clean

### Describe Your Changes

Some messages were written to `stdout` using `fmt.Printf` and
`fmt.Println`, while the other messages like import statistics were
written to `stderr` through the `log` package.

This led to ordering problems where the `Import finished!` +
`VictoriaMetrics importer stats` messages, which expected to be the last
messages, appeared before `Continue import process with filter`
messages, creating confusing output for users.

```
2025/08/20 13:07:26 Import finished!
2025/08/20 13:07:26 VictoriaMetrics importer stats:
  time spent while importing: 20h49m10.8497184s;
  total bytes: 277.1 GB;
  bytes/s: 3.7 MB;
  requests: 7978614;
  requests retries: 0;
2025/08/20 13:07:26 Total time: 20h49m10.851006088s
Continue import process with filter
        filter: match[]={__name__!=""}
        start: 2025-08-08T00:00:00Z
        end: 2025-08-15T00:00:00Z:
Continue import process with filter
        filter: match[]={__name__!=""}
        start: 2025-08-15T00:00:00Z
        end: 2025-08-19T16:18:15Z:
```


### Checklist

The following checks are **mandatory**:

- [x] My change adheres to [VictoriaMetrics contributing
guidelines](https://docs.victoriametrics.com/victoriametrics/contributing/#pull-request-checklist).
- [x] My change adheres to [VictoriaMetrics development
goals](https://docs.victoriametrics.com/victoriametrics/goals/).

.github/copilot-instructions.md

@@ -0,0 +1,23 @@
+# Project Overview
+
+VictoriaMetrics is a fast, cost-saving, and scalable solution for monitoring and managing time series data. It delivers high performance and reliability, making it an ideal choice for businesses of all sizes.
+
+## Folder Structure
+
+- `/app`: Contains the compilable binaries.
+- `/lib`: Contains the golang reusable libraries
+- `/docs/victoriametrics`: Contains documentation for the project.
+- `/apptest/tests`: Contains integration tests.
+
+## Libraries and Frameworks
+
+- Backend: Golang, no framework. Use third-party libraries sparingly.
+- Frontend: React.
+
+## Code review guidelines
+
+Ensure the feature or bugfix includes a changelog entry in /docs/victoriametrics/changelog/CHANGELOG.md.
+Verify the entry is under the ## tip section and matches the structure and style of existing entries.
+Chore-only changes may be omitted from the changelog.
+
+

.github/workflows/build.yml

@@ -31,43 +31,48 @@ concurrency:
 
 jobs:
   build:
-    name: Build
+    name: ${{ matrix.os }}-${{ matrix.arch }}
     runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - os: linux
+            arch: 386
+          - os: linux
+            arch: amd64
+          - os: linux
+            arch: arm64
+          - os: linux
+            arch: arm
+          - os: linux
+            arch: ppc64le
+          - os: darwin
+            arch: amd64
+          - os: darwin
+            arch: arm64
+          - os: freebsd
+            arch: amd64
+          - os: openbsd
+            arch: amd64
+          - os: windows
+            arch: amd64
     steps:
-      - name: Free space
-        run: |
-          # cleanup up space to free additional ~20GiB of memory
-          # which are lacking for multiplaform images build
-          formatByteCount() { echo $(numfmt --to=iec-i --suffix=B --padding=7 $1'000'); }
-          getAvailableSpace() { echo $(df -a $1 | awk 'NR > 1 {avail+=$4} END {print avail}'); }
-          BEFORE=$(getAvailableSpace)
-          sudo rm -rf /usr/local/lib/android || true
-          sudo rm -rf /usr/share/dotnet || true
-          sudo rm -rf /opt/ghc || true
-          sudo rm -rf /usr/local/.ghcup || true
-          AFTER=$(getAvailableSpace)
-          SAVED=$((AFTER-BEFORE))
-          echo "Saved $(formatByteCount $SAVED)"
-
       - name: Code checkout
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Setup Go
         id: go
         uses: actions/setup-go@v5
         with:
+          cache-dependency-path: |
+            go.sum
+            Makefile
+            app/**/Makefile
           go-version: stable
-          cache: false
 
-      - name: Cache Go artifacts
-        uses: actions/cache@v4
-        with:
-          path: |
-            ~/.cache/go-build
-            ~/go/bin
-            ~/go/pkg/mod
-          key: go-artifacts-${{ runner.os }}-crossbuild-${{ steps.go.outputs.go-version }}-${{ hashFiles('go.sum', 'Makefile', 'app/**/Makefile') }}
-          restore-keys: go-artifacts-${{ runner.os }}-crossbuild-
+      - name: Build victoria-metrics for ${{ matrix.os }}-${{ matrix.arch }}
+        run: make victoria-metrics-${{ matrix.os }}-${{ matrix.arch }}
 
-      - name: Run crossbuild
-        run: make crossbuild
+      - name: Build vmutils for ${{ matrix.os }}-${{ matrix.arch }}
+        run: make vmutils-${{ matrix.os }}-${{ matrix.arch }}

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

@@ -29,7 +29,7 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Set up Go
         id: go

.github/workflows/docs.yaml

@@ -16,12 +16,12 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
         with:
           path: __vm
 
       - name: Checkout private code
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
         with:
           repository: VictoriaMetrics/vmdocs
           token: ${{ secrets.VM_BOT_GH_TOKEN }}

.github/workflows/test.yml

@@ -1,4 +1,4 @@
-name: main
+name: test
 
 on:
   push:
@@ -25,39 +25,41 @@ concurrency:
   cancel-in-progress: true
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
 
+
 jobs:
   lint:
     name: lint
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Setup Go
         id: go
         uses: actions/setup-go@v5
         with:
-          cache: false
+          cache-dependency-path: |
+            go.sum
+            Makefile
+            app/**/Makefile
           go-version: stable
 
-      - name: Cache Go artifacts
+
+      - name: Cache golangci-lint
         uses: actions/cache@v4
         with:
           path: |
-            ~/.cache/go-build
+            ~/.cache/golangci-lint
             ~/go/bin
-            ~/go/pkg/mod
-          key: go-artifacts-${{ runner.os }}-check-all-${{ steps.go.outputs.go-version }}-${{ hashFiles('go.sum', 'Makefile', 'app/**/Makefile') }}
-          restore-keys: go-artifacts-${{ runner.os }}-check-all-
+          key: golangci-lint-${{ runner.os }}-${{ hashFiles('.golangci.yml') }}
 
       - name: Run check-all
         run: |
           make check-all
           git diff --exit-code
 
-  test:
-    name: test
-    needs: lint
+  unit:
+    name: unit
     runs-on: ubuntu-latest
 
     strategy:
@@ -69,25 +71,18 @@ jobs:
 
     steps:
       - name: Code checkout
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Setup Go
         id: go
         uses: actions/setup-go@v5
         with:
-          cache: false
+          cache-dependency-path: |
+            go.sum
+            Makefile
+            app/**/Makefile
           go-version: stable
 
-      - name: Cache Go artifacts
-        uses: actions/cache@v4
-        with:
-          path: |
-            ~/.cache/go-build
-            ~/go/bin
-            ~/go/pkg/mod
-          key: go-artifacts-${{ runner.os }}-${{ matrix.scenario }}-${{ steps.go.outputs.go-version }}-${{ hashFiles('go.sum', 'Makefile', 'app/**/Makefile') }}
-          restore-keys: go-artifacts-${{ runner.os }}-${{ matrix.scenario }}-
-
       - name: Run tests
         run: GOGC=10 make ${{ matrix.scenario}}
 
@@ -96,31 +91,23 @@ jobs:
         with:
           files: ./coverage.txt
 
-  integration-test:
-    name: integration-test
-    needs: [lint, test]
+  integration:
+    name: integration
     runs-on: ubuntu-latest
 
     steps:
       - name: Code checkout
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Setup Go
         id: go
         uses: actions/setup-go@v5
         with:
-          cache: false
+          cache-dependency-path: |
+            go.sum
+            Makefile
+            app/**/Makefile
           go-version: stable
 
-      - name: Cache Go artifacts
-        uses: actions/cache@v4
-        with:
-          path: |
-            ~/.cache/go-build
-            ~/go/bin
-            ~/go/pkg/mod
-          key: go-artifacts-${{ runner.os }}-${{ matrix.scenario }}-${{ steps.go.outputs.go-version }}-${{ hashFiles('go.sum', 'Makefile', 'app/**/Makefile') }}
-          restore-keys: go-artifacts-${{ runner.os }}-${{ matrix.scenario }}-
-
       - name: Run integration tests
         run: make integration-test

.github/workflows/vmui.yml

@@ -32,7 +32,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Code checkout
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Setup Node
         uses: actions/setup-node@v4

Makefile

@@ -12,11 +12,12 @@ PKG_TAG := $(BUILDINFO_TAG)
 endif
 
 EXTRA_DOCKER_TAG_SUFFIX ?=
+EXTRA_GO_BUILD_TAGS ?=
 
 GO_BUILDINFO = -X '$(PKG_PREFIX)/lib/buildinfo.Version=$(APP_NAME)-$(DATEINFO_TAG)-$(BUILDINFO_TAG)'
 TAR_OWNERSHIP ?= --owner=1000 --group=1000
 
-GOLANGCI_LINT_VERSION := 2.2.1
+GOLANGCI_LINT_VERSION := 2.4.0
 
 .PHONY: $(MAKECMDGOALS)
 
@@ -169,9 +170,11 @@ vmutils-windows-amd64: \
 	vmrestore-windows-amd64 \
 	vmctl-windows-amd64
 
+# When adding a new crossbuild target, please also add it to the .github/workflows/build.yml
 crossbuild:
 	$(MAKE_PARALLEL) victoria-metrics-crossbuild vmutils-crossbuild
 
+# When adding a new crossbuild target, please also add it to the .github/workflows/build.yml
 victoria-metrics-crossbuild: \
 	victoria-metrics-linux-386 \
 	victoria-metrics-linux-amd64 \
@@ -184,6 +187,7 @@ victoria-metrics-crossbuild: \
 	victoria-metrics-openbsd-amd64 \
 	victoria-metrics-windows-amd64
 
+# When adding a new crossbuild target, please also add it to the .github/workflows/build.yml
 vmutils-crossbuild: \
 	vmutils-linux-386 \
 	vmutils-linux-amd64 \
@@ -467,16 +471,16 @@ vendor-update:
 	go mod vendor
 
 app-local:
-	CGO_ENABLED=1 go build $(RACE) -ldflags "$(GO_BUILDINFO)" -o bin/$(APP_NAME)$(RACE) $(PKG_PREFIX)/app/$(APP_NAME)
+	CGO_ENABLED=1 go build $(RACE) -ldflags "$(GO_BUILDINFO)" -tags "$(EXTRA_GO_BUILD_TAGS)" -o bin/$(APP_NAME)$(RACE) $(PKG_PREFIX)/app/$(APP_NAME)
 
 app-local-pure:
-	CGO_ENABLED=0 go build $(RACE) -ldflags "$(GO_BUILDINFO)" -o bin/$(APP_NAME)-pure$(RACE) $(PKG_PREFIX)/app/$(APP_NAME)
+	CGO_ENABLED=0 go build $(RACE) -ldflags "$(GO_BUILDINFO)" -tags "$(EXTRA_GO_BUILD_TAGS)" -o bin/$(APP_NAME)-pure$(RACE) $(PKG_PREFIX)/app/$(APP_NAME)
 
 app-local-goos-goarch:
-	CGO_ENABLED=$(CGO_ENABLED) GOOS=$(GOOS) GOARCH=$(GOARCH) go build $(RACE) -ldflags "$(GO_BUILDINFO)" -o bin/$(APP_NAME)-$(GOOS)-$(GOARCH)$(RACE) $(PKG_PREFIX)/app/$(APP_NAME)
+	CGO_ENABLED=$(CGO_ENABLED) GOOS=$(GOOS) GOARCH=$(GOARCH) go build $(RACE) -ldflags "$(GO_BUILDINFO)" -tags "$(EXTRA_GO_BUILD_TAGS)" -o bin/$(APP_NAME)-$(GOOS)-$(GOARCH)$(RACE) $(PKG_PREFIX)/app/$(APP_NAME)
 
 app-local-windows-goarch:
-	CGO_ENABLED=0 GOOS=windows GOARCH=$(GOARCH) go build $(RACE) -ldflags "$(GO_BUILDINFO)" -o bin/$(APP_NAME)-windows-$(GOARCH)$(RACE).exe $(PKG_PREFIX)/app/$(APP_NAME)
+	CGO_ENABLED=0 GOOS=windows GOARCH=$(GOARCH) go build $(RACE) -ldflags "$(GO_BUILDINFO)" -tags "$(EXTRA_GO_BUILD_TAGS)" -o bin/$(APP_NAME)-windows-$(GOARCH)$(RACE).exe $(PKG_PREFIX)/app/$(APP_NAME)
 
 quicktemplate-gen: install-qtc
 	qtc

app/victoria-metrics/self_scraper.go

@@ -57,7 +57,8 @@ func selfScraper(scrapeInterval time.Duration) {
 		appmetrics.WritePrometheusMetrics(&bb)
 		s := bytesutil.ToUnsafeString(bb.B)
 		rows.Reset()
-		rows.Unmarshal(s)
+		// VictoriaMetrics components don't expose metadata yet, only need to parse samples
+		rows.UnmarshalWithErrLogger(s, nil)
 		mrs = mrs[:0]
 		for i := range rows.Rows {
 			r := &rows.Rows[i]

app/vmagent/opentelemetry/request_handler.go

@@ -8,6 +8,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/remotewrite"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentelemetry/firehose"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentelemetry/stream"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
@@ -16,9 +17,11 @@ import (
 )
 
 var (
-	rowsInserted       = metrics.NewCounter(`vmagent_rows_inserted_total{type="opentelemetry"}`)
-	rowsTenantInserted = tenantmetrics.NewCounterMap(`vmagent_tenant_inserted_rows_total{type="opentelemetry"}`)
-	rowsPerInsert      = metrics.NewHistogram(`vmagent_rows_per_insert{type="opentelemetry"}`)
+	rowsInserted           = metrics.NewCounter(`vmagent_rows_inserted_total{type="opentelemetry"}`)
+	metadataInserted       = metrics.NewCounter(`vmagent_metadata_inserted_total{type="opentelemetry"}`)
+	rowsTenantInserted     = tenantmetrics.NewCounterMap(`vmagent_tenant_inserted_rows_total{type="opentelemetry"}`)
+	metadataTenantInserted = tenantmetrics.NewCounterMap(`vmagent_tenant_inserted_metadata_total{type="opentelemetry"}`)
+	rowsPerInsert          = metrics.NewHistogram(`vmagent_rows_per_insert{type="opentelemetry"}`)
 )
 
 // InsertHandler processes opentelemetry metrics.
@@ -36,12 +39,12 @@ func InsertHandler(at *auth.Token, req *http.Request) error {
 			return fmt.Errorf("json encoding isn't supported for opentelemetry format. Use protobuf encoding")
 		}
 	}
-	return stream.ParseStream(req.Body, encoding, processBody, func(tss []prompb.TimeSeries) error {
-		return insertRows(at, tss, extraLabels)
+	return stream.ParseStream(req.Body, encoding, processBody, func(tss []prompb.TimeSeries, mms []prompb.MetricMetadata) error {
+		return insertRows(at, tss, mms, extraLabels)
 	})
 }
 
-func insertRows(at *auth.Token, tss []prompb.TimeSeries, extraLabels []prompb.Label) error {
+func insertRows(at *auth.Token, tss []prompb.TimeSeries, mms []prompb.MetricMetadata, extraLabels []prompb.Label) error {
 	ctx := common.GetPushCtx()
 	defer common.PutPushCtx(ctx)
 
@@ -63,14 +66,33 @@ func insertRows(at *auth.Token, tss []prompb.TimeSeries, extraLabels []prompb.La
 		})
 	}
 	ctx.WriteRequest.Timeseries = tssDst
+
+	var metadataTotal int
+	if promscrape.IsMetadataEnabled() {
+		var accountID, projectID uint32
+		if at != nil {
+			accountID = at.AccountID
+			projectID = at.ProjectID
+			for i := range mms {
+				mm := &mms[i]
+				mm.AccountID = accountID
+				mm.ProjectID = projectID
+			}
+		}
+		ctx.WriteRequest.Metadata = mms
+		metadataTotal = len(mms)
+	}
+
 	ctx.Labels = labels
 	ctx.Samples = samples
 	if !remotewrite.TryPush(at, &ctx.WriteRequest) {
 		return remotewrite.ErrQueueFullHTTPRetry
 	}
 	rowsInserted.Add(rowsTotal)
+	metadataInserted.Add(metadataTotal)
 	if at != nil {
 		rowsTenantInserted.Get(at).Add(rowsTotal)
+		metadataTenantInserted.Get(at).Add(metadataTotal)
 	}
 	rowsPerInsert.Update(float64(rowsTotal))
 	return nil

app/vmagent/prometheusimport/request_handler.go

@@ -8,6 +8,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/prometheus"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/prometheus/stream"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
@@ -16,9 +17,12 @@ import (
 )
 
 var (
-	rowsInserted       = metrics.NewCounter(`vmagent_rows_inserted_total{type="prometheus"}`)
-	rowsTenantInserted = tenantmetrics.NewCounterMap(`vmagent_tenant_inserted_rows_total{type="prometheus"}`)
-	rowsPerInsert      = metrics.NewHistogram(`vmagent_rows_per_insert{type="prometheus"}`)
+	rowsInserted           = metrics.NewCounter(`vmagent_rows_inserted_total{type="prometheus"}`)
+	metadataInserted       = metrics.NewCounter(`vmagent_metadata_inserted_total{type="prometheus"}`)
+	rowsTenantInserted     = tenantmetrics.NewCounterMap(`vmagent_tenant_inserted_rows_total{type="prometheus"}`)
+	metadataTenantInserted = tenantmetrics.NewCounterMap(`vmagent_tenant_inserted_metadata_total{type="prometheus"}`)
+
+	rowsPerInsert = metrics.NewHistogram(`vmagent_rows_per_insert{type="prometheus"}`)
 )
 
 // InsertHandler processes `/api/v1/import/prometheus` request.
@@ -32,18 +36,19 @@ func InsertHandler(at *auth.Token, req *http.Request) error {
 		return err
 	}
 	encoding := req.Header.Get("Content-Encoding")
-	return stream.Parse(req.Body, defaultTimestamp, encoding, true, func(rows []prometheus.Row) error {
-		return insertRows(at, rows, extraLabels)
+	return stream.Parse(req.Body, defaultTimestamp, encoding, true, promscrape.IsMetadataEnabled(), func(rows []prometheus.Row, mms []prometheus.Metadata) error {
+		return insertRows(at, rows, mms, extraLabels)
 	}, func(s string) {
 		httpserver.LogError(req, s)
 	})
 }
 
-func insertRows(at *auth.Token, rows []prometheus.Row, extraLabels []prompb.Label) error {
+func insertRows(at *auth.Token, rows []prometheus.Row, mms []prometheus.Metadata, extraLabels []prompb.Label) error {
 	ctx := common.GetPushCtx()
 	defer common.PutPushCtx(ctx)
 
 	tssDst := ctx.WriteRequest.Timeseries[:0]
+	mmsDst := ctx.WriteRequest.Metadata[:0]
 	labels := ctx.Labels[:0]
 	samples := ctx.Samples[:0]
 	for i := range rows {
@@ -70,15 +75,35 @@ func insertRows(at *auth.Token, rows []prometheus.Row, extraLabels []prompb.Labe
 			Samples: samples[len(samples)-1:],
 		})
 	}
+	var accountID, projectID uint32
+	if at != nil {
+		accountID = at.AccountID
+		projectID = at.ProjectID
+	}
+	for i := range mms {
+		mm := &mms[i]
+		mmsDst = append(mmsDst, prompb.MetricMetadata{
+			MetricFamilyName: mm.Metric,
+			Help:             mm.Help,
+			Type:             mm.Type,
+			// there is no unit in Prometheus exposition formats
+
+			AccountID: accountID,
+			ProjectID: projectID,
+		})
+	}
 	ctx.WriteRequest.Timeseries = tssDst
+	ctx.WriteRequest.Metadata = mmsDst
 	ctx.Labels = labels
 	ctx.Samples = samples
 	if !remotewrite.TryPush(at, &ctx.WriteRequest) {
 		return remotewrite.ErrQueueFullHTTPRetry
 	}
 	rowsInserted.Add(len(rows))
+	metadataInserted.Add(len(mms))
 	if at != nil {
 		rowsTenantInserted.Get(at).Add(len(rows))
+		metadataTenantInserted.Get(at).Add(len(mms))
 	}
 	rowsPerInsert.Update(float64(len(rows)))
 	return nil

app/vmagent/promremotewrite/request_handler.go

@@ -7,6 +7,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/remotewrite"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/promremotewrite/stream"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
@@ -14,9 +15,11 @@ import (
 )
 
 var (
-	rowsInserted       = metrics.NewCounter(`vmagent_rows_inserted_total{type="promremotewrite"}`)
-	rowsTenantInserted = tenantmetrics.NewCounterMap(`vmagent_tenant_inserted_rows_total{type="promremotewrite"}`)
-	rowsPerInsert      = metrics.NewHistogram(`vmagent_rows_per_insert{type="promremotewrite"}`)
+	rowsInserted           = metrics.NewCounter(`vmagent_rows_inserted_total{type="promremotewrite"}`)
+	metadataInserted       = metrics.NewCounter(`vmagent_metadata_inserted_total{type="promremotewrite"}`)
+	rowsTenantInserted     = tenantmetrics.NewCounterMap(`vmagent_tenant_inserted_rows_total{type="promremotewrite"}`)
+	metadataTenantInserted = tenantmetrics.NewCounterMap(`vmagent_tenant_inserted_metadata_total{type="promremotewrite"}`)
+	rowsPerInsert          = metrics.NewHistogram(`vmagent_rows_per_insert{type="promremotewrite"}`)
 )
 
 // InsertHandler processes remote write for prometheus.
@@ -26,17 +29,18 @@ func InsertHandler(at *auth.Token, req *http.Request) error {
 		return err
 	}
 	isVMRemoteWrite := req.Header.Get("Content-Encoding") == "zstd"
-	return stream.Parse(req.Body, isVMRemoteWrite, func(tss []prompb.TimeSeries) error {
-		return insertRows(at, tss, extraLabels)
+	return stream.Parse(req.Body, isVMRemoteWrite, func(tss []prompb.TimeSeries, mms []prompb.MetricMetadata) error {
+		return insertRows(at, tss, mms, extraLabels)
 	})
 }
 
-func insertRows(at *auth.Token, timeseries []prompb.TimeSeries, extraLabels []prompb.Label) error {
+func insertRows(at *auth.Token, timeseries []prompb.TimeSeries, mms []prompb.MetricMetadata, extraLabels []prompb.Label) error {
 	ctx := common.GetPushCtx()
 	defer common.PutPushCtx(ctx)
 
 	rowsTotal := 0
 	tssDst := ctx.WriteRequest.Timeseries[:0]
+	mmsDst := ctx.WriteRequest.Metadata[:0]
 	labels := ctx.Labels[:0]
 	samples := ctx.Samples[:0]
 	for i := range timeseries {
@@ -65,6 +69,30 @@ func insertRows(at *auth.Token, timeseries []prompb.TimeSeries, extraLabels []pr
 		})
 	}
 	ctx.WriteRequest.Timeseries = tssDst
+
+	var metadataTotal int
+	if promscrape.IsMetadataEnabled() {
+		var accountID, projectID uint32
+		if at != nil {
+			accountID = at.AccountID
+			projectID = at.ProjectID
+		}
+		for i := range mms {
+			mm := &mms[i]
+			mmsDst = append(mmsDst, prompb.MetricMetadata{
+				MetricFamilyName: mm.MetricFamilyName,
+				Help:             mm.Help,
+				Type:             mm.Type,
+				Unit:             mm.Unit,
+
+				AccountID: accountID,
+				ProjectID: projectID,
+			})
+		}
+		ctx.WriteRequest.Metadata = mmsDst
+		metadataTotal = len(mms)
+	}
+
 	ctx.Labels = labels
 	ctx.Samples = samples
 	if !remotewrite.TryPush(at, &ctx.WriteRequest) {
@@ -73,7 +101,9 @@ func insertRows(at *auth.Token, timeseries []prompb.TimeSeries, extraLabels []pr
 	rowsInserted.Add(rowsTotal)
 	if at != nil {
 		rowsTenantInserted.Get(at).Add(rowsTotal)
+		metadataTenantInserted.Get(at).Add(metadataTotal)
 	}
+	metadataInserted.Add(metadataTotal)
 	rowsPerInsert.Update(float64(rowsTotal))
 	return nil
 }

app/vmagent/remotewrite/client.go

@@ -463,12 +463,6 @@ again:
 		// - Real-world implementations of v1 use both 400 and 415 status codes.
 		// See more in research: https://github.com/VictoriaMetrics/VictoriaMetrics/pull/8462#issuecomment-2786918054
 	case 415, 400:
-		if c.canDowngradeVMProto.Swap(false) {
-			logger.Infof("received unsupported media type or bad request from remote storage at %q. Downgrading protocol from VictoriaMetrics to Prometheus remote write for all future requests. "+
-				"See https://docs.victoriametrics.com/victoriametrics/vmagent/#victoriametrics-remote-write-protocol", c.sanitizedURL)
-			c.useVMProto.Store(false)
-		}
-
 		if encoding.IsZstd(block) {
 			logger.Infof("received unsupported media type or bad request from remote storage at %q. Re-packing the block to Prometheus remote write and retrying."+
 				"See https://docs.victoriametrics.com/victoriametrics/vmagent/#victoriametrics-remote-write-protocol", c.sanitizedURL)

app/vmagent/remotewrite/pendingseries.go

@@ -24,9 +24,10 @@ import (
 
 var (
 	flushInterval = flag.Duration("remoteWrite.flushInterval", time.Second, "Interval for flushing the data to remote storage. "+
-		"This option takes effect only when less than 10K data points per second are pushed to -remoteWrite.url")
+		"This option takes effect only when less than -remoteWrite.maxRowsPerBlock data points per -remoteWrite.flushInterval are pushed to -remoteWrite.url")
 	maxUnpackedBlockSize = flagutil.NewBytes("remoteWrite.maxBlockSize", 8*1024*1024, "The maximum block size to send to remote storage. Bigger blocks may improve performance at the cost of the increased memory usage. See also -remoteWrite.maxRowsPerBlock")
 	maxRowsPerBlock      = flag.Int("remoteWrite.maxRowsPerBlock", 10000, "The maximum number of samples to send in each block to remote storage. Higher number may improve performance at the cost of the increased memory usage. See also -remoteWrite.maxBlockSize")
+	maxMetadataPerBlock  = flag.Int("remoteWrite.maxMetadataPerBlock", 5000, "The maximum number of metadata to send in each block to remote storage. Higher number may improve performance at the cost of the increased memory usage. See also -remoteWrite.maxBlockSize")
 	vmProtoCompressLevel = flag.Int("remoteWrite.vmProtoCompressLevel", 0, "The compression level for VictoriaMetrics remote write protocol. "+
 		"Higher values reduce network traffic at the cost of higher CPU usage. Negative values reduce CPU usage at the cost of increased network traffic. "+
 		"See https://docs.victoriametrics.com/victoriametrics/vmagent/#victoriametrics-remote-write-protocol")
@@ -60,9 +61,16 @@ func (ps *pendingSeries) MustStop() {
 	ps.periodicFlusherWG.Wait()
 }
 
-func (ps *pendingSeries) TryPush(tss []prompb.TimeSeries) bool {
+func (ps *pendingSeries) TryPushTimeSeries(tss []prompb.TimeSeries) bool {
 	ps.mu.Lock()
-	ok := ps.wr.tryPush(tss)
+	ok := ps.wr.tryPushTimeSeries(tss)
+	ps.mu.Unlock()
+	return ok
+}
+
+func (ps *pendingSeries) TryPushMetadata(mms []prompb.MetricMetadata) bool {
+	ps.mu.Lock()
+	ok := ps.wr.tryPushMetadata(mms)
 	ps.mu.Unlock()
 	return ok
 }
@@ -111,26 +119,34 @@ type writeRequest struct {
 	wr prompb.WriteRequest
 
 	tss     []prompb.TimeSeries
+	mms     []prompb.MetricMetadata
 	labels  []prompb.Label
 	samples []prompb.Sample
 
 	// buf holds labels data
 	buf []byte
+	// metadatabuf holds metadata data
+	metadatabuf []byte
 }
 
 func (wr *writeRequest) reset() {
 	// Do not reset lastFlushTime, fq, isVMRemoteWrite, significantFigures and roundDigits, since they are reused.
 
 	wr.wr.Timeseries = nil
+	wr.wr.Metadata = nil
 
 	clear(wr.tss)
 	wr.tss = wr.tss[:0]
 
+	clear(wr.mms)
+	wr.mms = wr.mms[:0]
+
 	promrelabel.CleanLabels(wr.labels)
 	wr.labels = wr.labels[:0]
 
 	wr.samples = wr.samples[:0]
 	wr.buf = wr.buf[:0]
+	wr.metadatabuf = wr.metadatabuf[:0]
 }
 
 // mustFlushOnStop force pushes wr data into wr.fq
@@ -138,6 +154,7 @@ func (wr *writeRequest) reset() {
 // This is needed in order to properly save in-memory data to persistent queue on graceful shutdown.
 func (wr *writeRequest) mustFlushOnStop() {
 	wr.wr.Timeseries = wr.tss
+	wr.wr.Metadata = wr.mms
 	if !tryPushWriteRequest(&wr.wr, wr.mustWriteBlock, wr.isVMRemoteWrite.Load()) {
 		logger.Panicf("BUG: final flush must always return true")
 	}
@@ -151,6 +168,7 @@ func (wr *writeRequest) mustWriteBlock(block []byte) bool {
 
 func (wr *writeRequest) tryFlush() bool {
 	wr.wr.Timeseries = wr.tss
+	wr.wr.Metadata = wr.mms
 	wr.lastFlushTime.Store(fasttime.UnixTimestamp())
 	if !tryPushWriteRequest(&wr.wr, wr.fq.TryWriteBlock, wr.isVMRemoteWrite.Load()) {
 		return false
@@ -174,7 +192,49 @@ func adjustSampleValues(samples []prompb.Sample, significantFigures, roundDigits
 	}
 }
 
-func (wr *writeRequest) tryPush(src []prompb.TimeSeries) bool {
+func (wr *writeRequest) tryPushMetadata(mms []prompb.MetricMetadata) bool {
+	mmdDst := wr.mms
+	maxMetadataPerBlock := *maxMetadataPerBlock
+	for i := range mms {
+		if len(wr.mms) >= maxMetadataPerBlock {
+			if !wr.tryFlush() {
+				return false
+			}
+			mmdDst = wr.mms
+		}
+		mmSrc := &mms[i]
+		mmdDst = append(mmdDst, prompb.MetricMetadata{})
+		wr.copyMetadata(&mmdDst[len(mmdDst)-1], mmSrc)
+	}
+	wr.mms = mmdDst
+	return true
+}
+
+func (wr *writeRequest) copyMetadata(dst, src *prompb.MetricMetadata) {
+	// Direct copy for non-string fields, which are safe by value.
+	dst.Type = src.Type
+	dst.Unit = src.Unit
+
+	// Pre-allocate memory for all string fields.
+	neededBufLen := len(src.MetricFamilyName) + len(src.Help)
+	bufLen := len(wr.metadatabuf)
+	wr.metadatabuf = slicesutil.SetLength(wr.metadatabuf, bufLen+neededBufLen)
+	buf := wr.metadatabuf[:bufLen]
+
+	// Copy MetricFamilyName
+	bufLen = len(buf)
+	buf = append(buf, src.MetricFamilyName...)
+	dst.MetricFamilyName = bytesutil.ToUnsafeString(buf[bufLen:])
+
+	// Copy Help
+	bufLen = len(buf)
+	buf = append(buf, src.Help...)
+	dst.Help = bytesutil.ToUnsafeString(buf[bufLen:])
+
+	wr.metadatabuf = buf
+}
+
+func (wr *writeRequest) tryPushTimeSeries(src []prompb.TimeSeries) bool {
 	tssDst := wr.tss
 	maxSamplesPerBlock := *maxRowsPerBlock
 	// Allow up to 10x of labels per each block on average.
@@ -241,7 +301,7 @@ func (wr *writeRequest) copyTimeSeries(dst, src *prompb.TimeSeries) {
 var marshalConcurrencyCh = make(chan struct{}, cgroup.AvailableCPUs())
 
 func tryPushWriteRequest(wr *prompb.WriteRequest, tryPushBlock func(block []byte) bool, isVMRemoteWrite bool) bool {
-	if len(wr.Timeseries) == 0 {
+	if wr.IsEmpty() {
 		// Nothing to push
 		return true
 	}
@@ -267,6 +327,7 @@ func tryPushWriteRequest(wr *prompb.WriteRequest, tryPushBlock func(block []byte
 			compressBufPool.Put(zb)
 			if ok {
 				blockSizeRows.Update(float64(len(wr.Timeseries)))
+				blockMetadataRows.Update(float64(len(wr.Metadata)))
 				blockSizeBytes.Update(float64(zbLen))
 			}
 			return ok
@@ -278,47 +339,86 @@ func tryPushWriteRequest(wr *prompb.WriteRequest, tryPushBlock func(block []byte
 		<-marshalConcurrencyCh
 	}
 
-	// Too big block. Recursively split it into smaller parts if possible.
-	if len(wr.Timeseries) == 1 {
-		// A single time series left. Recursively split its samples into smaller parts if possible.
+	// Split timeseries or metadata into two smaller blocks
+	switch len(wr.Timeseries) {
+	case 0:
+		if len(wr.Metadata) == 1 {
+			logger.Warnf("dropping a metadata exceeding -remoteWrite.maxBlockSize=%d bytes", maxUnpackedBlockSize.N)
+			return true
+		}
+		metadata := wr.Metadata
+		n := len(metadata) / 2
+		wr.Metadata = metadata[:n]
+		if !tryPushWriteRequest(wr, tryPushBlock, isVMRemoteWrite) {
+			wr.Metadata = metadata
+			return false
+		}
+		wr.Metadata = metadata[n:]
+		if !tryPushWriteRequest(wr, tryPushBlock, isVMRemoteWrite) {
+			wr.Metadata = metadata
+			return false
+		}
+		wr.Metadata = metadata
+		return true
+
+	case 1:
+		// A single time series left. Recursively split its samples and metadata into smaller parts if possible.
 		samples := wr.Timeseries[0].Samples
-		if len(samples) == 1 {
-			logger.Warnf("dropping a sample for metric with too long labels exceeding -remoteWrite.maxBlockSize=%d bytes", maxUnpackedBlockSize.N)
+		metaData := wr.Metadata
+		if len(samples) == 1 && len(metaData) <= 1 {
+			logger.Warnf("dropping a sample for metric and %d metadata which are exceeding -remoteWrite.maxBlockSize=%d bytes", len(metaData), maxUnpackedBlockSize.N)
 			return true
 		}
 		n := len(samples) / 2
+		m := len(metaData) / 2
 		wr.Timeseries[0].Samples = samples[:n]
+		wr.Metadata = metaData[:m]
 		if !tryPushWriteRequest(wr, tryPushBlock, isVMRemoteWrite) {
 			wr.Timeseries[0].Samples = samples
+			wr.Metadata = metaData
 			return false
 		}
 		wr.Timeseries[0].Samples = samples[n:]
+		wr.Metadata = metaData[m:]
 		if !tryPushWriteRequest(wr, tryPushBlock, isVMRemoteWrite) {
 			wr.Timeseries[0].Samples = samples
+			wr.Metadata = metaData
 			return false
 		}
 		wr.Timeseries[0].Samples = samples
+		wr.Metadata = metaData
+		return true
+
+	default:
+		// Split both timeseries and metadata.
+		timeseries := wr.Timeseries
+		metaData := wr.Metadata
+		n := len(timeseries) / 2
+		m := len(metaData) / 2
+		wr.Timeseries = timeseries[:n]
+		wr.Metadata = metaData[:m]
+		if !tryPushWriteRequest(wr, tryPushBlock, isVMRemoteWrite) {
+			wr.Timeseries = timeseries
+			wr.Metadata = metaData
+			return false
+		}
+		wr.Timeseries = timeseries[n:]
+		wr.Metadata = metaData[m:]
+		if !tryPushWriteRequest(wr, tryPushBlock, isVMRemoteWrite) {
+			wr.Timeseries = timeseries
+			wr.Metadata = metaData
+			return false
+		}
+		wr.Timeseries = timeseries
+		wr.Metadata = metaData
 		return true
 	}
-	timeseries := wr.Timeseries
-	n := len(timeseries) / 2
-	wr.Timeseries = timeseries[:n]
-	if !tryPushWriteRequest(wr, tryPushBlock, isVMRemoteWrite) {
-		wr.Timeseries = timeseries
-		return false
-	}
-	wr.Timeseries = timeseries[n:]
-	if !tryPushWriteRequest(wr, tryPushBlock, isVMRemoteWrite) {
-		wr.Timeseries = timeseries
-		return false
-	}
-	wr.Timeseries = timeseries
-	return true
 }
 
 var (
-	blockSizeBytes = metrics.NewHistogram(`vmagent_remotewrite_block_size_bytes`)
-	blockSizeRows  = metrics.NewHistogram(`vmagent_remotewrite_block_size_rows`)
+	blockSizeBytes    = metrics.NewHistogram(`vmagent_remotewrite_block_size_bytes`)
+	blockSizeRows     = metrics.NewHistogram(`vmagent_remotewrite_block_size_rows`)
+	blockMetadataRows = metrics.NewHistogram(`vmagent_remotewrite_block_metadata_rows`)
 )
 
 var (

app/vmagent/remotewrite/remotewrite.go

@@ -209,7 +209,7 @@ func Init() {
 	// In this case it is impossible to prevent from sending many duplicates of samples passed to TryPush() to all the configured -remoteWrite.url
 	// if these samples couldn't be sent to the -remoteWrite.url with the disabled persistent queue. So it is better sending samples
 	// to the remaining -remoteWrite.url and dropping them on the blocked queue.
-	dropSamplesOnFailureGlobal = *dropSamplesOnOverload || disableOnDiskQueueAny && len(disableOnDiskQueues) > 1
+	dropSamplesOnFailureGlobal = *dropSamplesOnOverload || disableOnDiskQueueAny && len(*remoteWriteURLs) > 1
 
 	dropDanglingQueues()
 
@@ -388,13 +388,7 @@ func TryPush(at *auth.Token, wr *prompb.WriteRequest) bool {
 
 func tryPush(at *auth.Token, wr *prompb.WriteRequest, forceDropSamplesOnFailure bool) bool {
 	tss := wr.Timeseries
-
-	var tenantRctx *relabelCtx
-	if at != nil {
-		// Convert at to (vm_account_id, vm_project_id) labels.
-		tenantRctx = getRelabelCtx()
-		defer putRelabelCtx(tenantRctx)
-	}
+	mms := wr.Metadata
 
 	// Quick check whether writes to configured remote storage systems are blocked.
 	// This allows saving CPU time spent on relabeling and block compression
@@ -411,6 +405,23 @@ func tryPush(at *auth.Token, wr *prompb.WriteRequest, forceDropSamplesOnFailure
 		return true
 	}
 
+	// Push metadata separately from time series, since it doesn't need sharding,
+	// relabeling, stream aggregation, deduplication, etc.
+	if !tryPushMetadataToRemoteStorages(rwctxs, mms, forceDropSamplesOnFailure) {
+		return false
+	}
+
+	if len(tss) == 0 {
+		return true
+	}
+
+	var tenantRctx *relabelCtx
+	if at != nil {
+		// Convert at to (vm_account_id, vm_project_id) labels.
+		tenantRctx = getRelabelCtx()
+		defer putRelabelCtx(tenantRctx)
+	}
+
 	var rctx *relabelCtx
 	rcs := allRelabelConfigs.Load()
 	pcsGlobal := rcs.global
@@ -481,7 +492,7 @@ func tryPush(at *auth.Token, wr *prompb.WriteRequest, forceDropSamplesOnFailure
 			deduplicatorGlobal.Push(tssBlock)
 			tssBlock = tssBlock[:0]
 		}
-		if !tryPushBlockToRemoteStorages(rwctxs, tssBlock, forceDropSamplesOnFailure) {
+		if !tryPushTimeSeriesToRemoteStorages(rwctxs, tssBlock, forceDropSamplesOnFailure) {
 			return false
 		}
 	}
@@ -520,18 +531,49 @@ func getEligibleRemoteWriteCtxs(tss []prompb.TimeSeries, forceDropSamplesOnFailu
 	return rwctxs, true
 }
 
-func pushToRemoteStoragesTrackDropped(tss []prompb.TimeSeries) {
+func pushTimeSeriesToRemoteStoragesTrackDropped(tss []prompb.TimeSeries) {
 	rwctxs, _ := getEligibleRemoteWriteCtxs(tss, true)
 	if len(rwctxs) == 0 {
 		return
 	}
 
-	if !tryPushBlockToRemoteStorages(rwctxs, tss, true) {
-		logger.Panicf("BUG: tryPushBlockToRemoteStorages() must return true when forceDropSamplesOnFailure=true")
+	if !tryPushTimeSeriesToRemoteStorages(rwctxs, tss, true) {
+		logger.Panicf("BUG: tryPushTimeSeriesToRemoteStorages() must return true when forceDropSamplesOnFailure=true")
 	}
 }
 
-func tryPushBlockToRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []prompb.TimeSeries, forceDropSamplesOnFailure bool) bool {
+func tryPushMetadataToRemoteStorages(rwctxs []*remoteWriteCtx, mms []prompb.MetricMetadata, forceDropSamplesOnFailure bool) bool {
+	if len(mms) == 0 {
+		// Nothing to push
+		return true
+	}
+	// 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.
+	//
+	// Push metadata to remote storage systems in parallel to reduce
+	// the time needed for sending the data to multiple remote storage systems.
+	var wg sync.WaitGroup
+	wg.Add(len(rwctxs))
+	var anyPushFailed atomic.Bool
+	for _, rwctx := range rwctxs {
+		go func(rwctx *remoteWriteCtx) {
+			defer wg.Done()
+			if !rwctx.tryPushMetadataInternal(mms) {
+				rwctx.pushFailures.Inc()
+				if forceDropSamplesOnFailure {
+					rwctx.metadataDroppedOnPushFailure.Add(len(mms))
+					return
+				}
+				anyPushFailed.Store(true)
+			}
+		}(rwctx)
+	}
+	wg.Wait()
+	return !anyPushFailed.Load()
+}
+
+func tryPushTimeSeriesToRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []prompb.TimeSeries, forceDropSamplesOnFailure bool) bool {
 	if len(tssBlock) == 0 {
 		// Nothing to push
 		return true
@@ -539,7 +581,7 @@ func tryPushBlockToRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []prompb.Ti
 
 	if len(rwctxs) == 1 {
 		// Fast path - just push data to the configured single remote storage
-		return rwctxs[0].TryPush(tssBlock, forceDropSamplesOnFailure)
+		return rwctxs[0].TryPushTimeSeries(tssBlock, forceDropSamplesOnFailure)
 	}
 
 	// We need to push tssBlock to multiple remote storages.
@@ -550,11 +592,11 @@ func tryPushBlockToRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []prompb.Ti
 		if replicas <= 0 {
 			replicas = 1
 		}
-		return tryShardingBlockAmongRemoteStorages(rwctxs, tssBlock, replicas, forceDropSamplesOnFailure)
+		return tryShardingTimeSeriesAmongRemoteStorages(rwctxs, tssBlock, replicas, forceDropSamplesOnFailure)
 	}
 
 	// Replicate tssBlock samples among rwctxs.
-	// Push tssBlock to remote storage systems in parallel in order to reduce
+	// Push tssBlock to remote storage systems in parallel to reduce
 	// the time needed for sending the data to multiple remote storage systems.
 	var wg sync.WaitGroup
 	wg.Add(len(rwctxs))
@@ -562,7 +604,7 @@ func tryPushBlockToRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []prompb.Ti
 	for _, rwctx := range rwctxs {
 		go func(rwctx *remoteWriteCtx) {
 			defer wg.Done()
-			if !rwctx.TryPush(tssBlock, forceDropSamplesOnFailure) {
+			if !rwctx.TryPushTimeSeries(tssBlock, forceDropSamplesOnFailure) {
 				anyPushFailed.Store(true)
 			}
 		}(rwctx)
@@ -571,7 +613,7 @@ func tryPushBlockToRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []prompb.Ti
 	return !anyPushFailed.Load()
 }
 
-func tryShardingBlockAmongRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []prompb.TimeSeries, replicas int, forceDropSamplesOnFailure bool) bool {
+func tryShardingTimeSeriesAmongRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []prompb.TimeSeries, replicas int, forceDropSamplesOnFailure bool) bool {
 	x := getTSSShards(len(rwctxs))
 	defer putTSSShards(x)
 
@@ -590,7 +632,7 @@ func tryShardingBlockAmongRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []pr
 		wg.Add(1)
 		go func(rwctx *remoteWriteCtx, tss []prompb.TimeSeries) {
 			defer wg.Done()
-			if !rwctx.TryPush(tss, forceDropSamplesOnFailure) {
+			if !rwctx.TryPushTimeSeries(tss, forceDropSamplesOnFailure) {
 				anyPushFailed.Store(true)
 			}
 		}(rwctx, shard)
@@ -797,8 +839,9 @@ type remoteWriteCtx struct {
 	rowsPushedAfterRelabel *metrics.Counter
 	rowsDroppedByRelabel   *metrics.Counter
 
-	pushFailures             *metrics.Counter
-	rowsDroppedOnPushFailure *metrics.Counter
+	pushFailures                 *metrics.Counter
+	metadataDroppedOnPushFailure *metrics.Counter
+	rowsDroppedOnPushFailure     *metrics.Counter
 }
 
 func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, maxInmemoryBlocks int, sanitizedURL string) *remoteWriteCtx {
@@ -862,8 +905,9 @@ func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, maxInmemoryBlocks in
 		rowsPushedAfterRelabel: metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_rows_pushed_after_relabel_total{path=%q,url=%q}`, queuePath, sanitizedURL)),
 		rowsDroppedByRelabel:   metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_relabel_metrics_dropped_total{path=%q,url=%q}`, queuePath, sanitizedURL)),
 
-		pushFailures:             metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_push_failures_total{path=%q,url=%q}`, queuePath, sanitizedURL)),
-		rowsDroppedOnPushFailure: metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_samples_dropped_total{path=%q,url=%q}`, queuePath, sanitizedURL)),
+		pushFailures:                 metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_push_failures_total{path=%q,url=%q}`, queuePath, sanitizedURL)),
+		metadataDroppedOnPushFailure: metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_metadata_dropped_total{path=%q,url=%q}`, queuePath, sanitizedURL)),
+		rowsDroppedOnPushFailure:     metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_samples_dropped_total{path=%q,url=%q}`, queuePath, sanitizedURL)),
 	}
 	rwctx.initStreamAggrConfig()
 
@@ -897,10 +941,10 @@ func (rwctx *remoteWriteCtx) MustStop() {
 	rwctx.rowsDroppedByRelabel = nil
 }
 
-// TryPush sends tss series to the configured remote write endpoint
+// TryPushTimeSeries sends tss series to the configured remote write endpoint
 //
-// TryPush doesn't modify tss, so tss can be passed concurrently to TryPush across distinct rwctx instances.
-func (rwctx *remoteWriteCtx) TryPush(tss []prompb.TimeSeries, forceDropSamplesOnFailure bool) bool {
+// TryPushTimeSeries doesn't modify tss, so tss can be passed concurrently to TryPush across distinct rwctx instances.
+func (rwctx *remoteWriteCtx) TryPushTimeSeries(tss []prompb.TimeSeries, forceDropSamplesOnFailure bool) bool {
 	var rctx *relabelCtx
 	var v *[]prompb.TimeSeries
 	defer func() {
@@ -953,7 +997,7 @@ func (rwctx *remoteWriteCtx) TryPush(tss []prompb.TimeSeries, forceDropSamplesOn
 	}
 
 	// Try pushing tss to remote storage
-	if rwctx.tryPushInternal(tss) {
+	if rwctx.tryPushTimeSeriesInternal(tss) {
 		return true
 	}
 
@@ -985,7 +1029,7 @@ func dropAggregatedSeries(src []prompb.TimeSeries, matchIdxs []byte, dropInput b
 }
 
 func (rwctx *remoteWriteCtx) pushInternalTrackDropped(tss []prompb.TimeSeries) {
-	if rwctx.tryPushInternal(tss) {
+	if rwctx.tryPushTimeSeriesInternal(tss) {
 		return
 	}
 	if !rwctx.fq.IsPersistentQueueDisabled() {
@@ -996,7 +1040,14 @@ func (rwctx *remoteWriteCtx) pushInternalTrackDropped(tss []prompb.TimeSeries) {
 	rwctx.rowsDroppedOnPushFailure.Add(rowsCount)
 }
 
-func (rwctx *remoteWriteCtx) tryPushInternal(tss []prompb.TimeSeries) bool {
+func (rwctx *remoteWriteCtx) tryPushMetadataInternal(mms []prompb.MetricMetadata) bool {
+	pss := rwctx.pss
+	idx := rwctx.pssNextIdx.Add(1) % uint64(len(pss))
+
+	return pss[idx].TryPushMetadata(mms)
+}
+
+func (rwctx *remoteWriteCtx) tryPushTimeSeriesInternal(tss []prompb.TimeSeries) bool {
 	var rctx *relabelCtx
 	var v *[]prompb.TimeSeries
 	defer func() {
@@ -1020,7 +1071,7 @@ func (rwctx *remoteWriteCtx) tryPushInternal(tss []prompb.TimeSeries) bool {
 	pss := rwctx.pss
 	idx := rwctx.pssNextIdx.Add(1) % uint64(len(pss))
 
-	return pss[idx].TryPush(tss)
+	return pss[idx].TryPushTimeSeries(tss)
 }
 
 var tssPool = &sync.Pool{

app/vmagent/remotewrite/remotewrite_test.go

@@ -106,7 +106,7 @@ func TestRemoteWriteContext_TryPush_ImmutableTimeseries(t *testing.T) {
 
 		// copy inputTss to make sure it is not mutated during TryPush call
 		copy(expectedTss, inputTss)
-		if !rwctx.TryPush(inputTss, false) {
+		if !rwctx.TryPushTimeSeries(inputTss, false) {
 			t.Fatalf("cannot push samples to rwctx")
 		}
 

app/vmagent/remotewrite/streamaggr.go

@@ -141,7 +141,7 @@ func initStreamAggrConfigGlobal() {
 	}
 	dedupInterval := *streamAggrGlobalDedupInterval
 	if dedupInterval > 0 {
-		deduplicatorGlobal = streamaggr.NewDeduplicator(pushToRemoteStoragesTrackDropped, *streamAggrGlobalEnableWindows, dedupInterval, *streamAggrGlobalDropInputLabels, "dedup-global")
+		deduplicatorGlobal = streamaggr.NewDeduplicator(pushTimeSeriesToRemoteStoragesTrackDropped, *streamAggrGlobalEnableWindows, dedupInterval, *streamAggrGlobalDropInputLabels, "dedup-global")
 	}
 }
 
@@ -216,7 +216,7 @@ func newStreamAggrConfigGlobal() (*streamaggr.Aggregators, error) {
 		EnableWindows:        *streamAggrGlobalEnableWindows,
 	}
 
-	sas, err := streamaggr.LoadFromFile(path, pushToRemoteStoragesTrackDropped, opts, "global")
+	sas, err := streamaggr.LoadFromFile(path, pushTimeSeriesToRemoteStoragesTrackDropped, opts, "global")
 	if err != nil {
 		return nil, fmt.Errorf("cannot load -streamAggr.config=%q: %w", *streamAggrGlobalConfig, err)
 	}

app/vmalert-tool/unittest/testdata/rules.yaml

@@ -15,6 +15,10 @@ groups:
       - alert: SameAlertNameWithDifferentGroup
         expr: absent(test)
         for: 1m
+      - alert: AlertWithTemplate
+        expr: test
+        annotations:
+          queryAnno: '{{ query "foo" | first | value }}'
 
   - name: group2
     rules:

app/vmalert-tool/unittest/testdata/test1.yaml

@@ -10,7 +10,9 @@ tests:
     input_series:
       - series: "test"
         values: "_x5 1x5 _ stale"
-
+      - series: "foo"
+        values: "1x20"
+  
     alert_rule_test:
       - eval_time: 1m
         groupname: group1
@@ -32,6 +34,14 @@ tests:
         groupname: group1
         alertname: SameAlertNameWithDifferentGroup
         exp_alerts: []
+      - eval_time: 6m
+        groupname: group1
+        alertname: AlertWithTemplate
+        exp_alerts:
+          - exp_labels:
+              cluster: prod
+            exp_annotations:
+              queryAnno: '1'
 
     metricsql_expr_test:
       - expr: test
@@ -50,6 +60,8 @@ tests:
         values: "0+0x1440"
       - series: "test"
         values: "0+1x1440"
+      - series: "foo"
+        values: "1x20"
 
     metricsql_expr_test:
       - expr: count(ALERTS) by (alertgroup, alertname, alertstate)
@@ -59,6 +71,8 @@ tests:
             value: 1
           - labels: '{alertgroup="group1", alertname="InstanceDown", alertstate="pending"}'
             value: 1
+          - labels: '{alertgroup="group1", alertname="AlertWithTemplate", alertstate="firing"}'
+            value: 1
       - expr: t1
         eval_time: 4m
         exp_samples:

app/vmalert-tool/unittest/unittest.go

@@ -366,6 +366,7 @@ func (tg *testGroup) test(evalInterval time.Duration, groupOrderMap map[string]i
 			mergedExternalLabels[k] = v
 		}
 		ng := rule.NewGroup(group, q, time.Minute, mergedExternalLabels)
+		ng.Init()
 		groups = append(groups, ng)
 	}
 

app/vmalert/config/config.go

@@ -295,10 +295,7 @@ func parse(files map[string][]byte, validateTplFn ValidateTplFn, validateExpress
 }
 
 func parseConfig(data []byte) ([]Group, error) {
-	data, err := envtemplate.ReplaceBytes(data)
-	if err != nil {
-		return nil, fmt.Errorf("cannot expand environment vars: %w", err)
-	}
+	data = envtemplate.ReplaceBytes(data)
 
 	var result []Group
 	type cfgFile struct {
@@ -310,13 +307,13 @@ func parseConfig(data []byte) ([]Group, error) {
 	decoder := yaml.NewDecoder(bytes.NewReader(data))
 	for {
 		var cf cfgFile
-		if err = decoder.Decode(&cf); err != nil {
+		if err := decoder.Decode(&cf); err != nil {
 			if err == io.EOF { // EOF indicates no more documents to read
 				break
 			}
 			return nil, err
 		}
-		if err = checkOverflow(cf.XXX, "config"); err != nil {
+		if err := checkOverflow(cf.XXX, "config"); err != nil {
 			return nil, err
 		}
 		result = append(result, cf.Groups...)

app/vmalert/remotewrite/client_test.go

@@ -182,7 +182,7 @@ func (rw *rwServer) handler(w http.ResponseWriter, r *http.Request) {
 		rw.err(w, fmt.Errorf("decode err: %w", err))
 		return
 	}
-	wru := &prompb.WriteRequestUnmarshaller{}
+	wru := &prompb.WriteRequestUnmarshaler{}
 	wr, err := wru.UnmarshalProtobuf(b)
 	if err != nil {
 		rw.err(w, fmt.Errorf("unmarhsal err: %w", err))

app/vmalert/replay.go

@@ -28,8 +28,8 @@ var (
 		"Defines how many retries to make before giving up on rule if request for it returns an error.")
 	disableProgressBar = flag.Bool("replay.disableProgressBar", false, "Whether to disable rendering progress bars during the replay. "+
 		"Progress bar rendering might be verbose or break the logs parsing, so it is recommended to be disabled when not used in interactive mode.")
-	ruleEvaluationConcurrency = flag.Int("replay.ruleEvaluationConcurrency", 1, "The maximum number of concurrent `/query_range` requests for a single rule. "+
-		"Increasing this value when replaying for a long time and a single request range is limited by `-replay.maxDatapointsPerQuery`.")
+	ruleEvaluationConcurrency = flag.Int("replay.ruleEvaluationConcurrency", 1, "The maximum number of concurrent '/query_range' requests when replay recording rule or alerting rule with for=0. "+
+		"Increasing this value when replaying for a long time, since each request is limited by -replay.maxDatapointsPerQuery.")
 )
 
 func replay(groupsCfg []config.Group, qb datasource.QuerierBuilder, rw remotewrite.RWClient) (totalRows, droppedRows int, err error) {

app/vmalert/replay_test.go

@@ -246,24 +246,33 @@ func TestReplay(t *testing.T) {
 
 	// multiple rules + rule concurrency + group concurrency
 	f("2021-01-01T12:00:00.000Z", "2021-01-01T12:02:30.000Z", 1, 3, 0, []config.Group{
-		{Rules: []config.Rule{{Alert: "foo-group-single-concurrent", Expr: "sum(up) > 1"}, {Alert: "bar-group-single-concurrent", Expr: "max(up) < 1"}}, Concurrency: 2}}, &fakeReplayQuerier{
+		{Rules: []config.Rule{{Alert: "foo-group-single-concurrent", For: promutil.NewDuration(30 * time.Second), Expr: "sum(up) > 1"}, {Alert: "bar-group-single-concurrent", Expr: "max(up) < 1"}}, Concurrency: 2}}, &fakeReplayQuerier{
 		registry: map[string]map[string][]datasource.Metric{
 			"sum(up) > 1": {
-				"12:00:00+12:01:00": {},
-				"12:01:00+12:02:00": {{
-					Timestamps: []int64{1},
+				"12:00:00+12:01:00": {{
+					Timestamps: []int64{1609502460},
+					Values:     []float64{1},
+				}},
+				"12:01:00+12:02:00": {{
+					Timestamps: []int64{1609502520},
+					Values:     []float64{1},
+				}},
+				"12:02:00+12:02:30": {{
+					Timestamps: []int64{1609502580},
 					Values:     []float64{1},
 				}},
-				"12:02:00+12:02:30": {},
 			},
 			"max(up) < 1": {
-				"12:00:00+12:01:00": {},
+				"12:00:00+12:01:00": {{
+					Timestamps: []int64{1609502460},
+					Values:     []float64{1},
+				}},
 				"12:01:00+12:02:00": {{
-					Timestamps: []int64{1},
+					Timestamps: []int64{1609502520},
 					Values:     []float64{1},
 				}},
 				"12:02:00+12:02:30": {},
 			},
 		},
-	}, 4)
+	}, 10)
 }

app/vmalert/rule/alerting.go

@@ -341,11 +341,15 @@ func (ar *AlertingRule) execRange(ctx context.Context, start, end time.Time) ([]
 		return []datasource.Metric{{Timestamps: []int64{0}, Values: []float64{math.NaN()}}}, nil
 	}
 	for _, s := range res.Data {
-		ls, as, err := ar.expandTemplates(s, qFn, time.Time{})
+		ls, err := ar.expandLabelTemplates(s)
 		if err != nil {
-			return nil, fmt.Errorf("failed to expand templates: %s", err)
+			return nil, err
 		}
 		alertID := hash(ls.processed)
+		as, err := ar.expandAnnotationTemplates(s, qFn, time.Time{}, ls)
+		if err != nil {
+			return nil, err
+		}
 		a := ar.newAlert(s, time.Time{}, ls.processed, as) // initial alert
 
 		prevT := time.Time{}
@@ -363,7 +367,7 @@ func (ar *AlertingRule) execRange(ctx context.Context, start, end time.Time) ([]
 				a.State = notifier.StatePending
 				a.ActiveAt = at
 				// re-template the annotations as active timestamp is changed
-				_, a.Annotations, _ = ar.expandTemplates(s, qFn, at)
+				a.Annotations, _ = ar.expandAnnotationTemplates(s, qFn, at, ls)
 				a.Start = time.Time{}
 			} else if at.Sub(a.ActiveAt) >= ar.For && a.State != notifier.StateFiring {
 				a.State = notifier.StateFiring
@@ -376,13 +380,15 @@ func (ar *AlertingRule) execRange(ctx context.Context, start, end time.Time) ([]
 			}
 			result = append(result, ar.alertToTimeSeries(a, s.Timestamps[i])...)
 
-			// save alert's state on last iteration, so it can be used on the next execRange call
-			if at.Equal(end) {
+			// if for>0, save alert's state on last iteration, so it can be used on the next execRange call
+			if ar.For > 0 && at.Equal(end) {
 				holdAlertState[alertID] = a
 			}
 		}
 	}
-	ar.alerts = holdAlertState
+	if len(holdAlertState) > 0 {
+		ar.alerts = holdAlertState
+	}
 	return result, nil
 }
 
@@ -428,9 +434,22 @@ func (ar *AlertingRule) exec(ctx context.Context, ts time.Time, limit int) ([]pr
 	expandedLabels := make([]*labelSet, len(res.Data))
 	expandedAnnotations := make([]map[string]string, len(res.Data))
 	for i, m := range res.Data {
-		ls, as, err := ar.expandTemplates(m, qFn, ts)
+		ls, err := ar.expandLabelTemplates(m)
 		if err != nil {
-			curState.Err = fmt.Errorf("failed to expand templates: %w", err)
+			curState.Err = err
+			return nil, curState.Err
+		}
+		at := ts
+		alertID := hash(ls.processed)
+		if a, ok := ar.alerts[alertID]; ok {
+			// modify activeAt for annotation templating if the alert has already triggered(in state Pending or Firing)
+			if a.State != notifier.StateInactive {
+				at = a.ActiveAt
+			}
+		}
+		as, err := ar.expandAnnotationTemplates(m, qFn, at, ls)
+		if err != nil {
+			curState.Err = err
 			return nil, curState.Err
 		}
 		expandedLabels[i] = ls
@@ -473,6 +492,7 @@ func (ar *AlertingRule) exec(ctx context.Context, ts time.Time, limit int) ([]pr
 			a.KeepFiringSince = time.Time{}
 			continue
 		}
+
 		a := ar.newAlert(m, ts, labels.processed, annotations)
 		a.ID = alertID
 		a.State = notifier.StatePending
@@ -536,12 +556,18 @@ func (ar *AlertingRule) exec(ctx context.Context, ts time.Time, limit int) ([]pr
 	return append(tss, ar.toTimeSeries(ts.Unix())...), nil
 }
 
-func (ar *AlertingRule) expandTemplates(m datasource.Metric, qFn templates.QueryFn, ts time.Time) (*labelSet, map[string]string, error) {
+func (ar *AlertingRule) expandLabelTemplates(m datasource.Metric) (*labelSet, error) {
+	qFn := func(_ string) ([]datasource.Metric, error) {
+		return nil, fmt.Errorf("`query` template isn't supported in rule label")
+	}
 	ls, err := ar.toLabels(m, qFn)
 	if err != nil {
-		return nil, nil, fmt.Errorf("failed to expand labels: %w", err)
+		return nil, fmt.Errorf("failed to expand label templates: %s", err)
 	}
+	return ls, nil
+}
 
+func (ar *AlertingRule) expandAnnotationTemplates(m datasource.Metric, qFn templates.QueryFn, activeAt time.Time, ls *labelSet) (map[string]string, error) {
 	tplData := notifier.AlertTplData{
 		Value:    m.Values[0],
 		Type:     ar.Type.String(),
@@ -549,14 +575,14 @@ func (ar *AlertingRule) expandTemplates(m datasource.Metric, qFn templates.Query
 		Expr:     ar.Expr,
 		AlertID:  hash(ls.processed),
 		GroupID:  ar.GroupID,
-		ActiveAt: ts,
+		ActiveAt: activeAt,
 		For:      ar.For,
 	}
 	as, err := notifier.ExecTemplate(qFn, ar.Annotations, tplData)
 	if err != nil {
-		return nil, nil, fmt.Errorf("failed to template annotations: %w", err)
+		return nil, fmt.Errorf("failed to expand annotation templates: %s", err)
 	}
-	return ls, as, nil
+	return as, nil
 }
 
 // toTimeSeries creates `ALERTS` and `ALERTS_FOR_STATE` for active alerts

app/vmalert/rule/alerting_test.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"reflect"
 	"sort"
+	"strconv"
 	"strings"
 	"sync"
 	"testing"
@@ -267,8 +268,15 @@ func TestAlertingRule_Exec(t *testing.T) {
 				if got.State != exp.State {
 					t.Fatalf("evalIndex %d: expected state %d; got %d", i, exp.State, got.State)
 				}
+				if rule.Annotations != nil && exp.Annotations != nil {
+					if !reflect.DeepEqual(got.Annotations, exp.Annotations) {
+						t.Fatalf("evalIndex %d: expected annotations %v; got %v", i, exp.Annotations, got.Annotations)
+					}
+				}
 			}
 		}
+		// reset ts for next test
+		ts, _ = time.Parse(time.RFC3339, "2024-10-29T00:00:00Z")
 	}
 
 	f(newTestAlertingRule("empty", 0), [][]datasource.Metric{}, nil, nil)
@@ -522,7 +530,7 @@ func TestAlertingRule_Exec(t *testing.T) {
 		},
 	})
 
-	f(newTestAlertingRule("for-pending=>firing=>inactive=>pending=>firing", defaultStep), [][]datasource.Metric{
+	f(newTestAlertingRuleWithCustomFields("for-pending=>firing=>inactive=>pending=>firing", defaultStep, 0, 0, map[string]string{"activeAt": "{{ $activeAt.UnixMilli }}"}), [][]datasource.Metric{
 		{metricWithLabels(t, "name", "foo")},
 		{metricWithLabels(t, "name", "foo")},
 		// empty step to set alert inactive
@@ -530,11 +538,11 @@ func TestAlertingRule_Exec(t *testing.T) {
 		{metricWithLabels(t, "name", "foo")},
 		{metricWithLabels(t, "name", "foo")},
 	}, map[int][]testAlert{
-		0: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
-		1: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
-		2: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}}},
-		3: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
-		4: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
+		0: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending, Annotations: map[string]string{"activeAt": strconv.FormatInt(ts.UnixMilli(), 10)}}}},
+		1: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring, Annotations: map[string]string{"activeAt": strconv.FormatInt(ts.UnixMilli(), 10)}}}},
+		2: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive, Annotations: map[string]string{"activeAt": strconv.FormatInt(ts.UnixMilli(), 10)}}}},
+		3: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending, Annotations: map[string]string{"activeAt": strconv.FormatInt(ts.Add(defaultStep*3).UnixMilli(), 10)}}}},
+		4: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring, Annotations: map[string]string{"activeAt": strconv.FormatInt(ts.Add(defaultStep*3).UnixMilli(), 10)}}}},
 	}, nil)
 
 	f(newTestAlertingRuleWithCustomFields("for-pending=>firing=>keepfiring=>firing", defaultStep, 0, defaultStep, nil), [][]datasource.Metric{

app/vmalert/rule/group.go

@@ -587,6 +587,11 @@ func (g *Group) Replay(start, end time.Time, rw remotewrite.RWClient, maxDataPoi
 
 func replayRuleRange(r Rule, ri rangeIterator, bar *pb.ProgressBar, rw remotewrite.RWClient, replayRuleRetryAttempts, ruleEvaluationConcurrency int) int {
 	fmt.Printf("> Rule %q (ID: %d)\n", r, r.ID())
+	// alerting rule with for>0 can't be replayed concurrently, since the status change might depend on the previous evaluation
+	// see https://github.com/VictoriaMetrics/VictoriaMetrics/commit/abcb21aa5ee918ba9a4e9cde495dba06e1e9564c
+	if r, ok := r.(*AlertingRule); ok && r.For > 0 {
+		ruleEvaluationConcurrency = 1
+	}
 	sem := make(chan struct{}, ruleEvaluationConcurrency)
 	wg := sync.WaitGroup{}
 	res := make(chan int, int(ri.end.Sub(ri.start)/ri.step)+1)

app/vmalert/rule/recording_test.go

@@ -437,7 +437,7 @@ func TestRecordingRuleExec_Negative(t *testing.T) {
 
 	_, err = rr.exec(context.TODO(), time.Now(), 0)
 	if err != nil {
-		t.Fatalf("cannot execute recroding rule: %s", err)
+		t.Fatalf("cannot execute recording rule: %s", err)
 	}
 }
 

app/vmauth/auth_config.go

@@ -723,14 +723,11 @@ func reloadAuthConfigData(data []byte) (bool, error) {
 }
 
 func parseAuthConfig(data []byte) (*AuthConfig, error) {
-	data, err := envtemplate.ReplaceBytes(data)
-	if err != nil {
-		return nil, fmt.Errorf("cannot expand environment vars: %w", err)
-	}
+	data = envtemplate.ReplaceBytes(data)
 	ac := &AuthConfig{
 		ms: metrics.NewSet(),
 	}
-	if err = yaml.UnmarshalStrict(data, ac); err != nil {
+	if err := yaml.UnmarshalStrict(data, ac); err != nil {
 		return nil, fmt.Errorf("cannot unmarshal AuthConfig data: %w", err)
 	}
 

app/vmbackup/Makefile

@@ -1,106 +1,110 @@
 # All these commands must run from repository root.
 
+# special tag to reduce resulting binary size
+# See this issue https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8008
+VMBACKUP_GO_BUILD_TAGS=disable_grpc_modules
+
 vmbackup:
-	APP_NAME=vmbackup $(MAKE) app-local
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-local
 
 vmbackup-race:
-	APP_NAME=vmbackup RACE=-race $(MAKE) app-local
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) RACE=-race $(MAKE) app-local
 
 vmbackup-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker
 
 vmbackup-pure-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-pure
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-pure
 
 vmbackup-linux-amd64-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-linux-amd64
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-linux-amd64
 
 vmbackup-linux-arm-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-linux-arm
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-linux-arm
 
 vmbackup-linux-arm64-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-linux-arm64
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-linux-arm64
 
 vmbackup-linux-ppc64le-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-linux-ppc64le
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-linux-ppc64le
 
 vmbackup-linux-386-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-linux-386
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-linux-386
 
 vmbackup-darwin-amd64-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-darwin-amd64
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-darwin-amd64
 
 vmbackup-darwin-arm64-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-darwin-arm64
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-darwin-arm64
 
 vmbackup-freebsd-amd64-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-freebsd-amd64
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-freebsd-amd64
 
 vmbackup-openbsd-amd64-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-openbsd-amd64
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-openbsd-amd64
 
 vmbackup-windows-amd64-prod:
-	APP_NAME=vmbackup $(MAKE) app-via-docker-windows-amd64
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-via-docker-windows-amd64
 
 package-vmbackup:
-	APP_NAME=vmbackup $(MAKE) package-via-docker
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) package-via-docker
 
 package-vmbackup-pure:
-	APP_NAME=vmbackup $(MAKE) package-via-docker-pure
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) package-via-docker-pure
 
 package-vmbackup-amd64:
-	APP_NAME=vmbackup $(MAKE) package-via-docker-amd64
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) package-via-docker-amd64
 
 package-vmbackup-arm:
-	APP_NAME=vmbackup $(MAKE) package-via-docker-arm
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) package-via-docker-arm
 
 package-vmbackup-arm64:
-	APP_NAME=vmbackup $(MAKE) package-via-docker-arm64
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) package-via-docker-arm64
 
 package-vmbackup-ppc64le:
-	APP_NAME=vmbackup $(MAKE) package-via-docker-ppc64le
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) package-via-docker-ppc64le
 
 package-vmbackup-386:
-	APP_NAME=vmbackup $(MAKE) package-via-docker-386
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) package-via-docker-386
 
 publish-vmbackup:
-	APP_NAME=vmbackup $(MAKE) publish-via-docker
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) publish-via-docker
 
 vmbackup-linux-amd64:
-	APP_NAME=vmbackup CGO_ENABLED=1 GOOS=linux GOARCH=amd64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=1 GOOS=linux GOARCH=amd64 $(MAKE) app-local-goos-goarch
 
 vmbackup-linux-arm:
-	APP_NAME=vmbackup CGO_ENABLED=0 GOOS=linux GOARCH=arm $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=arm $(MAKE) app-local-goos-goarch
 
 vmbackup-linux-arm64:
-	APP_NAME=vmbackup CGO_ENABLED=0 GOOS=linux GOARCH=arm64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=arm64 $(MAKE) app-local-goos-goarch
 
 vmbackup-linux-ppc64le:
-	APP_NAME=vmbackup CGO_ENABLED=0 GOOS=linux GOARCH=ppc64le $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=ppc64le $(MAKE) app-local-goos-goarch
 
 vmbackup-linux-s390x:
-	APP_NAME=vmbackup CGO_ENABLED=0 GOOS=linux GOARCH=s390x $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=s390x $(MAKE) app-local-goos-goarch
 
 vmbackup-linux-loong64:
-	APP_NAME=vmbackup CGO_ENABLED=0 GOOS=linux GOARCH=loong64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=loong64 $(MAKE) app-local-goos-goarch
 
 vmbackup-linux-386:
-	APP_NAME=vmbackup CGO_ENABLED=0 GOOS=linux GOARCH=386 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=386 $(MAKE) app-local-goos-goarch
 
 vmbackup-darwin-amd64:
-	APP_NAME=vmbackup CGO_ENABLED=0 GOOS=darwin GOARCH=amd64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=darwin GOARCH=amd64 $(MAKE) app-local-goos-goarch
 
 vmbackup-darwin-arm64:
-	APP_NAME=vmbackup CGO_ENABLED=0 GOOS=darwin GOARCH=arm64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=darwin GOARCH=arm64 $(MAKE) app-local-goos-goarch
 
 vmbackup-freebsd-amd64:
-	APP_NAME=vmbackup CGO_ENABLED=0 GOOS=freebsd GOARCH=amd64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=freebsd GOARCH=amd64 $(MAKE) app-local-goos-goarch
 
 vmbackup-openbsd-amd64:
-	APP_NAME=vmbackup CGO_ENABLED=0 GOOS=openbsd GOARCH=amd64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=openbsd GOARCH=amd64 $(MAKE) app-local-goos-goarch
 
 vmbackup-windows-amd64:
-	GOARCH=amd64 APP_NAME=vmbackup $(MAKE) app-local-windows-goarch
+	GOARCH=amd64 APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-local-windows-goarch
 
 vmbackup-pure:
-	APP_NAME=vmbackup $(MAKE) app-local-pure
+	APP_NAME=vmbackup EXTRA_GO_BUILD_TAGS=$(VMBACKUP_GO_BUILD_TAGS) $(MAKE) app-local-pure

app/vmctl/vm_native.go

@@ -121,7 +121,7 @@ func (p *vmNativeProcessor) runSingle(ctx context.Context, f native.Filter, srcU
 		pr := bar.NewProxyReader(reader)
 		if pr != nil {
 			reader = pr
-			fmt.Printf("Continue import process with filter %s:\n", f.String())
+			fmt.Fprintf(log.Writer(), "Continue import process with filter %s:\n", f.String())
 		}
 	}
 
@@ -191,7 +191,7 @@ func (p *vmNativeProcessor) runBackfilling(ctx context.Context, tenantID string,
 		initParams = []any{srcURL, dstURL, p.filter.String(), tenantID}
 	}
 
-	fmt.Println("") // extra line for better output formatting
+	fmt.Fprintln(log.Writer(), "") // extra line for better output formatting
 	log.Printf(initMessage, initParams...)
 	if len(ranges) > 1 {
 		log.Printf("Selected time range will be split into %d ranges according to %q step", len(ranges), p.filter.Chunk)

app/vminsert/opentelemetry/request_handler.go

@@ -33,7 +33,7 @@ func InsertHandler(req *http.Request) error {
 			return fmt.Errorf("json encoding isn't supported for opentelemetry format. Use protobuf encoding")
 		}
 	}
-	return stream.ParseStream(req.Body, encoding, processBody, func(tss []prompb.TimeSeries) error {
+	return stream.ParseStream(req.Body, encoding, processBody, func(tss []prompb.TimeSeries, _ []prompb.MetricMetadata) error {
 		return insertRows(tss, extraLabels)
 	})
 }

app/vminsert/prometheusimport/request_handler.go

@@ -7,6 +7,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vminsert/relabel"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/prometheus"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/prometheus/stream"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
@@ -29,7 +30,7 @@ func InsertHandler(req *http.Request) error {
 		return err
 	}
 	encoding := req.Header.Get("Content-Encoding")
-	return stream.Parse(req.Body, defaultTimestamp, encoding, true, func(rows []prometheus.Row) error {
+	return stream.Parse(req.Body, defaultTimestamp, encoding, true, promscrape.IsMetadataEnabled(), func(rows []prometheus.Row, _ []prometheus.Metadata) error {
 		return insertRows(rows, extraLabels)
 	}, func(s string) {
 		httpserver.LogError(req, s)

app/vminsert/promremotewrite/request_handler.go

@@ -23,7 +23,7 @@ func InsertHandler(req *http.Request) error {
 		return err
 	}
 	isVMRemoteWrite := req.Header.Get("Content-Encoding") == "zstd"
-	return stream.Parse(req.Body, isVMRemoteWrite, func(tss []prompb.TimeSeries) error {
+	return stream.Parse(req.Body, isVMRemoteWrite, func(tss []prompb.TimeSeries, _ []prompb.MetricMetadata) error {
 		return insertRows(tss, extraLabels)
 	})
 }

app/vmrestore/Makefile

@@ -1,106 +1,110 @@
 # All these commands must run from repository root.
 
+# special tag to reduce resulting binary size
+# See this issue https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8008
+VMRESTORE_GO_BUILD_TAGS=disable_grpc_modules
+
 vmrestore:
-	APP_NAME=vmrestore $(MAKE) app-local
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-local
 
 vmrestore-race:
-	APP_NAME=vmrestore RACE=-race $(MAKE) app-local
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) RACE=-race $(MAKE) app-local
 
 vmrestore-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker
 
 vmrestore-pure-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-pure
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-pure
 
 vmrestore-linux-amd64-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-linux-amd64
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-linux-amd64
 
 vmrestore-linux-arm-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-linux-arm
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-linux-arm
 
 vmrestore-linux-arm64-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-linux-arm64
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-linux-arm64
 
 vmrestore-linux-ppc64le-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-linux-ppc64le
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-linux-ppc64le
 
 vmrestore-linux-386-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-linux-386
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-linux-386
 
 vmrestore-darwin-amd64-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-darwin-amd64
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-darwin-amd64
 
 vmrestore-darwin-arm64-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-darwin-arm64
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-darwin-arm64
 
 vmrestore-freebsd-amd64-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-freebsd-amd64
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-freebsd-amd64
 
 vmrestore-openbsd-amd64-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-openbsd-amd64
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-openbsd-amd64
 
 vmrestore-windows-amd64-prod:
-	APP_NAME=vmrestore $(MAKE) app-via-docker-windows-amd64
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-via-docker-windows-amd64
 
 package-vmrestore:
-	APP_NAME=vmrestore $(MAKE) package-via-docker
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) package-via-docker
 
 package-vmrestore-pure:
-	APP_NAME=vmrestore $(MAKE) package-via-docker-pure
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) package-via-docker-pure
 
 package-vmrestore-amd64:
-	APP_NAME=vmrestore $(MAKE) package-via-docker-amd64
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) package-via-docker-amd64
 
 package-vmrestore-arm:
-	APP_NAME=vmrestore $(MAKE) package-via-docker-arm
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) package-via-docker-arm
 
 package-vmrestore-arm64:
-	APP_NAME=vmrestore $(MAKE) package-via-docker-arm64
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) package-via-docker-arm64
 
 package-vmrestore-ppc64le:
-	APP_NAME=vmrestore $(MAKE) package-via-docker-ppc64le
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) package-via-docker-ppc64le
 
 package-vmrestore-386:
-	APP_NAME=vmrestore $(MAKE) package-via-docker-386
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) package-via-docker-386
 
 publish-vmrestore:
-	APP_NAME=vmrestore $(MAKE) publish-via-docker
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) publish-via-docker
 
 vmrestore-linux-amd64:
-	APP_NAME=vmrestore CGO_ENABLED=1 GOOS=linux GOARCH=amd64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=1 GOOS=linux GOARCH=amd64 $(MAKE) app-local-goos-goarch
 
 vmrestore-linux-arm:
-	APP_NAME=vmrestore CGO_ENABLED=0 GOOS=linux GOARCH=arm $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=arm $(MAKE) app-local-goos-goarch
 
 vmrestore-linux-arm64:
-	APP_NAME=vmrestore CGO_ENABLED=0 GOOS=linux GOARCH=arm64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=arm64 $(MAKE) app-local-goos-goarch
 
 vmrestore-linux-ppc64le:
-	APP_NAME=vmrestore CGO_ENABLED=0 GOOS=linux GOARCH=ppc64le $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=ppc64le $(MAKE) app-local-goos-goarch
 
 vmrestore-linux-s390x:
-	APP_NAME=vmrestore CGO_ENABLED=0 GOOS=linux GOARCH=s390x $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=s390x $(MAKE) app-local-goos-goarch
 
 vmrestore-linux-loong64:
-	APP_NAME=vmrestore CGO_ENABLED=0 GOOS=linux GOARCH=loong64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=loong64 $(MAKE) app-local-goos-goarch
 
 vmrestore-linux-386:
-	APP_NAME=vmrestore CGO_ENABLED=0 GOOS=linux GOARCH=386 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=linux GOARCH=386 $(MAKE) app-local-goos-goarch
 
 vmrestore-darwin-amd64:
-	APP_NAME=vmrestore CGO_ENABLED=0 GOOS=darwin GOARCH=amd64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=darwin GOARCH=amd64 $(MAKE) app-local-goos-goarch
 
 vmrestore-darwin-arm64:
-	APP_NAME=vmrestore CGO_ENABLED=0 GOOS=darwin GOARCH=arm64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=darwin GOARCH=arm64 $(MAKE) app-local-goos-goarch
 
 vmrestore-freebsd-amd64:
-	APP_NAME=vmrestore CGO_ENABLED=0 GOOS=freebsd GOARCH=amd64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=freebsd GOARCH=amd64 $(MAKE) app-local-goos-goarch
 
 vmrestore-openbsd-amd64:
-	APP_NAME=vmrestore CGO_ENABLED=0 GOOS=openbsd GOARCH=amd64 $(MAKE) app-local-goos-goarch
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) CGO_ENABLED=0 GOOS=openbsd GOARCH=amd64 $(MAKE) app-local-goos-goarch
 
 vmrestore-windows-amd64:
-	GOARCH=amd64 APP_NAME=vmrestore $(MAKE) app-local-windows-goarch
+	GOARCH=amd64 APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-local-windows-goarch
 
 vmrestore-pure:
-	APP_NAME=vmrestore $(MAKE) app-local-pure
+	APP_NAME=vmrestore EXTRA_GO_BUILD_TAGS=$(VMRESTORE_GO_BUILD_TAGS) $(MAKE) app-local-pure

app/vmselect/graphite/eval.go

@@ -17,6 +17,9 @@ import (
 var maxGraphiteSeries = flag.Int("search.maxGraphiteSeries", 300e3, "The maximum number of time series, which can be scanned during queries to Graphite Render API. "+
 	"See https://docs.victoriametrics.com/victoriametrics/integrations/graphite/#render-api")
 
+var maxGraphitePathExpressionLen = flag.Int("search.maxGraphitePathExpressionLen", 1024, "The maximum length of pathExpression field in Graphite series. "+
+	"Longer expressions are truncated to prevent memory exhaustion on complex nested queries. Set to 0 to disable truncation.")
+
 type evalConfig struct {
 	startTime   int64
 	endTime     int64
@@ -53,6 +56,21 @@ func (ec *evalConfig) newTimestamps(step int64) []int64 {
 	return timestamps
 }
 
+// safePathExpression creates a pathExpression string from the given expression,
+// truncating it if it exceeds the maximum allowed length to prevent memory exhaustion.
+func safePathExpression(expr graphiteql.Expr) string {
+	if expr == nil {
+		return ""
+	}
+
+	pathExpr := string(expr.AppendString(nil))
+	maxLen := *maxGraphitePathExpressionLen
+	if maxLen > 0 && len(pathExpr) > maxLen {
+		return pathExpr[:maxLen] + "..."
+	}
+	return pathExpr
+}
+
 type series struct {
 	Name       string
 	Tags       map[string]string
@@ -169,7 +187,7 @@ func newNextSeriesForSearchQuery(ec *evalConfig, sq *storage.SearchQuery, expr g
 				Timestamps:     append([]int64{}, rs.Timestamps...),
 				Values:         append([]float64{}, rs.Values...),
 				expr:           expr,
-				pathExpression: string(expr.AppendString(nil)),
+				pathExpression: safePathExpression(expr),
 			}
 			s.summarize(aggrAvg, ec.startTime, ec.endTime, ec.storageStep, 0)
 			t := timerpool.Get(30 * time.Second)

app/vmselect/graphite/eval_test.go

@@ -4178,3 +4178,170 @@ func formatTimestamps(tss []int64) string {
 	fmt.Fprintf(&sb, " ]")
 	return sb.String()
 }
+
+func TestSafePathExpression(t *testing.T) {
+	// Save original value and restore after test
+	originalMaxLen := *maxGraphitePathExpressionLen
+	defer func() {
+		*maxGraphitePathExpressionLen = originalMaxLen
+	}()
+
+	t.Run("nil expression", func(t *testing.T) {
+		result := safePathExpression(nil)
+		if result != "" {
+			t.Fatalf("expected empty string for nil expression, got: %q", result)
+		}
+	})
+
+	t.Run("short expression - no truncation", func(t *testing.T) {
+		*maxGraphitePathExpressionLen = 50
+		expr := &graphiteql.MetricExpr{Query: "metric.cpu.usage"}
+		result := safePathExpression(expr)
+		expected := "metric.cpu.usage"
+		if result != expected {
+			t.Fatalf("expected %q, got %q", expected, result)
+		}
+	})
+
+	t.Run("long expression - with truncation", func(t *testing.T) {
+		*maxGraphitePathExpressionLen = 20
+		longQuery := "vertica.metrics.fr4.verticamultitenant-eon.request_resource_consumption.very_long_metric_name"
+		expr := &graphiteql.MetricExpr{Query: longQuery}
+		result := safePathExpression(expr)
+		expectedPrefix := longQuery[:20]
+		expectedSuffix := "..."
+		expected := expectedPrefix + expectedSuffix
+
+		if result != expected {
+			t.Fatalf("expected %q, got %q", expected, result)
+		}
+		if len(result) != 23 { // 20 + 3 for "..."
+			t.Fatalf("expected result length 23, got %d", len(result))
+		}
+		if !strings.HasSuffix(result, "...") {
+			t.Fatalf("expected result to end with '...', got %q", result)
+		}
+	})
+
+	t.Run("truncation disabled", func(t *testing.T) {
+		*maxGraphitePathExpressionLen = 0 // Disable truncation
+		longQuery := "very.long.metric.name.that.would.normally.be.truncated.but.should.not.be"
+		expr := &graphiteql.MetricExpr{Query: longQuery}
+		result := safePathExpression(expr)
+
+		if result != longQuery {
+			t.Fatalf("expected full string %q when truncation disabled, got %q", longQuery, result)
+		}
+	})
+
+	t.Run("function expression", func(t *testing.T) {
+		*maxGraphitePathExpressionLen = 30
+		// Create a function expression: sum(metric.cpu.usage)
+		args := []*graphiteql.ArgExpr{
+			{Expr: &graphiteql.MetricExpr{Query: "metric.cpu.usage"}},
+		}
+		funcExpr := &graphiteql.FuncExpr{
+			FuncName: "sum",
+			Args:     args,
+		}
+
+		result := safePathExpression(funcExpr)
+		expected := "sum(metric.cpu.usage)"
+
+		if result != expected {
+			t.Fatalf("expected %q, got %q", expected, result)
+		}
+	})
+
+	t.Run("complex nested function - truncated", func(t *testing.T) {
+		*maxGraphitePathExpressionLen = 15
+		// Create nested functions: sum(avg(metric.cpu.usage))
+		innerArgs := []*graphiteql.ArgExpr{
+			{Expr: &graphiteql.MetricExpr{Query: "metric.cpu.usage"}},
+		}
+		innerFunc := &graphiteql.FuncExpr{
+			FuncName: "avg",
+			Args:     innerArgs,
+		}
+		outerArgs := []*graphiteql.ArgExpr{
+			{Expr: innerFunc},
+		}
+		outerFunc := &graphiteql.FuncExpr{
+			FuncName: "sum",
+			Args:     outerArgs,
+		}
+
+		result := safePathExpression(outerFunc)
+
+		if len(result) != 18 { // 15 + 3 for "..."
+			t.Fatalf("expected result length 18, got %d", len(result))
+		}
+		if !strings.HasSuffix(result, "...") {
+			t.Fatalf("expected result to end with '...', got %q", result)
+		}
+		if !strings.HasPrefix(result, "sum(avg(metric") {
+			t.Fatalf("expected result to start with 'sum(avg(metric', got %q", result)
+		}
+	})
+
+	t.Run("boundary case - exact length", func(t *testing.T) {
+		*maxGraphitePathExpressionLen = 10
+		expr := &graphiteql.MetricExpr{Query: "metric.cpu"} // Exactly 10 characters
+		result := safePathExpression(expr)
+		expected := "metric.cpu"
+
+		if result != expected {
+			t.Fatalf("expected %q, got %q", expected, result)
+		}
+	})
+
+	t.Run("boundary case - one character over", func(t *testing.T) {
+		*maxGraphitePathExpressionLen = 10
+		expr := &graphiteql.MetricExpr{Query: "metric.cpu.x"} // 11 characters
+		result := safePathExpression(expr)
+		expected := "metric.cpu..."
+
+		if result != expected {
+			t.Fatalf("expected %q, got %q", expected, result)
+		}
+	})
+}
+
+func TestSafePathExpressionFromString(t *testing.T) {
+	// Save original value and restore after test
+	originalMaxLen := *maxGraphitePathExpressionLen
+	defer func() {
+		*maxGraphitePathExpressionLen = originalMaxLen
+	}()
+
+	t.Run("short string - no truncation", func(t *testing.T) {
+		*maxGraphitePathExpressionLen = 50
+		input := "sumSeries(metric1,metric2)"
+		result := safePathExpressionFromString(input)
+
+		if result != input {
+			t.Fatalf("expected %q, got %q", input, result)
+		}
+	})
+
+	t.Run("long string - with truncation", func(t *testing.T) {
+		*maxGraphitePathExpressionLen = 20
+		input := "sumSeries(very.long.metric.name.that.exceeds.limit,another.metric)"
+		result := safePathExpressionFromString(input)
+		expected := "sumSeries(very.long...."
+
+		if result != expected {
+			t.Fatalf("expected %q, got %q", expected, result)
+		}
+	})
+
+	t.Run("truncation disabled", func(t *testing.T) {
+		*maxGraphitePathExpressionLen = 0
+		input := "very.long.string.that.would.normally.be.truncated"
+		result := safePathExpressionFromString(input)
+
+		if result != input {
+			t.Fatalf("expected full string when truncation disabled, got %q", result)
+		}
+	})
+}

app/vmselect/graphite/render_api_test.go

@@ -10,7 +10,7 @@ func TestParseIntervalSuccess(t *testing.T) {
 		t.Helper()
 		interval, err := parseInterval(s)
 		if err != nil {
-			t.Fatalf("unexpected error in parseInterva(%q): %s", s, err)
+			t.Fatalf("unexpected error in parseInterval(%q): %s", s, err)
 		}
 		if interval != intervalExpected {
 			t.Fatalf("unexpected result for parseInterval(%q); got %d; want %d", s, interval, intervalExpected)

app/vmselect/graphite/transform.go

@@ -17,6 +17,16 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
 )
 
+// safePathExpressionFromString truncates a pathExpression string if it exceeds
+// the maximum allowed length to prevent memory exhaustion.
+func safePathExpressionFromString(pathExpr string) string {
+	maxLen := *maxGraphitePathExpressionLen
+	if maxLen > 0 && len(pathExpr) > maxLen {
+		return pathExpr[:maxLen] + "..."
+	}
+	return pathExpr
+}
+
 // nextSeriesFunc must return the next series to process.
 //
 // nextSeriesFunc must release all the occupied resources before returning non-nil error.
@@ -319,7 +329,7 @@ func aggregateSeries(ec *evalConfig, expr graphiteql.Expr, nextSeries nextSeries
 		Tags:           tags,
 		Timestamps:     ec.newTimestamps(step),
 		Values:         as.Finalize(xFilesFactor),
-		pathExpression: name,
+		pathExpression: safePathExpressionFromString(name),
 		expr:           expr,
 		step:           step,
 	}
@@ -1124,7 +1134,7 @@ func constantLine(ec *evalConfig, expr graphiteql.Expr, n float64) nextSeriesFun
 		Timestamps:     []int64{ec.startTime, ec.startTime + step, ec.startTime + 2*step},
 		Values:         []float64{n, n, n},
 		expr:           expr,
-		pathExpression: string(expr.AppendString(nil)),
+		pathExpression: safePathExpression(expr),
 		step:           step,
 	}
 	return singleSeriesFunc(s)

app/vmselect/graphiteql/lexer_test.go

@@ -17,7 +17,7 @@ func TestScanStringSuccess(t *testing.T) {
 			t.Fatalf("unexpected string scanned from %s; got %s; want %s", s, result, sExpected)
 		}
 		if !strings.HasPrefix(s, result) {
-			t.Fatalf("invalid prefix for scanne string %s: %s", s, result)
+			t.Fatalf("invalid prefix for scanned string %s: %s", s, result)
 		}
 	}
 	f(`""`, `""`)

app/vmselect/graphiteql/parser.go

@@ -210,7 +210,7 @@ func (p *parser) parseMetricExprOrFuncCall() (Expr, error) {
 		}
 		return fe, nil
 	default:
-		// Metric epxression or bool expression or None.
+		// Metric expression or bool expression or None.
 		if isBool(ident) {
 			be := &BoolExpr{
 				B: strings.EqualFold(ident, "true"),

app/vmselect/main.go

@@ -262,6 +262,13 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 			return true
 		}
 		return true
+	case "/api/v1/config":
+		httpserver.EnableCORS(w, r)
+		if err := prometheus.ConfigHandler(qt, startTime, w, r); err != nil {
+			httpserver.SendPrometheusError(w, r, err)
+			return true
+		}
+		return true
 	case "/api/v1/export":
 		exportRequests.Inc()
 		if err := prometheus.ExportHandler(startTime, w, r); err != nil {
@@ -538,6 +545,13 @@ func handleStaticAndSimpleRequests(w http.ResponseWriter, r *http.Request, path
 		expandWithExprsRequests.Inc()
 		prometheus.ExpandWithExprs(w, r)
 		return true
+	case "/extract-metric-exprs":
+		startTime := time.Now()
+		if err := prometheus.ExtractMetricExprsHandler(startTime, w, r); err != nil {
+			httpserver.Errorf(w, r, "%s", err)
+			return true
+		}
+		return true
 	case "/prettify-query":
 		prettifyQueryRequests.Inc()
 		prometheus.PrettifyQuery(w, r)

app/vmselect/netstorage/netstorage.go

@@ -63,10 +63,18 @@ type Results struct {
 	packedTimeseries []packedTimeseries
 	sr               *storage.Search
 	tbf              *tmpBlocksFile
+
+	// the result is simulated
+	isSimulated     bool
+	simulatedSeries []*storage.SimulatedSamples
 }
 
 // Len returns the number of results in rss.
 func (rss *Results) Len() int {
+	if rss.isSimulated {
+		return len(rss.simulatedSeries)
+	}
+
 	return len(rss.packedTimeseries)
 }
 
@@ -218,6 +226,10 @@ var defaultMaxWorkersPerQuery = func() int {
 //
 // rss becomes unusable after the call to RunParallel.
 func (rss *Results) RunParallel(qt *querytracer.Tracer, f func(rs *Result, workerID uint) error) error {
+	if rss.isSimulated {
+		return rss.runParallelSimulated(qt, f)
+	}
+
 	qt = qt.NewChild("parallel process of fetched data")
 	defer rss.mustClose()
 
@@ -233,6 +245,87 @@ func (rss *Results) RunParallel(qt *querytracer.Tracer, f func(rs *Result, worke
 	return err
 }
 
+func (rss *Results) runParallelSimulated(qt *querytracer.Tracer, f func(rs *Result, workerID uint) error) error {
+	qt = qt.NewChild("parallel process of fetched data")
+
+	cb := f
+	tmpResult := getTmpResult()
+	defer putTmpResult(tmpResult)
+
+	// For simplicity, let's process serially first. Parallelization can be added if needed.
+	// If parallelization is desired, it would mirror the worker pool logic of the original runParallel,
+	// but iterating over rss.simulatedSamples entries.
+	workerID := uint(0)
+	var firstErr error
+	for _, metric := range rss.simulatedSeries {
+		r := &tmpResult.rs
+		r.reset()
+		r.MetricName.CopyFrom(&metric.Name)
+		for i, ts := range metric.Timestamps {
+			if ts >= rss.tr.MinTimestamp && ts <= rss.tr.MaxTimestamp {
+				r.Values = append(r.Values, metric.Value[i])
+				r.Timestamps = append(r.Timestamps, ts)
+			}
+		}
+
+		// Sort timestamps chronologically to match real storage behavior.
+		// Real storage ensures chronological order through:
+		// 1. Block-level sorting by MinTimestamp
+		// 2. Within-block timestamp ordering via encoding.EnsureNonDecreasingSequence()
+		if len(r.Timestamps) > 1 {
+			// Create pairs for sorting
+			type timestampValue struct {
+				timestamp int64
+				value     float64
+			}
+			pairs := make([]timestampValue, len(r.Timestamps))
+			for i := range r.Timestamps {
+				pairs[i] = timestampValue{
+					timestamp: r.Timestamps[i],
+					value:     r.Values[i],
+				}
+			}
+
+			// Sort by timestamp
+			sort.Slice(pairs, func(i, j int) bool {
+				return pairs[i].timestamp < pairs[j].timestamp
+			})
+
+			// Extract back to separate slices
+			for i := range pairs {
+				r.Timestamps[i] = pairs[i].timestamp
+				r.Values[i] = pairs[i].value
+			}
+		}
+
+		// The input from the client is most likely already deduplicated, since it's emitted by
+		// vmselect. However, the client may modify the input instead of using the returned one.
+		dedupInterval := storage.GetDedupInterval()
+		if dedupInterval > 0 && len(r.Timestamps) > 0 {
+			r.Timestamps, r.Values = storage.DeduplicateSamples(r.Timestamps, r.Values, dedupInterval)
+		}
+
+		rowProcessed := len(r.Timestamps)
+
+		if rowProcessed > 0 {
+			err := cb(r, workerID)
+			if err != nil {
+				firstErr = err
+				break
+			}
+		}
+	}
+
+	// Count total samples across all series
+	totalSamples := 0
+	for _, metric := range rss.simulatedSeries {
+		totalSamples += len(metric.Timestamps)
+	}
+	qt.Donef("series=%d, samples=%d", len(rss.simulatedSeries), totalSamples)
+
+	return firstErr
+}
+
 func (rss *Results) runParallel(qt *querytracer.Tracer, f func(rs *Result, workerID uint) error) (int, error) {
 	tswsLen := len(rss.packedTimeseries)
 	if tswsLen == 0 {
@@ -269,7 +362,7 @@ func (rss *Results) runParallel(qt *querytracer.Tracer, f func(rs *Result, worke
 	}
 
 	// Slow path - spin up multiple local workers for parallel data processing.
-	// Do not use global workers pool, since it increases inter-CPU memory ping-poing,
+	// Do not use global workers pool, since it increases inter-CPU memory ping-pong,
 	// which reduces the scalability on systems with many CPU cores.
 
 	// Prepare the work for workers.
@@ -485,7 +578,7 @@ func (pts *packedTimeseries) unpackTo(dst []*sortBlock, tbf *tmpBlocksFile, tr s
 	}
 
 	// Slow path - spin up multiple local workers for parallel data unpacking.
-	// Do not use global workers pool, since it increases inter-CPU memory ping-poing,
+	// Do not use global workers pool, since it increases inter-CPU memory ping-pong,
 	// which reduces the scalability on systems with many CPU cores.
 
 	// Prepare the work for workers.
@@ -1119,6 +1212,10 @@ func SearchMetricNames(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline
 //
 // Results.RunParallel or Results.Cancel must be called on the returned Results.
 func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline searchutil.Deadline) (*Results, error) {
+	if len(sq.SimulatedSeries) > 0 {
+		return processSearchSimulated(qt, sq, deadline)
+	}
+
 	qt = qt.NewChild("fetch matching series: %s", sq)
 	defer qt.Done()
 	if deadline.Exceeded() {
@@ -1291,6 +1388,41 @@ func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadlin
 	return &rss, nil
 }
 
+func processSearchSimulated(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline searchutil.Deadline) (*Results, error) {
+	qt = qt.NewChild("fetch matching series (simulated): %s", sq)
+	defer qt.Done()
+	if deadline.Exceeded() {
+		return nil, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
+	}
+
+	tr := storage.TimeRange{
+		MinTimestamp: sq.MinTimestamp,
+		MaxTimestamp: sq.MaxTimestamp,
+	}
+
+	// Process simulated samples.
+	matchedSamples, err := storage.MatchSimulatedSamples(sq.SimulatedSeries, sq.TagFilterss)
+	if err != nil {
+		return nil, fmt.Errorf("cannot match simulated samples: %w", err)
+	}
+
+	// Create a result set similar to ProcessSearchQuery
+	rss := &Results{
+		tr:              tr,
+		deadline:        deadline,
+		isSimulated:     true,
+		simulatedSeries: matchedSamples,
+	}
+
+	if len(matchedSamples) == 0 {
+		qt.Printf("no matching series found")
+	} else {
+		qt.Printf("found %d series", len(rss.simulatedSeries))
+	}
+
+	return rss, nil
+}
+
 type blockRef struct {
 	partRef storage.PartRef
 	addr    tmpBlockAddr

app/vmselect/netstorage/tmp_blocks_file.go

@@ -135,7 +135,7 @@ func (tbf *tmpBlocksFile) WriteBlockRefData(b []byte) (tmpBlockAddr, error) {
 	return addr, nil
 }
 
-// Len() returnt tbf size in bytes.
+// Len() return tbf size in bytes.
 func (tbf *tmpBlocksFile) Len() uint64 {
 	return tbf.offset
 }

app/vmselect/prometheus/config_response.qtpl

@@ -0,0 +1,20 @@
+{% import (
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
+) %}
+
+{% stripspace %}
+
+ConfigResponse generates response for /api/v1/config .
+{% func ConfigResponse(config *ConfigData, qt *querytracer.Tracer) %}
+{
+	"status":"success",
+	"data":{
+		"minStalenessInterval": {%q= config.MinStalenessInterval %},
+		"maxStalenessInterval": {%q= config.MaxStalenessInterval %}
+	}
+	{% code	qt.Done() %}
+	{%= dumpQueryTrace(qt) %}
+}
+{% endfunc %}
+
+{% endstripspace %} 

app/vmselect/prometheus/config_response.qtpl.go

@@ -0,0 +1,73 @@
+// Code generated by qtc from "config_response.qtpl". DO NOT EDIT.
+// See https://github.com/valyala/quicktemplate for details.
+
+//line app/vmselect/prometheus/config_response.qtpl:1
+package prometheus
+
+//line app/vmselect/prometheus/config_response.qtpl:1
+import (
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
+)
+
+// ConfigResponse generates response for /api/v1/config .
+
+//line app/vmselect/prometheus/config_response.qtpl:8
+import (
+	qtio422016 "io"
+
+	qt422016 "github.com/valyala/quicktemplate"
+)
+
+//line app/vmselect/prometheus/config_response.qtpl:8
+var (
+	_ = qtio422016.Copy
+	_ = qt422016.AcquireByteBuffer
+)
+
+//line app/vmselect/prometheus/config_response.qtpl:8
+func StreamConfigResponse(qw422016 *qt422016.Writer, config *ConfigData, qt *querytracer.Tracer) {
+//line app/vmselect/prometheus/config_response.qtpl:8
+	qw422016.N().S(`{"status":"success","data":{"minStalenessInterval":`)
+//line app/vmselect/prometheus/config_response.qtpl:12
+	qw422016.N().Q(config.MinStalenessInterval)
+//line app/vmselect/prometheus/config_response.qtpl:12
+	qw422016.N().S(`,"maxStalenessInterval":`)
+//line app/vmselect/prometheus/config_response.qtpl:13
+	qw422016.N().Q(config.MaxStalenessInterval)
+//line app/vmselect/prometheus/config_response.qtpl:13
+	qw422016.N().S(`}`)
+//line app/vmselect/prometheus/config_response.qtpl:15
+	qt.Done()
+
+//line app/vmselect/prometheus/config_response.qtpl:16
+	streamdumpQueryTrace(qw422016, qt)
+//line app/vmselect/prometheus/config_response.qtpl:16
+	qw422016.N().S(`}`)
+//line app/vmselect/prometheus/config_response.qtpl:18
+}
+
+//line app/vmselect/prometheus/config_response.qtpl:18
+func WriteConfigResponse(qq422016 qtio422016.Writer, config *ConfigData, qt *querytracer.Tracer) {
+//line app/vmselect/prometheus/config_response.qtpl:18
+	qw422016 := qt422016.AcquireWriter(qq422016)
+//line app/vmselect/prometheus/config_response.qtpl:18
+	StreamConfigResponse(qw422016, config, qt)
+//line app/vmselect/prometheus/config_response.qtpl:18
+	qt422016.ReleaseWriter(qw422016)
+//line app/vmselect/prometheus/config_response.qtpl:18
+}
+
+//line app/vmselect/prometheus/config_response.qtpl:18
+func ConfigResponse(config *ConfigData, qt *querytracer.Tracer) string {
+//line app/vmselect/prometheus/config_response.qtpl:18
+	qb422016 := qt422016.AcquireByteBuffer()
+//line app/vmselect/prometheus/config_response.qtpl:18
+	WriteConfigResponse(qb422016, config, qt)
+//line app/vmselect/prometheus/config_response.qtpl:18
+	qs422016 := string(qb422016.B)
+//line app/vmselect/prometheus/config_response.qtpl:18
+	qt422016.ReleaseByteBuffer(qb422016)
+//line app/vmselect/prometheus/config_response.qtpl:18
+	return qs422016
+//line app/vmselect/prometheus/config_response.qtpl:18
+}

app/vmselect/prometheus/extract_metric_exprs_response.qtpl

@@ -0,0 +1,18 @@
+{% stripspace %}
+
+ExtractMetricExprsResponse generates response for /extract-metric-exprs .
+{% func ExtractMetricExprsResponse(metrics []string) %}
+{
+	"status":"success",
+	"data":[
+		{% if len(metrics) > 0 %}
+			{%q= metrics[0] %}
+			{% for i := 1; i < len(metrics); i++ %}
+				,{%q= metrics[i] %}
+			{% endfor %}
+		{% endif %}
+	]
+}
+{% endfunc %}
+
+{% endstripspace %} 

app/vmselect/prometheus/extract_metric_exprs_response.qtpl.go

@@ -0,0 +1,69 @@
+// Code generated by qtc from "extract_metric_exprs_response.qtpl". DO NOT EDIT.
+// See https://github.com/valyala/quicktemplate for details.
+
+// ExtractMetricExprsResponse generates response for /extract-metric-exprs .
+
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:4
+package prometheus
+
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:4
+import (
+	qtio422016 "io"
+
+	qt422016 "github.com/valyala/quicktemplate"
+)
+
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:4
+var (
+	_ = qtio422016.Copy
+	_ = qt422016.AcquireByteBuffer
+)
+
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:4
+func StreamExtractMetricExprsResponse(qw422016 *qt422016.Writer, metrics []string) {
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:4
+	qw422016.N().S(`{"status":"success","data":[`)
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:8
+	if len(metrics) > 0 {
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:9
+		qw422016.N().Q(metrics[0])
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:10
+		for i := 1; i < len(metrics); i++ {
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:10
+			qw422016.N().S(`,`)
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:11
+			qw422016.N().Q(metrics[i])
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:12
+		}
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:13
+	}
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:13
+	qw422016.N().S(`]}`)
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+}
+
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+func WriteExtractMetricExprsResponse(qq422016 qtio422016.Writer, metrics []string) {
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+	qw422016 := qt422016.AcquireWriter(qq422016)
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+	StreamExtractMetricExprsResponse(qw422016, metrics)
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+	qt422016.ReleaseWriter(qw422016)
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+}
+
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+func ExtractMetricExprsResponse(metrics []string) string {
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+	qb422016 := qt422016.AcquireByteBuffer()
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+	WriteExtractMetricExprsResponse(qb422016, metrics)
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+	qs422016 := string(qb422016.B)
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+	qt422016.ReleaseByteBuffer(qb422016)
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+	return qs422016
+//line app/vmselect/prometheus/extract_metric_exprs_response.qtpl:16
+}

app/vmselect/prometheus/prometheus.go

@@ -1,8 +1,10 @@
 package prometheus
 
 import (
+	"encoding/json"
 	"flag"
 	"fmt"
+	"io"
 	"math"
 	"net/http"
 	"runtime"
@@ -20,6 +22,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/promql"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/querystats"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/searchutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bufferedwriter"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
@@ -42,6 +45,9 @@ var (
 	maxLookback = flag.Duration("search.maxLookback", 0, "Synonym to -query.lookback-delta from Prometheus. "+
 		"The value is dynamically detected from interval between time series datapoints if not set. It can be overridden on per-query basis via max_lookback arg. "+
 		"See also '-search.maxStalenessInterval' flag, which has the same meaning due to historical reasons")
+	minStalenessInterval = flag.Duration("search.minStalenessInterval", 0, "The minimum interval for staleness calculations. "+
+		"This flag could be useful for removing gaps on graphs generated from time series with irregular intervals between samples. "+
+		"See also '-search.maxStalenessInterval'")
 	maxStalenessInterval = flag.Duration("search.maxStalenessInterval", 0, "The maximum interval for staleness calculations. "+
 		"By default, it is automatically calculated from the median interval between samples. This flag could be useful for tuning "+
 		"Prometheus data model closer to Influx-style data model. See https://prometheus.io/docs/prometheus/latest/querying/basics/#staleness for details. "+
@@ -116,7 +122,7 @@ func FederateHandler(startTime time.Time, w http.ResponseWriter, r *http.Request
 	if err != nil {
 		return err
 	}
-	lookbackDelta, err := getMaxLookback(r)
+	lookbackDelta, err := getMaxLookback(r, *maxStalenessInterval)
 	if err != nil {
 		return err
 	}
@@ -611,6 +617,55 @@ func TSDBStatusHandler(qt *querytracer.Tracer, startTime time.Time, w http.Respo
 
 var tsdbStatusDuration = metrics.NewSummary(`vm_request_duration_seconds{path="/api/v1/status/tsdb"}`)
 
+// ConfigData holds the current configuration values for search-related flags
+type ConfigData struct {
+	MinStalenessInterval string
+	MaxStalenessInterval string
+}
+
+// ConfigHandler processes /api/v1/config request.
+//
+// It returns the current configuration for search-related flags.
+func ConfigHandler(qt *querytracer.Tracer, startTime time.Time, w http.ResponseWriter, _ *http.Request) error {
+	config := &ConfigData{
+		MinStalenessInterval: (*minStalenessInterval).String(),
+		MaxStalenessInterval: (*maxStalenessInterval).String(),
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	bw := bufferedwriter.Get(w)
+	defer bufferedwriter.Put(bw)
+	WriteConfigResponse(bw, config, qt)
+	if err := bw.Flush(); err != nil {
+		return fmt.Errorf("cannot send config response to remote client: %w", err)
+	}
+	return nil
+}
+
+// ExtractMetricExprsHandler processes /extract-metric-exprs request.
+//
+// It extracts metric expressions from a given PromQL query.
+func ExtractMetricExprsHandler(startTime time.Time, w http.ResponseWriter, r *http.Request) error {
+	query := r.FormValue("query")
+	if len(query) == 0 {
+		return fmt.Errorf("missing `query` arg")
+	}
+
+	metrics, err := promql.ExtractMetricsFromQuery(query)
+	if err != nil {
+		return fmt.Errorf("cannot extract metrics from query: %w", err)
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	bw := bufferedwriter.Get(w)
+	defer bufferedwriter.Put(bw)
+	WriteExtractMetricExprsResponse(bw, metrics)
+	if err := bw.Flush(); err != nil {
+		return fmt.Errorf("cannot send extract metric exprs response to remote client: %w", err)
+	}
+	return nil
+}
+
 // LabelsHandler processes /api/v1/labels request.
 //
 // See https://prometheus.io/docs/prometheus/latest/querying/api/#getting-label-names
@@ -712,7 +767,8 @@ func QueryHandler(qt *querytracer.Tracer, startTime time.Time, w http.ResponseWr
 
 	ct := startTime.UnixNano() / 1e6
 	deadline := searchutil.GetDeadlineForQuery(r, startTime)
-	mayCache := !httputil.GetBool(r, "nocache")
+	isDebug := httputil.GetBool(r, "debug")
+	noCache := httputil.GetBool(r, "nocache") || isDebug
 	query := r.FormValue("query")
 	if len(query) == 0 {
 		return fmt.Errorf("missing `query` arg")
@@ -721,7 +777,7 @@ func QueryHandler(qt *querytracer.Tracer, startTime time.Time, w http.ResponseWr
 	if err != nil {
 		return err
 	}
-	lookbackDelta, err := getMaxLookback(r)
+	lookbackDelta, err := getMaxLookback(r, *maxStalenessInterval)
 	if err != nil {
 		return err
 	}
@@ -807,23 +863,14 @@ func QueryHandler(qt *querytracer.Tracer, startTime time.Time, w http.ResponseWr
 	} else {
 		queryOffset = 0
 	}
-	ec := &promql.EvalConfig{
-		Start:               start,
-		End:                 start,
-		Step:                step,
-		MaxPointsPerSeries:  *maxPointsPerTimeseries,
-		MaxSeries:           GetMaxUniqueTimeSeries(),
-		QuotedRemoteAddr:    httpserver.GetQuotedRemoteAddr(r),
-		Deadline:            deadline,
-		MayCache:            mayCache,
-		LookbackDelta:       lookbackDelta,
-		RoundDigits:         getRoundDigits(r),
-		EnforcedTagFilterss: etfs,
-		CacheTagFilters:     etfs,
-		GetRequestURI: func() string {
-			return httpserver.GetRequestURI(r)
-		},
+	ec := newEvalConfig(r, start, start, step, deadline, noCache, lookbackDelta, isDebug, etfs)
+	if isDebug {
+		if err := populateSimulatedData(r, nil, ec); err != nil {
+			_ = r.Body.Close()
+			return fmt.Errorf("cannot read simulated samples: %w", err)
+		}
 	}
+	_ = r.Body.Close()
 	qs := promql.NewQueryStats(query, nil, ec)
 	ec.QueryStats = qs
 
@@ -897,8 +944,9 @@ func QueryRangeHandler(qt *querytracer.Tracer, startTime time.Time, w http.Respo
 func queryRangeHandler(qt *querytracer.Tracer, startTime time.Time, w http.ResponseWriter, query string,
 	start, end, step int64, r *http.Request, ct int64, etfs [][]storage.TagFilter) error {
 	deadline := searchutil.GetDeadlineForQuery(r, startTime)
-	mayCache := !httputil.GetBool(r, "nocache")
-	lookbackDelta, err := getMaxLookback(r)
+	isDebug := httputil.GetBool(r, "debug")
+	noCache := httputil.GetBool(r, "nocache") || isDebug
+	lookbackDelta, err := getMaxLookback(r, *maxStalenessInterval)
 	if err != nil {
 		return err
 	}
@@ -913,27 +961,19 @@ func queryRangeHandler(qt *querytracer.Tracer, startTime time.Time, w http.Respo
 	if err := promql.ValidateMaxPointsPerSeries(start, end, step, *maxPointsPerTimeseries); err != nil {
 		return fmt.Errorf("%w; (see -search.maxPointsPerTimeseries command-line flag)", err)
 	}
-	if mayCache {
+	if !noCache {
 		start, end = promql.AdjustStartEnd(start, end, step)
 	}
 
-	ec := &promql.EvalConfig{
-		Start:               start,
-		End:                 end,
-		Step:                step,
-		MaxPointsPerSeries:  *maxPointsPerTimeseries,
-		MaxSeries:           GetMaxUniqueTimeSeries(),
-		QuotedRemoteAddr:    httpserver.GetQuotedRemoteAddr(r),
-		Deadline:            deadline,
-		MayCache:            mayCache,
-		LookbackDelta:       lookbackDelta,
-		RoundDigits:         getRoundDigits(r),
-		EnforcedTagFilterss: etfs,
-		CacheTagFilters:     etfs,
-		GetRequestURI: func() string {
-			return httpserver.GetRequestURI(r)
-		},
+	ec := newEvalConfig(r, start, end, step, deadline, noCache, lookbackDelta, isDebug, etfs)
+	if isDebug {
+		if err := populateSimulatedData(r, nil, ec); err != nil {
+			_ = r.Body.Close()
+			return fmt.Errorf("cannot read simulated samples: %w", err)
+		}
 	}
+	_ = r.Body.Close()
+
 	qs := promql.NewQueryStats(query, nil, ec)
 	ec.QueryStats = qs
 
@@ -969,6 +1009,93 @@ func queryRangeHandler(qt *querytracer.Tracer, startTime time.Time, w http.Respo
 	return nil
 }
 
+func newEvalConfig(r *http.Request, start, end, step int64, deadline searchutil.Deadline, noCache bool, lookbackDelta int64, isDebug bool, etfs [][]storage.TagFilter) *promql.EvalConfig {
+	ec := &promql.EvalConfig{
+		Start:                start,
+		End:                  end,
+		Step:                 step,
+		MaxPointsPerSeries:   *maxPointsPerTimeseries,
+		MaxSeries:            GetMaxUniqueTimeSeries(),
+		MinStalenessInterval: *minStalenessInterval,
+		QuotedRemoteAddr:     httpserver.GetQuotedRemoteAddr(r),
+		Deadline:             deadline,
+		MayCache:             !noCache,
+		LookbackDelta:        lookbackDelta,
+		RoundDigits:          getRoundDigits(r),
+		EnforcedTagFilterss:  etfs,
+		CacheTagFilters:      etfs,
+		GetRequestURI: func() string {
+			return httpserver.GetRequestURI(r)
+		},
+	}
+
+	return ec
+}
+
+func populateSimulatedData(r *http.Request, at *auth.Token, evalConfig *promql.EvalConfig) error {
+	type jsonExportBlockInput struct {
+		Metric     map[string]string `json:"metric"`
+		Values     []float64         `json:"values"`
+		Timestamps []int64           `json:"timestamps"`
+	}
+
+	// --- Read and Parse Input Samples from r.Body ---
+	var simulatedSeries []*storage.SimulatedSamples
+	decoder := json.NewDecoder(r.Body)
+	lineNum := 0
+	for {
+		var jeb jsonExportBlockInput
+		if err := decoder.Decode(&jeb); err == io.EOF {
+			break
+		} else if err != nil {
+			return fmt.Errorf("error decoding input JSON on line %d: %w", lineNum, err)
+		}
+
+		// Validate that values and timestamps arrays have the same length
+		if len(jeb.Values) != len(jeb.Timestamps) {
+			return fmt.Errorf("mismatched values and timestamps arrays length in debug data on line %d: values=%d, timestamps=%d", lineNum, len(jeb.Values), len(jeb.Timestamps))
+		}
+
+		var mn = storage.GetMetricName()
+		defer storage.PutMetricName(mn)
+		for k, v := range jeb.Metric {
+			mn.AddTag(k, v)
+		}
+
+		ss := &storage.SimulatedSamples{
+			Value:      jeb.Values,
+			Timestamps: jeb.Timestamps,
+		}
+		ss.Name.CopyFrom(mn)
+		simulatedSeries = append(simulatedSeries, ss)
+		lineNum++
+	}
+
+	// It doesn't make sense to debug with empty samples
+	if len(simulatedSeries) == 0 {
+		return fmt.Errorf("no simulated samples found")
+	}
+
+	minStalenessInterval, err := httputil.GetDurationRaw(r, "min_staleness_interval", evalConfig.MinStalenessInterval)
+	if err != nil {
+		return fmt.Errorf("cannot parse `min_staleness_interval` arg: %w", err)
+	}
+
+	maxStalenessInterval, err := httputil.GetDurationRaw(r, "max_staleness_interval", *maxStalenessInterval)
+	if err != nil {
+		return fmt.Errorf("cannot parse `max_staleness_interval` arg: %w", err)
+	}
+
+	evalConfig.SimulatedSamples = simulatedSeries
+	evalConfig.MinStalenessInterval = minStalenessInterval
+	evalConfig.LookbackDelta, err = getMaxLookback(r, maxStalenessInterval)
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
 func removeEmptyValuesAndTimeseries(tss []netstorage.Result) []netstorage.Result {
 	dst := tss[:0]
 	for i := range tss {
@@ -1044,7 +1171,7 @@ func adjustLastPoints(tss []netstorage.Result, start, end int64) []netstorage.Re
 	return tss
 }
 
-func getMaxLookback(r *http.Request) (int64, error) {
+func getMaxLookback(r *http.Request, maxStalenessInterval time.Duration) (int64, error) {
 	d := maxLookback.Milliseconds()
 	if d == 0 {
 		d = maxStalenessInterval.Milliseconds()

app/vmselect/promql/binary_op.go

@@ -188,7 +188,7 @@ func newBinaryOpFunc(bf func(left, right float64, isBool bool) float64) binaryOp
 			rightValues := right[i].Values
 			dstValues := dst[i].Values
 			if len(leftValues) != len(rightValues) || len(leftValues) != len(dstValues) {
-				logger.Panicf("BUG: len(leftVaues) must match len(rightValues) and len(dstValues); got %d vs %d vs %d",
+				logger.Panicf("BUG: len(leftValues) must match len(rightValues) and len(dstValues); got %d vs %d vs %d",
 					len(leftValues), len(rightValues), len(dstValues))
 			}
 			for j, a := range leftValues {

app/vmselect/promql/eval.go

@@ -134,6 +134,10 @@ type EvalConfig struct {
 	// LookbackDelta is analog to `-query.lookback-delta` from Prometheus.
 	LookbackDelta int64
 
+	// MaxStalenessInterval corresponds to -search.maxStalenessInterval,
+	// but customized per query request.
+	MinStalenessInterval time.Duration
+
 	// How many decimal digits after the point to leave in response.
 	RoundDigits int
 
@@ -158,6 +162,9 @@ type EvalConfig struct {
 
 	timestamps     []int64
 	timestampsOnce sync.Once
+
+	// Simulated samples
+	SimulatedSamples []*storage.SimulatedSamples
 }
 
 // copyEvalConfig returns src copy.
@@ -176,6 +183,8 @@ func copyEvalConfig(src *EvalConfig) *EvalConfig {
 	ec.CacheTagFilters = src.CacheTagFilters
 	ec.GetRequestURI = src.GetRequestURI
 	ec.QueryStats = src.QueryStats
+	ec.MinStalenessInterval = src.MinStalenessInterval
+	ec.SimulatedSamples = src.SimulatedSamples
 
 	// do not copy src.timestamps - they must be generated again.
 	return &ec
@@ -929,7 +938,7 @@ func evalRollupFuncWithSubquery(qt *querytracer.Tracer, ec *EvalConfig, funcName
 	}
 
 	ecSQ := copyEvalConfig(ec)
-	ecSQ.Start -= window + step + maxSilenceInterval()
+	ecSQ.Start -= window + step + maxSilenceInterval(ec.MinStalenessInterval)
 	ecSQ.End += step
 	ecSQ.Step = step
 	ecSQ.MaxPointsPerSeries = *maxPointsSubqueryPerTimeseries
@@ -946,7 +955,7 @@ func evalRollupFuncWithSubquery(qt *querytracer.Tracer, ec *EvalConfig, funcName
 		return nil, nil
 	}
 	sharedTimestamps := getTimestamps(ec.Start, ec.End, ec.Step, ec.MaxPointsPerSeries)
-	preFunc, rcs, err := getRollupConfigs(funcName, rf, expr, ec.Start, ec.End, ec.Step, ec.MaxPointsPerSeries, window, ec.LookbackDelta, sharedTimestamps)
+	preFunc, rcs, err := getRollupConfigs(funcName, rf, expr, ec.Start, ec.End, ec.Step, ec.MaxPointsPerSeries, window, ec.LookbackDelta, sharedTimestamps, ec.MinStalenessInterval)
 	if err != nil {
 		return nil, err
 	}
@@ -1684,7 +1693,7 @@ func evalRollupFuncNoCache(qt *querytracer.Tracer, ec *EvalConfig, funcName stri
 	}
 	// Obtain rollup configs before fetching data from db, so type errors could be caught earlier.
 	sharedTimestamps := getTimestamps(ec.Start, ec.End, ec.Step, ec.MaxPointsPerSeries)
-	preFunc, rcs, err := getRollupConfigs(funcName, rf, expr, ec.Start, ec.End, ec.Step, ec.MaxPointsPerSeries, window, ec.LookbackDelta, sharedTimestamps)
+	preFunc, rcs, err := getRollupConfigs(funcName, rf, expr, ec.Start, ec.End, ec.Step, ec.MaxPointsPerSeries, window, ec.LookbackDelta, sharedTimestamps, ec.MinStalenessInterval)
 	if err != nil {
 		return nil, err
 	}
@@ -1694,7 +1703,7 @@ func evalRollupFuncNoCache(qt *querytracer.Tracer, ec *EvalConfig, funcName stri
 	tfss = searchutil.JoinTagFilterss(tfss, ec.EnforcedTagFilterss)
 	minTimestamp := ec.Start
 	if needSilenceIntervalForRollupFunc[funcName] {
-		minTimestamp -= maxSilenceInterval()
+		minTimestamp -= maxSilenceInterval(ec.MinStalenessInterval)
 	}
 	if window > ec.Step {
 		minTimestamp -= window
@@ -1702,6 +1711,8 @@ func evalRollupFuncNoCache(qt *querytracer.Tracer, ec *EvalConfig, funcName stri
 		minTimestamp -= ec.Step
 	}
 	sq := storage.NewSearchQuery(minTimestamp, ec.End, tfss, ec.MaxSeries)
+	sq.SimulatedSeries = ec.SimulatedSamples
+
 	rss, err := netstorage.ProcessSearchQuery(qt, sq, ec.Deadline)
 	if err != nil {
 		return nil, err
@@ -1787,7 +1798,7 @@ func getRollupMemoryLimiter() *memoryLimiter {
 	return &rollupMemoryLimiter
 }
 
-func maxSilenceInterval() int64 {
+func maxSilenceInterval(minStalenessInterval time.Duration) int64 {
 	d := minStalenessInterval.Milliseconds()
 	if d <= 0 {
 		d = 5 * 60 * 1000

app/vmselect/promql/eval_test.go

@@ -55,7 +55,7 @@ func TestValidateMaxPointsPerSeriesFailure(t *testing.T) {
 	f := func(start, end, step int64, maxPoints int) {
 		t.Helper()
 		if err := ValidateMaxPointsPerSeries(start, end, step, maxPoints); err == nil {
-			t.Fatalf("expecint non-nil error for ValidateMaxPointsPerSeries(start=%d, end=%d, step=%d, maxPoints=%d)", start, end, step, maxPoints)
+			t.Fatalf("expecting non-nil error for ValidateMaxPointsPerSeries(start=%d, end=%d, step=%d, maxPoints=%d)", start, end, step, maxPoints)
 		}
 	}
 	// zero step

app/vmselect/promql/exec.go

@@ -61,12 +61,15 @@ func Exec(qt *querytracer.Tracer, ec *EvalConfig, q string, isFirstPointOnly boo
 		}
 	}
 
+	var rv []*timeseries
+
 	qid := activeQueriesV.Add(ec, q)
-	rv, err := evalExpr(qt, ec, e)
+	rv, err = evalExpr(qt, ec, e)
 	activeQueriesV.Remove(qid)
 	if err != nil {
 		return nil, err
 	}
+
 	if isFirstPointOnly {
 		// Remove all the points except the first one from every time series.
 		for _, ts := range rv {
@@ -325,3 +328,23 @@ func escapeDots(s string) string {
 	}
 	return string(result)
 }
+
+// ExtractMetricsFromQuery visits all the expressions in query and returns all the metrics found in the query.
+func ExtractMetricsFromQuery(query string) ([]string, error) {
+	expr, err := metricsql.Parse(query)
+	if err != nil {
+		return nil, fmt.Errorf("error parsing query: %w", err)
+	}
+
+	var metrics []string
+	metricsql.VisitAll(expr, func(e metricsql.Expr) {
+		if me, ok := e.(*metricsql.MetricExpr); ok {
+			metricStr := string(me.AppendString(nil))
+			if metricStr != "" {
+				metrics = append(metrics, metricStr)
+			}
+		}
+	})
+
+	return metrics, nil
+}

app/vmselect/promql/exec_debug_test.go

@@ -0,0 +1,313 @@
+package promql
+
+import (
+	"math"
+	"slices"
+	"testing"
+	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/netstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/searchutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+)
+
+func TestSimulatedExec(t *testing.T) {
+	accountID := uint32(123)
+	projectID := uint32(567)
+	start := int64(1000e3)
+	end := int64(2000e3)
+	step := int64(200e3)
+
+	// Base EvalConfig that will be copied for each test
+	baseEC := EvalConfig{
+		Start:              start,
+		End:                end,
+		Step:               step,
+		MaxPointsPerSeries: 1e4,
+		MaxSeries:          1000,
+		Deadline:           searchutil.NewDeadline(time.Now(), time.Hour, ""),
+		RoundDigits:        100,
+		MayCache:           false,
+	}
+
+	t.Run(`simple_metric_exact_match`, func(t *testing.T) {
+		t.Skip()
+		ec := copyEvalConfig(&baseEC)
+		mn := newMetric(accountID, projectID,
+			"__name__", "test_metric",
+			"a", "b",
+		)
+
+		ec.SimulatedSamples = []*storage.SimulatedSamples{mn.build()}
+
+		q := `test_metric{a="b"}`
+		result, err := Exec(nil, ec, q, false)
+		if err != nil {
+			t.Fatalf(`unexpected error when executing %q: %s`, q, err)
+		}
+
+		// Expected result
+		expectedMN := storage.MetricName{
+			MetricGroup: []byte("test_metric"),
+			Tags: []storage.Tag{
+				{
+					Key:   []byte("a"),
+					Value: []byte("b"),
+				},
+			},
+		}
+		expectedResult := []netstorage.Result{
+			{
+				MetricName: expectedMN,
+				Values:     mn.Value,
+				Timestamps: mn.Timestamps,
+			},
+		}
+
+		testResultsEqual(t, result, expectedResult)
+	})
+
+	t.Run(`filtered_by_tag_value`, func(t *testing.T) {
+		t.Skip()
+
+		// Create a copy of base EvalConfig
+		ec := copyEvalConfig(&baseEC)
+		mn := metricBuilders{
+			newMetric(accountID, projectID,
+				"__name__", "test_metric",
+				"a", "b",
+				"region", "us-west",
+			),
+			newMetric(accountID, projectID,
+				"__name__", "test_metric",
+				"a", "b",
+				"region", "us-east",
+			),
+		}
+		ec.SimulatedSamples = mn.build()
+
+		q := `test_metric{region="us-west"}`
+		result, err := Exec(nil, ec, q, false)
+		if err != nil {
+			t.Fatalf(`unexpected error when executing %q: %s`, q, err)
+		}
+
+		// Expected result
+		expectedMN := storage.MetricName{
+			MetricGroup: []byte("test_metric"),
+			Tags: []storage.Tag{
+				{
+					Key:   []byte("a"),
+					Value: []byte("b"),
+				},
+				{
+					Key:   []byte("region"),
+					Value: []byte("us-west"),
+				},
+			},
+		}
+		expectedResult := []netstorage.Result{
+			{
+				MetricName: expectedMN,
+				Values:     mn[0].Value,
+				Timestamps: mn[0].Timestamps,
+			},
+		}
+
+		testResultsEqual(t, result, expectedResult)
+	})
+
+	t.Run(`regex_match_on_tag`, func(t *testing.T) {
+		ec := copyEvalConfig(&baseEC)
+		mn := metricBuilders{
+			newMetric(accountID, projectID,
+				"__name__", "test_metric",
+				"env", "prod",
+			),
+			newMetric(accountID, projectID,
+				"__name__", "test_metric",
+				"env", "staging",
+			),
+			newMetric(accountID, projectID,
+				"__name__", "test_metric",
+				"env", "dev",
+			),
+		}
+		ec.SimulatedSamples = mn.build()
+
+		q := `test_metric{env=~"prod|staging"}`
+		result, err := Exec(nil, ec, q, false)
+		if err != nil {
+			t.Fatalf(`unexpected error when executing %q: %s`, q, err)
+		}
+
+		expectedResult := []netstorage.Result{mn[0].toResult(), mn[1].toResult()}
+		testResultsEqual(t, result, expectedResult)
+	})
+}
+
+func TestSumOverTime(t *testing.T) {
+	accountID := uint32(123)
+	projectID := uint32(567)
+	start := int64(1000e3)
+	end := int64(1300e3)
+	step := int64(30e3)
+
+	baseEC := EvalConfig{
+		Start:              start,
+		End:                end,
+		Step:               step,
+		MaxPointsPerSeries: 1e4,
+		MaxSeries:          1000,
+		Deadline:           searchutil.NewDeadline(time.Now(), time.Hour, ""),
+		RoundDigits:        100,
+		MayCache:           false,
+	}
+
+	t.Run(`basic_sum_over_time`, func(t *testing.T) {
+		ec := copyEvalConfig(&baseEC)
+
+		metric := newMetric(accountID, projectID,
+			"__name__", "test_metric",
+			"app", "api-server",
+		).withValues(1, 2, 3, 4, 5, 6).withUnix(1000, 1015, 1030, 1045, 1060, 1075)
+		ec.SimulatedSamples = []*storage.SimulatedSamples{metric.build()}
+
+		q := `sum_over_time(test_metric[30s])`
+		result, err := Exec(nil, ec, q, false)
+		if err != nil {
+			t.Fatalf(`unexpected error when executing %q: %s`, q, err)
+		}
+
+		expectedResult := []netstorage.Result{
+			newMetric(accountID, projectID,
+				"app", "api-server",
+			).withValues(1, 5, 9, 6).withUnix(1000, 1030, 1060, 1090).toResult(),
+		}
+
+		testSimulatedResultsEqual(t, result, expectedResult)
+	})
+}
+
+type metricBuilder storage.SimulatedSamples
+
+func newMetric(accountID uint32, projectID uint32, pairs ...string) *metricBuilder {
+	mn := storage.MetricName{}
+	for i := 0; i < len(pairs); i += 2 {
+		mn.AddTag(pairs[i], pairs[i+1])
+	}
+	return &metricBuilder{
+		Name:       mn,
+		Value:      []float64{10, 20, 30, 40, 50, 60},
+		Timestamps: []int64{1000e3, 1200e3, 1400e3, 1600e3, 1800e3, 2000e3},
+	}
+}
+
+func (b *metricBuilder) withUnix(unix ...int64) *metricBuilder {
+	b.Timestamps = make([]int64, len(unix))
+	for i := range unix {
+		b.Timestamps[i] = unix[i] * 1e3
+	}
+	return b
+}
+
+func (b *metricBuilder) withValues(values ...float64) *metricBuilder {
+	b.Value = values
+	return b
+}
+
+func (b *metricBuilder) build() *storage.SimulatedSamples {
+	return (*storage.SimulatedSamples)(b)
+}
+
+func (b *metricBuilder) toResult() netstorage.Result {
+	return netstorage.Result{
+		MetricName: b.Name,
+		Values:     b.Value,
+		Timestamps: b.Timestamps,
+	}
+}
+
+type metricBuilders []*metricBuilder
+
+func (b metricBuilders) build() []*storage.SimulatedSamples {
+	ss := make([]*storage.SimulatedSamples, len(b))
+	for i := range b {
+		ss[i] = b[i].build()
+	}
+	return ss
+}
+
+func testSimulatedResultsEqual(t *testing.T, result, resultExpected []netstorage.Result) {
+	t.Helper()
+	result = removeEmptyValuesAndTimeseries(result)
+
+	if len(result) != len(resultExpected) {
+		t.Fatalf(`unexpected timeseries count; got %d; want %d`, len(result), len(resultExpected))
+	}
+	for i := range result {
+		r := &result[i]
+		rExpected := &resultExpected[i]
+		testMetricNamesEqual(t, &r.MetricName, &rExpected.MetricName, i)
+		testRowsEqual(t, r.Values, r.Timestamps, rExpected.Values, rExpected.Timestamps)
+	}
+}
+
+func removeEmptyValuesAndTimeseries(tss []netstorage.Result) []netstorage.Result {
+	dst := tss[:0]
+	for i := range tss {
+		ts := &tss[i]
+		hasNaNs := slices.ContainsFunc(ts.Values, math.IsNaN)
+		if !hasNaNs {
+			// Fast path: nothing to remove.
+			if len(ts.Values) > 0 {
+				dst = append(dst, *ts)
+			}
+			continue
+		}
+
+		// Slow path: remove NaNs.
+		srcTimestamps := ts.Timestamps
+		dstValues := ts.Values[:0]
+		// Do not reuse ts.Timestamps for dstTimestamps, since ts.Timestamps
+		// may be shared among multiple time series.
+		dstTimestamps := make([]int64, 0, len(ts.Timestamps))
+		for j, v := range ts.Values {
+			if math.IsNaN(v) {
+				continue
+			}
+			dstValues = append(dstValues, v)
+			dstTimestamps = append(dstTimestamps, srcTimestamps[j])
+		}
+		ts.Values = dstValues
+		ts.Timestamps = dstTimestamps
+		if len(ts.Values) > 0 {
+			dst = append(dst, *ts)
+		}
+	}
+	return dst
+}
+
+func TestExtractMetricsFromQuery(t *testing.T) {
+	query := `(vm_free_disk_space_bytes{job=~"$job", instance=~"$instance"}-vm_free_disk_space_limit_bytes{job=~"$job", instance=~"$instance"}) 
+/ 
+ignoring(path) (
+    (rate(vm_rows_added_to_storage_total{job=~"$job", instance=~"$instance"}[1d]) - 
+        sum(rate(vm_deduplicated_samples_total{job=~"$job", instance=~"$instance"}[1d])) without (type)) * 
+    (
+        sum(vm_data_size_bytes{job=~"$job", instance=~"$instance", type!~"indexdb.*"}) without(type) /
+        sum(vm_rows{job=~"$job", instance=~"$instance", type!~"indexdb.*"}) without(type)
+    )
+    +
+    rate(vm_new_timeseries_created_total{job=~"$job", instance=~"$instance"}[1d]) * 
+    (
+        sum(vm_data_size_bytes{job=~"$job", instance=~"$instance", type="indexdb/file"}) /
+        sum(vm_rows{job=~"$job", instance=~"$instance", type="indexdb/file"})
+    )
+)`
+	metrics, err := ExtractMetricsFromQuery(query)
+	if err != nil {
+		t.Fatalf(`unexpected error when extracting metrics from query: %s`, err)
+	}
+	t.Logf(`metrics: %v`, metrics)
+}

app/vmselect/promql/rollup.go

@@ -1,12 +1,12 @@
 package promql
 
 import (
-	"flag"
 	"fmt"
 	"math"
 	"strconv"
 	"strings"
 	"sync"
+	"time"
 
 	"github.com/VictoriaMetrics/metrics"
 	"github.com/VictoriaMetrics/metricsql"
@@ -17,10 +17,6 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
 )
 
-var minStalenessInterval = flag.Duration("search.minStalenessInterval", 0, "The minimum interval for staleness calculations. "+
-	"This flag could be useful for removing gaps on graphs generated from time series with irregular intervals between samples. "+
-	"See also '-search.maxStalenessInterval'")
-
 var rollupFuncs = map[string]newRollupFunc{
 	"absent_over_time":        newRollupFuncOneArg(rollupAbsent),
 	"aggr_over_time":          newRollupFuncTwoArgs(rollupFake),
@@ -372,7 +368,7 @@ func getRollupTag(expr metricsql.Expr) (string, error) {
 }
 
 func getRollupConfigs(funcName string, rf rollupFunc, expr metricsql.Expr, start, end, step int64, maxPointsPerSeries int,
-	window, lookbackDelta int64, sharedTimestamps []int64) (
+	window, lookbackDelta int64, sharedTimestamps []int64, minStalenessInterval time.Duration) (
 	func(values []float64, timestamps []int64), []*rollupConfig, error) {
 	preFunc := func(_ []float64, _ []int64) {}
 	funcName = strings.ToLower(funcName)
@@ -408,6 +404,7 @@ func getRollupConfigs(funcName string, rf rollupFunc, expr metricsql.Expr, start
 			Timestamps:            sharedTimestamps,
 			isDefaultRollup:       funcName == "default_rollup",
 			samplesScannedPerCall: samplesScannedPerCall,
+			minStalenessInterval:  minStalenessInterval,
 		}
 	}
 
@@ -600,6 +597,9 @@ type rollupConfig struct {
 	//
 	// If zero, then it is considered that Func scans all the samples passed to it.
 	samplesScannedPerCall int
+
+	// The minimum interval for staleness calculations.
+	minStalenessInterval time.Duration
 }
 
 func (rc *rollupConfig) getTimestamps() []int64 {
@@ -723,8 +723,8 @@ func (rc *rollupConfig) doInternal(dstValues []float64, tsm *timeseriesMap, valu
 	if rc.LookbackDelta > 0 && maxPrevInterval > rc.LookbackDelta {
 		maxPrevInterval = rc.LookbackDelta
 	}
-	if *minStalenessInterval > 0 {
-		if msi := minStalenessInterval.Milliseconds(); msi > 0 && maxPrevInterval < msi {
+	if rc.minStalenessInterval > 0 {
+		if msi := rc.minStalenessInterval.Milliseconds(); msi > 0 && maxPrevInterval < msi {
 			maxPrevInterval = msi
 		}
 	}
@@ -2443,13 +2443,14 @@ func rollupFake(_ *rollupFuncArg) float64 {
 	return 0
 }
 
+// getScalar expects result from a [scalar](https://prometheus.io/docs/prometheus/latest/querying/basics/#expression-language-data-types).
 func getScalar(arg any, argNum int) ([]float64, error) {
 	ts, ok := arg.([]*timeseries)
 	if !ok {
-		return nil, fmt.Errorf(`unexpected type for arg #%d; got %T; want %T`, argNum+1, arg, ts)
+		return nil, fmt.Errorf(`arg #%d must be a scalar`, argNum+1)
 	}
 	if len(ts) != 1 {
-		return nil, fmt.Errorf(`arg #%d must contain a single timeseries; got %d timeseries`, argNum+1, len(ts))
+		return nil, fmt.Errorf(`arg #%d must be a scalar`, argNum+1)
 	}
 	return ts[0].Values, nil
 }
@@ -2466,14 +2467,15 @@ func getIntNumber(arg any, argNum int) (int, error) {
 	return n, nil
 }
 
+// getString expects result from a string expression, which contains a single timeseries with only NaN values.
 func getString(tss []*timeseries, argNum int) (string, error) {
 	if len(tss) != 1 {
-		return "", fmt.Errorf(`arg #%d must contain a single timeseries; got %d timeseries`, argNum+1, len(tss))
+		return "", fmt.Errorf(`arg #%d must be a string`, argNum+1)
 	}
 	ts := tss[0]
 	for _, v := range ts.Values {
 		if !math.IsNaN(v) {
-			return "", fmt.Errorf(`arg #%d contains non-string timeseries`, argNum+1)
+			return "", fmt.Errorf(`arg #%d must be a string`, argNum+1)
 		}
 	}
 	return string(ts.MetricName.MetricGroup), nil

app/vmselect/promql/transform.go

@@ -903,7 +903,6 @@ func transformHistogramQuantile(tfa *transformFuncArg) ([]*timeseries, error) {
 
 	// Convert buckets with `vmrange` labels to buckets with `le` labels.
 	tss := vmrangeBucketsToLE(args[1])
-
 	// Parse boundsLabel. See https://github.com/prometheus/prometheus/issues/5706 for details.
 	var boundsLabel string
 	if len(args) > 2 {
@@ -1050,9 +1049,15 @@ func fixBrokenBuckets(i int, xss []leTimeseries) {
 		return
 	}
 
+	vNext := 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
+	}
 	// Substitute upper bucket values with lower bucket values if the upper values are NaN
 	// or are bigger than the lower bucket values.
-	vNext := xss[0].ts.Values[i]
 	for j := 1; j < len(xss); j++ {
 		v := xss[j].ts.Values[i]
 		if math.IsNaN(v) || vNext > v {

app/vmselect/promql/transform_test.go

@@ -37,6 +37,9 @@ func TestFixBrokenBuckets(t *testing.T) {
 	f([]float64{5, 1, 2, 3, nan}, []float64{5, 5, 5, 5, 5})
 	f([]float64{1, 5, 2, nan, 6, 3}, []float64{1, 5, 5, 5, 6, 6})
 	f([]float64{5, 10, 4, 3}, []float64{5, 10, 10, 10})
+	f([]float64{nan, 2, nan, 5}, []float64{0, 2, 2, 5})
+	f([]float64{nan, nan, 4, 5}, []float64{0, 0, 4, 5})
+	f([]float64{nan, nan, nan, 4}, []float64{0, 0, 0, 4})
 }
 
 func TestFixBrokenBucketsMultipleValues(t *testing.T) {
@@ -44,12 +47,11 @@ func TestFixBrokenBucketsMultipleValues(t *testing.T) {
 		t.Helper()
 		xss := make([]leTimeseries, len(values))
 		for i, v := range values {
-
 			xss[i].ts = &timeseries{
 				Values: v,
 			}
 		}
-		for i := range len(values) - 1 {
+		for i := range len(values[0]) {
 			fixBrokenBuckets(i, xss)
 		}
 		result := make([][]float64, len(values))
@@ -61,6 +63,8 @@ func TestFixBrokenBucketsMultipleValues(t *testing.T) {
 		}
 	}
 	f([][]float64{{10, 1}, {11, 2}, {13, 3}}, [][]float64{{10, 1}, {11, 2}, {13, 3}})
+	f([][]float64{{nan, nan}, {11, 2}, {13, 3}}, [][]float64{{0, 0}, {11, 2}, {13, 3}})
+	f([][]float64{{nan, nan, nan}, {11, 2, 3}, {13, 3, 4}}, [][]float64{{0, 0, 0}, {11, 2, 3}, {13, 3, 4}})
 }
 
 func TestVmrangeBucketsToLE(t *testing.T) {

app/vmselect/vmui/index.html

@@ -36,7 +36,7 @@
   <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-BT5pWGkz.js"></script>
+  <script type="module" crossorigin src="./assets/index-Ck5nH8JI.js"></script>
   <link rel="modulepreload" crossorigin href="./assets/vendor-BVRvRxZ2.js">
   <link rel="stylesheet" crossorigin href="./assets/vendor-D1GxaB_c.css">
   <link rel="stylesheet" crossorigin href="./assets/index-BHg4iVVe.css">

app/vmstorage/main.go

@@ -486,6 +486,7 @@ func writeStorageMetrics(w io.Writer, strg *storage.Storage) {
 
 	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_free_disk_space_bytes{path=%q}`, *DataPath), fs.MustGetFreeSpace(*DataPath))
 	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_free_disk_space_limit_bytes{path=%q}`, *DataPath), uint64(minFreeDiskSpaceBytes.N))
+	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_total_disk_space_bytes{path=%q}`, *DataPath), fs.MustGetTotalSpace(*DataPath))
 
 	isReadOnly := 0
 	if strg.IsReadOnly() {

app/vmui/Dockerfile-web

@@ -1,4 +1,4 @@
-FROM golang:1.24.5 AS build-web-stage
+FROM golang:1.25.0 AS build-web-stage
 COPY build /build
 
 WORKDIR /build

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

@@ -15,3 +15,24 @@ export const getExportDataUrl = (server: string, query: string, period: TimePara
   if (reduceMemUsage) params.set("reduce_mem_usage", "1");
   return `${server}/api/v1/export?${params}`;
 };
+
+export const getExportCSVDataUrl = (server: string, query: string[], period: TimeParams, reduceMemUsage: boolean): string => {
+  const params = new URLSearchParams({
+    start: period.start.toString(),
+    end: period.end.toString(),
+    format: "__name__,__value__,__timestamp__:unix_ms",
+  });
+  query.forEach((q => params.append("match[]", q)));
+  if (reduceMemUsage) params.set("reduce_mem_usage", "1");
+  return `${server}/api/v1/export/csv?${params}`;
+};
+
+export const getExportJSONDataUrl = (server: string, query: string[], period: TimeParams, reduceMemUsage: boolean): string => {
+  const params = new URLSearchParams({
+    start: period.start.toString(),
+    end: period.end.toString(),
+  });
+  query.forEach((q => params.append("match[]", q)));
+  if (reduceMemUsage) params.set("reduce_mem_usage", "1");
+  return `${server}/api/v1/export?${params}`;
+};

app/vmui/packages/vmui/src/components/DownloadButton/DownloadButton.tsx

@@ -1,20 +1,18 @@
-import { FC, useCallback } from "preact/compat";
+import { useCallback, useRef } from "preact/compat";
 import Tooltip from "../Main/Tooltip/Tooltip";
 import Button from "../Main/Button/Button";
 import { DownloadIcon } from "../Main/Icons";
 import Popper from "../Main/Popper/Popper";
-import { useRef } from "react";
 import "./style.scss";
 import useBoolean from "../../hooks/useBoolean";
 
-interface DownloadButtonProps {
+interface DownloadButtonProps<T extends string> {
   title: string;
-  downloadFormatOptions?: string[];
-  onDownload: (format?: string) => void;
+  downloadFormatOptions?: T[];
+  onDownload: (format?: T) => void;
 }
 
-/** TODO: Currently unused, later will be added for the exporting metrics */
-const DownloadButton: FC<DownloadButtonProps> = ({ title, downloadFormatOptions, onDownload }) => {
+const DownloadButton = <T extends string>({ title, downloadFormatOptions, onDownload }: DownloadButtonProps<T>) => {
   const {
     value: isPopupOpen,
     setTrue: onOpenPopup,
@@ -35,9 +33,19 @@ const DownloadButton: FC<DownloadButtonProps> = ({ title, downloadFormatOptions,
     }
   }, [onDownload, onClosePopup, isPopupOpen, onOpenPopup]);
 
+  const isDownloadFormat = useCallback((format: string): format is T => {
+    return (downloadFormatOptions as string[])?.includes(format);
+  }, [downloadFormatOptions]);
+
   const onDownloadFormatClick = useCallback((event: Event) => {
     const button = event.currentTarget as HTMLButtonElement;
-    onDownload(button.textContent ?? undefined);
+    const format = button.textContent;
+    if (format && isDownloadFormat(format)) {
+      onDownload(format);
+    } else {
+      onDownload();
+    }
+    onClosePopup();
   }, [onDownload]);
 
   return (

app/vmui/packages/vmui/src/components/Main/Icons/index.tsx

@@ -578,97 +578,13 @@ export const CommentIcon = () => (
   </svg>
 );
 
-export const FilterIcon = () => (
+export const DebugIcon = () => (
   <svg
     viewBox="0 0 24 24"
     fill="currentColor"
   >
     <path
-      d="M4.25 5.61C6.27 8.2 10 13 10 13v6c0 .55.45 1 1 1h2c.55 0 1-.45 1-1v-6s3.72-4.8 5.74-7.39c.51-.66.04-1.61-.79-1.61H5.04c-.83 0-1.3.95-.79 1.61"
-    ></path>
-  </svg>
-);
-
-export const FilterOffIcon = () => (
-  <svg
-    viewBox="0 0 24 24"
-    fill="currentColor"
-  >
-    <path
-      d="M19.79 5.61C20.3 4.95 19.83 4 19 4H6.83l7.97 7.97zM2.81 2.81 1.39 4.22 10 13v6c0 .55.45 1 1 1h2c.55 0 1-.45 1-1v-2.17l5.78 5.78 1.41-1.41z"
-    ></path>
-  </svg>
-);
-
-export const OpenNewIcon = () => (
-  <svg
-    viewBox="0 0 24 24"
-    fill="currentColor"
-  >
-    <path
-      d="M19 19H5V5h7V3H5c-1.11 0-2 .9-2 2v14c0 1.1.89 2 2 2h14c1.1 0 2-.9 2-2v-7h-2zM14 3v2h3.59l-9.83 9.83 1.41 1.41L19 6.41V10h2V3z"
-    ></path>
-  </svg>
-);
-
-export const ModalIcon = () => (
-  <svg
-    viewBox="0 0 24 24"
-    fill="currentColor"
-  >
-    <path d="M19 4H5c-1.11 0-2 .9-2 2v12c0 1.1.89 2 2 2h14c1.1 0 2-.9 2-2V6c0-1.1-.89-2-2-2m0 14H5V8h14z"></path>
-  </svg>
-);
-
-export const PauseIcon = () => (
-  <svg
-    viewBox="0 0 24 24"
-    fill="currentColor"
-  >
-    <path d="M6 19h4V5H6v14zm8-14v14h4V5h-4z" />
-  </svg>
-);
-
-export const ScrollToTopIcon = () => (
-  <svg
-    viewBox="0 0 24 24"
-    fill="currentColor"
-  >
-    <path
-      d="M8 12l4-4 4 4m-4-4v12"
-      strokeWidth="2"
-      stroke="currentColor"
-      fill="none"
+      d="M20 8h-2.81c-.45-.78-1.07-1.45-1.82-1.96L17 4.41 15.59 3l-2.17 2.17C12.96 5.06 12.49 5 12 5c-.49 0-.96.06-1.41.17L8.41 3 7 4.41l1.62 1.63C7.88 6.55 7.26 7.22 6.81 8H4v2h2.09c-.05.33-.09.66-.09 1v1H4v2h2v1c0 .34.04.67.09 1H4v2h2.81c1.04 1.79 2.97 3 5.19 3s4.15-1.21 5.19-3H20v-2h-2.09c.05-.33.09-.66.09-1v-1h2v-2h-2v-1c0-.34-.04-.67-.09-1H20V8zm-6 8h-4v-2h4v2zm0-4h-4v-2h4v2z"
     />
   </svg>
 );
-
-export const SortIcon = () => (
-  <svg
-    viewBox="0 0 24 24"
-    fill="currentColor"
-  >
-    <path d="M4 3 L4 15 L1.5 15 L5.5 21 L9.5 15 L7 15 L7 3 Z"/>
-    <path d="M13 21 L13 9 L10.5 9 L14.5 3 L18.5 9 L16 9 L16 21 Z"/>
-  </svg>
-);
-
-export const SortArrowDownIcon = () => (
-  <svg
-    viewBox="0 0 24 24"
-    fill="currentColor"
-  >
-    <path d="M10.5 3 L10.5 15 L8 15 L12 21 L16 15 L13.5 15 L13.5 3 Z"/>
-  </svg>
-);
-
-export const SortArrowUpIcon = () => (
-  <svg
-    viewBox="0 0 24 24"
-    fill="currentColor"
-  >
-    <path d="M10.5 21 L10.5 9 L8 9 L12 3 L16 9 L13.5 9 L13.5 21 Z"/>
-  </svg>
-);
-
-

app/vmui/packages/vmui/src/hooks/useFetchQuery.ts

@@ -152,7 +152,7 @@ export const useFetchQuery = ({
         counter++;
       }
 
-      const limitText = `Showing ${tempData.length} series out of ${totalLength} series due to performance reasons. Please narrow down the query, so it returns less series`;
+      const limitText = `Showing ${tempData.length} series out of ${totalLength} series due to performance reasons. Please narrow down the query, so it returns fewer series`;
       setWarning(totalLength > seriesLimit ? limitText : "");
       isDisplayChart ? setGraphData(tempData as MetricResult[]) : setLiveData(tempData as InstantMetricResult[]);
       setTraces(tempTraces);

app/vmui/packages/vmui/src/pages/CustomPanel/DownloadReport/DownloadReport.tsx

@@ -1,5 +1,5 @@
 import { FC, useCallback, useEffect, useRef, useState } from "preact/compat";
-import { DownloadIcon } from "../../../components/Main/Icons";
+import { DebugIcon } from "../../../components/Main/Icons";
 import Button from "../../../components/Main/Button/Button";
 import Tooltip from "../../../components/Main/Tooltip/Tooltip";
 import useBoolean from "../../../hooks/useBoolean";
@@ -217,17 +217,17 @@ const DownloadReport: FC<Props> = ({ fetchUrl, reportType = ReportType.QUERY_DAT
 
   return (
     <>
-      <Tooltip title={"Export query"}>
+      <Tooltip title={"Debug query"}>
         <Button
           variant="text"
-          startIcon={<DownloadIcon/>}
+          startIcon={<DebugIcon />}
           onClick={toggleOpen}
-          ariaLabel="export query"
+          ariaLabel="Debug query"
         />
       </Tooltip>
       {openModal && (
         <Modal
-          title={"Export query"}
+          title={"Debug query"}
           onClose={handleClose}
           isOpen={openModal}
         >

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

@@ -1,4 +1,4 @@
-import { FC, useEffect, useState } from "preact/compat";
+import { FC, useEffect, useState, useMemo, useRef, useCallback } from "preact/compat";
 import QueryConfigurator from "./QueryConfigurator/QueryConfigurator";
 import { useFetchQuery } from "../../hooks/useFetchQuery";
 import { DisplayTypeSwitch } from "./DisplayTypeSwitch";
@@ -12,13 +12,17 @@ import Alert from "../../components/Main/Alert/Alert";
 import classNames from "classnames";
 import useDeviceDetect from "../../hooks/useDeviceDetect";
 import InstantQueryTip from "./InstantQueryTip/InstantQueryTip";
-import { useRef } from "react";
 import CustomPanelTraces from "./CustomPanelTraces/CustomPanelTraces";
 import WarningLimitSeries from "./WarningLimitSeries/WarningLimitSeries";
 import CustomPanelTabs from "./CustomPanelTabs";
 import { DisplayType } from "../../types";
 import DownloadReport from "./DownloadReport/DownloadReport";
 import WarningHeatmapToLine from "./WarningHeatmapToLine/WarningHeatmapToLine";
+import DownloadButton from "../../components/DownloadButton/DownloadButton";
+import { downloadCSV, downloadJSON } from "../../utils/file";
+import { convertMetricsDataToCSV } from "./utils";
+
+type ExportFormats = "csv" | "json";
 
 const CustomPanel: FC = () => {
   useSetQueryParams();
@@ -55,6 +59,27 @@ const CustomPanel: FC = () => {
     showAllSeries
   });
 
+  const fileDownloaders = useMemo(() => {
+    const getFilename = (format: ExportFormats) => {
+      return `vmui_export_${query.join("_")}.${format}`;
+    };
+
+    return {
+      csv: async () => {
+        if(!liveData) return;
+        const csvData = convertMetricsDataToCSV(liveData);
+        downloadCSV(csvData, getFilename("csv"));
+      },
+      json: async () => {
+        downloadJSON(JSON.stringify(liveData), getFilename("json"));
+      },
+    };
+  }, [liveData, query]);
+
+  const onDownloadClick = useCallback((format?: ExportFormats) => {
+    format && fileDownloaders[format]();
+  }, [fileDownloaders]);
+
   const showInstantQueryTip = !liveData?.length && (displayType !== DisplayType.chart);
   const showError = !hideError && error;
 
@@ -110,7 +135,7 @@ const CustomPanel: FC = () => {
           "vm-block_mobile": isMobile,
         })}
       >
-        {isLoading && <LineLoader />}
+        {isLoading && <LineLoader/>}
         <div
           className="vm-custom-panel-body-header"
           ref={controlsRef}
@@ -118,7 +143,13 @@ const CustomPanel: FC = () => {
           <div className="vm-custom-panel-body-header__tabs">
             <DisplayTypeSwitch/>
           </div>
-          {(graphData || liveData) && <DownloadReport fetchUrl={fetchUrl}/>}
+          {displayType === "table" && (
+            <DownloadButton
+              title={"Export query"}
+              onDownload={onDownloadClick}
+              downloadFormatOptions={["json", "csv"]}
+            />)}
+          {(graphData || liveData) && displayType !== "code" && <DownloadReport fetchUrl={fetchUrl}/>}
         </div>
         <CustomPanelTabs
           graphData={graphData}

app/vmui/packages/vmui/src/pages/CustomPanel/utils.test.ts

@@ -0,0 +1,86 @@
+import { describe, expect, it } from "vitest";
+import { convertMetricsDataToCSV } from "./utils";
+import { InstantMetricResult } from "../../api/types";
+
+describe("convertMetricsDataToCSV", () => {
+  it("should return an empty string if headers are empty", () => {
+    const data: InstantMetricResult[] = [];
+    expect(convertMetricsDataToCSV(data)).toBe("");
+  });
+
+  it("should return a valid CSV string for single metric entry with value", () => {
+    const data: InstantMetricResult[] = [
+      {
+        value: [1623945600, "123"],
+        group: 0,
+        metric: {
+          header1: "123",
+          header2: "value2"
+        }
+      },
+    ];
+    const result = convertMetricsDataToCSV(data);
+    expect(result).toBe("header1,header2\n123,value2");
+  });
+
+  it("should return a valid CSV string for multiple metric entries with values", () => {
+    const data: InstantMetricResult[] = [
+      {
+        value: [1623945600, "123"],
+        group: 0,
+        metric: {
+          header1: "123",
+          header2: "value2"
+        }
+      },
+      {
+        value: [1623949200, "456"],
+        group: 0,
+        metric: {
+          header1: "456",
+          header2: "value4"
+        }
+      },
+    ];
+    const result = convertMetricsDataToCSV(data);
+    expect(result).toBe("header1,header2\n123,value2\n456,value4");
+  });
+
+  it("should handle metric entries with multiple values field", () => {
+    const data: InstantMetricResult[] = [
+      { 
+        values: [[1623945600, "123"], [1623949200, "456"]], 
+        group: 0, 
+        metric: {
+          header1: "123-456",
+          header2: "values"
+        } 
+      },
+    ];
+    const result = convertMetricsDataToCSV(data);
+    expect(result).toBe("header1,header2\n123-456,values");
+  });
+
+  it("should handle a combination of metric entries with value and values", () => {
+    const data: InstantMetricResult[] = [
+      { 
+        value: [1623945600, "123"], 
+        group: 0, 
+        metric: {
+          header1: "123",
+          header2: "first"
+        } 
+      },
+      { 
+        values: [[1623949200, "456"], [1623952800, "789"]], 
+        group: 0, 
+        metric: {
+          header1: "456-789",
+          header2: "second"
+        } 
+      },
+    ];
+    const result = convertMetricsDataToCSV(data);
+    expect(result).toBe("header1,header2\n123,first\n456-789,second");
+  });
+});

app/vmui/packages/vmui/src/pages/CustomPanel/utils.ts

@@ -0,0 +1,18 @@
+import { InstantMetricResult } from "../../api/types";
+import { getColumns, MetricCategory } from "../../hooks/useSortedCategories";
+import { formatValueToCSV } from "../../utils/csv";
+
+const getHeaders = (data: InstantMetricResult[]): string => {
+  return getColumns(data).map(({ key }) => key).join(",");
+};
+
+const getRows = (data: InstantMetricResult[], headers: MetricCategory[]) => {
+  return data?.map(d => headers.map(c => formatValueToCSV(d.metric[c.key] || "-")).join(","));
+};
+
+export const convertMetricsDataToCSV = (data: InstantMetricResult[]): string => {
+  const headers = getHeaders(data);
+  if (!headers.length) return "";
+  const rows = getRows(data, getColumns(data));
+  return [headers, ...rows].join("\n");
+};

app/vmui/packages/vmui/src/pages/RawQueryPage/hooks/useFetchExport.ts

@@ -1,13 +1,15 @@
 import { Dispatch, SetStateAction, useCallback, useEffect, useMemo, useRef, useState } from "preact/compat";
 import { MetricBase, MetricResult, ExportMetricResult } from "../../../api/types";
-import { ErrorTypes, SeriesLimits } from "../../../types";
+import { ErrorTypes, SeriesLimits, TimeParams } from "../../../types";
 import { useQueryState } from "../../../state/query/QueryStateContext";
 import { useTimeState } from "../../../state/time/TimeStateContext";
 import { useAppState } from "../../../state/common/StateContext";
 import { useCustomPanelState } from "../../../state/customPanel/CustomPanelStateContext";
 import { isValidHttpUrl } from "../../../utils/url";
-import { getExportDataUrl } from "../../../api/query-range";
+import { getExportCSVDataUrl, getExportDataUrl, getExportJSONDataUrl } from "../../../api/query-range";
 import { parseLineToJSON } from "../../../utils/json";
+import { downloadCSV, downloadJSON } from "../../../utils/file";
+import { useSnack } from "../../../contexts/Snackbar";
 
 interface FetchQueryParams {
   hideQuery?: number[];
@@ -16,6 +18,7 @@ interface FetchQueryParams {
 
 interface FetchQueryReturn {
   fetchUrl?: string[],
+  exportData: (format: ExportFormats) => void,
   isLoading: boolean,
   data?: MetricResult[],
   error?: ErrorTypes | string,
@@ -25,11 +28,16 @@ interface FetchQueryReturn {
   abortFetch: () => void
 }
 
+type ExportFormats = "csv" | "json";
+type FormatDownloader = (serverUrl: string, query: string[], period: TimeParams, reduceMemUsage: boolean) => void;
+type DownloadFileFormats = Record<ExportFormats, FormatDownloader>
+
 export const useFetchExport = ({ hideQuery, showAllSeries }: FetchQueryParams): FetchQueryReturn => {
   const { query } = useQueryState();
   const { period } = useTimeState();
   const { displayType, reduceMemUsage, seriesLimits: stateSeriesLimits } = useCustomPanelState();
   const { serverUrl } = useAppState();
+  const { showInfoMessage } = useSnack();
 
   const [isLoading, setIsLoading] = useState(false);
   const [data, setData] = useState<MetricResult[]>();
@@ -55,6 +63,35 @@ export const useFetchExport = ({ hideQuery, showAllSeries }: FetchQueryParams):
     }
   }, [serverUrl, period, hideQuery, reduceMemUsage]);
 
+  const fileDownloaders: DownloadFileFormats = useMemo(() => {
+    const getFilename = (format: ExportFormats) => `vmui_export_${query.join("_")}_${period.start}_${period.end}.${format}`;
+    return {
+      csv: async () => {
+        const url = getExportCSVDataUrl(serverUrl, query, period, reduceMemUsage);
+        const response = await fetch(url);
+        try {
+          let text = await response.text();
+          text = "name,value,timestamp\n" + text;
+          downloadCSV(text, getFilename("csv"));
+        } catch (e) {
+          console.error(e);
+          showInfoMessage({ text: "Couldn't fetch data for CSV export. Please try again", type: "error" });
+        }
+      },
+      json: async () => {
+        const url = getExportJSONDataUrl(serverUrl, query, period, reduceMemUsage);
+        try {
+          const response = await fetch(url);
+          const text = await response.text();
+          downloadJSON(text, getFilename("json"));
+        } catch (e) {
+          console.error(e);
+          showInfoMessage({ text: "Couldn't fetch data for JSON export. Please try again", type: "error" });
+        }
+      }
+    };
+  }, [query, period, serverUrl, reduceMemUsage]);
+
   const fetchData = useCallback(async ({ fetchUrl, stateSeriesLimits, showAllSeries }: {
     fetchUrl: string[];
     stateSeriesLimits: SeriesLimits;
@@ -131,7 +168,7 @@ export const useFetchExport = ({ hideQuery, showAllSeries }: FetchQueryParams):
 
         counter++;
       }
-      const limitText = `Showing ${tempData.length} series out of ${totalLength} series due to performance reasons. Please narrow down the query, so it returns less series`;
+      const limitText = `Showing ${tempData.length} series out of ${totalLength} series due to performance reasons. Please narrow down the query, so it returns fewer series`;
       setWarning(totalLength > seriesLimit ? limitText : "");
       setData(tempData as MetricResult[]);
       setIsLoading(false);
@@ -144,6 +181,12 @@ export const useFetchExport = ({ hideQuery, showAllSeries }: FetchQueryParams):
     }
   }, [displayType, hideQuery]);
 
+  const exportData = useCallback((format: ExportFormats) => {
+    if (error) return;
+    const updatedPeriod = { ...period };
+    fileDownloaders[format](serverUrl, query, updatedPeriod, reduceMemUsage);
+  }, [serverUrl, query, period, reduceMemUsage, error, fileDownloaders]);
+
   const abortFetch = useCallback(() => {
     abortControllerRef.current.abort();
     setData([]);
@@ -167,5 +210,6 @@ export const useFetchExport = ({ hideQuery, showAllSeries }: FetchQueryParams):
     setQueryErrors,
     warning,
     abortFetch,
+    exportData
   };
 };

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

@@ -1,4 +1,4 @@
-import { FC, useState } from "preact/compat";
+import { FC, useCallback, useState } from "preact/compat";
 import LineLoader from "../../components/Main/LineLoader/LineLoader";
 import { useCustomPanelState } from "../../state/customPanel/CustomPanelStateContext";
 import { useQueryState } from "../../state/query/QueryStateContext";
@@ -17,7 +17,7 @@ import { DisplayType } from "../../types";
 import Hyperlink from "../../components/Main/Hyperlink/Hyperlink";
 import { CloseIcon } from "../../components/Main/Icons";
 import Button from "../../components/Main/Button/Button";
-import DownloadReport, { ReportType } from "../CustomPanel/DownloadReport/DownloadReport";
+import DownloadButton from "../../components/DownloadButton/DownloadButton";
 
 const RawSamplesLink = () => (
   <Hyperlink
@@ -66,7 +66,7 @@ const RawQueryPage: FC = () => {
     queryErrors,
     setQueryErrors,
     abortFetch,
-    fetchUrl,
+    exportData
   } = useFetchExport({ hideQuery, showAllSeries });
 
   const controlsRef = useRef<HTMLDivElement>(null);
@@ -85,6 +85,11 @@ const RawQueryPage: FC = () => {
     setShowPageDescription(false);
   };
 
+  const onExportClick = useCallback(async (format?: "csv" | "json") => {
+    if (!format) return;
+    exportData(format);
+  }, [exportData]);
+
   return (
     <div
       className={classNames({
@@ -159,9 +164,10 @@ const RawQueryPage: FC = () => {
             <DisplayTypeSwitch tabFilter={(tab) => (tab.value !== DisplayType.table)}/>
           </div>
           {data && (
-            <DownloadReport
-              fetchUrl={fetchUrl}
-              reportType={ReportType.RAW_DATA}
+            <DownloadButton
+              title={"Export query"}
+              downloadFormatOptions={["json", "csv"]}
+              onDownload={onExportClick}
             />
           )}
         </div>

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

@@ -0,0 +1,34 @@
+import { describe, expect, it } from "vitest";
+import { formatValueToCSV } from "./csv";
+
+describe("formatValueToCSV", () => {
+  it("should wrap value in quotes if it contains a comma", () => {
+    const value = "hello,world";
+    const result = formatValueToCSV(value);
+    expect(result).toBe("\"hello,world\"");
+  });
+
+  it("should wrap value in quotes if it contains a newline", () => {
+    const value = "hello\nworld";
+    const result = formatValueToCSV(value);
+    expect(result).toBe("\"hello\nworld\"");
+  });
+
+  it("should escape quotes and wrap in quotes if value contains a double quote", () => {
+    const value = "hello \"world\"";
+    const result = formatValueToCSV(value);
+    expect(result).toBe("\"hello \"\"world\"\"\"");
+  });
+
+  it("should return the same value if it does not contain special characters", () => {
+    const value = "hello world";
+    const result = formatValueToCSV(value);
+    expect(result).toBe("hello world");
+  });
+
+  it("should handle empty strings correctly", () => {
+    const value = "";
+    const result = formatValueToCSV(value);
+    expect(result).toBe("");
+  });
+});

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

@@ -0,0 +1,4 @@
+export const formatValueToCSV= (value: string) =>
+  (value.includes(",") || value.includes("\n") || value.includes("\""))
+    ? "\"" + value.replace(/"/g, "\"\"") + "\""
+    : value;

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

@@ -11,38 +11,12 @@ export const downloadFile = (data: Blob, filename: string) => {
   URL.revokeObjectURL(url);
 };
 
-export const downloadCSV = (data: Record<string, string>[], filename: string) => {
-  const getHeader = (data: Record<string, string>[]) => {
-    const headersObj = data.reduce<Record<string, boolean>>((headers, row) => {
-      Object.keys(row).forEach((key) => {
-        if(key && !headers[key]){
-          headers[key] = true;
-        }
-      });
-      return headers;
-    }, {});
-    return Object.keys(headersObj);
-  };
-
-  const formatValueToCSV= (value: string) =>
-    (value.includes(",") || value.includes("\n") || value.includes("\""))
-      ? "\"" + value.replace(/"/g, "\"\"") + "\""
-      : value;
-
-  const convertToCSV = (data: Record<string, string>[]): string => {
-    const header = getHeader(data);
-    const rows = data.map(item =>
-      header.map(fieldName => item[fieldName] ? formatValueToCSV(item[fieldName]): "").join(",")
-    );
-    return [header.map(formatValueToCSV).join(","), ...rows].join("\r\n");
-  };
-
-  const csvContent = convertToCSV(data);
-  const blob = new Blob([csvContent], { type: "text/csv;charset=utf-8;" });
+export const downloadCSV = (data: string, filename: string) => {
+  const blob = new Blob([data],  { type: "text/csv;charset=utf-8;" });
   downloadFile(blob, filename);
 };
 
 export const downloadJSON = (data: string, filename: string) => {
   const blob = new Blob([data], { type: "application/json" });
   downloadFile(blob, filename);
-};
+};

apptest/client.go

@@ -6,6 +6,7 @@ import (
 	"net"
 	"net/http"
 	"net/url"
+	"regexp"
 	"strconv"
 	"strings"
 	"testing"
@@ -183,3 +184,32 @@ func (app *ServesMetrics) GetMetricsByPrefix(t *testing.T, prefix string) []floa
 	}
 	return values
 }
+
+func (app *ServesMetrics) GetMetricsByRegexp(t *testing.T, re *regexp.Regexp) []float64 {
+	t.Helper()
+
+	values := []float64{}
+
+	metrics, statusCode := app.cli.Get(t, app.metricsURL)
+	if statusCode != http.StatusOK {
+		t.Fatalf("unexpected status code: got %d, want %d", statusCode, http.StatusOK)
+	}
+	for _, metric := range strings.Split(metrics, "\n") {
+		if !re.MatchString(metric) {
+			continue
+		}
+
+		parts := strings.Split(metric, " ")
+		if len(parts) < 2 {
+			t.Fatalf("unexpected record format: got %q, want metric name and value separated by a space", metric)
+		}
+
+		value, err := strconv.ParseFloat(parts[len(parts)-1], 64)
+		if err != nil {
+			t.Fatalf("could not parse metric value %s: %v", metric, err)
+		}
+
+		values = append(values, value)
+	}
+	return values
+}

apptest/testcase.go

@@ -173,7 +173,7 @@ func (tc *TestCase) MustStartVmagent(instance string, flags []string, promScrape
 // vminsert, and one vmselect.
 //
 // Both Vmsingle and Vmcluster implement the PrometheusWriteQuerier used in
-// business logic tests to abstract out the infrasture.
+// business logic tests to abstract out the infrastructure.
 //
 // This type is not suitable for infrastructure tests where custom cluster
 // setups are often required.

apptest/tests/multilevel_test.go

@@ -17,7 +17,7 @@ func TestClusterMultilevelSelect(t *testing.T) {
 	//
 	// vmselect (L2) -> vmselect (L1) -> vmstorage <- vminsert
 	//
-	// vmisert writes data into vmstorage.
+	// vminsert writes data into vmstorage.
 	// vmselect (L2) reads that data via vmselect (L1).
 
 	vmstorage := tc.MustStartVmstorage("vmstorage", []string{

apptest/tests/vmagent_remotewrite_test.go

@@ -6,6 +6,7 @@ import (
 	"net/http/httptest"
 	"sync"
 	"testing"
+	"time"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/apptest"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
@@ -248,3 +249,88 @@ func TestSingleVMAgentDowngradeRemoteWriteProtocol(t *testing.T) {
 		t.Fatalf("unexpected number of dropped packets; got %d, want %d", actualPacketsDroppedCount, expectedPacketsDroppedTotal)
 	}
 }
+
+func TestSingleVMAgentDropOnOverload(t *testing.T) {
+	tc := apptest.NewTestCase(t)
+	defer tc.Stop()
+
+	remoteWriteSrv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusNoContent)
+	}))
+	defer remoteWriteSrv.Close()
+
+	remoteWriteSrv2 := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusServiceUnavailable)
+	}))
+	defer remoteWriteSrv2.Close()
+
+	vmagent := tc.MustStartVmagent("vmagent", []string{
+		`-remoteWrite.flushInterval=50ms`,
+		fmt.Sprintf(`-remoteWrite.url=%s/api/v1/write`, remoteWriteSrv.URL),
+		fmt.Sprintf(`-remoteWrite.url=%s/api/v1/write`, remoteWriteSrv2.URL),
+		"-remoteWrite.disableOnDiskQueue=true",
+		// use only 1 worker to get a full queue faster
+		"-remoteWrite.queues=1",
+		// fastqueue size is roughly memory.Allowed() / len(urls) / *maxRowsPerBlock / 100
+		// Use very large maxRowsPerBlock to get fastqueue of minimal length(2).
+		// See initRemoteWriteCtxs function in remotewrite.go for details.
+		"-remoteWrite.maxRowsPerBlock=1000000000",
+		"-remoteWrite.tmpDataPath=" + tc.Dir() + "/vmagent",
+	}, ``)
+
+	const (
+		retries = 20
+		period  = 100 * time.Millisecond
+	)
+
+	waitFor := func(f func() bool) {
+		t.Helper()
+		for i := 0; i < retries; i++ {
+			if f() {
+				return
+			}
+			time.Sleep(period)
+		}
+		t.Fatalf("timed out waiting for retry #%d", retries)
+	}
+
+	// Real remote write URLs are hidden in metrics
+	url1 := "1:secret-url"
+	url2 := "2:secret-url"
+
+	// Wait until first request got flushed to remote write server
+	vmagent.APIV1ImportPrometheusNoWaitFlush(t, []string{
+		"foo_bar 1 1652169600000", // 2022-05-10T08:00:00Z
+	}, apptest.QueryOpts{})
+
+	waitFor(
+		func() bool {
+			return vmagent.RemoteWriteRequests(t, url1) == 1 && vmagent.RemoteWriteRequests(t, url2) == 1
+		},
+	)
+
+	// Send 2 more requests, the first RW endpoint should receive everything, the second should add them to the queue
+	// since worker is busy with the first request.
+	for i := 0; i < 2; i++ {
+		vmagent.APIV1ImportPrometheusNoWaitFlush(t, []string{
+			"foo_bar 1 1652169600000", // 2022-05-10T08:00:00Z
+		}, apptest.QueryOpts{})
+
+		waitFor(
+			func() bool {
+				return vmagent.RemoteWriteRequests(t, url1) == 2+i && vmagent.RemoteWritePendingInmemoryBlocks(t, url2) == 1+i
+			},
+		)
+	}
+
+	// Send one more request.
+	vmagent.APIV1ImportPrometheusNoWaitFlush(t, []string{
+		"foo_bar 1 1652169600000", // 2022-05-10T08:00:00Z
+	}, apptest.QueryOpts{})
+
+	waitFor(
+		func() bool {
+			return vmagent.RemoteWriteRequests(t, url1) == 4 && vmagent.RemoteWriteSamplesDropped(t, url2) > 0
+		},
+	)
+}

apptest/tests/vmauth_routing_test.go

@@ -1,8 +1,10 @@
 package tests
 
 import (
+	"context"
 	"fmt"
 	"io"
+	"net"
 	"net/http"
 	"net/http/httptest"
 	"net/url"
@@ -301,3 +303,82 @@ unauthorized_user:
 	assertBackendsRequestsCount(1)
 
 }
+
+func TestSingleVMAuthUseProxyProtocol(t *testing.T) {
+	tc := apptest.NewTestCase(t)
+	defer tc.Stop()
+
+	var requestsCount int
+	var actualForwardedForHeader string
+	backend := httptest.NewServer(http.HandlerFunc(func(_ http.ResponseWriter, r *http.Request) {
+		actualForwardedForHeader = r.Header.Get("X-Forwarded-For")
+		requestsCount++
+	}))
+	defer backend.Close()
+
+	authConfig := fmt.Sprintf(`
+unauthorized_user:
+  url_prefix: %s
+  `, backend.URL)
+
+	vmauth := tc.MustStartVmauth("vmauth", []string{
+		"-httpListenAddr.useProxyProtocol=true",
+	}, authConfig)
+
+	req, err := http.NewRequest("GET", fmt.Sprintf("http://%s/backend", vmauth.GetHTTPListenAddr()), nil)
+	if err != nil {
+		t.Fatalf("cannot build http.Request: %s", err)
+	}
+
+	// make request using proxy protocol
+	c := &http.Client{
+		Transport: &http.Transport{
+			DialContext: func(_ context.Context, network, addr string) (net.Conn, error) {
+				conn, err := net.Dial(network, addr)
+				if err != nil {
+					return nil, err
+				}
+
+				// Write a proxy protocol header to the connection
+				if _, err := conn.Write([]byte{
+					0x0D, 0x0A, 0x0D, 0x0A, 0x00, 0x0D, 0x0A, 0x51, 0x55, 0x49, 0x54, 0x0A, // signature
+					0x21,       // version 2
+					0x11,       // family IPv4
+					0x00, 0x0C, // length: 12 bytes (IPv4 + ports)
+					192, 168, 1, 100, // source IP
+					10, 0, 0, 1, // destination IP
+					0x1F, 0x90, // source port 8080
+					0x00, 0x50, // destination port 80
+				}); err != nil {
+					t.Fatalf("cannot send proxy protocol header: %s", err)
+				}
+
+				return conn, nil
+			},
+		},
+	}
+
+	resp, err := c.Do(req)
+	if err != nil {
+		t.Fatalf("cannot make http.Get request for target=%q: %s", req.URL, err)
+	}
+	responseText, err := io.ReadAll(resp.Body)
+	if err != nil {
+		t.Fatalf("cannot read response body: %s", err)
+	}
+	resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("unexpected http response code: %d, want: %d, response text: %s", resp.StatusCode, http.StatusOK, responseText)
+	}
+
+	// ensure that request was proxied
+	if requestsCount != 1 {
+		t.Fatalf("expected to have %d unauthorized proxied requests, got: %d", 1, requestsCount)
+	}
+
+	// ensure that X-Forwarded-For header is set to the source IP from proxy protocol
+	expectedForwardedForHeader := "192.168.1.100"
+	if actualForwardedForHeader != expectedForwardedForHeader {
+		t.Fatalf("expected X-Forwarded-For header to be equal to proxy source IP, got: %s, want: %s'", actualForwardedForHeader, expectedForwardedForHeader)
+	}
+}

apptest/vmagent.go

@@ -104,6 +104,36 @@ func (app *Vmagent) RemoteWritePacketsDroppedTotal(t *testing.T) int {
 	return int(total)
 }
 
+// RemoteWriteSamplesDropped sums up the total number of dropped remote write samples for given remote write URL.
+func (app *Vmagent) RemoteWriteSamplesDropped(t *testing.T, url string) int {
+	re := regexp.MustCompile(fmt.Sprintf("vmagent_remotewrite_samples_dropped_total{.*url=%q.*}", url))
+	total := 0.0
+	for _, v := range app.GetMetricsByRegexp(t, re) {
+		total += v
+	}
+	return int(total)
+}
+
+// RemoteWritePendingInmemoryBlocks sums up the total number of pending in-memory blocks for given remote write URL.
+func (app *Vmagent) RemoteWritePendingInmemoryBlocks(t *testing.T, url string) int {
+	re := regexp.MustCompile(fmt.Sprintf("vmagent_remotewrite_pending_inmemory_blocks{.*url=%q.*}", url))
+	total := 0.0
+	for _, v := range app.GetMetricsByRegexp(t, re) {
+		total += v
+	}
+	return int(total)
+}
+
+// RemoteWriteRequests sums up the total number of sending requests for given remote write URL.
+func (app *Vmagent) RemoteWriteRequests(t *testing.T, url string) int {
+	re := regexp.MustCompile(fmt.Sprintf("vmagent_remotewrite_requests_total{.*url=%q.*}", url))
+	total := 0.0
+	for _, v := range app.GetMetricsByRegexp(t, re) {
+		total += v
+	}
+	return int(total)
+}
+
 // ReloadRelabelConfigs sends SIGHUP to trigger relabel config reload
 // and waits until vmagent_relabel_config_reloads_total increases.
 // Fails the test if no reload is detected within 3 seconds.
@@ -123,9 +153,7 @@ func (app *Vmagent) ReloadRelabelConfigs(t *testing.T) {
 		time.Sleep(100 * time.Millisecond)
 	}
 
-	if currTotal <= prevTotal {
-		t.Fatalf("relabel configs were not reloaded after SIGHUP signal; previous total: %f, current total: %f", prevTotal, currTotal)
-	}
+	t.Fatalf("relabel configs were not reloaded after SIGHUP signal; previous total: %f, current total: %f", prevTotal, currTotal)
 }
 
 // sendBlocking sends the data to vmstorage by executing `send` function and

apptest/vmauth.go

@@ -56,25 +56,32 @@ func StartVmauth(instance string, flags []string, cli *Client, configFilePath st
 	}, nil
 }
 
-// UpdateConfiguration performs configuration file reload for app and waits for configuration apply
-//
-// Due to second prescision of config reload metric, config cannot be reloaded more than 1 time in a second
+// UpdateConfiguration updates the vmauth configuration file with the provided YAML content,
+// sends SIGHUP to trigger config reload
+// and waits until vmauth_config_last_reload_total increases.
+// Fails the test if no reload is detected within 2 seconds.
 func (app *Vmauth) UpdateConfiguration(t *testing.T, configFileYAML string) {
 	t.Helper()
-	ct := int(time.Now().Unix())
+
 	fs.MustWriteSync(app.configFilePath, []byte(configFileYAML))
+
+	prevTotal := app.GetIntMetric(t, "vmauth_config_last_reload_total")
+
 	if err := app.process.Signal(syscall.SIGHUP); err != nil {
 		t.Fatalf("unexpected signal error: %s", err)
 	}
-	for range 10 {
-		ts := app.GetIntMetric(t, "vmauth_config_last_reload_success_timestamp_seconds")
-		if ts < ct {
-			time.Sleep(time.Millisecond * 100)
-			continue
+
+	var currTotal int
+	for range 20 {
+		currTotal = app.GetIntMetric(t, "vmauth_config_last_reload_total")
+		if currTotal > prevTotal {
+			return
 		}
-		return
+
+		time.Sleep(time.Millisecond * 100)
 	}
-	t.Fatalf("timeout waiting for config reload success")
+
+	t.Fatalf("config were not reloaded after SIGHUP signal; previous total: %d, current total: %d", prevTotal, currTotal)
 }
 
 // GetHTTPListenAddr returns listen http addr

apptest/vminsert.go

@@ -49,7 +49,7 @@ func StartVminsert(instance string, flags []string, cli *Client, output io.Write
 		graphiteListenAddrRE,
 		openTSDBListenAddrRE,
 	}
-	// Add storateNode REs to block until vminsert establishes connections with
+	// Add storageNode REs to block until vminsert establishes connections with
 	// all storage nodes. The extracted values are unused.
 	for _, sn := range storageNodes(flags) {
 		logRecord := fmt.Sprintf("successfully dialed -storageNode=\"%s\"", sn)

benchmarks/Makefile

@@ -34,26 +34,28 @@
 # for details
 tsbs: tsbs-build tsbs-generate-data tsbs-load-data tsbs-generate-queries tsbs-run-queries
 
-TSBS_SCALE := 100000
-# If GNU date is available, use it; otherwise, fall back to the standard date command
-# User can install GNU date on macOS via `brew install coreutils`
-DATE_CMD := $(shell which gdate 2>/dev/null || echo date)
-TSBS_START := $(shell $(DATE_CMD) -u -d "1 day ago 00:00:00" +"%Y-%m-%dT%H:%M:%SZ")
-TSBS_END   := $(shell $(DATE_CMD) -u -d "00:00:00" +"%Y-%m-%dT%H:%M:%SZ")
-TSBS_STEP := 80s
-TSBS_QUERIES := 1000
-TSBS_WORKERS := 4
+TSBS_SCALE ?= 100000
+TSBS_END ?= $(shell date -u +%Y-%m-%dT00:00:00Z)
+TSBS_START ?= $(shell \
+  NOW=$$(date -u +%s); \
+  START=$$((NOW - 86400)); \
+  date -u -d "@$$START" +%Y-%m-%dT00:00:00Z 2>/dev/null || \
+  date -u -r $$START +%Y-%m-%dT00:00:00Z 2>/dev/null \
+)
+TSBS_STEP ?= 80s
+TSBS_QUERIES ?= 1000
+TSBS_WORKERS ?= 4
 TSBS_DATA_FILE := /tmp/tsbs-data-$(TSBS_SCALE)-$(TSBS_START)-$(TSBS_END)-$(TSBS_STEP).gz
 TSBS_QUERY_FILE := /tmp/tsbs-queries-$(TSBS_SCALE)-$(TSBS_START)-$(TSBS_END)-$(TSBS_QUERIES).gz
 # For cluster setup use http://vminsert:8480/insert/0/influx/write
-TSBS_WRITE_URLS := http://localhost:8428/write
+TSBS_WRITE_URLS ?= http://localhost:8428/write
 # For cluster setup use http://vmselect:8481/select/0/prometheus
-TSBS_READ_URLS := http://localhost:8428
-TSBS_METRICS_URL := http://localhost:8428/metrics
+TSBS_READ_URLS ?= http://localhost:8428
+TSBS_METRICS_URL ?= http://localhost:8428/metrics
 
 # Build TSBS tools
 tsbs-build:
-	test -d /tmp/tsbs || (git clone https://github.com/timescale/tsbs.git /tmp/tsbs && \
+	test -d /tmp/tsbs/cmd/tsbs_run_queries_victoriametrics || (git clone https://github.com/timescale/tsbs.git /tmp/tsbs && \
 		cd /tmp/tsbs/cmd/tsbs_generate_data && GOBIN=/tmp/tsbs/bin go install && \
 		cd /tmp/tsbs/cmd/tsbs_generate_queries && GOBIN=/tmp/tsbs/bin go install && \
 		cd /tmp/tsbs/cmd/tsbs_load_victoriametrics && GOBIN=/tmp/tsbs/bin go install && \

dashboards/victoriametrics-cluster.json

@@ -7962,7 +7962,7 @@
             "type": "prometheus",
             "uid": "$ds"
           },
-          "description": "99th percentile of the number of time series read per query.",
+          "description": "Shows the max 99th percentile of time series read per query across all vmselects.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -8051,7 +8051,7 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(histogram_quantile(0.99, sum(rate(vm_series_read_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
+              "expr": "max(histogram_quantile(0.99, sum(rate(vm_series_read_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
               "format": "time_series",
               "interval": "",
               "intervalFactor": 1,
@@ -8068,7 +8068,7 @@
             "type": "prometheus",
             "uid": "$ds"
           },
-          "description": "99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) read per queried time series.",
+          "description": "Shows the max 99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) read per time series.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -8157,7 +8157,7 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(histogram_quantile(0.99, sum(rate(vm_rows_read_per_series_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
+              "expr": "max(histogram_quantile(0.99, sum(rate(vm_rows_read_per_series_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
               "format": "time_series",
               "interval": "",
               "intervalFactor": 1,
@@ -8174,7 +8174,7 @@
             "type": "prometheus",
             "uid": "$ds"
           },
-          "description": "99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) read per query.",
+          "description": "Shows the max 99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) read per query.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -8263,7 +8263,7 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(histogram_quantile(0.99, sum(rate(vm_rows_read_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
+              "expr": "max(histogram_quantile(0.99, sum(rate(vm_rows_read_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
               "format": "time_series",
               "interval": "",
               "intervalFactor": 1,
@@ -8280,7 +8280,7 @@
             "type": "prometheus",
             "uid": "$ds"
           },
-          "description": "99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) scanner per query.\n\nThis number can exceed number of DatapointsReadPerQuery if `step` query arg passed to [/api/v1/query_range](https://prometheus.io/docs/prometheus/latest/querying/api/#range-queries) is smaller than the lookbehind window set in square brackets of [rollup function](https://docs.victoriametrics.com/victoriametrics/metricsql/#rollup-functions). For example, if `increase(some_metric[1h])` is executed with the `step=5m`, then the same [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) on a hour time range are scanned `1h/5m=12` times. See [this article](https://valyala.medium.com/how-to-optimize-promql-and-metricsql-queries-85a1b75bf986) for details.",
+          "description": "Shows the max 99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) scanner per query.\n\nThis number can exceed number of DatapointsReadPerQuery if `step` query arg passed to [/api/v1/query_range](https://prometheus.io/docs/prometheus/latest/querying/api/#range-queries) is smaller than the lookbehind window set in square brackets of [rollup function](https://docs.victoriametrics.com/victoriametrics/metricsql/#rollup-functions). For example, if `increase(some_metric[1h])` is executed with the `step=5m`, then the same [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) on a hour time range are scanned `1h/5m=12` times. See [this article](https://valyala.medium.com/how-to-optimize-promql-and-metricsql-queries-85a1b75bf986) for details.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -8369,7 +8369,7 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(histogram_quantile(0.99, sum(rate(vm_rows_scanned_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
+              "expr": "max(histogram_quantile(0.99, sum(rate(vm_rows_scanned_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
               "format": "time_series",
               "interval": "",
               "intervalFactor": 1,

dashboards/vm/victoriametrics-cluster.json

@@ -7963,7 +7963,7 @@
             "type": "victoriametrics-metrics-datasource",
             "uid": "$ds"
           },
-          "description": "99th percentile of the number of time series read per query.",
+          "description": "Shows the max 99th percentile of time series read per query across all vmselects.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -8052,7 +8052,7 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(histogram_quantile(0.99, sum(rate(vm_series_read_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
+              "expr": "max(histogram_quantile(0.99, sum(rate(vm_series_read_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
               "format": "time_series",
               "interval": "",
               "intervalFactor": 1,
@@ -8069,7 +8069,7 @@
             "type": "victoriametrics-metrics-datasource",
             "uid": "$ds"
           },
-          "description": "99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) read per queried time series.",
+          "description": "Shows the max 99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) read per time series.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -8158,7 +8158,7 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(histogram_quantile(0.99, sum(rate(vm_rows_read_per_series_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
+              "expr": "max(histogram_quantile(0.99, sum(rate(vm_rows_read_per_series_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
               "format": "time_series",
               "interval": "",
               "intervalFactor": 1,
@@ -8175,7 +8175,7 @@
             "type": "victoriametrics-metrics-datasource",
             "uid": "$ds"
           },
-          "description": "99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) read per query.",
+          "description": "Shows the max 99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) read per query.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -8264,7 +8264,7 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(histogram_quantile(0.99, sum(rate(vm_rows_read_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
+              "expr": "max(histogram_quantile(0.99, sum(rate(vm_rows_read_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
               "format": "time_series",
               "interval": "",
               "intervalFactor": 1,
@@ -8281,7 +8281,7 @@
             "type": "victoriametrics-metrics-datasource",
             "uid": "$ds"
           },
-          "description": "99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) scanner per query.\n\nThis number can exceed number of DatapointsReadPerQuery if `step` query arg passed to [/api/v1/query_range](https://victoriametrics-metrics-datasource.io/docs/victoriametrics-metrics-datasource/latest/querying/api/#range-queries) is smaller than the lookbehind window set in square brackets of [rollup function](https://docs.victoriametrics.com/victoriametrics/metricsql/#rollup-functions). For example, if `increase(some_metric[1h])` is executed with the `step=5m`, then the same [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) on a hour time range are scanned `1h/5m=12` times. See [this article](https://valyala.medium.com/how-to-optimize-promql-and-metricsql-queries-85a1b75bf986) for details.",
+          "description": "Shows the max 99th percentile of number of [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) scanner per query.\n\nThis number can exceed number of DatapointsReadPerQuery if `step` query arg passed to [/api/v1/query_range](https://victoriametrics-metrics-datasource.io/docs/victoriametrics-metrics-datasource/latest/querying/api/#range-queries) is smaller than the lookbehind window set in square brackets of [rollup function](https://docs.victoriametrics.com/victoriametrics/metricsql/#rollup-functions). For example, if `increase(some_metric[1h])` is executed with the `step=5m`, then the same [data samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) on a hour time range are scanned `1h/5m=12` times. See [this article](https://valyala.medium.com/how-to-optimize-promql-and-metricsql-queries-85a1b75bf986) for details.",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -8370,7 +8370,7 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(histogram_quantile(0.99, sum(rate(vm_rows_scanned_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
+              "expr": "max(histogram_quantile(0.99, sum(rate(vm_rows_scanned_per_query_bucket{job=~\"$job_select\", instance=~\"$instance\"}[$__rate_interval])) by (instance, vmrange)))",
               "format": "time_series",
               "interval": "",
               "intervalFactor": 1,

dashboards/vm/vmagent.json

@@ -7966,7 +7966,7 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(rate(vm_promscrape_scraped_samples_sum{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)\n+ sum(rate(vmagent_rows_inserted_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)",
+              "expr": "sum(rate({__name__=~\"vm_promscrape_scraped_samples_sum|vmagent_rows_inserted_total\",job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)",
               "hide": false,
               "interval": "",
               "legendFormat": "in {{instance}} {{job}}",
@@ -8244,4 +8244,4 @@
   "uid": "G7Z9GzMGz_vm",
   "version": 1,
   "weekStart": ""
-}
+}

dashboards/vmagent.json

@@ -7965,7 +7965,7 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(rate(vm_promscrape_scraped_samples_sum{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)\n+ sum(rate(vmagent_rows_inserted_total{job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)",
+              "expr": "sum(rate({__name__=~\"vm_promscrape_scraped_samples_sum|vmagent_rows_inserted_total\",job=~\"$job\", instance=~\"$instance\"}[$__rate_interval])) by(job, instance)",
               "hide": false,
               "interval": "",
               "legendFormat": "in {{instance}} {{job}}",
@@ -8243,4 +8243,4 @@
   "uid": "G7Z9GzMGz",
   "version": 1,
   "weekStart": ""
-}
+}

deployment/docker/Makefile

@@ -7,7 +7,7 @@ ROOT_IMAGE ?= alpine:3.22.1
 ROOT_IMAGE_SCRATCH ?= scratch
 CERTS_IMAGE := alpine:3.22.1
 
-GO_BUILDER_IMAGE := golang:1.24.5-alpine
+GO_BUILDER_IMAGE := golang:1.25.0-alpine
 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 :/ __)
 DOCKER ?= docker
@@ -43,7 +43,7 @@ app-via-docker: package-builder
 		$(BUILDER_IMAGE) \
 		go build $(RACE) -trimpath -buildvcs=false \
 			-ldflags "-extldflags '-static' $(GO_BUILDINFO)" \
-			-tags 'netgo osusergo musl' \
+			-tags 'netgo osusergo musl $(EXTRA_GO_BUILD_TAGS)' \
 			-o bin/$(APP_NAME)$(APP_SUFFIX)-prod $(PKG_PREFIX)/app/$(APP_NAME)
 
 app-via-docker-windows: package-builder
@@ -58,7 +58,7 @@ app-via-docker-windows: package-builder
 		$(BUILDER_IMAGE) \
 		go build $(RACE) -trimpath -buildvcs=false \
 			-ldflags "-s -w -extldflags '-static' $(GO_BUILDINFO)" \
-			-tags 'netgo osusergo' \
+			-tags 'netgo osusergo $(EXTRA_GO_BUILD_TAGS)' \
 			-o bin/$(APP_NAME)-windows$(APP_SUFFIX)-prod.exe $(PKG_PREFIX)/app/$(APP_NAME)
 
 package-via-docker: package-base

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:v1.122.0
+    image: victoriametrics/vmagent:v1.124.0
     depends_on:
       - "vmauth"
     ports:
@@ -35,14 +35,14 @@ 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:v1.122.0-cluster
+    image: victoriametrics/vmstorage:v1.124.0-cluster
     volumes:
       - strgdata-1:/storage
     command:
       - "--storageDataPath=/storage"
     restart: always
   vmstorage-2:
-    image: victoriametrics/vmstorage:v1.122.0-cluster
+    image: victoriametrics/vmstorage:v1.124.0-cluster
     volumes:
       - strgdata-2:/storage
     command:
@@ -52,7 +52,7 @@ 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:v1.122.0-cluster
+    image: victoriametrics/vminsert:v1.124.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -61,7 +61,7 @@ services:
       - "--storageNode=vmstorage-2:8400"
     restart: always
   vminsert-2:
-    image: victoriametrics/vminsert:v1.122.0-cluster
+    image: victoriametrics/vminsert:v1.124.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -73,7 +73,7 @@ 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:v1.122.0-cluster
+    image: victoriametrics/vmselect:v1.124.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -83,7 +83,7 @@ services:
       - "--vmalert.proxyURL=http://vmalert:8880"
     restart: always
   vmselect-2:
-    image: victoriametrics/vmselect:v1.122.0-cluster
+    image: victoriametrics/vmselect:v1.124.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -98,7 +98,7 @@ services:
   # read requests from Grafana, vmui, vmalert among vmselects.
   # It can be used as an authentication proxy.
   vmauth:
-    image: victoriametrics/vmauth:v1.122.0
+    image: victoriametrics/vmauth:v1.124.0
     depends_on:
       - "vmselect-1"
       - "vmselect-2"
@@ -112,7 +112,7 @@ services:
 
   # vmalert executes alerting and recording rules
   vmalert:
-    image: victoriametrics/vmalert:v1.122.0
+    image: victoriametrics/vmalert:v1.124.0
     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:v1.122.0
+    image: victoriametrics/vmagent:v1.124.0
     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:v1.122.0
+    image: victoriametrics/victoria-metrics:v1.124.0
     ports:
       - 8428:8428
       - 8089:8089
@@ -54,7 +54,7 @@ services:
 
   # vmalert executes alerting and recording rules
   vmalert:
-    image: victoriametrics/vmalert:v1.122.0
+    image: victoriametrics/vmalert:v1.124.0
     depends_on:
       - "victoriametrics"
       - "alertmanager"

deployment/docker/rules/alerts-health.yml

@@ -7,7 +7,7 @@ groups:
     # note the `job` filter and update accordingly to your setup
     rules:
       - alert: TooManyRestarts
-        expr: changes(process_start_time_seconds{job=~".*(victoriametrics|vmselect|vminsert|vmstorage|vmagent|vmalert|vmsingle|vmalertmanager|vmauth|victorialogs|vlstorage|vlselect|vlinsert).*"}[15m]) > 2
+        expr: changes(process_start_time_seconds{job=~".*(victoriametrics|vmselect|vminsert|vmstorage|vmagent|vmalert|vmsingle|vmalertmanager|vmauth).*"}[15m]) > 2
         labels:
           severity: critical
         annotations:
@@ -17,7 +17,7 @@ groups:
             It might be crashlooping.
 
       - alert: ServiceDown
-        expr: up{job=~".*(victoriametrics|vmselect|vminsert|vmstorage|vmagent|vmalert|vmsingle|vmalertmanager|vmauth|victorialogs|vlstorage|vlselect|vlinsert).*"} == 0
+        expr: up{job=~".*(victoriametrics|vmselect|vminsert|vmstorage|vmagent|vmalert|vmsingle|vmalertmanager|vmauth).*"} == 0
         for: 2m
         labels:
           severity: critical
@@ -59,7 +59,7 @@ groups:
             Consider to either increase available CPU resources or decrease the load on the process.
 
       - alert: TooHighGoroutineSchedulingLatency
-        expr: histogram_quantile(0.99, sum(rate(go_sched_latencies_seconds_bucket[5m])) by (le, job, instance)) > 0.1
+        expr: histogram_quantile(0.99, sum(rate(go_sched_latencies_seconds_bucket{job=~".*(victoriametrics|vmselect|vminsert|vmstorage|vmagent|vmalert|vmsingle|vmalertmanager|vmauth).*"}[5m])) by (le, job, instance)) > 0.1
         for: 15m
         labels:
           severity: critical

deployment/docker/rules/alerts.yml

@@ -100,7 +100,7 @@ groups:
           summary: "Churn rate is more than 10% on \"{{ $labels.instance }}\" for the last 15m"
           description: "VM constantly creates new time series on \"{{ $labels.instance }}\".\n
             This effect is known as Churn Rate.\n
-            High Churn Rate tightly connected with database performance and may
+            High Churn Rate is tightly connected with database performance and may
             result in unexpected OOM's or slow queries."
 
       - alert: TooHighChurnRate24h
@@ -117,7 +117,7 @@ groups:
           description: "The number of created new time series over last 24h is 3x times higher than
             current number of active series on \"{{ $labels.instance }}\".\n
             This effect is known as Churn Rate.\n
-            High Churn Rate tightly connected with database performance and may
+            High Churn Rate is tightly connected with database performance and may
             result in unexpected OOM's or slow queries."
 
       - alert: TooHighSlowInsertsRate
@@ -135,4 +135,4 @@ groups:
           summary: "Percentage of slow inserts is more than 5% on \"{{ $labels.instance }}\" for the last 15m"
           description: "High rate of slow inserts on \"{{ $labels.instance }}\" may be a sign of resource exhaustion
             for the current load. It is likely more RAM is needed for optimal handling of the current number of active time series.
-            See also https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3976#issuecomment-1476883183"
+            See also https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3976#issuecomment-1476883183"

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

@@ -1,6 +1,6 @@
 services:
   vmagent:
-    image: victoriametrics/vmagent:v1.122.0
+    image: victoriametrics/vmagent:v1.124.0
     depends_on:
       - "victoriametrics"
     ports:
@@ -14,7 +14,7 @@ services:
     restart: always
 
   victoriametrics:
-    image: victoriametrics/victoria-metrics:v1.122.0
+    image: victoriametrics/victoria-metrics:v1.124.0
     ports:
       - 8428:8428
     volumes:
@@ -40,7 +40,7 @@ services:
     restart: always
 
   vmalert:
-    image: victoriametrics/vmalert:v1.122.0
+    image: victoriametrics/vmalert:v1.124.0
     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:v1.25.2
+    image: victoriametrics/vmanomaly:v1.25.3
     depends_on:
       - "victoriametrics"
     ports:

deployment/docker/vmanomaly/vmanomaly-node-exporter-preset/dashboard.json

@@ -1005,7 +1005,7 @@
           "refId": "A"
         }
       ],
-      "title": "Anoamlies: Read Latency",
+      "title": "Anomalies: Read Latency",
       "type": "state-timeline"
     },
     {

deployment/logs-benchmark/docker-compose-elk.yml

@@ -18,7 +18,7 @@ services:
       - vlogs
 
   generator:
-    image: golang:1.24.5-alpine
+    image: golang:1.25.0-alpine
     restart: always
     working_dir: /go/src/app
     volumes:

deployment/logs-benchmark/docker-compose-loki.yml

@@ -2,7 +2,7 @@ version: "3"
 
 services:
   generator:
-    image: golang:1.24.5-alpine
+    image: golang:1.25.0-alpine
     restart: always
     working_dir: /go/src/app
     volumes:

deployment/logs-benchmark/readme.md

@@ -4,12 +4,12 @@ Benchmark compares VictoriaLogs with ELK stack and Grafana Loki.
 
 Benchmark is based on:
 
-- Logs from this repository - https://github.com/logpai/loghub
+- Logs from this repository - [https://github.com/logpai/loghub](https://github.com/logpai/loghub)
 - [logs generator](./generator)
 
 For ELK suite it uses:
 
-- filebeat - https://www.elastic.co/beats/filebeat
+- filebeat - [https://www.elastic.co/beats/filebeat](https://www.elastic.co/beats/filebeat)
 - elastic + kibana
 
 For Grafana Loki suite it uses:
@@ -24,7 +24,7 @@ For Grafana Loki suite it uses:
 
 - VictoriaLogs instance
 - vmsingle - port forwarded to `localhost:8428` to see UI
-- exporters for system metris
+- exporters for system metrics
 
 ELK suite uses [docker-compose-elk.yml](./docker-compose-elk.yml) with the following services:
 
@@ -54,7 +54,7 @@ Each filebeat than writes logs to elastic and VictoriaLogs via elasticsearch-com
 1. Download and unarchive logs by running:
 
 ```shell
-cd source_logs 
+cd source_logs
 bash download.sh
 ```
 
@@ -74,11 +74,11 @@ Unarchived logs size per file for reference:
  13G hadoop-*.log
 ```
 
-2. (optional) If needed, adjust amount of logs sent by generator by modifying `-outputRateLimitItems` and
+1. (optional) If needed, adjust amount of logs sent by generator by modifying `-outputRateLimitItems` and
    `outputRateLimitPeriod` parameters in [docker-compose.yml](./docker-compose.yml). By default, it is configured to
    send 10000 logs per second.
 
-3. (optional) Build victoria-logs image and adjust `image` parameter in [docker-compose.yml](./docker-compose.yml):
+1. (optional) Build victoria-logs image and adjust `image` parameter in [docker-compose.yml](./docker-compose.yml):
 
 ```shell
 make package-victoria-logs
@@ -95,26 +95,27 @@ output.elasticsearch:
   hosts: [ "http://vlogs:9428/insert/elasticsearch/" ]
 ```
 
-4. Choose a suite to run.
+1. Choose a suite to run.
 
    In order to run ELK suite use the following command:
-   ```
+
+   ```sh
    make docker-up-elk
    ```
+
    In order to run Loki suite use the following command:
-   ```
+
+   ```sh
    make docker-up-loki
    ```
 
-
-5. Navigate to `http://localhost:3000/` to see Grafana dashboards with resource usage
+1. Navigate to `http://localhost:3000/` to see Grafana dashboards with resource usage
    comparison.
 
    Navigate to `http://localhost:3000/d/hkm6P6_4z/elastic-vs-vlogs` to see ELK suite results.
 
    Navigate to `http://localhost:3000/d/hkm6P6_4y/loki-vs-vlogs` to see Loki suite results.
 
-
 Example results vs ELK:
 
 ![elk-grafana-dashboard.png](results/elk-grafana-dashboard.png)

docs/anomaly-detection/CHANGELOG.md

@@ -14,6 +14,19 @@ aliases:
 ---
 Please find the changelog for VictoriaMetrics Anomaly Detection below.
 
+## v1.25.3
+Released: 2025-08-19
+
+- FEATURE: Added forecasting capabilities to the [`ProphetModel`](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet) this allows users to generate *future* (point-wise and interval) predictions with offsets defined by `forecast_at` argument (e.g. `['1d', '1w']`) at *current* timestamp and store these in respective series, e.g. `yhat_1d`, `yhat_lower_1d`, `yhat_upper_1d`, etc. This feature is particularly useful for scenarios where future predictions are needed, such as capacity planning or trend analysis. See [FAQ](https://docs.victoriametrics.com/anomaly-detection/faq/#forecasting) for more details.
+
+- IMPROVEMENT: Added `logger_levels` argument to `settings` [config section](https://docs.victoriametrics.com/anomaly-detection/components/settings/#logger-levels) to allow setting specific log levels for individual components. Useful for debugging specific components. For example, `logger_levels: { "reader.vm": "DEBUG" }` will set the log level for the `VmReader` component to `DEBUG`, while leaving other components at their default log levels. Also is supported in [hot reload](https://docs.victoriametrics.com/anomaly-detection/components/#hot-reload) mode, allowing for dynamic log level changes without service restarts.
+
+- IMPROVEMENT: Added logging of URLs used for querying VictoriaMetrics TSDB in [`VmReader`](https://docs.victoriametrics.com/anomaly-detection/components/reader/#vm-reader) to ease the debugging of incomplete data retrieval, incorrect endpoints, or misconfigured tenant IDs. The URLs are logged at the `DEBUG` level, so you can control their verbosity using the `--loggerLevelComponents` argument with `reader.vm=DEBUG` or `reader=DEBUG` to see the URLs in the logs.
+
+- IMPROVEMENT: Added `offset` [argument](https://docs.victoriametrics.com/anomaly-detection/components/reader/#vm-reader) to `VmReader` on reader and query levels to allow for flexible time offset adjustments in the reader. Useful for correcting for data collection delays. The `offset` can be specified as a string (e.g., "15s", "-20s") and will be applied to all queries processed by the reader. See [FAQ](https://docs.victoriametrics.com/anomaly-detection/faq/#using-offsets) for more details.
+
+- BUGFIX: Resolved the issue where symlink-ed configuration files were not properly processed by [hot reload](https://docs.victoriametrics.com/anomaly-detection/components/#hot-reload) mechanism, leading to the service not picking up changes made to the original files. Now it properly resolves symlinks and reloads the configuration when the original file is modified.
+
 ## v1.25.2
 Released: 2025-07-30
 

docs/anomaly-detection/FAQ.md

@@ -54,6 +54,25 @@ Respective config is defined in a [`reader`](https://docs.victoriametrics.com/an
 ## Handling noisy input data
 `vmanomaly` operates on data fetched from VictoriaMetrics using [MetricsQL](https://docs.victoriametrics.com/victoriametrics/metricsql/) queries, so the initial data quality can be fine-tuned with aggregation, grouping, and filtering to reduce noise and improve anomaly detection accuracy.
 
+## Using offsets
+`vmanomaly` supports {{% available_from "v1.25.3" anomaly %}} the use of offsets in the [`reader`](https://docs.victoriametrics.com/anomaly-detection/components/reader/#vm-reader) section to adjust the time range of the data being queried. This can be particularly useful for correcting for data collection delays or other timing issues. It can be also defined or overridden on [per-query basis](https://docs.victoriametrics.com/anomaly-detection/components/reader/#per-query-parameters).
+
+For example, if you want to query data with a 60-second delay (e.g. data collection happened 1 sec ago, however, timestamps written to VictoriaMetrics are 60 seconds in the past), you can set the `offset` argument to `-60s` in the reader section:
+
+```yaml
+reader:
+  class: 'vm'
+  datasource_url: 'http://localhost:8428'
+  sampling_period: '10s'
+  offset: '-60s'
+  queries:
+    vmb:
+      expr: 'avg(vm_blocks)'
+    cpu_custom_offset:
+      expr: 'avg(rate(vm_cpu_usage[5m]))'
+      offset: '-30s'  # this will override the global offset for this query only
+```
+
 ## Handling timezones
 
 `vmanomaly` supports timezone-aware anomaly detection {{% available_from "v1.18.0" anomaly %}} through a `tz` argument, available both at the [reader level](https://docs.victoriametrics.com/anomaly-detection/components/reader#vm-reader) and at the [query level](https://docs.victoriametrics.com/anomaly-detection/components/reader/#per-query-parameters).
@@ -179,6 +198,22 @@ While `vmanomaly` detects anomalies and produces scores, it *does not directly g
 
 <img src="https://docs.victoriametrics.com/anomaly-detection/guides/guide-vmanomaly-vmalert/guide-vmanomaly-vmalert_overview.webp" alt="node_exporter_example_diagram" style="width:60%"/>
 
+Once anomaly scores are written back to VictoriaMetrics, you can use [MetricsQL](https://docs.victoriametrics.com/victoriametrics/metricsql/) expressions subset in `vmalert` to define alerting rules based on these scores. Reasonable defaults are `anomaly_score > 1`:
+
+```yaml
+groups:
+- name: vmanomaly_alerts
+  rules:
+  - alert: HighAnomalyScore
+    expr: anomaly_score > 1  # or similar expressions, like `min(anomaly_score{...}) by (...) > 1`
+    for: 5m
+    labels:
+      severity: warning
+    annotations:
+      summary: "Anomaly score > 1 for {{ $labels.for }} query"
+      description: "Anomaly score is {{ $value }} for query {{ $labels.for }}. Value: {{ $value }}."
+```
+
 ## Preventing alert fatigue
 Produced anomaly scores are designed in such a way that values from 0.0 to 1.0 indicate non-anomalous data, while a value greater than 1.0 is generally classified as an anomaly. However, there are no perfect models for anomaly detection, that's why reasonable defaults expressions like `anomaly_score > 1` may not work 100% of the time. However, anomaly scores, produced by `vmanomaly` are written back as metrics to VictoriaMetrics, where tools like [`vmalert`](https://docs.victoriametrics.com/victoriametrics/vmalert/) can use [MetricsQL](https://docs.victoriametrics.com/victoriametrics/metricsql/) expressions to fine-tune alerting thresholds and conditions, balancing between avoiding [false negatives](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#false-negative) and reducing [false positives](https://victoriametrics.com/blog/victoriametrics-anomaly-detection-handbook-chapter-1/#false-positive).
 
@@ -228,6 +263,117 @@ writer:
 
 Configuration above will produce N intervals of full length (`fit_window`=14d + `fit_every`=1h) until `to_iso` timestamp is reached to run N consecutive `fit` calls to train models; Then these models will be used to produce `M = [fit_every / sampling_frequency]` infer datapoints for `fit_every` range at the end of each such interval, imitating M consecutive calls of `infer_every` in `PeriodicScheduler` [config](https://docs.victoriametrics.com/anomaly-detection/components/scheduler#periodic-scheduler). These datapoints then will be written back to VictoriaMetrics TSDB, defined in `writer` [section](https://docs.victoriametrics.com/anomaly-detection/components/writer#vm-writer) for further visualization (i.e. in VMUI or Grafana)
 
+## Forecasting
+
+Not intended for forecasting in its core, `vmanomaly` can still be used to produce forecasts using [ProphetModel](https://docs.victoriametrics.com/anomaly-detection/components/models#prophet) {{% available_from "v1.25.3" anomaly %}}, which can be helpful in scenarios like capacity planning, resource allocation, or trend analysis, if the underlying data is complex and can't be handled by inline MetricsQL queries, including [predict_linear](https://docs.victoriametrics.com/victoriametrics/metricsql/#predict_linear).
+
+> 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**.
+
+Here's an example of how to produce forecasts using `vmanomaly` and combine it with the regular model, e.g. to estimate daily outcomes for a disk usage metric:
+
+```yaml
+# https://docs.victoriametrics.com/anomaly-detection/components/scheduler/#periodic-scheduler
+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_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_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
+reader:
+  class: 'vm'
+  datasource_url: 'http://play.victoriametrics.com'
+  tenant_id: '0:0'
+  sampling_period: '5m'
+  # other reader params ...
+  queries:
+    disk_usage_perc_5m:
+      expr: |
+        max_over_time(
+          1 - (node_filesystem_avail_bytes{mountpoint="/",fstype!="rootfs"} 
+          / 
+          node_filesystem_size_bytes{mountpoint="/",fstype!="rootfs"}),
+          1h
+        )
+      data_range: [0, 1]
+      # step: '1m'  # default will be inherited from sampling_period
+    disk_usage_perc_1d:
+      expr: |
+        max_over_time(
+          1 - (node_filesystem_avail_bytes{mountpoint="/",fstype!="rootfs"} 
+          / 
+          node_filesystem_size_bytes{mountpoint="/",fstype!="rootfs"}),
+          24h
+        )
+      step: '1d'  # override default step to 1d, as we want to produce daily forecasts
+      data_range: [0, 1]
+# https://docs.victoriametrics.com/anomaly-detection/components/models/
+models:
+  quantile_5m:
+    class: 'quantile_online'  # online model, which updates itself each infer call
+    queries: ['disk_usage_perc_5m']
+    schedulers: ['periodic_5m']
+    clip_predictions: True
+    detection_direction: 'above_expected'  # as we are interested in spikes in capacity planning
+    quantiles: [0.25, 0.5, 0.75]  # to produce median and upper quartiles
+    iqr_threshold: 2.0
+
+  prophet_1d:
+    class: 'prophet'
+    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
+    # https://facebook.github.io/prophet/docs/quick_start#python-api
+    args:
+      interval_width: 0.98  # see https://facebook.github.io/prophet/docs/uncertainty_intervals
+      country_holidays: 'US'
+# https://docs.victoriametrics.com/anomaly-detection/components/writer/#vm-writer
+writer:
+  class: 'vm'
+  datasource_url: '{your_victoriametrics_url_for_writing}'
+  # tenant_id: '0:0' # or your tenant ID if using clustered VictoriaMetrics
+  # other writer params ...
+  # https://docs.victoriametrics.com/anomaly-detection/components/writer/#metrics-formatting
+  metric_format:
+    __name__: $VAR
+    for: $QUERY_KEY
+```
+
+Then, respective alerts can be configured in [`vmalert`](https://docs.victoriametrics.com/victoriametrics/vmalert/) to notify disk exhaustion risks, e.g. if the forecasted disk usage exceeds 90% in the next 3 days:
+
+```yaml
+groups:
+- name: disk_usage_alerts
+  rules:
+  - alert: DiskUsageHigh
+    expr: |
+      yhat_7d{for="disk_usage_perc_1d"} > 0.9
+    for: 24h
+    labels:
+      severity: critical
+    annotations:
+      summary: "Disk usage is forecasted to exceed 90% in the next 3 days"
+      description: "Disk usage is forecasted to exceed 90% in the next 3 days for instance {{ $labels.instance }}. Forecasted value: {{ $value }}."
+  - alert: DiskUsageCritical
+    expr: |
+      yhat_3d{for="disk_usage_perc_1d"} > 0.95
+    for: 24h
+    labels:
+      severity: critical
+    annotations:
+      summary: "Disk usage is forecasted to exceed 95% in the next 3 days"
+      description: "Disk usage is forecasted to exceed 95% in the next 3 days for instance {{ $labels.instance }}. Forecasted value: {{ $value }}."
+```
+
 ## Resource consumption of vmanomaly
 `vmanomaly` itself is a lightweight service, resource usage is primarily dependent on [scheduling](https://docs.victoriametrics.com/anomaly-detection/components/scheduler) (how often and on what data to fit/infer your models), [# and size of timeseries returned by your queries](https://docs.victoriametrics.com/anomaly-detection/components/reader/#vm-reader), and the complexity of the employed [models](https://docs.victoriametrics.com/anomaly-detection/components/models). Its resource usage is directly related to these factors, making it adaptable to various operational scales. Various optimizations are available to balance between RAM usage, processing speed, and model capacity. These options are described in the sections below.
 
@@ -243,7 +389,7 @@ services:
   # ...
   vmanomaly:
     container_name: vmanomaly
-    image: victoriametrics/vmanomaly:v1.25.2
+    image: victoriametrics/vmanomaly:v1.25.3
     # ...
     ports:
       - "8490:8490"
@@ -456,7 +602,7 @@ 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:v1.25.2 && docker image tag victoriametrics/vmanomaly:v1.25.2 vmanomaly
+docker pull victoriametrics/vmanomaly:v1.25.3 && docker image tag victoriametrics/vmanomaly:v1.25.3 vmanomaly
 ```
 
 ```sh

docs/anomaly-detection/QuickStart.md

@@ -121,13 +121,13 @@ Below are the steps to get `vmanomaly` up and running inside a Docker container:
 1. Pull Docker image:
 
 ```sh
-docker pull victoriametrics/vmanomaly:v1.25.2
+docker pull victoriametrics/vmanomaly:v1.25.3
 ```
 
 2. (Optional step) tag the `vmanomaly` Docker image:
 
 ```sh
-docker image tag victoriametrics/vmanomaly:v1.25.2 vmanomaly
+docker image tag victoriametrics/vmanomaly:v1.25.3 vmanomaly
 ```
 
 3. Start the `vmanomaly` Docker container with a *license file*, use the command below.
@@ -163,7 +163,7 @@ docker run -it --user 1000:1000 \
 services:
   # ...
   vmanomaly:
-    image: victoriametrics/vmanomaly:v1.25.2
+    image: victoriametrics/vmanomaly:v1.25.3
     volumes:
         $YOUR_LICENSE_FILE_PATH:/license
         $YOUR_CONFIG_FILE_PATH:/config.yml
@@ -220,6 +220,14 @@ settings:
   n_workers: 4  # number of workers to run workload in parallel, set to 0 or negative number to use all available CPU cores
   anomaly_score_outside_data_range: 5.0  # default anomaly score for anomalies outside expected data range
   restore_state: True  # restore state from previous run, available since v1.24.0
+  # https://docs.victoriametrics.com/anomaly-detection/components/settings/#logger-levels
+  # to override service-global logger levels, use the `logger_levels` section
+  logger_levels:  
+    # vmanomaly: info
+    # scheduler: info
+    # reader: info
+    # writer: info
+    model.prophet: warning
 
 schedulers:
   1d_1m:
@@ -299,6 +307,9 @@ For optimal service behavior, consider the following tweaks when configuring `vm
 - Set up [anomaly score dashboard](https://docs.victoriametrics.com/anomaly-detection/presets/#grafana-dashboard) to visualize the results of anomaly detection.
 - Set up [self-monitoring dashboard](https://docs.victoriametrics.com/anomaly-detection/self-monitoring/) to monitor the health of `vmanomaly` service and its components.
 
+**Logging**:
+- Tune logging levels in the `settings.logger_levels` [section](https://docs.victoriametrics.com/anomaly-detection/components/settings/#logger-levels) to control the verbosity of logs. This can help in debugging and monitoring the service behavior, as well as in disabling excessive logging for production environments.
+
 ## Check also
 
 Please refer to the following links for a deeper understanding of Anomaly Detection and `vmanomaly`:

docs/anomaly-detection/components/models.md

@@ -652,7 +652,7 @@ models:
 
 > `ProphetModel` is a [univariate](#univariate-models), [non-rolling](#non-rolling-models), [offline](#offline-models) model.
 
-> {{% available_from "v1.18.2" anomaly %}} the format for `tz_seasonalities` has been updated to enhance flexibility. Previously, it accepted a list of strings (e.g., `['hod', 'minute']`). Now, it follows the same structure as custom seasonalities defined in the `seasonalities` argument (e.g., `{"name": "hod", "fourier_order": 5, "mode": "additive"}`). This change is backward-compatible, so older configurations will be automatically converted to the new format using default values.
+> {{% 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*:
 
@@ -661,7 +661,11 @@ models:
 - `scale`{{% available_from "v1.18.0" anomaly %}} (float): 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.
 - `tz_aware`{{% available_from "v1.18.0" anomaly %}} (bool): Enables handling of timezone-aware timestamps. Default is `False`. Should be used with `tz_seasonalities` and `tz_use_cyclical_encoding` parameters.
 - `tz_seasonalities`{{% available_from "v1.18.0" anomaly %}} (list[dict]): Specifies timezone-aware seasonal components. Requires `tz_aware=True`. Supported options include `minute`, `hod` (hour of day), `dow` (day of week), and `month` (month of year). {{% available_from "v1.18.2" anomaly %}} users can configure additional parameters for each seasonality, such as `fourier_order`, `prior_scale`, and `mode`. For more details, please refer to the **Timezone-unaware** configuration example below.
+  > {{% available_from "v1.18.2" anomaly %}} the format for `tz_seasonalities` has been updated to enhance flexibility. Previously, it accepted a list of strings (e.g., `['hod', 'minute']`). Now, it follows the same structure as custom seasonalities defined in the `seasonalities` argument (e.g., `{"name": "hod", "fourier_order": 5, "mode": "additive"}`). This change is backward-compatible, so older configurations will be automatically converted to the new format using default values.
 - `tz_use_cyclical_encoding`{{% available_from "v1.18.0" anomaly %}} (bool): If set to `True`, applies [cyclical encoding technique](https://www.kaggle.com/code/avanwyk/encoding-cyclical-features-for-deep-learning) to timezone-aware seasonalities. Should be used with `tz_aware=True` and `tz_seasonalities`.
+- `forecast_at`{{% available_from "v1.25.3" anomaly %}} (list[str]): Specifies future relative offsets for which forecasts should be generated (e.g., `['1h', '1d']`). Works similarly to [predict_linear](https://docs.victoriametrics.com/victoriametrics/metricsql/#predict_linear) in MetricQL, but with more flexibility and seasonality support - produced series will have *the same timestamp* as the other [output](#vmanomaly-output) series, but with the forecasted value for the *future timestamp*. Defaults to `[]` (empty list, meaning no future forecasts are produced). If set, `provide_series` must include at least `yhat` for point-wise forecasts (and `yhat_lower` or/and `yhat_upper` for respective confidence intervals). For example, if `forecast_at` is set to `['1h', '1d']`, the model will produce forecasts for both the next hour and the next day, and these series can be accessed by `yhat_1h`, `yhat_lower_1h`, `yhat_upper_1h`, `yhat_1d`, `yhat_lower_1d`, and `yhat_upper_1d` in the output, respectively. See [FAQ](https://docs.victoriametrics.com/anomaly-detection/faq/#forecasting) for more details.
+
+> `forecast_at` parameter can lead to **significant increase in active timeseries** if you have a lot of time series returned by your queries, as it will produce additional series for each of the future timestamps specified in `forecast_at` (optionally multiplied by 1-3 if interval forecasts are included). For example, if you have 1000 time series returned by your query and set `forecast_at` to `[1h, 1d, 1w]`, and `provide_series` includes `yhat_lower` and `yhat_upper`, it will produce 1000 (series) * 3 (intervals) * 3 (predictions, point + interval) = 9000 additional timeseries. Consider using it only on small subset of metrics (e.g. grouped by `host` or `region`) to avoid this issue, as it also **proportionally (to the number of `forecast_at` elements) increases the timings of inference calls**.
 
 > Apart from standard [`vmanomaly` output](#vmanomaly-output), Prophet model can provide additional metrics.
 
@@ -1308,7 +1312,7 @@ monitoring:
 Let's pull the docker image for `vmanomaly`:
 
 ```sh
-docker pull victoriametrics/vmanomaly:v1.25.2
+docker pull victoriametrics/vmanomaly:v1.25.3
 ```
 
 Now we can run the docker container putting as volumes both config and model file:
@@ -1322,7 +1326,7 @@ docker run -it \
 -v $(PWD)/license:/license \
 -v $(PWD)/custom_model.py:/vmanomaly/model/custom.py \
 -v $(PWD)/custom.yaml:/config.yaml \
-victoriametrics/vmanomaly:v1.25.2 /config.yaml \
+victoriametrics/vmanomaly:v1.25.3 /config.yaml \
 --licenseFile=/license
 ```