS_KEY_ID=YOUR_AWS_ACCESS_KEY
+export AWS_SECRET_ACCESS_KEY=YOUR_SECRET_AWS_ACCESS_KEY
+
+vmalert \
+  -rule=s3://my-alert-bucket/rules/alerts_ \
+  -datasource.url=http://vmselect:8481/select/0/prometheus \
+  -notifier.url=http://alertmanager:9093 \
+  -licenseFile=/etc/vm-license
+```
+
+> To use non-AWS S3 buckets, you must [supply the `-s3.customEndpoint` argument](#s3-compatible).
+
+To read rules from Google Cloud Storage:
+
+```sh
+vmalert \
+  -rule=gs://my-alert-bucket/rules/alerts_ \
+  -s3.credsFilePath=/etc/vmalert/gcp-service-account.json \
+  -datasource.url=http://vmselect:8481/select/0/prometheus \
+  -notifier.url=http://alertmanager:9093 \
+  -licenseFile=/etc/vm-license
+```
+
+If you prefer, you can supply the path to the JSON credential file with the `GOOGLE_APPLICATION_CREDENTIALS` environment variable:
+
+```sh
+export GOOGLE_APPLICATION_CREDENTIALS=/etc/credentials
+
+vmalert \
+  -rule=gs://my-alert-bucket/rules/alerts_ \
+  -datasource.url=http://vmselect:8481/select/0/prometheus \
+  -notifier.url=http://alertmanager:9093 \
+  -licenseFile=/etc/vm-license
+```
+
+## Credentials files format
+
+The file format depends on the storage provider.
+
+### S3 credentials
+
+The file uses the standard AWS shared credentials format used by the [AWS CLI](https://docs.aws.amazon.com/cli/v1/userguide/cli-configure-files.html) and [AWS SDKs](https://docs.aws.amazon.com/sdkref/latest/guide/file-format.html):
+
+```ini
+[default]
+aws_access_key_id=YOUR_AWS_ACCESS_KEY
+aws_secret_access_key=YOUR_AWS_SECRET_ACCESS_KEY
+```
+
+You can define multiple profiles in a single file:
+
+```ini
+[default]
+aws_access_key_id=DEFAULT_ACCESS_KEY
+aws_secret_access_key=DEFAULT_SECRET_KEY
+
+[monitoring]
+aws_access_key_id=MONITORING_ACCESS_KEY
+aws_secret_access_key=MONITORING_SECRET_KEY
+
+[alerts]
+aws_access_key_id=ALERTS_ACCESS_KEY
+aws_secret_access_key=ALERTS_SECRET_KEY
+```
+
+Use the `-configProfile` flag (or `-s3.configProfile` in vmalert) to select a non-default profile:
+
+```sh
+-configProfile=alerts
+```
+
+You can separate credentials from other configuration settings. Put credentials in one file:
+
+```ini
+[default]
+aws_access_key_id=DEFAULT_ACCESS_KEY
+aws_secret_access_key=DEFAULT_SECRET_KEY
+```
+
+And non-sensitive settings in another:
+
+```ini
+[default]
+region=us-east-1
+```
+
+Then pass both files:
+
+```sh
+-configFilePath=/etc/aws-config \
+-credsFilePath=/etc/credentials
+```
+
+### GCS credentials file format
+
+The file is the JSON key downloaded from Google Cloud Console. Its content looks like this:
+
+```json
+{
+  "type": "service_account",
+  "project_id": "project-id",
+  "private_key_id": "key-id",
+  "private_key": "-----BEGIN PRIVATE KEY-----\nprivate-key\n-----END PRIVATE KEY-----\n",
+  "client_email": "service-account-email",
+  "client_id": "client-id",
+  "auth_uri": "https://accounts.google.com/o/oauth2/auth",
+  "token_uri": "https://accounts.google.com/o/oauth2/token",
+  "auth_provider_x509_cert_url": "https://www.googleapis.com/oauth2/v1/certs",
+  "client_x509_cert_url": "https://www.googleapis.com/robot/v1/metadata/x509/service-account-email"
+}
+```
+
+This is the standard service account key format defined by [Google Cloud IAM](https://developers.google.com/workspace/guides/create-credentials).
+
+### Azure Blob Storage
+
+Azure does not support credentials via file. Use environment variables instead.
+
+```sh
+export AZURE_STORAGE_ACCOUNT_NAME=mystorageaccount
+export AZURE_STORAGE_ACCOUNT_KEY=myaccountkey
+```
+
+## Self-hosted and S3-compatible endpoints
+
+For S3-compatible storage such as MinIO or Ceph, set a custom endpoint with the `-customS3Endpoint` flag  for vmbackup, vmrestore, and vmbackupmanager. For example:
+
+```sh
+vmbackup \
+  -storageDataPath=/data \
+  -snapshot.createURL=http://localhost:8428/snapshot/create \
+  -dst=s3://victoriametrics-backup/backup01 \
+  -customS3Endpoint=http://minio.example.local:9000
+```
+
+On vmalert, use the `-s3.customEndpoint` flag instead:
+
+```sh
+vmalert \
+  -rule=s3://my-alert-bucket/rules/alerts_ \
+  -s3.customEndpoint=http://minio.example.local:9000 \
+  -s3.credsFilePath=/etc/vmalert/aws-credentials \
+  -datasource.url=http://vmselect:8481/select/0/prometheus \
+  -notifier.url=http://alertmanager:9093 \
+  -licenseFile=/etc/vm-license
+```
+
+### Addressing S3-compatible buckets {#s3-compatible}
+
+When connecting to non-AWS S3-compatible buckets, there is an additional flag you might need to configure:
+
+- `-s3ForcePathStyle`: on vmbackupmanager, vmbackup, and vmrestore.
+- `-s3.forcePathStyle`: on vmalert.
+
+The flag changes the expected URL pattern for a bucket.
+
+| Flag value | Address-style | Example | Use with |
+|------------|---------------|---------|----------|
+| `true` (default) | Path-style | `https://endpoint/bucket/key` |  MinIO, Ceph, most S3-compatible storages |
+| `false`        | Virtual host-style | `https://bucket.endpoint/key` | [Aliyun OSS](https://www.aliyun.com/product/oss) and other endpoints that require it |
+
+> The flag only takes effect when you use a custom endpoint (`-customS3Endpoint` or `-s3.customEndpoint` on vmalert). When connecting to real AWS S3, the SDK handles addressing automatically.
+
+## Per-component flag reference
+
+The table below shows how the same concept maps to different flag names across components.
+
+| Concept | vmalert | vmbackup, vmrestore, and vmbackupmanager |
+|---|---|---|---|---|
+| Credentials file | `-s3.credsFilePath` | `-credsFilePath` |
+| Config file | `-s3.configFilePath` | `-configFilePath` |
+| Profile selection | `-s3.configProfile` | `-configProfile` |
+| Custom endpoint | `-s3.customEndpoint` | `-customS3Endpoint` |
+| Force path style | `-s3.forcePathStyle` | `-s3ForcePathStyle` |
+| TLS insecure | N/A | `-s3TLSInsecureSkipVerify` |
+| Storage class | N/A | `-s3StorageClass` |
+
+
+

docs/guides/connecting-vm-components-to-cloud-storage/_index.md

@@ -0,0 +1,12 @@
+---
+weight: 5
+title: Connecting VictoriaMetrics components to cloud storage
+menu:
+  docs:
+    parent: guides
+    weight: 5
+tags:
+  - metrics
+  - guide
+---
+{{% content "README.md" %}}

docs/guides/grafana-vmauth-openid-configuration/README.md

@@ -240,23 +240,23 @@ vmagent will write data into VictoriaMetrics single-node and cluster (with tenan
 # compose.yaml
 services:
   vmsingle:
-    image: victoriametrics/victoria-metrics:v1.146.0
+    image: victoriametrics/victoria-metrics:v1.148.0
 
   vmstorage:
-    image: victoriametrics/vmstorage:v1.146.0-cluster
+    image: victoriametrics/vmstorage:v1.148.0-cluster
 
   vminsert:
-    image: victoriametrics/vminsert:v1.146.0-cluster
+    image: victoriametrics/vminsert:v1.148.0-cluster
     command:
       - -storageNode=vmstorage:8400
 
   vmselect:
-    image: victoriametrics/vmselect:v1.146.0-cluster
+    image: victoriametrics/vmselect:v1.148.0-cluster
     command:
       - -storageNode=vmstorage:8401
 
   vmagent:
-    image: victoriametrics/vmagent:v1.146.0
+    image: victoriametrics/vmagent:v1.148.0
     volumes:
       - ./scrape.yaml:/etc/vmagent/config.yaml
     command:
@@ -308,7 +308,7 @@ Now add the vmauth service to `compose.yaml`:
 # compose.yaml
 services:
   vmauth:
-    image: docker.io/victoriametrics/vmauth:v1.146.0
+    image: docker.io/victoriametrics/vmauth:v1.148.0
     ports:
       - 8427:8427
     volumes:

docs/guides/vm-architectures/README.md

@@ -39,7 +39,7 @@ Each subsequent section of this guide presents an architecture designed to handl
 ### The decision tree
 
 <p align="center">
-<img src="decision-tree.webp" alt="Decision Tree" width="80%">
+<img src="/guides/vm-architectures/decision-tree.webp" alt="Decision Tree" width="80%">
 </p>
 
 ## Basic
@@ -62,7 +62,7 @@ Installation guide reference: [VictoriaMetrics Single](https://docs.victoriametr
 **Schema:**
 
 <p align="center">
-<img src="basic-architecture.webp" alt="Basic Architecture" width="40%">
+<img src="/guides/vm-architectures/basic-architecture.webp" alt="Basic Architecture" width="40%">
 </p>
 
 ### Unavailability Scenarios
@@ -93,7 +93,7 @@ High availability implementation: [HA VictoriaMetrics Cluster](https://docs.vict
 **Schema:**
 
 <p align="center">
-<img src="single-az-architecture.webp" alt="Single AZ Architecture" width="60%">
+<img src="/guides/vm-architectures/single-az-architecture.webp" alt="Single AZ Architecture" width="60%">
 </p>
 
 ### Application vs. Storage Replication
@@ -210,7 +210,7 @@ To ensure reliability, vmagent implements the bulkhead pattern: each destination
 **Schema:**
 
 <p align="center">
-<img src="multi-az-architecture.webp" alt="Multi-AZ Architecture" width="65%">
+<img src="/guides/vm-architectures/multi-az-architecture.webp" alt="Multi-AZ Architecture" width="65%">
 </p>
 
 ### Unavailability Scenarios
@@ -257,7 +257,7 @@ For complete disaster recovery, this entire cell-based architecture is duplicate
 A global, stateless layer of routing cells (vmagent, vmauth) sits on top. It routes traffic to several logical groups of storage cells. Each storage group contains multiple AZs, and data is replicated or sharded across them. There are several approaches to implementing it.
 
 <p align="center">
-<img src="hyperscale-architecture.webp" alt="Hyperscale Architecture" width="85%">
+<img src="/guides/vm-architectures/hyperscale-architecture.webp" alt="Hyperscale Architecture" width="85%">
 </p>
 
 ### Choosing Your Read Path Strategy
@@ -375,7 +375,7 @@ This multitenancy approach gives us another trade-off in the isolation implement
 **Schema:**
 
 <p align="center">
-<img src="logical-layers-architecture.webp" alt="Logical Layers Architecture" width="80%">
+<img src="/guides/vm-architectures/logical-layers-architecture.webp" alt="Logical Layers Architecture" width="80%">
 </p>
 
 **Path A: Shared resources.** We have a single, shared pool of all cluster components.

docs/guides/vmagent-openid-configuration/README.md

@@ -155,15 +155,15 @@ These services will store and query the metrics scraped by vmagent.
 # compose.yaml
 services:
   vmstorage:
-    image: victoriametrics/vmstorage:v1.146.0-cluster
+    image: victoriametrics/vmstorage:v1.148.0-cluster
 
   vminsert:
-    image: victoriametrics/vminsert:v1.146.0-cluster
+    image: victoriametrics/vminsert:v1.148.0-cluster
     command:
       - -storageNode=vmstorage:8400
 
   vmselect:
-    image: victoriametrics/vmselect:v1.146.0-cluster
+    image: victoriametrics/vmselect:v1.148.0-cluster
     command:
       - -storageNode=vmstorage:8401
     ports:
@@ -196,7 +196,7 @@ Add the vmauth service to `compose.yaml`:
 # compose.yaml
 services:
   vmauth:
-    image: victoriametrics/vmauth:v1.146.0-enterprise
+    image: victoriametrics/vmauth:v1.148.0-enterprise
     ports:
       - 8427:8427
     volumes:
@@ -251,7 +251,7 @@ Add the vmagent service to `compose.yaml` with OAuth2 configuration:
 # compose.yaml
 services:
   vmagent:
-    image: victoriametrics/vmagent:v1.146.0
+    image: victoriametrics/vmagent:v1.148.0
     volumes:
       - ./scrape.yaml:/etc/vmagent/config.yaml
     command:

docs/guides/vmalert-datasource-managed-alerts-grafana/README.md

@@ -107,7 +107,7 @@ The final piece is the Docker Compose file. This ties all the services together
 # compose.yml
 services:
   victoriametrics:
-    image: victoriametrics/victoria-metrics:v1.146.0
+    image: victoriametrics/victoria-metrics:v1.148.0
     command:
       - "--storageDataPath=/victoria-metrics-data"
       - "--selfScrapeInterval=10s"
@@ -128,7 +128,7 @@ services:
       - ./alertmanager.yml:/etc/alertmanager/alertmanager.yml:ro
 
   vmalert:
-    image: victoriametrics/vmalert:v1.146.0
+    image: victoriametrics/vmalert:v1.148.0
     depends_on:
       - victoriametrics
       - alertmanager

docs/victoriametrics/Articles.md

@@ -49,7 +49,7 @@ See also [case studies](https://docs.victoriametrics.com/victoriametrics/casestu
 * [Brewblox: InfluxDB to Victoria Metrics](https://www.brewblox.com/dev/decisions/20210718_victoria_metrics.html)
 * [Techetio: Evaluating Backend Options For Prometheus Metrics](https://www.techetio.com/2022/08/21/evaluating-backend-options-for-prometheus-metrics/)
 * [Asserts: Announcing Asserts](https://www.asserts.ai/blog/announcing-asserts/)
-* [Optimizing Linkerd metrics in Prometheus](https://aatarasoff.medium.com/optimizing-linkerd-metrics-in-prometheus-de607ec10f6b)
+* [Optimizing Linkerd metrics in Prometheus](https://itnext.io/optimizing-linkerd-metrics-in-prometheus-de607ec10f6b)
 * [VictoriaMetrics vs. OpenTSDB](https://blg.robot-house.us/posts/tsdbs-grow/)
 * [Monitoring of multiple OpenShift clusters with VictoriaMetrics](https://medium.com/ibm-garage/monitoring-of-multiple-openshift-clusters-with-victoriametrics-d4f0979e2544)
 * [Ultra Monitoring with Victoria Metrics](https://dev.to/aws-builders/ultra-monitoring-with-victoria-metrics-1p2)
@@ -114,6 +114,8 @@ See also [case studies](https://docs.victoriametrics.com/victoriametrics/casestu
 * [QCon London 2026: Wrangling Telemetry at Scale, a Guide to Self-Hosted Observability](https://www.infoq.com/news/2026/03/self-hosted-observability/)
 * [How We Made Telemetry Queries 10x Faster: Chunk-Split Caching for Metrics, Logs, and Traces](https://mirastacklabs.ai/blog/chunk-split-caching/)
 * [Claude Code: creating Kubernetes debugging AI Agent for VictoriaMetrics](https://rtfm.co.ua/en/claude-code-creating-kubernetes-debugging-ai-agent-for-victoriametrics/)
+* [OpenTelemetry: OTel Collectors in Kubernetes and VictoriaMetrics Stack integration](https://itnext.io/opentelemetry-otel-collectors-in-kubernetes-and-victoriametrics-stack-integration-d907ed0a15a0)
+* [VictoriaMetrics vs Prometheus: my default, and when I still pick Prometheus](https://jorijn.com/en/blog/victoriametrics-vs-prometheus/)
 
 ## Third-party articles and slides about VictoriaLogs