docs: improve wording of query stats

Signed-off-by: hagen1778 <roman@victoriametrics.com>

.github/pull_request_template.md

@@ -6,4 +6,4 @@ Please provide a brief description of the changes you made. Be as specific as po
 
 The following checks are **mandatory**:
 
-- [ ] My change adheres [VictoriaMetrics contributing guidelines](https://docs.victoriametrics.com/contributing/).
+- [ ] My change adheres to [VictoriaMetrics contributing guidelines](https://docs.victoriametrics.com/contributing/).

.github/workflows/docs.yaml

@@ -40,9 +40,8 @@ jobs:
       - name: Copy docs
         id: update
         run: |
-          rsync -zarv \
-            --exclude="Makefile" \
-            docs/ ../__vm-docs/content/victoriametrics
+          find docs -type d -maxdepth 1 -mindepth 1 -exec \
+            sh -c 'rsync -zarvh --delete {}/ ../__vm-docs/content/$(basename {})/' \;
           echo "SHORT_SHA=$(git rev-parse --short $GITHUB_SHA)" >> $GITHUB_OUTPUT
         working-directory: __vm
 

.gitignore

@@ -27,3 +27,4 @@ _site
 coverage.txt
 cspell.json
 *~
+deployment/docker/provisioning/plugins/

Makefile

@@ -18,7 +18,7 @@ TAR_OWNERSHIP ?= --owner=1000 --group=1000
 .PHONY: $(MAKECMDGOALS)
 
 include app/*/Makefile
-include cspell/Makefile
+include codespell/Makefile
 include docs/Makefile
 include deployment/*/Makefile
 include dashboards/Makefile
@@ -504,7 +504,7 @@ fmt:
 	gofmt -l -w -s ./apptest
 
 vet:
-	go vet ./lib/...
+	GOEXPERIMENT=synctest go vet ./lib/...
 	go vet ./app/...
 	go vet ./apptest/...
 
@@ -513,35 +513,35 @@ check-all: fmt vet golangci-lint govulncheck
 clean-checkers: remove-golangci-lint remove-govulncheck
 
 test:
-	go test ./lib/... ./app/...
+	GOEXPERIMENT=synctest go test ./lib/... ./app/...
 
 test-race:
-	go test -race ./lib/... ./app/...
+	GOEXPERIMENT=synctest go test -race ./lib/... ./app/...
 
 test-pure:
-	CGO_ENABLED=0 go test ./lib/... ./app/...
+	GOEXPERIMENT=synctest CGO_ENABLED=0 go test ./lib/... ./app/...
 
 test-full:
-	go test -coverprofile=coverage.txt -covermode=atomic ./lib/... ./app/...
+	GOEXPERIMENT=synctest go test -coverprofile=coverage.txt -covermode=atomic ./lib/... ./app/...
 
 test-full-386:
-	GOARCH=386 go test -coverprofile=coverage.txt -covermode=atomic ./lib/... ./app/...
+	GOEXPERIMENT=synctest GOARCH=386 go test -coverprofile=coverage.txt -covermode=atomic ./lib/... ./app/...
 
 integration-test: victoria-metrics vmagent vmalert vmauth
 	go test ./apptest/... -skip="^TestCluster.*"
 
 benchmark:
-	go test -bench=. ./lib/...
+	GOEXPERIMENT=synctest go test -bench=. ./lib/...
 	go test -bench=. ./app/...
 
 benchmark-pure:
-	CGO_ENABLED=0 go test -bench=. ./lib/...
+	GOEXPERIMENT=synctest CGO_ENABLED=0 go test -bench=. ./lib/...
 	CGO_ENABLED=0 go test -bench=. ./app/...
 
 vendor-update:
 	go get -u ./lib/...
 	go get -u ./app/...
-	go mod tidy -compat=1.23
+	go mod tidy -compat=1.24
 	go mod vendor
 
 app-local:
@@ -564,10 +564,10 @@ install-qtc:
 
 
 golangci-lint: install-golangci-lint
-	golangci-lint run
+	GOEXPERIMENT=synctest golangci-lint run
 
 install-golangci-lint:
-	which golangci-lint || curl -sSfL https://raw.githubusercontent.com/golangci/golangci-lint/master/install.sh | sh -s -- -b $(shell go env GOPATH)/bin v1.64.5
+	which golangci-lint || curl -sSfL https://raw.githubusercontent.com/golangci/golangci-lint/master/install.sh | sh -s -- -b $(shell go env GOPATH)/bin v1.64.7
 
 remove-golangci-lint:
 	rm -rf `which golangci-lint`

app/vlinsert/datadog/datadog.go

@@ -3,7 +3,6 @@ package datadog
 import (
 	"bytes"
 	"fmt"
-	"io"
 	"net/http"
 	"strconv"
 	"time"
@@ -11,20 +10,22 @@ import (
 	"github.com/VictoriaMetrics/metrics"
 	"github.com/valyala/fastjson"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/writeconcurrencylimiter"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 )
 
 var (
-	datadogStreamFields = flagutil.NewArrayString("datadog.streamFields", "Datadog tags to be used as stream fields.")
-	datadogIgnoreFields = flagutil.NewArrayString("datadog.ignoreFields", "Datadog tags to ignore.")
+	datadogStreamFields = flagutil.NewArrayString("datadog.streamFields", "Comma-separated list of fields to use as log stream fields for logs ingested via DataDog protocol. "+
+		"See https://docs.victoriametrics.com/victorialogs/data-ingestion/datadog-agent/#stream-fields")
+	datadogIgnoreFields = flagutil.NewArrayString("datadog.ignoreFields", "Comma-separated list of fields to ignore for logs ingested via DataDog protocol. "+
+		"See https://docs.victoriametrics.com/victorialogs/data-ingestion/datadog-agent/#dropping-fields")
+
+	maxRequestSize = flagutil.NewBytes("datadog.maxRequestSize", 64*1024*1024, "The maximum size in bytes of a single DataDog request")
 )
 
 var parserPool fastjson.ParserPool
@@ -46,7 +47,6 @@ func datadogLogsIngestion(w http.ResponseWriter, r *http.Request) bool {
 	w.Header().Add("Content-Type", "application/json")
 	startTime := time.Now()
 	v2LogsRequestsTotal.Inc()
-	reader := r.Body
 
 	var ts int64
 	if tsValue := r.Header.Get("dd-message-timestamp"); tsValue != "" && tsValue != "0" {
@@ -61,25 +61,7 @@ func datadogLogsIngestion(w http.ResponseWriter, r *http.Request) bool {
 		ts = startTime.UnixNano()
 	}
 
-	if r.Header.Get("Content-Encoding") == "gzip" {
-		zr, err := common.GetGzipReader(reader)
-		if err != nil {
-			httpserver.Errorf(w, r, "cannot read gzipped logs request: %s", err)
-			return true
-		}
-		defer common.PutGzipReader(zr)
-		reader = zr
-	}
-
-	wcr := writeconcurrencylimiter.GetReader(reader)
-	data, err := io.ReadAll(wcr)
-	writeconcurrencylimiter.PutReader(wcr)
-	if err != nil {
-		httpserver.Errorf(w, r, "cannot read request body: %s", err)
-		return true
-	}
-
-	cp, err := insertutils.GetCommonParams(r)
+	cp, err := insertutil.GetCommonParams(r)
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return true
@@ -97,11 +79,15 @@ func datadogLogsIngestion(w http.ResponseWriter, r *http.Request) bool {
 		return true
 	}
 
-	lmp := cp.NewLogMessageProcessor("datadog")
-	err = readLogsRequest(ts, data, lmp)
-	lmp.MustClose()
+	encoding := r.Header.Get("Content-Encoding")
+	err = protoparserutil.ReadUncompressedData(r.Body, encoding, maxRequestSize, func(data []byte) error {
+		lmp := cp.NewLogMessageProcessor("datadog", false)
+		err := readLogsRequest(ts, data, lmp)
+		lmp.MustClose()
+		return err
+	})
 	if err != nil {
-		logger.Warnf("cannot decode log message in /api/v2/logs request: %s, stream fields: %s", err, cp.StreamFields)
+		httpserver.Errorf(w, r, "cannot read DataDog protocol data: %s", err)
 		return true
 	}
 
@@ -121,7 +107,7 @@ var (
 // datadog message field has two formats:
 //   - regular log message with string text
 //   - nested json format for serverless plugins
-//     which has folowing format:
+//     which has the following format:
 //     {"message": {"message": "text","lamdba": {"arn": "string","requestID": "string"}, "timestamp": int64} }
 //
 // See https://github.com/DataDog/datadog-lambda-extension/blob/28b90c7e4e985b72d60b5f5a5147c69c7ac693c4/bottlecap/src/logs/lambda/mod.rs#L24
@@ -184,7 +170,7 @@ func appendMsgFields(fields []logstorage.Field, v *fastjson.Value) ([]logstorage
 
 // readLogsRequest parses data according to DataDog logs format
 // https://docs.datadoghq.com/api/latest/logs/#send-logs
-func readLogsRequest(ts int64, data []byte, lmp insertutils.LogMessageProcessor) error {
+func readLogsRequest(ts int64, data []byte, lmp insertutil.LogMessageProcessor) error {
 	p := parserPool.Get()
 	defer parserPool.Put(p)
 	v, err := p.ParseBytes(data)

app/vlinsert/datadog/datadog_test.go

@@ -4,7 +4,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 )
 
 func TestReadLogsRequestFailure(t *testing.T) {
@@ -13,7 +13,7 @@ func TestReadLogsRequestFailure(t *testing.T) {
 
 		ts := time.Now().UnixNano()
 
-		lmp := &insertutils.TestLogMessageProcessor{}
+		lmp := &insertutil.TestLogMessageProcessor{}
 		if err := readLogsRequest(ts, []byte(data), lmp); err == nil {
 			t.Fatalf("expecting non-empty error")
 		}
@@ -37,7 +37,7 @@ func TestReadLogsRequestSuccess(t *testing.T) {
 		for i := 0; i < rowsExpected; i++ {
 			timestampsExpected = append(timestampsExpected, ts)
 		}
-		lmp := &insertutils.TestLogMessageProcessor{}
+		lmp := &insertutil.TestLogMessageProcessor{}
 		if err := readLogsRequest(ts, []byte(data), lmp); err != nil {
 			t.Fatalf("unexpected error: %s", err)
 		}

app/vlinsert/elasticsearch/elasticsearch.go

@@ -10,14 +10,14 @@ import (
 
 	"github.com/VictoriaMetrics/metrics"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bufferedwriter"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/writeconcurrencylimiter"
 )
 
@@ -90,7 +90,7 @@ func RequestHandler(path string, w http.ResponseWriter, r *http.Request) bool {
 		startTime := time.Now()
 		bulkRequestsTotal.Inc()
 
-		cp, err := insertutils.GetCommonParams(r)
+		cp, err := insertutil.GetCommonParams(r)
 		if err != nil {
 			httpserver.Errorf(w, r, "%s", err)
 			return true
@@ -99,10 +99,10 @@ func RequestHandler(path string, w http.ResponseWriter, r *http.Request) bool {
 			httpserver.Errorf(w, r, "%s", err)
 			return true
 		}
-		lmp := cp.NewLogMessageProcessor("elasticsearch_bulk")
-		isGzip := r.Header.Get("Content-Encoding") == "gzip"
+		lmp := cp.NewLogMessageProcessor("elasticsearch_bulk", true)
+		encoding := r.Header.Get("Content-Encoding")
 		streamName := fmt.Sprintf("remoteAddr=%s, requestURI=%q", httpserver.GetQuotedRemoteAddr(r), r.RequestURI)
-		n, err := readBulkRequest(streamName, r.Body, isGzip, cp.TimeField, cp.MsgFields, lmp)
+		n, err := readBulkRequest(streamName, r.Body, encoding, cp.TimeField, cp.MsgFields, lmp)
 		lmp.MustClose()
 		if err != nil {
 			logger.Warnf("cannot decode log message #%d in /_bulk request: %s, stream fields: %s", n, err, cp.StreamFields)
@@ -131,22 +131,19 @@ var (
 	bulkRequestDuration = metrics.NewHistogram(`vl_http_request_duration_seconds{path="/insert/elasticsearch/_bulk"}`)
 )
 
-func readBulkRequest(streamName string, r io.Reader, isGzip bool, timeField string, msgFields []string, lmp insertutils.LogMessageProcessor) (int, error) {
+func readBulkRequest(streamName string, r io.Reader, encoding string, timeField string, msgFields []string, lmp insertutil.LogMessageProcessor) (int, error) {
 	// See https://www.elastic.co/guide/en/elasticsearch/reference/current/docs-bulk.html
 
-	if isGzip {
-		zr, err := common.GetGzipReader(r)
-		if err != nil {
-			return 0, fmt.Errorf("cannot read gzipped _bulk request: %w", err)
-		}
-		defer common.PutGzipReader(zr)
-		r = zr
+	reader, err := protoparserutil.GetUncompressedReader(r, encoding)
+	if err != nil {
+		return 0, fmt.Errorf("cannot decode Elasticsearch protocol data: %w", err)
 	}
+	defer protoparserutil.PutUncompressedReader(reader)
 
-	wcr := writeconcurrencylimiter.GetReader(r)
+	wcr := writeconcurrencylimiter.GetReader(reader)
 	defer writeconcurrencylimiter.PutReader(wcr)
 
-	lr := insertutils.NewLineReader(streamName, wcr)
+	lr := insertutil.NewLineReader(streamName, wcr)
 
 	n := 0
 	for {
@@ -159,7 +156,7 @@ func readBulkRequest(streamName string, r io.Reader, isGzip bool, timeField stri
 	}
 }
 
-func readBulkLine(lr *insertutils.LineReader, timeField string, msgFields []string, lmp insertutils.LogMessageProcessor) (bool, error) {
+func readBulkLine(lr *insertutil.LineReader, timeField string, msgFields []string, lmp insertutil.LogMessageProcessor) (bool, error) {
 	var line []byte
 
 	// Read the command, must be "create" or "index"
@@ -231,7 +228,7 @@ func parseElasticsearchTimestamp(s string) (int64, error) {
 	}
 	if len(s) < len("YYYY-MM-DD") || s[len("YYYY")] != '-' {
 		// Try parsing timestamp in seconds or milliseconds
-		return insertutils.ParseUnixTimestamp(s)
+		return insertutil.ParseUnixTimestamp(s)
 	}
 	if len(s) == len("YYYY-MM-DD") {
 		t, err := time.Parse("2006-01-02", s)

app/vlinsert/elasticsearch/elasticsearch_test.go

@@ -2,20 +2,24 @@ package elasticsearch
 
 import (
 	"bytes"
-	"compress/gzip"
 	"fmt"
+	"github.com/golang/snappy"
+	"github.com/klauspost/compress/gzip"
+	"github.com/klauspost/compress/zlib"
+	"github.com/klauspost/compress/zstd"
+	"io"
 	"testing"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 )
 
 func TestReadBulkRequest_Failure(t *testing.T) {
 	f := func(data string) {
 		t.Helper()
 
-		tlp := &insertutils.TestLogMessageProcessor{}
+		tlp := &insertutil.TestLogMessageProcessor{}
 		r := bytes.NewBufferString(data)
-		rows, err := readBulkRequest("test", r, false, "_time", []string{"_msg"}, tlp)
+		rows, err := readBulkRequest("test", r, "", "_time", []string{"_msg"}, tlp)
 		if err == nil {
 			t.Fatalf("expecting non-empty error")
 		}
@@ -33,15 +37,15 @@ foobar`)
 }
 
 func TestReadBulkRequest_Success(t *testing.T) {
-	f := func(data, timeField, msgField string, timestampsExpected []int64, resultExpected string) {
+	f := func(data, encoding, timeField, msgField string, timestampsExpected []int64, resultExpected string) {
 		t.Helper()
 
 		msgFields := []string{"non_existing_foo", msgField, "non_exiting_bar"}
-		tlp := &insertutils.TestLogMessageProcessor{}
+		tlp := &insertutil.TestLogMessageProcessor{}
 
 		// Read the request without compression
 		r := bytes.NewBufferString(data)
-		rows, err := readBulkRequest("test", r, false, timeField, msgFields, tlp)
+		rows, err := readBulkRequest("test", r, "", timeField, msgFields, tlp)
 		if err != nil {
 			t.Fatalf("unexpected error: %s", err)
 		}
@@ -53,10 +57,12 @@ func TestReadBulkRequest_Success(t *testing.T) {
 		}
 
 		// Read the request with compression
-		tlp = &insertutils.TestLogMessageProcessor{}
-		compressedData := compressData(data)
-		r = bytes.NewBufferString(compressedData)
-		rows, err = readBulkRequest("test", r, true, timeField, msgFields, tlp)
+		tlp = &insertutil.TestLogMessageProcessor{}
+		if encoding != "" {
+			data = compressData(data, encoding)
+		}
+		r = bytes.NewBufferString(data)
+		rows, err = readBulkRequest("test", r, encoding, timeField, msgFields, tlp)
 		if err != nil {
 			t.Fatalf("unexpected error: %s", err)
 		}
@@ -69,9 +75,9 @@ func TestReadBulkRequest_Success(t *testing.T) {
 	}
 
 	// Verify an empty data
-	f("", "_time", "_msg", nil, "")
-	f("\n", "_time", "_msg", nil, "")
-	f("\n\n", "_time", "_msg", nil, "")
+	f("", "gzip", "_time", "_msg", nil, "")
+	f("\n", "gzip", "_time", "_msg", nil, "")
+	f("\n\n", "gzip", "_time", "_msg", nil, "")
 
 	// Verify non-empty data
 	data := `{"create":{"_index":"filebeat-8.8.0"}}
@@ -82,20 +88,35 @@ func TestReadBulkRequest_Success(t *testing.T) {
 {"message":"xyz","@timestamp":"1686026893735","x":"y"}
 {"create":{"_index":"filebeat-8.8.0"}}
 {"message":"qwe rty","@timestamp":"1686026893"}
+{"create":{"_index":"filebeat-8.8.0"}}
+{"message":"qwe rty float","@timestamp":"1686026123.62"}
 `
 	timeField := "@timestamp"
 	msgField := "message"
-	timestampsExpected := []int64{1686026891735000000, 1686023292735000000, 1686026893735000000, 1686026893000000000}
+	timestampsExpected := []int64{1686026891735000000, 1686023292735000000, 1686026893735000000, 1686026893000000000, 1686026123620000000}
 	resultExpected := `{"log.offset":"71770","log.file.path":"/var/log/auth.log","_msg":"foobar"}
 {"_msg":"baz"}
 {"_msg":"xyz","x":"y"}
-{"_msg":"qwe rty"}`
-	f(data, timeField, msgField, timestampsExpected, resultExpected)
+{"_msg":"qwe rty"}
+{"_msg":"qwe rty float"}`
+	f(data, "zstd", timeField, msgField, timestampsExpected, resultExpected)
 }
 
-func compressData(s string) string {
+func compressData(s string, encoding string) string {
 	var bb bytes.Buffer
-	zw := gzip.NewWriter(&bb)
+	var zw io.WriteCloser
+	switch encoding {
+	case "gzip":
+		zw = gzip.NewWriter(&bb)
+	case "zstd":
+		zw, _ = zstd.NewWriter(&bb)
+	case "snappy":
+		zw = snappy.NewBufferedWriter(&bb)
+	case "deflate":
+		zw = zlib.NewWriter(&bb)
+	default:
+		panic(fmt.Errorf("%q encoding is not supported", encoding))
+	}
 	if _, err := zw.Write([]byte(s)); err != nil {
 		panic(fmt.Errorf("unexpected error when compressing data: %w", err))
 	}

app/vlinsert/elasticsearch/elasticsearch_timing_test.go

@@ -5,20 +5,29 @@ import (
 	"fmt"
 	"testing"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 )
 
 func BenchmarkReadBulkRequest(b *testing.B) {
-	b.Run("gzip:off", func(b *testing.B) {
-		benchmarkReadBulkRequest(b, false)
+	b.Run("encoding:none", func(b *testing.B) {
+		benchmarkReadBulkRequest(b, "")
 	})
-	b.Run("gzip:on", func(b *testing.B) {
-		benchmarkReadBulkRequest(b, true)
+	b.Run("encoding:gzip", func(b *testing.B) {
+		benchmarkReadBulkRequest(b, "gzip")
+	})
+	b.Run("encoding:zstd", func(b *testing.B) {
+		benchmarkReadBulkRequest(b, "zstd")
+	})
+	b.Run("encoding:deflate", func(b *testing.B) {
+		benchmarkReadBulkRequest(b, "deflate")
+	})
+	b.Run("encoding:snappy", func(b *testing.B) {
+		benchmarkReadBulkRequest(b, "snappy")
 	})
 }
 
-func benchmarkReadBulkRequest(b *testing.B, isGzip bool) {
+func benchmarkReadBulkRequest(b *testing.B, encoding string) {
 	data := `{"create":{"_index":"filebeat-8.8.0"}}
 {"@timestamp":"2023-06-06T04:48:11.735Z","log":{"offset":71770,"file":{"path":"/var/log/auth.log"}},"message":"foobar"}
 {"create":{"_index":"filebeat-8.8.0"}}
@@ -26,14 +35,14 @@ func benchmarkReadBulkRequest(b *testing.B, isGzip bool) {
 {"create":{"_index":"filebeat-8.8.0"}}
 {"message":"xyz","@timestamp":"2023-06-06T04:48:13.735Z","x":"y"}
 `
-	if isGzip {
-		data = compressData(data)
+	if encoding != "" {
+		data = compressData(data, encoding)
 	}
 	dataBytes := bytesutil.ToUnsafeBytes(data)
 
 	timeField := "@timestamp"
 	msgFields := []string{"message"}
-	blp := &insertutils.BenchmarkLogMessageProcessor{}
+	blp := &insertutil.BenchmarkLogMessageProcessor{}
 
 	b.ReportAllocs()
 	b.SetBytes(int64(len(data)))
@@ -41,7 +50,7 @@ func benchmarkReadBulkRequest(b *testing.B, isGzip bool) {
 		r := &bytes.Reader{}
 		for pb.Next() {
 			r.Reset(dataBytes)
-			_, err := readBulkRequest("test", r, isGzip, timeField, msgFields, blp)
+			_, err := readBulkRequest("test", r, encoding, timeField, msgFields, blp)
 			if err != nil {
 				panic(fmt.Errorf("unexpected error: %w", err))
 			}

app/vlinsert/insertutil/common_params.go

@@ -1,4 +1,4 @@
-package insertutils
+package insertutil
 
 import (
 	"flag"
@@ -13,7 +13,7 @@ import (
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeutil"
@@ -49,13 +49,13 @@ func GetCommonParams(r *http.Request) (*CommonParams, error) {
 	}
 
 	timeField := "_time"
-	if tf := httputils.GetRequestValue(r, "_time_field", "VL-Time-Field"); tf != "" {
+	if tf := httputil.GetRequestValue(r, "_time_field", "VL-Time-Field"); tf != "" {
 		timeField = tf
 	}
 
-	msgFields := httputils.GetArray(r, "_msg_field", "VL-Msg-Field")
-	streamFields := httputils.GetArray(r, "_stream_fields", "VL-Stream-Fields")
-	ignoreFields := httputils.GetArray(r, "ignore_fields", "VL-Ignore-Fields")
+	msgFields := httputil.GetArray(r, "_msg_field", "VL-Msg-Field")
+	streamFields := httputil.GetArray(r, "_stream_fields", "VL-Stream-Fields")
+	ignoreFields := httputil.GetArray(r, "ignore_fields", "VL-Ignore-Fields")
 
 	extraFields, err := getExtraFields(r)
 	if err != nil {
@@ -63,7 +63,7 @@ func GetCommonParams(r *http.Request) (*CommonParams, error) {
 	}
 
 	debug := false
-	if dv := httputils.GetRequestValue(r, "debug", "VL-Debug"); dv != "" {
+	if dv := httputil.GetRequestValue(r, "debug", "VL-Debug"); dv != "" {
 		debug, err = strconv.ParseBool(dv)
 		if err != nil {
 			return nil, fmt.Errorf("cannot parse debug=%q: %w", dv, err)
@@ -92,7 +92,7 @@ func GetCommonParams(r *http.Request) (*CommonParams, error) {
 }
 
 func getExtraFields(r *http.Request) ([]logstorage.Field, error) {
-	efs := httputils.GetArray(r, "extra_fields", "VL-Extra-Fields")
+	efs := httputil.GetArray(r, "extra_fields", "VL-Extra-Fields")
 	if len(efs) == 0 {
 		return nil, nil
 	}
@@ -141,7 +141,7 @@ type LogMessageProcessor interface {
 	//
 	// If streamFields is non-nil, then the given streamFields must be used as log stream fields instead of pre-configured fields.
 	//
-	// The LogMessageProcessor implementation cannot hold references to fields, since the caller can re-use them.
+	// The LogMessageProcessor implementation cannot hold references to fields, since the caller can reuse them.
 	AddRow(timestamp int64, fields, streamFields []logstorage.Field)
 
 	// MustClose() must flush all the remaining fields and free up resources occupied by LogMessageProcessor.
@@ -191,9 +191,6 @@ func (lmp *logMessageProcessor) initPeriodicFlush() {
 //
 // If streamFields is non-nil, then it is used as log stream fields instead of the pre-configured stream fields.
 func (lmp *logMessageProcessor) AddRow(timestamp int64, fields, streamFields []logstorage.Field) {
-	lmp.mu.Lock()
-	defer lmp.mu.Unlock()
-
 	lmp.rowsIngestedTotal.Inc()
 	n := logstorage.EstimatedJSONRowLen(fields)
 	lmp.bytesIngestedTotal.Add(n)
@@ -205,7 +202,47 @@ func (lmp *logMessageProcessor) AddRow(timestamp int64, fields, streamFields []l
 		return
 	}
 
+	lmp.mu.Lock()
+	defer lmp.mu.Unlock()
+
 	lmp.lr.MustAdd(lmp.cp.TenantID, timestamp, fields, streamFields)
+
+	if lmp.cp.Debug {
+		s := lmp.lr.GetRowString(0)
+		lmp.lr.ResetKeepSettings()
+		logger.Infof("remoteAddr=%s; requestURI=%s; ignoring log entry because of `debug` arg: %s", lmp.cp.DebugRemoteAddr, lmp.cp.DebugRequestURI, s)
+		rowsDroppedTotalDebug.Inc()
+		return
+	}
+	if lmp.lr.NeedFlush() {
+		lmp.flushLocked()
+	}
+}
+
+// InsertRowProcessor is used by native data ingestion protocol parser.
+type InsertRowProcessor interface {
+	// AddInsertRow must add r to the underlying storage.
+	AddInsertRow(r *logstorage.InsertRow)
+}
+
+// AddInsertRow adds r to lmp.
+func (lmp *logMessageProcessor) AddInsertRow(r *logstorage.InsertRow) {
+	lmp.rowsIngestedTotal.Inc()
+	n := logstorage.EstimatedJSONRowLen(r.Fields)
+	lmp.bytesIngestedTotal.Add(n)
+
+	if len(r.Fields) > *MaxFieldsPerLine {
+		line := logstorage.MarshalFieldsToJSON(nil, r.Fields)
+		logger.Warnf("dropping log line with %d fields; it exceeds -insert.maxFieldsPerLine=%d; %s", len(r.Fields), *MaxFieldsPerLine, line)
+		rowsDroppedTotalTooManyFields.Inc()
+		return
+	}
+
+	lmp.mu.Lock()
+	defer lmp.mu.Unlock()
+
+	lmp.lr.MustAddInsertRow(r)
+
 	if lmp.cp.Debug {
 		s := lmp.lr.GetRowString(0)
 		lmp.lr.ResetKeepSettings()
@@ -238,7 +275,7 @@ func (lmp *logMessageProcessor) MustClose() {
 // NewLogMessageProcessor returns new LogMessageProcessor for the given cp.
 //
 // MustClose() must be called on the returned LogMessageProcessor when it is no longer needed.
-func (cp *CommonParams) NewLogMessageProcessor(protocolName string) LogMessageProcessor {
+func (cp *CommonParams) NewLogMessageProcessor(protocolName string, isStreamMode bool) LogMessageProcessor {
 	lr := logstorage.GetLogRows(cp.StreamFields, cp.IgnoreFields, cp.ExtraFields, *defaultMsgValue)
 	rowsIngestedTotal := metrics.GetOrCreateCounter(fmt.Sprintf("vl_rows_ingested_total{type=%q}", protocolName))
 	bytesIngestedTotal := metrics.GetOrCreateCounter(fmt.Sprintf("vl_bytes_ingested_total{type=%q}", protocolName))
@@ -251,7 +288,10 @@ func (cp *CommonParams) NewLogMessageProcessor(protocolName string) LogMessagePr
 
 		stopCh: make(chan struct{}),
 	}
-	lmp.initPeriodicFlush()
+
+	if isStreamMode {
+		lmp.initPeriodicFlush()
+	}
 
 	return lmp
 }

app/vlinsert/insertutil/flags.go

@@ -1,4 +1,4 @@
-package insertutils
+package insertutil
 
 import (
 	"flag"

app/vlinsert/insertutil/line_reader.go

@@ -1,4 +1,4 @@
-package insertutils
+package insertutil
 
 import (
 	"bytes"

app/vlinsert/insertutil/line_reader_test.go

@@ -1,4 +1,4 @@
-package insertutils
+package insertutil
 
 import (
 	"bytes"

app/vlinsert/insertutil/testutils.go

@@ -1,4 +1,4 @@
-package insertutils
+package insertutil
 
 import (
 	"fmt"

app/vlinsert/insertutil/timestamp.go

@@ -1,8 +1,10 @@
-package insertutils
+package insertutil
 
 import (
 	"fmt"
+	"math"
 	"strconv"
+	"strings"
 	"time"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
@@ -49,6 +51,35 @@ func parseTimestamp(s string) (int64, error) {
 
 // ParseUnixTimestamp parses s as unix timestamp in seconds, milliseconds, microseconds or nanoseconds and returns the parsed timestamp in nanoseconds.
 func ParseUnixTimestamp(s string) (int64, error) {
+	if strings.IndexByte(s, '.') >= 0 {
+		// Parse timestamp as floating-point value
+		f, err := strconv.ParseFloat(s, 64)
+		if err != nil {
+			return 0, fmt.Errorf("cannot parse unix timestamp from %q: %w", s, err)
+		}
+		if f < (1<<31) && f >= (-1<<31) {
+			// The timestamp is in seconds.
+			return int64(f * 1e9), nil
+		}
+		if f < 1e3*(1<<31) && f >= 1e3*(-1<<31) {
+			// The timestamp is in milliseconds.
+			return int64(f * 1e6), nil
+		}
+		if f < 1e6*(1<<31) && f >= 1e6*(-1<<31) {
+			// The timestamp is in microseconds.
+			return int64(f * 1e3), nil
+		}
+		// The timestamp is in nanoseconds
+		if f > math.MaxInt64 {
+			return 0, fmt.Errorf("too big timestamp in nanoseconds: %v; mustn't exceed %v", f, int64(math.MaxInt64))
+		}
+		if f < math.MinInt64 {
+			return 0, fmt.Errorf("too small timestamp in nanoseconds: %v; must be bigger or equal to %v", f, int64(math.MinInt64))
+		}
+		return int64(f), nil
+	}
+
+	// Parse timestamp as integer
 	n, err := strconv.ParseInt(s, 10, 64)
 	if err != nil {
 		return 0, fmt.Errorf("cannot parse unix timestamp from %q: %w", s, err)

app/vlinsert/insertutil/timestamp_test.go

@@ -1,4 +1,4 @@
-package insertutils
+package insertutil
 
 import (
 	"testing"
@@ -6,6 +6,63 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
 )
 
+func TestParseUnixTimestamp_Success(t *testing.T) {
+	f := func(s string, timestampExpected int64) {
+		t.Helper()
+
+		timestamp, err := ParseUnixTimestamp(s)
+		if err != nil {
+			t.Fatalf("unexpected error in ParseUnixTimestamp(%q): %s", s, err)
+		}
+		if timestamp != timestampExpected {
+			t.Fatalf("unexpected timestamp returned from ParseUnixTimestamp(%q); got %d; want %d", s, timestamp, timestampExpected)
+		}
+	}
+
+	f("0", 0)
+
+	// nanoseconds
+	f("-1234567890123456789", -1234567890123456789)
+	f("1234567890123456789", 1234567890123456789)
+
+	// microseconds
+	f("-1234567890123456", -1234567890123456000)
+	f("1234567890123456", 1234567890123456000)
+	f("1234567890123456.789", 1234567890123456768)
+
+	// milliseconds
+	f("-1234567890123", -1234567890123000000)
+	f("1234567890123", 1234567890123000000)
+	f("1234567890123.456", 1234567890123456000)
+
+	// seconds
+	f("-1234567890", -1234567890000000000)
+	f("1234567890", 1234567890000000000)
+	f("-1234567890.123456", -1234567890123456000)
+}
+
+func TestParseUnixTimestamp_Failure(t *testing.T) {
+	f := func(s string) {
+		t.Helper()
+
+		_, err := ParseUnixTimestamp(s)
+		if err == nil {
+			t.Fatalf("expecting non-nil error in ParseUnixTimestamp(%q)", s)
+		}
+	}
+
+	// non-numeric timestamp
+	f("")
+	f("foobar")
+	f("foo.bar")
+
+	// too big timestamp
+	f("12345678901234567890")
+	f("-12345678901234567890")
+	f("12345678901234567890.235424")
+	f("-12345678901234567890.235424")
+}
+
 func TestExtractTimestampFromFields_Success(t *testing.T) {
 	f := func(timeField string, fields []logstorage.Field, nsecsExpected int64) {
 		t.Helper()

app/vlinsert/internalinsert/internalinsert.go

@@ -0,0 +1,96 @@
+package internalinsert
+
+import (
+	"flag"
+	"fmt"
+	"net/http"
+	"time"
+
+	"github.com/VictoriaMetrics/metrics"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage/netinsert"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
+)
+
+var (
+	disableInsert  = flag.Bool("internalinsert.disable", false, "Whether to disable /internal/insert HTTP endpoint")
+	maxRequestSize = flagutil.NewBytes("internalinsert.maxRequestSize", 64*1024*1024, "The maximum size in bytes of a single request, which can be accepted at /internal/insert HTTP endpoint")
+)
+
+// RequestHandler processes /internal/insert requests.
+func RequestHandler(w http.ResponseWriter, r *http.Request) {
+	if *disableInsert {
+		httpserver.Errorf(w, r, "requests to /internal/insert are disabled with -internalinsert.disable command-line flag")
+		return
+	}
+
+	startTime := time.Now()
+	if r.Method != "POST" {
+		w.WriteHeader(http.StatusMethodNotAllowed)
+		return
+	}
+	version := r.FormValue("version")
+	if version != netinsert.ProtocolVersion {
+		httpserver.Errorf(w, r, "unsupported protocol version=%q; want %q", version, netinsert.ProtocolVersion)
+		return
+	}
+
+	requestsTotal.Inc()
+
+	cp, err := insertutil.GetCommonParams(r)
+	if err != nil {
+		httpserver.Errorf(w, r, "%s", err)
+		return
+	}
+	if err := vlstorage.CanWriteData(); err != nil {
+		httpserver.Errorf(w, r, "%s", err)
+		return
+	}
+
+	encoding := r.Header.Get("Content-Encoding")
+	err = protoparserutil.ReadUncompressedData(r.Body, encoding, maxRequestSize, func(data []byte) error {
+		lmp := cp.NewLogMessageProcessor("internalinsert", false)
+		irp := lmp.(insertutil.InsertRowProcessor)
+		err := parseData(irp, data)
+		lmp.MustClose()
+		return err
+	})
+	if err != nil {
+		errorsTotal.Inc()
+		httpserver.Errorf(w, r, "cannot parse internal insert request: %s", err)
+		return
+	}
+
+	requestDuration.UpdateDuration(startTime)
+}
+
+func parseData(irp insertutil.InsertRowProcessor, data []byte) error {
+	r := logstorage.GetInsertRow()
+	src := data
+	i := 0
+	for len(src) > 0 {
+		tail, err := r.UnmarshalInplace(src)
+		if err != nil {
+			return fmt.Errorf("cannot parse row #%d: %s", i, err)
+		}
+		src = tail
+		i++
+
+		irp.AddInsertRow(r)
+	}
+	logstorage.PutInsertRow(r)
+
+	return nil
+}
+
+var (
+	requestsTotal = metrics.NewCounter(`vl_http_requests_total{path="/internal/insert"}`)
+	errorsTotal   = metrics.NewCounter(`vl_http_errors_total{path="/internal/insert"}`)
+
+	requestDuration = metrics.NewHistogram(`vl_http_request_duration_seconds{path="/internal/insert"}`)
+)

app/vlinsert/journald/journald.go

@@ -5,7 +5,6 @@ import (
 	"encoding/binary"
 	"flag"
 	"fmt"
-	"io"
 	"net/http"
 	"regexp"
 	"slices"
@@ -13,39 +12,37 @@ import (
 	"strings"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding/zstd"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/writeconcurrencylimiter"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/metrics"
 )
 
-const (
-	journaldEntryMaxNameLen = 64
-)
+// See https://github.com/systemd/systemd/blob/main/src/libsystemd/sd-journal/journal-file.c#L1703
+const journaldEntryMaxNameLen = 64
+
+var allowedJournaldEntryNameChars = regexp.MustCompile(`^[A-Z_][A-Z0-9_]*`)
 
 var (
-	bodyBufferPool                bytesutil.ByteBufferPool
-	allowedJournaldEntryNameChars = regexp.MustCompile(`^[A-Z_][A-Z0-9_]*`)
-)
-
-var (
-	journaldStreamFields = flagutil.NewArrayString("journald.streamFields", "Journal fields to be used as stream fields. "+
-		"See the list of allowed fields at https://www.freedesktop.org/software/systemd/man/latest/systemd.journal-fields.html.")
-	journaldIgnoreFields = flagutil.NewArrayString("journald.ignoreFields", "Journal fields to ignore. "+
-		"See the list of allowed fields at https://www.freedesktop.org/software/systemd/man/latest/systemd.journal-fields.html.")
-	journaldTimeField = flag.String("journald.timeField", "__REALTIME_TIMESTAMP", "Journal field to be used as time field. "+
-		"See the list of allowed fields at https://www.freedesktop.org/software/systemd/man/latest/systemd.journal-fields.html.")
-	journaldTenantID             = flag.String("journald.tenantID", "0:0", "TenantID for logs ingested via the Journald endpoint.")
+	journaldStreamFields = flagutil.NewArrayString("journald.streamFields", "Comma-separated list of fields to use as log stream fields for logs ingested over journald protocol. "+
+		"See https://docs.victoriametrics.com/victorialogs/data-ingestion/journald/#stream-fields")
+	journaldIgnoreFields = flagutil.NewArrayString("journald.ignoreFields", "Comma-separated list of fields to ignore for logs ingested over journald protocol. "+
+		"See https://docs.victoriametrics.com/victorialogs/data-ingestion/journald/#dropping-fields")
+	journaldTimeField = flag.String("journald.timeField", "__REALTIME_TIMESTAMP", "Field to use as a log timestamp for logs ingested via journald protocol. "+
+		"See https://docs.victoriametrics.com/victorialogs/data-ingestion/journald/#time-field")
+	journaldTenantID = flag.String("journald.tenantID", "0:0", "TenantID for logs ingested via the Journald endpoint. "+
+		"See https://docs.victoriametrics.com/victorialogs/data-ingestion/journald/#multitenancy")
 	journaldIncludeEntryMetadata = flag.Bool("journald.includeEntryMetadata", false, "Include journal entry fields, which with double underscores.")
+
+	maxRequestSize = flagutil.NewBytes("journald.maxRequestSize", 64*1024*1024, "The maximum size in bytes of a single journald request")
 )
 
-func getCommonParams(r *http.Request) (*insertutils.CommonParams, error) {
-	cp, err := insertutils.GetCommonParams(r)
+func getCommonParams(r *http.Request) (*insertutil.CommonParams, error) {
+	cp, err := insertutil.GetCommonParams(r)
 	if err != nil {
 		return nil, err
 	}
@@ -89,43 +86,29 @@ func handleJournald(r *http.Request, w http.ResponseWriter) {
 	startTime := time.Now()
 	requestsJournaldTotal.Inc()
 
-	if err := vlstorage.CanWriteData(); err != nil {
-		httpserver.Errorf(w, r, "%s", err)
-		return
-	}
-
-	reader := r.Body
-	var err error
-
-	wcr := writeconcurrencylimiter.GetReader(reader)
-	data, err := io.ReadAll(wcr)
-	if err != nil {
-		httpserver.Errorf(w, r, "cannot read request body: %s", err)
-		return
-	}
-	writeconcurrencylimiter.PutReader(wcr)
-	bb := bodyBufferPool.Get()
-	defer bodyBufferPool.Put(bb)
-	if r.Header.Get("Content-Encoding") == "zstd" {
-		bb.B, err = zstd.Decompress(bb.B[:0], data)
-		if err != nil {
-			httpserver.Errorf(w, r, "cannot decompress zstd-encoded request with length %d: %s", len(data), err)
-			return
-		}
-		data = bb.B
-	}
 	cp, err := getCommonParams(r)
 	if err != nil {
+		errorsTotal.Inc()
 		httpserver.Errorf(w, r, "cannot parse common params from request: %s", err)
 		return
 	}
 
-	lmp := cp.NewLogMessageProcessor("journald")
-	err = parseJournaldRequest(data, lmp, cp)
-	lmp.MustClose()
+	if err := vlstorage.CanWriteData(); err != nil {
+		errorsTotal.Inc()
+		httpserver.Errorf(w, r, "%s", err)
+		return
+	}
+
+	encoding := r.Header.Get("Content-Encoding")
+	err = protoparserutil.ReadUncompressedData(r.Body, encoding, maxRequestSize, func(data []byte) error {
+		lmp := cp.NewLogMessageProcessor("journald", false)
+		err := parseJournaldRequest(data, lmp, cp)
+		lmp.MustClose()
+		return err
+	})
 	if err != nil {
 		errorsTotal.Inc()
-		httpserver.Errorf(w, r, "cannot parse Journald protobuf request: %s", err)
+		httpserver.Errorf(w, r, "cannot read journald protocol data: %s", err)
 		return
 	}
 
@@ -148,7 +131,7 @@ var (
 )
 
 // See https://systemd.io/JOURNAL_EXPORT_FORMATS/#journal-export-format
-func parseJournaldRequest(data []byte, lmp insertutils.LogMessageProcessor, cp *insertutils.CommonParams) error {
+func parseJournaldRequest(data []byte, lmp insertutil.LogMessageProcessor, cp *insertutil.CommonParams) error {
 	var fields []logstorage.Field
 	var ts int64
 	var size uint64
@@ -198,7 +181,7 @@ func parseJournaldRequest(data []byte, lmp insertutils.LogMessageProcessor, cp *
 			if err != nil {
 				return fmt.Errorf("failed to extract binary field %q value size: %w", name, err)
 			}
-			// skip binary data sise
+			// skip binary data size
 			data = data[idx:]
 			if size == 0 {
 				return fmt.Errorf("unexpected zero binary data size decoded %d", size)
@@ -218,7 +201,6 @@ func parseJournaldRequest(data []byte, lmp insertutils.LogMessageProcessor, cp *
 			}
 			data = data[1:]
 		}
-		// https://github.com/systemd/systemd/blob/main/src/libsystemd/sd-journal/journal-file.c#L1703
 		if len(name) > journaldEntryMaxNameLen {
 			return fmt.Errorf("journald entry name should not exceed %d symbols, got: %q", journaldEntryMaxNameLen, name)
 		}

app/vlinsert/journald/journald_test.go

@@ -3,14 +3,14 @@ package journald
 import (
 	"testing"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 )
 
 func TestPushJournaldOk(t *testing.T) {
 	f := func(src string, timestampsExpected []int64, resultExpected string) {
 		t.Helper()
-		tlp := &insertutils.TestLogMessageProcessor{}
-		cp := &insertutils.CommonParams{
+		tlp := &insertutil.TestLogMessageProcessor{}
+		cp := &insertutil.CommonParams{
 			TimeField: "__REALTIME_TIMESTAMP",
 			MsgFields: []string{"MESSAGE"},
 		}
@@ -44,8 +44,8 @@ func TestPushJournaldOk(t *testing.T) {
 func TestPushJournald_Failure(t *testing.T) {
 	f := func(data string) {
 		t.Helper()
-		tlp := &insertutils.TestLogMessageProcessor{}
-		cp := &insertutils.CommonParams{
+		tlp := &insertutil.TestLogMessageProcessor{}
+		cp := &insertutil.CommonParams{
 			TimeField: "__REALTIME_TIMESTAMP",
 			MsgFields: []string{"MESSAGE"},
 		}

app/vlinsert/jsonline/jsonline.go

@@ -6,12 +6,12 @@ import (
 	"net/http"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/writeconcurrencylimiter"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -28,7 +28,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) {
 
 	requestsTotal.Inc()
 
-	cp, err := insertutils.GetCommonParams(r)
+	cp, err := insertutil.GetCommonParams(r)
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
@@ -38,18 +38,15 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	reader := r.Body
-	if r.Header.Get("Content-Encoding") == "gzip" {
-		zr, err := common.GetGzipReader(reader)
-		if err != nil {
-			logger.Errorf("cannot read gzipped jsonline request: %s", err)
-			return
-		}
-		defer common.PutGzipReader(zr)
-		reader = zr
+	encoding := r.Header.Get("Content-Encoding")
+	reader, err := protoparserutil.GetUncompressedReader(r.Body, encoding)
+	if err != nil {
+		logger.Errorf("cannot decode jsonline request: %s", err)
+		return
 	}
+	defer protoparserutil.PutUncompressedReader(reader)
 
-	lmp := cp.NewLogMessageProcessor("jsonline")
+	lmp := cp.NewLogMessageProcessor("jsonline", true)
 	streamName := fmt.Sprintf("remoteAddr=%s, requestURI=%q", httpserver.GetQuotedRemoteAddr(r), r.RequestURI)
 	processStreamInternal(streamName, reader, cp.TimeField, cp.MsgFields, lmp)
 	lmp.MustClose()
@@ -57,11 +54,11 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) {
 	requestDuration.UpdateDuration(startTime)
 }
 
-func processStreamInternal(streamName string, r io.Reader, timeField string, msgFields []string, lmp insertutils.LogMessageProcessor) {
+func processStreamInternal(streamName string, r io.Reader, timeField string, msgFields []string, lmp insertutil.LogMessageProcessor) {
 	wcr := writeconcurrencylimiter.GetReader(r)
 	defer writeconcurrencylimiter.PutReader(wcr)
 
-	lr := insertutils.NewLineReader(streamName, wcr)
+	lr := insertutil.NewLineReader(streamName, wcr)
 
 	n := 0
 	for {
@@ -78,7 +75,7 @@ func processStreamInternal(streamName string, r io.Reader, timeField string, msg
 	}
 }
 
-func readLine(lr *insertutils.LineReader, timeField string, msgFields []string, lmp insertutils.LogMessageProcessor) (bool, error) {
+func readLine(lr *insertutil.LineReader, timeField string, msgFields []string, lmp insertutil.LogMessageProcessor) (bool, error) {
 	var line []byte
 	for len(line) == 0 {
 		if !lr.NextLine() {
@@ -94,7 +91,7 @@ func readLine(lr *insertutils.LineReader, timeField string, msgFields []string,
 	if err := p.ParseLogMessage(line); err != nil {
 		return true, fmt.Errorf("cannot parse json-encoded line: %w; line contents: %q", err, line)
 	}
-	ts, err := insertutils.ExtractTimestampFromFields(timeField, p.Fields)
+	ts, err := insertutil.ExtractTimestampFromFields(timeField, p.Fields)
 	if err != nil {
 		return true, fmt.Errorf("cannot get timestamp from json-encoded line: %w; line contents: %q", err, line)
 	}

app/vlinsert/jsonline/jsonline_test.go

@@ -4,7 +4,7 @@ import (
 	"bytes"
 	"testing"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 )
 
 func TestProcessStreamInternal(t *testing.T) {
@@ -12,7 +12,7 @@ func TestProcessStreamInternal(t *testing.T) {
 		t.Helper()
 
 		msgFields := []string{msgField}
-		tlp := &insertutils.TestLogMessageProcessor{}
+		tlp := &insertutil.TestLogMessageProcessor{}
 		r := bytes.NewBufferString(data)
 		processStreamInternal("test", r, timeField, msgFields, tlp)
 

app/vlinsert/loki/loki.go

@@ -1,12 +1,18 @@
 package loki
 
 import (
+	"flag"
+	"fmt"
 	"net/http"
+	"strconv"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
 )
 
+var disableMessageParsing = flag.Bool("loki.disableMessageParsing", false, "Whether to disable automatic parsing of JSON-encoded log fields inside Loki log message into distinct log fields")
+
 // RequestHandler processes Loki insert requests
 func RequestHandler(path string, w http.ResponseWriter, r *http.Request) bool {
 	switch path {
@@ -35,8 +41,17 @@ func handleInsert(r *http.Request, w http.ResponseWriter) {
 	}
 }
 
-func getCommonParams(r *http.Request) (*insertutils.CommonParams, error) {
-	cp, err := insertutils.GetCommonParams(r)
+type commonParams struct {
+	cp *insertutil.CommonParams
+
+	// Whether to parse JSON inside plaintext log message.
+	//
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8486
+	parseMessage bool
+}
+
+func getCommonParams(r *http.Request) (*commonParams, error) {
+	cp, err := insertutil.GetCommonParams(r)
 	if err != nil {
 		return nil, err
 	}
@@ -55,5 +70,17 @@ func getCommonParams(r *http.Request) (*insertutils.CommonParams, error) {
 
 	}
 
-	return cp, nil
+	parseMessage := !*disableMessageParsing
+	if rv := httputil.GetRequestValue(r, "disable_message_parsing", "VL-Loki-Disable-Message-Parsing"); rv != "" {
+		bv, err := strconv.ParseBool(rv)
+		if err != nil {
+			return nil, fmt.Errorf("cannot parse dusable_message_parsing=%q: %s", rv, err)
+		}
+		parseMessage = !bv
+	}
+
+	return &commonParams{
+		cp:           cp,
+		parseMessage: parseMessage,
+	}, nil
 }

app/vlinsert/loki/loki_json.go

@@ -2,47 +2,28 @@ package loki
 
 import (
 	"fmt"
-	"io"
-	"math"
 	"net/http"
-	"strconv"
 	"time"
 
 	"github.com/VictoriaMetrics/metrics"
 	"github.com/valyala/fastjson"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/writeconcurrencylimiter"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 )
 
+var maxRequestSize = flagutil.NewBytes("loki.maxRequestSize", 64*1024*1024, "The maximum size in bytes of a single Loki request")
+
 var parserPool fastjson.ParserPool
 
 func handleJSON(r *http.Request, w http.ResponseWriter) {
 	startTime := time.Now()
 	requestsJSONTotal.Inc()
-	reader := r.Body
-	if r.Header.Get("Content-Encoding") == "gzip" {
-		zr, err := common.GetGzipReader(reader)
-		if err != nil {
-			httpserver.Errorf(w, r, "cannot initialize gzip reader: %s", err)
-			return
-		}
-		defer common.PutGzipReader(zr)
-		reader = zr
-	}
-
-	wcr := writeconcurrencylimiter.GetReader(reader)
-	data, err := io.ReadAll(wcr)
-	writeconcurrencylimiter.PutReader(wcr)
-	if err != nil {
-		httpserver.Errorf(w, r, "cannot read request body: %s", err)
-		return
-	}
 
 	cp, err := getCommonParams(r)
 	if err != nil {
@@ -53,12 +34,17 @@ func handleJSON(r *http.Request, w http.ResponseWriter) {
 		httpserver.Errorf(w, r, "%s", err)
 		return
 	}
-	lmp := cp.NewLogMessageProcessor("loki_json")
-	useDefaultStreamFields := len(cp.StreamFields) == 0
-	err = parseJSONRequest(data, lmp, useDefaultStreamFields)
-	lmp.MustClose()
+
+	encoding := r.Header.Get("Content-Encoding")
+	err = protoparserutil.ReadUncompressedData(r.Body, encoding, maxRequestSize, func(data []byte) error {
+		lmp := cp.cp.NewLogMessageProcessor("loki_json", false)
+		useDefaultStreamFields := len(cp.cp.StreamFields) == 0
+		err := parseJSONRequest(data, lmp, cp.cp.MsgFields, useDefaultStreamFields, cp.parseMessage)
+		lmp.MustClose()
+		return err
+	})
 	if err != nil {
-		httpserver.Errorf(w, r, "cannot parse Loki json request: %s; data=%s", err, data)
+		httpserver.Errorf(w, r, "cannot read Loki json data: %s", err)
 		return
 	}
 
@@ -66,6 +52,9 @@ func handleJSON(r *http.Request, w http.ResponseWriter) {
 	// There is no need in updating requestJSONDuration for request errors,
 	// since their timings are usually much smaller than the timing for successful request parsing.
 	requestJSONDuration.UpdateDuration(startTime)
+
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8505
+	w.WriteHeader(http.StatusNoContent)
 }
 
 var (
@@ -73,9 +62,10 @@ var (
 	requestJSONDuration = metrics.NewHistogram(`vl_http_request_duration_seconds{path="/insert/loki/api/v1/push",format="json"}`)
 )
 
-func parseJSONRequest(data []byte, lmp insertutils.LogMessageProcessor, useDefaultStreamFields bool) error {
+func parseJSONRequest(data []byte, lmp insertutil.LogMessageProcessor, msgFields []string, useDefaultStreamFields, parseMessage bool) error {
 	p := parserPool.Get()
 	defer parserPool.Put(p)
+
 	v, err := p.ParseBytes(data)
 	if err != nil {
 		return fmt.Errorf("cannot parse JSON request body: %w", err)
@@ -90,11 +80,20 @@ func parseJSONRequest(data []byte, lmp insertutils.LogMessageProcessor, useDefau
 		return fmt.Errorf("`streams` item in the parsed JSON must contain an array; got %q", streamsV)
 	}
 
+	fields := getFields()
+	defer putFields(fields)
+
+	var msgParser *logstorage.JSONParser
+	if parseMessage {
+		msgParser = logstorage.GetJSONParser()
+		defer logstorage.PutJSONParser(msgParser)
+	}
+
 	currentTimestamp := time.Now().UnixNano()
-	var commonFields []logstorage.Field
+
 	for _, stream := range streams {
 		// populate common labels from `stream` dict
-		commonFields = commonFields[:0]
+		fields.fields = fields.fields[:0]
 		labelsV := stream.Get("stream")
 		var labels *fastjson.Object
 		if labelsV != nil {
@@ -110,7 +109,7 @@ func parseJSONRequest(data []byte, lmp insertutils.LogMessageProcessor, useDefau
 				err = fmt.Errorf("unexpected label value type for %q:%q; want string", k, v)
 				return
 			}
-			commonFields = append(commonFields, logstorage.Field{
+			fields.fields = append(fields.fields, logstorage.Field{
 				Name:  bytesutil.ToUnsafeString(k),
 				Value: bytesutil.ToUnsafeString(vStr),
 			})
@@ -129,8 +128,10 @@ func parseJSONRequest(data []byte, lmp insertutils.LogMessageProcessor, useDefau
 			return fmt.Errorf("`values` item in the parsed JSON must contain an array; got %q", linesV)
 		}
 
-		fields := commonFields
+		commonFieldsLen := len(fields.fields)
 		for _, line := range lines {
+			fields.fields = fields.fields[:commonFieldsLen]
+
 			lineA, err := line.Array()
 			if err != nil {
 				return fmt.Errorf("unexpected contents of `values` item; want array; got %q", line)
@@ -152,17 +153,6 @@ func parseJSONRequest(data []byte, lmp insertutils.LogMessageProcessor, useDefau
 				ts = currentTimestamp
 			}
 
-			// parse log message
-			msg, err := lineA[1].StringBytes()
-			if err != nil {
-				return fmt.Errorf("unexpected log message type for %q; want string", lineA[1])
-			}
-
-			fields = append(fields[:len(commonFields)], logstorage.Field{
-				Name:  "_msg",
-				Value: bytesutil.ToUnsafeString(msg),
-			})
-
 			// parse structured metadata - see https://grafana.com/docs/loki/latest/reference/loki-http-api/#ingest-logs
 			if len(lineA) > 2 {
 				structuredMetadata, err := lineA[2].Object()
@@ -177,7 +167,7 @@ func parseJSONRequest(data []byte, lmp insertutils.LogMessageProcessor, useDefau
 						return
 					}
 
-					fields = append(fields, logstorage.Field{
+					fields.fields = append(fields.fields, logstorage.Field{
 						Name:  bytesutil.ToUnsafeString(k),
 						Value: bytesutil.ToUnsafeString(vStr),
 					})
@@ -186,39 +176,51 @@ func parseJSONRequest(data []byte, lmp insertutils.LogMessageProcessor, useDefau
 					return fmt.Errorf("error when parsing `structuredMetadata` object: %w", err)
 				}
 			}
+
+			// parse log message
+			msg, err := lineA[1].StringBytes()
+			if err != nil {
+				return fmt.Errorf("unexpected log message type for %q; want string", lineA[1])
+			}
+			allowMsgRenaming := false
+			fields.fields, allowMsgRenaming = addMsgField(fields.fields, msgParser, bytesutil.ToUnsafeString(msg))
+
 			var streamFields []logstorage.Field
 			if useDefaultStreamFields {
-				streamFields = commonFields
+				streamFields = fields.fields[:commonFieldsLen]
 			}
-			lmp.AddRow(ts, fields, streamFields)
+			if allowMsgRenaming {
+				logstorage.RenameField(fields.fields[commonFieldsLen:], msgFields, "_msg")
+			}
+			lmp.AddRow(ts, fields.fields, streamFields)
 		}
 	}
 
 	return nil
 }
 
+func addMsgField(dst []logstorage.Field, msgParser *logstorage.JSONParser, msg string) ([]logstorage.Field, bool) {
+	if msgParser == nil || len(msg) < 2 || msg[0] != '{' || msg[len(msg)-1] != '}' {
+		return append(dst, logstorage.Field{
+			Name:  "_msg",
+			Value: msg,
+		}), false
+	}
+	if msgParser != nil && len(msg) >= 2 && msg[0] == '{' && msg[len(msg)-1] == '}' {
+		if err := msgParser.ParseLogMessage(bytesutil.ToUnsafeBytes(msg)); err == nil {
+			return append(dst, msgParser.Fields...), true
+		}
+	}
+	return append(dst, logstorage.Field{
+		Name:  "_msg",
+		Value: msg,
+	}), false
+}
+
 func parseLokiTimestamp(s string) (int64, error) {
 	if s == "" {
 		// Special case - an empty timestamp must be substituted with the current time by the caller.
 		return 0, nil
 	}
-	n, err := strconv.ParseInt(s, 10, 64)
-	if err != nil {
-		// Fall back to parsing floating-point value
-		f, err := strconv.ParseFloat(s, 64)
-		if err != nil {
-			return 0, err
-		}
-		if f > math.MaxInt64 {
-			return 0, fmt.Errorf("too big timestamp in nanoseconds: %v; mustn't exceed %v", f, int64(math.MaxInt64))
-		}
-		if f < math.MinInt64 {
-			return 0, fmt.Errorf("too small timestamp in nanoseconds: %v; must be bigger or equal to %v", f, int64(math.MinInt64))
-		}
-		n = int64(f)
-	}
-	if n < 0 {
-		return 0, fmt.Errorf("too small timestamp in nanoseconds: %d; must be bigger than 0", n)
-	}
-	return n, nil
+	return insertutil.ParseUnixTimestamp(s)
 }

app/vlinsert/loki/loki_json_test.go

@@ -3,15 +3,15 @@ package loki
 import (
 	"testing"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 )
 
 func TestParseJSONRequest_Failure(t *testing.T) {
 	f := func(s string) {
 		t.Helper()
 
-		tlp := &insertutils.TestLogMessageProcessor{}
-		if err := parseJSONRequest([]byte(s), tlp, false); err == nil {
+		tlp := &insertutil.TestLogMessageProcessor{}
+		if err := parseJSONRequest([]byte(s), tlp, nil, false, false); err == nil {
 			t.Fatalf("expecting non-nil error")
 		}
 		if err := tlp.Verify(nil, ""); err != nil {
@@ -63,9 +63,9 @@ func TestParseJSONRequest_Success(t *testing.T) {
 	f := func(s string, timestampsExpected []int64, resultExpected string) {
 		t.Helper()
 
-		tlp := &insertutils.TestLogMessageProcessor{}
+		tlp := &insertutil.TestLogMessageProcessor{}
 
-		if err := parseJSONRequest([]byte(s), tlp, false); err != nil {
+		if err := parseJSONRequest([]byte(s), tlp, nil, false, false); err != nil {
 			t.Fatalf("unexpected error: %s", err)
 		}
 		if err := tlp.Verify(timestampsExpected, resultExpected); err != nil {
@@ -89,9 +89,9 @@ func TestParseJSONRequest_Success(t *testing.T) {
 	"label2": "value2"
 },"values":[
 	["1577836800000000001", "foo bar"],
-	["1477836900005000002", "abc"],
+	["1686026123.62", "abc"],
 	["147.78369e9", "foobar"]
-]}]}`, []int64{1577836800000000001, 1477836900005000002, 147783690000}, `{"label1":"value1","label2":"value2","_msg":"foo bar"}
+]}]}`, []int64{1577836800000000001, 1686026123620000000, 147783690000000000}, `{"label1":"value1","label2":"value2","_msg":"foo bar"}
 {"label1":"value1","label2":"value2","_msg":"abc"}
 {"label1":"value1","label2":"value2","_msg":"foobar"}`)
 
@@ -122,6 +122,48 @@ func TestParseJSONRequest_Success(t *testing.T) {
 {"x":"y","_msg":"yx"}`)
 
 	// values with metadata
-	f(`{"streams":[{"values":[["1577836800000000001", "foo bar", {"metadata_1": "md_value"}]]}]}`, []int64{1577836800000000001}, `{"_msg":"foo bar","metadata_1":"md_value"}`)
+	f(`{"streams":[{"values":[["1577836800000000001", "foo bar", {"metadata_1": "md_value"}]]}]}`, []int64{1577836800000000001}, `{"metadata_1":"md_value","_msg":"foo bar"}`)
 	f(`{"streams":[{"values":[["1577836800000000001", "foo bar", {}]]}]}`, []int64{1577836800000000001}, `{"_msg":"foo bar"}`)
 }
+
+func TestParseJSONRequest_ParseMessage(t *testing.T) {
+	f := func(s string, msgFields []string, timestampsExpected []int64, resultExpected string) {
+		t.Helper()
+
+		tlp := &insertutil.TestLogMessageProcessor{}
+
+		if err := parseJSONRequest([]byte(s), tlp, msgFields, false, true); err != nil {
+			t.Fatalf("unexpected error: %s", err)
+		}
+		if err := tlp.Verify(timestampsExpected, resultExpected); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	f(`{
+	"streams": [
+		{
+			"stream": {
+				"foo": "bar",
+				"a": "b"
+			},
+			"values": [
+				["1577836800000000001", "{\"user_id\":\"123\"}"],
+				["1577836900005000002", "abc", {"trace_id":"pqw"}],
+				["1577836900005000003", "{def}"]
+			]
+		},
+		{
+			"stream": {
+				"x": "y"
+			},
+			"values": [
+				["1877836900005000004", "{\"trace_id\":\"111\",\"parent_id\":\"abc\"}"]
+			]
+		}
+	]
+}`, []string{"a", "trace_id"}, []int64{1577836800000000001, 1577836900005000002, 1577836900005000003, 1877836900005000004}, `{"foo":"bar","a":"b","user_id":"123"}
+{"foo":"bar","a":"b","trace_id":"pqw","_msg":"abc"}
+{"foo":"bar","a":"b","_msg":"{def}"}
+{"x":"y","_msg":"111","parent_id":"abc"}`)
+}

app/vlinsert/loki/loki_json_timing_test.go

@@ -6,7 +6,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 )
 
 func BenchmarkParseJSONRequest(b *testing.B) {
@@ -22,13 +22,13 @@ func BenchmarkParseJSONRequest(b *testing.B) {
 }
 
 func benchmarkParseJSONRequest(b *testing.B, streams, rows, labels int) {
-	blp := &insertutils.BenchmarkLogMessageProcessor{}
+	blp := &insertutil.BenchmarkLogMessageProcessor{}
 	b.ReportAllocs()
 	b.SetBytes(int64(streams * rows))
 	b.RunParallel(func(pb *testing.PB) {
 		data := getJSONBody(streams, rows, labels)
 		for pb.Next() {
-			if err := parseJSONRequest(data, blp, false); err != nil {
+			if err := parseJSONRequest(data, blp, nil, false, true); err != nil {
 				panic(fmt.Errorf("unexpected error: %w", err))
 			}
 		}

app/vlinsert/loki/loki_protobuf.go

@@ -2,38 +2,27 @@ package loki
 
 import (
 	"fmt"
-	"io"
 	"net/http"
 	"strconv"
 	"strings"
 	"sync"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/writeconcurrencylimiter"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/metrics"
-	"github.com/golang/snappy"
 )
 
 var (
-	bytesBufPool bytesutil.ByteBufferPool
 	pushReqsPool sync.Pool
 )
 
 func handleProtobuf(r *http.Request, w http.ResponseWriter) {
 	startTime := time.Now()
 	requestsProtobufTotal.Inc()
-	wcr := writeconcurrencylimiter.GetReader(r.Body)
-	data, err := io.ReadAll(wcr)
-	writeconcurrencylimiter.PutReader(wcr)
-	if err != nil {
-		httpserver.Errorf(w, r, "cannot read request body: %s", err)
-		return
-	}
 
 	cp, err := getCommonParams(r)
 	if err != nil {
@@ -44,12 +33,22 @@ func handleProtobuf(r *http.Request, w http.ResponseWriter) {
 		httpserver.Errorf(w, r, "%s", err)
 		return
 	}
-	lmp := cp.NewLogMessageProcessor("loki_protobuf")
-	useDefaultStreamFields := len(cp.StreamFields) == 0
-	err = parseProtobufRequest(data, lmp, useDefaultStreamFields)
-	lmp.MustClose()
+
+	encoding := r.Header.Get("Content-Encoding")
+	if encoding == "" {
+		// Loki protocol uses snappy compression by default.
+		// See https://grafana.com/docs/loki/latest/reference/loki-http-api/#ingest-logs
+		encoding = "snappy"
+	}
+	err = protoparserutil.ReadUncompressedData(r.Body, encoding, maxRequestSize, func(data []byte) error {
+		lmp := cp.cp.NewLogMessageProcessor("loki_protobuf", false)
+		useDefaultStreamFields := len(cp.cp.StreamFields) == 0
+		err := parseProtobufRequest(data, lmp, cp.cp.MsgFields, useDefaultStreamFields, cp.parseMessage)
+		lmp.MustClose()
+		return err
+	})
 	if err != nil {
-		httpserver.Errorf(w, r, "cannot parse Loki protobuf request: %s", err)
+		httpserver.Errorf(w, r, "cannot read Loki protobuf data: %s", err)
 		return
 	}
 
@@ -57,6 +56,9 @@ func handleProtobuf(r *http.Request, w http.ResponseWriter) {
 	// There is no need in updating requestProtobufDuration for request errors,
 	// since their timings are usually much smaller than the timing for successful request parsing.
 	requestProtobufDuration.UpdateDuration(startTime)
+
+	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8505
+	w.WriteHeader(http.StatusNoContent)
 }
 
 var (
@@ -64,20 +66,11 @@ var (
 	requestProtobufDuration = metrics.NewHistogram(`vl_http_request_duration_seconds{path="/insert/loki/api/v1/push",format="protobuf"}`)
 )
 
-func parseProtobufRequest(data []byte, lmp insertutils.LogMessageProcessor, useDefaultStreamFields bool) error {
-	bb := bytesBufPool.Get()
-	defer bytesBufPool.Put(bb)
-
-	buf, err := snappy.Decode(bb.B[:cap(bb.B)], data)
-	if err != nil {
-		return fmt.Errorf("cannot decode snappy-encoded request body: %w", err)
-	}
-	bb.B = buf
-
+func parseProtobufRequest(data []byte, lmp insertutil.LogMessageProcessor, msgFields []string, useDefaultStreamFields, parseMessage bool) error {
 	req := getPushRequest()
 	defer putPushRequest(req)
 
-	err = req.UnmarshalProtobuf(bb.B)
+	err := req.UnmarshalProtobuf(data)
 	if err != nil {
 		return fmt.Errorf("cannot parse request body: %w", err)
 	}
@@ -85,8 +78,15 @@ func parseProtobufRequest(data []byte, lmp insertutils.LogMessageProcessor, useD
 	fields := getFields()
 	defer putFields(fields)
 
+	var msgParser *logstorage.JSONParser
+	if parseMessage {
+		msgParser = logstorage.GetJSONParser()
+		defer logstorage.PutJSONParser(msgParser)
+	}
+
 	streams := req.Streams
 	currentTimestamp := time.Now().UnixNano()
+
 	for i := range streams {
 		stream := &streams[i]
 		// st.Labels contains labels for the stream.
@@ -109,10 +109,8 @@ func parseProtobufRequest(data []byte, lmp insertutils.LogMessageProcessor, useD
 				})
 			}
 
-			fields.fields = append(fields.fields, logstorage.Field{
-				Name:  "_msg",
-				Value: e.Line,
-			})
+			allowMsgRenaming := false
+			fields.fields, allowMsgRenaming = addMsgField(fields.fields, msgParser, e.Line)
 
 			ts := e.Timestamp.UnixNano()
 			if ts == 0 {
@@ -123,6 +121,9 @@ func parseProtobufRequest(data []byte, lmp insertutils.LogMessageProcessor, useD
 			if useDefaultStreamFields {
 				streamFields = fields.fields[:commonFieldsLen]
 			}
+			if allowMsgRenaming {
+				logstorage.RenameField(fields.fields[commonFieldsLen:], msgFields, "_msg")
+			}
 			lmp.AddRow(ts, fields.fields, streamFields)
 		}
 	}

app/vlinsert/loki/loki_protobuf_test.go

@@ -6,9 +6,8 @@ import (
 	"testing"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
-	"github.com/golang/snappy"
 )
 
 type testLogMessageProcessor struct {
@@ -53,7 +52,7 @@ func TestParseProtobufRequest_Success(t *testing.T) {
 		t.Helper()
 
 		tlp := &testLogMessageProcessor{}
-		if err := parseJSONRequest([]byte(s), tlp, false); err != nil {
+		if err := parseJSONRequest([]byte(s), tlp, nil, false, false); err != nil {
 			t.Fatalf("unexpected error: %s", err)
 		}
 		if len(tlp.pr.Streams) != len(timestampsExpected) {
@@ -61,10 +60,9 @@ func TestParseProtobufRequest_Success(t *testing.T) {
 		}
 
 		data := tlp.pr.MarshalProtobuf(nil)
-		encodedData := snappy.Encode(nil, data)
 
-		tlp2 := &insertutils.TestLogMessageProcessor{}
-		if err := parseProtobufRequest(encodedData, tlp2, false); err != nil {
+		tlp2 := &insertutil.TestLogMessageProcessor{}
+		if err := parseProtobufRequest(data, tlp2, nil, false, false); err != nil {
 			t.Fatalf("unexpected error: %s", err)
 		}
 		if err := tlp2.Verify(timestampsExpected, resultExpected); err != nil {
@@ -90,7 +88,7 @@ func TestParseProtobufRequest_Success(t *testing.T) {
 	["1577836800000000001", "foo bar"],
 	["1477836900005000002", "abc"],
 	["147.78369e9", "foobar"]
-]}]}`, []int64{1577836800000000001, 1477836900005000002, 147783690000}, `{"label1":"value1","label2":"value2","_msg":"foo bar"}
+]}]}`, []int64{1577836800000000001, 1477836900005000002, 147783690000000000}, `{"label1":"value1","label2":"value2","_msg":"foo bar"}
 {"label1":"value1","label2":"value2","_msg":"abc"}
 {"label1":"value1","label2":"value2","_msg":"foobar"}`)
 
@@ -121,6 +119,57 @@ func TestParseProtobufRequest_Success(t *testing.T) {
 {"x":"y","_msg":"yx"}`)
 }
 
+func TestParseProtobufRequest_ParseMessage(t *testing.T) {
+	f := func(s string, msgFields []string, timestampsExpected []int64, resultExpected string) {
+		t.Helper()
+
+		tlp := &testLogMessageProcessor{}
+		if err := parseJSONRequest([]byte(s), tlp, nil, false, false); err != nil {
+			t.Fatalf("unexpected error: %s", err)
+		}
+		if len(tlp.pr.Streams) != len(timestampsExpected) {
+			t.Fatalf("unexpected number of streams; got %d; want %d", len(tlp.pr.Streams), len(timestampsExpected))
+		}
+
+		data := tlp.pr.MarshalProtobuf(nil)
+
+		tlp2 := &insertutil.TestLogMessageProcessor{}
+		if err := parseProtobufRequest(data, tlp2, msgFields, false, true); err != nil {
+			t.Fatalf("unexpected error: %s", err)
+		}
+		if err := tlp2.Verify(timestampsExpected, resultExpected); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	f(`{
+	"streams": [
+		{
+			"stream": {
+				"foo": "bar",
+				"a": "b"
+			},
+			"values": [
+				["1577836800000000001", "{\"user_id\":\"123\"}"],
+				["1577836900005000002", "abc", {"trace_id":"pqw"}],
+				["1577836900005000003", "{def}"]
+			]
+		},
+		{
+			"stream": {
+				"x": "y"
+			},
+			"values": [
+				["1877836900005000004", "{\"trace_id\":\"432\",\"parent_id\":\"qwerty\"}"]
+			]
+		}
+	]
+}`, []string{"a", "trace_id"}, []int64{1577836800000000001, 1577836900005000002, 1577836900005000003, 1877836900005000004}, `{"foo":"bar","a":"b","user_id":"123"}
+{"foo":"bar","a":"b","trace_id":"pqw","_msg":"abc"}
+{"foo":"bar","a":"b","_msg":"{def}"}
+{"x":"y","_msg":"432","parent_id":"qwerty"}`)
+}
+
 func TestParsePromLabels_Success(t *testing.T) {
 	f := func(s string) {
 		t.Helper()

app/vlinsert/loki/loki_protobuf_timing_test.go

@@ -6,9 +6,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/golang/snappy"
-
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 )
 
@@ -25,13 +23,13 @@ func BenchmarkParseProtobufRequest(b *testing.B) {
 }
 
 func benchmarkParseProtobufRequest(b *testing.B, streams, rows, labels int) {
-	blp := &insertutils.BenchmarkLogMessageProcessor{}
+	blp := &insertutil.BenchmarkLogMessageProcessor{}
 	b.ReportAllocs()
 	b.SetBytes(int64(streams * rows))
 	b.RunParallel(func(pb *testing.PB) {
 		body := getProtobufBody(streams, rows, labels)
 		for pb.Next() {
-			if err := parseProtobufRequest(body, blp, false); err != nil {
+			if err := parseProtobufRequest(body, blp, nil, false, true); err != nil {
 				panic(fmt.Errorf("unexpected error: %w", err))
 			}
 		}
@@ -78,8 +76,5 @@ func getProtobufBody(streamsCount, rowsCount, labelsCount int) []byte {
 		Streams: streams,
 	}
 
-	body := pr.MarshalProtobuf(nil)
-	encodedBody := snappy.Encode(nil, body)
-
-	return encodedBody
+	return pr.MarshalProtobuf(nil)
 }

app/vlinsert/main.go

@@ -7,6 +7,7 @@ import (
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/datadog"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/elasticsearch"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/internalinsert"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/journald"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/jsonline"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/loki"
@@ -28,6 +29,11 @@ func Stop() {
 func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 	path := r.URL.Path
 
+	if path == "/internal/insert" {
+		internalinsert.RequestHandler(w, r)
+		return true
+	}
+
 	if !strings.HasPrefix(path, "/insert/") {
 		// Skip requests, which do not start with /insert/, since these aren't our requests.
 		return false

app/vlinsert/opentelemetry/opentelemetry.go

@@ -2,21 +2,22 @@ package opentelemetry
 
 import (
 	"fmt"
-	"io"
 	"net/http"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentelemetry/pb"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/slicesutil"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/writeconcurrencylimiter"
 	"github.com/VictoriaMetrics/metrics"
 )
 
+var maxRequestSize = flagutil.NewBytes("opentelemetry.maxRequestSize", 64*1024*1024, "The maximum size in bytes of a single OpenTelemetry request")
+
 // RequestHandler processes Opentelemetry insert requests
 func RequestHandler(path string, w http.ResponseWriter, r *http.Request) bool {
 	switch path {
@@ -37,26 +38,8 @@ func RequestHandler(path string, w http.ResponseWriter, r *http.Request) bool {
 func handleProtobuf(r *http.Request, w http.ResponseWriter) {
 	startTime := time.Now()
 	requestsProtobufTotal.Inc()
-	reader := r.Body
-	if r.Header.Get("Content-Encoding") == "gzip" {
-		zr, err := common.GetGzipReader(reader)
-		if err != nil {
-			httpserver.Errorf(w, r, "cannot initialize gzip reader: %s", err)
-			return
-		}
-		defer common.PutGzipReader(zr)
-		reader = zr
-	}
 
-	wcr := writeconcurrencylimiter.GetReader(reader)
-	data, err := io.ReadAll(wcr)
-	writeconcurrencylimiter.PutReader(wcr)
-	if err != nil {
-		httpserver.Errorf(w, r, "cannot read request body: %s", err)
-		return
-	}
-
-	cp, err := insertutils.GetCommonParams(r)
+	cp, err := insertutil.GetCommonParams(r)
 	if err != nil {
 		httpserver.Errorf(w, r, "cannot parse common params from request: %s", err)
 		return
@@ -66,12 +49,16 @@ func handleProtobuf(r *http.Request, w http.ResponseWriter) {
 		return
 	}
 
-	lmp := cp.NewLogMessageProcessor("opentelelemtry_protobuf")
-	useDefaultStreamFields := len(cp.StreamFields) == 0
-	err = pushProtobufRequest(data, lmp, useDefaultStreamFields)
-	lmp.MustClose()
+	encoding := r.Header.Get("Content-Encoding")
+	err = protoparserutil.ReadUncompressedData(r.Body, encoding, maxRequestSize, func(data []byte) error {
+		lmp := cp.NewLogMessageProcessor("opentelelemtry_protobuf", false)
+		useDefaultStreamFields := len(cp.StreamFields) == 0
+		err := pushProtobufRequest(data, lmp, useDefaultStreamFields)
+		lmp.MustClose()
+		return err
+	})
 	if err != nil {
-		httpserver.Errorf(w, r, "cannot parse OpenTelemetry protobuf request: %s", err)
+		httpserver.Errorf(w, r, "cannot read OpenTelemetry protocol data: %s", err)
 		return
 	}
 
@@ -88,7 +75,7 @@ var (
 	requestProtobufDuration = metrics.NewHistogram(`vl_http_request_duration_seconds{path="/insert/opentelemetry/v1/logs",format="protobuf"}`)
 )
 
-func pushProtobufRequest(data []byte, lmp insertutils.LogMessageProcessor, useDefaultStreamFields bool) error {
+func pushProtobufRequest(data []byte, lmp insertutil.LogMessageProcessor, useDefaultStreamFields bool) error {
 	var req pb.ExportLogsServiceRequest
 	if err := req.UnmarshalProtobuf(data); err != nil {
 		errorsTotal.Inc()
@@ -101,7 +88,7 @@ func pushProtobufRequest(data []byte, lmp insertutils.LogMessageProcessor, useDe
 		commonFields = slicesutil.SetLength(commonFields, len(attributes))
 		for i, attr := range attributes {
 			commonFields[i].Name = attr.Key
-			commonFields[i].Value = attr.Value.FormatString()
+			commonFields[i].Value = attr.Value.FormatString(true)
 		}
 		commonFieldsLen := len(commonFields)
 		for _, sc := range rl.ScopeLogs {
@@ -112,18 +99,18 @@ func pushProtobufRequest(data []byte, lmp insertutils.LogMessageProcessor, useDe
 	return nil
 }
 
-func pushFieldsFromScopeLogs(sc *pb.ScopeLogs, commonFields []logstorage.Field, lmp insertutils.LogMessageProcessor, useDefaultStreamFields bool) []logstorage.Field {
+func pushFieldsFromScopeLogs(sc *pb.ScopeLogs, commonFields []logstorage.Field, lmp insertutil.LogMessageProcessor, useDefaultStreamFields bool) []logstorage.Field {
 	fields := commonFields
 	for _, lr := range sc.LogRecords {
 		fields = fields[:len(commonFields)]
 		fields = append(fields, logstorage.Field{
 			Name:  "_msg",
-			Value: lr.Body.FormatString(),
+			Value: lr.Body.FormatString(true),
 		})
 		for _, attr := range lr.Attributes {
 			fields = append(fields, logstorage.Field{
 				Name:  attr.Key,
-				Value: attr.Value.FormatString(),
+				Value: attr.Value.FormatString(true),
 			})
 		}
 		if len(lr.TraceID) > 0 {

app/vlinsert/opentelemetry/opentelemetry_test.go

@@ -3,7 +3,7 @@ package opentelemetry
 import (
 	"testing"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentelemetry/pb"
 )
 
@@ -15,7 +15,7 @@ func TestPushProtoOk(t *testing.T) {
 		}
 
 		pData := lr.MarshalProtobuf(nil)
-		tlp := &insertutils.TestLogMessageProcessor{}
+		tlp := &insertutil.TestLogMessageProcessor{}
 		if err := pushProtobufRequest(pData, tlp, false); err != nil {
 			t.Fatalf("unexpected error: %s", err)
 		}
@@ -24,6 +24,7 @@ func TestPushProtoOk(t *testing.T) {
 			t.Fatal(err)
 		}
 	}
+
 	// single line without resource attributes
 	f([]pb.ResourceLogs{
 		{
@@ -39,6 +40,27 @@ func TestPushProtoOk(t *testing.T) {
 		[]int64{1234},
 		`{"_msg":"log-line-message","severity":"Trace"}`,
 	)
+
+	// severities mapping
+	f([]pb.ResourceLogs{
+		{
+			ScopeLogs: []pb.ScopeLogs{
+				{
+					LogRecords: []pb.LogRecord{
+						{Attributes: []*pb.KeyValue{}, TimeUnixNano: 1234, SeverityNumber: 1, Body: pb.AnyValue{StringValue: ptrTo("log-line-message")}},
+						{Attributes: []*pb.KeyValue{}, TimeUnixNano: 1234, SeverityNumber: 13, Body: pb.AnyValue{StringValue: ptrTo("log-line-message")}},
+						{Attributes: []*pb.KeyValue{}, TimeUnixNano: 1234, SeverityNumber: 24, Body: pb.AnyValue{StringValue: ptrTo("log-line-message")}},
+					},
+				},
+			},
+		},
+	},
+		[]int64{1234, 1234, 1234},
+		`{"_msg":"log-line-message","severity":"Trace"}
+{"_msg":"log-line-message","severity":"Warn"}
+{"_msg":"log-line-message","severity":"Fatal4"}`,
+	)
+
 	// multi-line with resource attributes
 	f([]pb.ResourceLogs{
 		{
@@ -58,7 +80,7 @@ func TestPushProtoOk(t *testing.T) {
 				{
 					LogRecords: []pb.LogRecord{
 						{Attributes: []*pb.KeyValue{}, TimeUnixNano: 1234, SeverityNumber: 1, Body: pb.AnyValue{StringValue: ptrTo("log-line-message")}},
-						{Attributes: []*pb.KeyValue{}, TimeUnixNano: 1235, SeverityNumber: 21, Body: pb.AnyValue{StringValue: ptrTo("log-line-message-msg-2")}},
+						{Attributes: []*pb.KeyValue{}, TimeUnixNano: 1235, SeverityNumber: 25, Body: pb.AnyValue{StringValue: ptrTo("log-line-message-msg-2")}},
 						{Attributes: []*pb.KeyValue{}, TimeUnixNano: 1236, SeverityNumber: -1, Body: pb.AnyValue{StringValue: ptrTo("log-line-message-msg-2")}},
 					},
 				},
@@ -66,9 +88,9 @@ func TestPushProtoOk(t *testing.T) {
 		},
 	},
 		[]int64{1234, 1235, 1236},
-		`{"logger":"context","instance_id":"10","node_taints":"[{\"Key\":\"role\",\"Value\":{\"StringValue\":\"dev\",\"BoolValue\":null,\"IntValue\":null,\"DoubleValue\":null,\"ArrayValue\":null,\"KeyValueList\":null,\"BytesValue\":null}},{\"Key\":\"cluster_load_percent\",\"Value\":{\"StringValue\":null,\"BoolValue\":null,\"IntValue\":null,\"DoubleValue\":0.55,\"ArrayValue\":null,\"KeyValueList\":null,\"BytesValue\":null}}]","_msg":"log-line-message","severity":"Trace"}
-{"logger":"context","instance_id":"10","node_taints":"[{\"Key\":\"role\",\"Value\":{\"StringValue\":\"dev\",\"BoolValue\":null,\"IntValue\":null,\"DoubleValue\":null,\"ArrayValue\":null,\"KeyValueList\":null,\"BytesValue\":null}},{\"Key\":\"cluster_load_percent\",\"Value\":{\"StringValue\":null,\"BoolValue\":null,\"IntValue\":null,\"DoubleValue\":0.55,\"ArrayValue\":null,\"KeyValueList\":null,\"BytesValue\":null}}]","_msg":"log-line-message-msg-2","severity":"Unspecified"}
-{"logger":"context","instance_id":"10","node_taints":"[{\"Key\":\"role\",\"Value\":{\"StringValue\":\"dev\",\"BoolValue\":null,\"IntValue\":null,\"DoubleValue\":null,\"ArrayValue\":null,\"KeyValueList\":null,\"BytesValue\":null}},{\"Key\":\"cluster_load_percent\",\"Value\":{\"StringValue\":null,\"BoolValue\":null,\"IntValue\":null,\"DoubleValue\":0.55,\"ArrayValue\":null,\"KeyValueList\":null,\"BytesValue\":null}}]","_msg":"log-line-message-msg-2","severity":"Unspecified"}`,
+		`{"logger":"context","instance_id":"10","node_taints":"{\"role\":\"dev\",\"cluster_load_percent\":0.55}","_msg":"log-line-message","severity":"Trace"}
+{"logger":"context","instance_id":"10","node_taints":"{\"role\":\"dev\",\"cluster_load_percent\":0.55}","_msg":"log-line-message-msg-2","severity":"Unspecified"}
+{"logger":"context","instance_id":"10","node_taints":"{\"role\":\"dev\",\"cluster_load_percent\":0.55}","_msg":"log-line-message-msg-2","severity":"Unspecified"}`,
 	)
 
 	// multi-scope with resource attributes and multi-line
@@ -106,19 +128,21 @@ func TestPushProtoOk(t *testing.T) {
 				{
 					LogRecords: []pb.LogRecord{
 						{TimeUnixNano: 2347, SeverityNumber: 12, Body: pb.AnyValue{StringValue: ptrTo("log-line-resource-scope-1-1-0")}},
-						{ObservedTimeUnixNano: 2348, SeverityNumber: 12, Body: pb.AnyValue{StringValue: ptrTo("log-line-resource-scope-1-1-1")}},
+						{TraceID: "1234", SpanID: "45", ObservedTimeUnixNano: 2348, SeverityNumber: 12, Body: pb.AnyValue{StringValue: ptrTo("log-line-resource-scope-1-1-1")}},
+						{TraceID: "4bf92f3577b34da6a3ce929d0e0e4736", SpanID: "00f067aa0ba902b7", ObservedTimeUnixNano: 3333, Body: pb.AnyValue{StringValue: ptrTo("log-line-resource-scope-1-1-2")}},
 					},
 				},
 			},
 		},
 	},
-		[]int64{1234, 1235, 2345, 2346, 2347, 2348},
-		`{"logger":"context","instance_id":"10","node_taints":"[{\"Key\":\"role\",\"Value\":{\"StringValue\":\"dev\",\"BoolValue\":null,\"IntValue\":null,\"DoubleValue\":null,\"ArrayValue\":null,\"KeyValueList\":null,\"BytesValue\":null}},{\"Key\":\"cluster_load_percent\",\"Value\":{\"StringValue\":null,\"BoolValue\":null,\"IntValue\":null,\"DoubleValue\":0.55,\"ArrayValue\":null,\"KeyValueList\":null,\"BytesValue\":null}}]","_msg":"log-line-message","severity":"Trace"}
-{"logger":"context","instance_id":"10","node_taints":"[{\"Key\":\"role\",\"Value\":{\"StringValue\":\"dev\",\"BoolValue\":null,\"IntValue\":null,\"DoubleValue\":null,\"ArrayValue\":null,\"KeyValueList\":null,\"BytesValue\":null}},{\"Key\":\"cluster_load_percent\",\"Value\":{\"StringValue\":null,\"BoolValue\":null,\"IntValue\":null,\"DoubleValue\":0.55,\"ArrayValue\":null,\"KeyValueList\":null,\"BytesValue\":null}}]","_msg":"log-line-message-msg-2","severity":"Debug"}
+		[]int64{1234, 1235, 2345, 2346, 2347, 2348, 3333},
+		`{"logger":"context","instance_id":"10","node_taints":"{\"role\":\"dev\",\"cluster_load_percent\":0.55}","_msg":"log-line-message","severity":"Trace"}
+{"logger":"context","instance_id":"10","node_taints":"{\"role\":\"dev\",\"cluster_load_percent\":0.55}","_msg":"log-line-message-msg-2","severity":"Debug"}
 {"_msg":"log-line-resource-scope-1-0-0","severity":"Info2"}
 {"_msg":"log-line-resource-scope-1-0-1","severity":"Info2"}
 {"_msg":"log-line-resource-scope-1-1-0","severity":"Info4"}
-{"_msg":"log-line-resource-scope-1-1-1","severity":"Info4"}`,
+{"_msg":"log-line-resource-scope-1-1-1","trace_id":"1234","span_id":"45","severity":"Info4"}
+{"_msg":"log-line-resource-scope-1-1-2","trace_id":"4bf92f3577b34da6a3ce929d0e0e4736","span_id":"00f067aa0ba902b7","severity":"Unspecified"}`,
 	)
 }
 

app/vlinsert/opentelemetry/opentelemetry_timing_test.go

@@ -4,7 +4,7 @@ import (
 	"fmt"
 	"testing"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentelemetry/pb"
 )
 
@@ -21,7 +21,7 @@ func BenchmarkParseProtobufRequest(b *testing.B) {
 }
 
 func benchmarkParseProtobufRequest(b *testing.B, streams, rows, labels int) {
-	blp := &insertutils.BenchmarkLogMessageProcessor{}
+	blp := &insertutil.BenchmarkLogMessageProcessor{}
 	b.ReportAllocs()
 	b.SetBytes(int64(streams * rows))
 	b.RunParallel(func(pb *testing.PB) {

app/vlinsert/syslog/syslog.go

@@ -16,9 +16,7 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/klauspost/compress/gzip"
-
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
@@ -27,7 +25,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/netutil"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/slicesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/writeconcurrencylimiter"
 	"github.com/VictoriaMetrics/metrics"
@@ -274,14 +272,14 @@ func runTCPListener(addr string, argIdx int) {
 
 func checkCompressMethod(compressMethod, addr, protocol string) {
 	switch compressMethod {
-	case "", "none", "gzip", "deflate":
+	case "", "none", "zstd", "gzip", "deflate":
 		return
 	default:
-		logger.Fatalf("unsupported -syslog.compressMethod.%s=%q for -syslog.listenAddr.%s=%q; supported values: 'none', 'gzip', 'deflate'", protocol, compressMethod, protocol, addr)
+		logger.Fatalf("unsupported -syslog.compressMethod.%s=%q for -syslog.listenAddr.%s=%q; supported values: 'none', 'zstd', 'gzip', 'deflate'", protocol, compressMethod, protocol, addr)
 	}
 }
 
-func serveUDP(ln net.PacketConn, tenantID logstorage.TenantID, compressMethod string, useLocalTimestamp bool, streamFields, ignoreFields []string, extraFields []logstorage.Field) {
+func serveUDP(ln net.PacketConn, tenantID logstorage.TenantID, encoding string, useLocalTimestamp bool, streamFields, ignoreFields []string, extraFields []logstorage.Field) {
 	gomaxprocs := cgroup.AvailableCPUs()
 	var wg sync.WaitGroup
 	localAddr := ln.LocalAddr()
@@ -289,7 +287,7 @@ func serveUDP(ln net.PacketConn, tenantID logstorage.TenantID, compressMethod st
 		wg.Add(1)
 		go func() {
 			defer wg.Done()
-			cp := insertutils.GetCommonParamsForSyslog(tenantID, streamFields, ignoreFields, extraFields)
+			cp := insertutil.GetCommonParamsForSyslog(tenantID, streamFields, ignoreFields, extraFields)
 			var bb bytesutil.ByteBuffer
 			bb.B = bytesutil.ResizeNoCopyNoOverallocate(bb.B, 64*1024)
 			for {
@@ -314,7 +312,7 @@ func serveUDP(ln net.PacketConn, tenantID logstorage.TenantID, compressMethod st
 				}
 				bb.B = bb.B[:n]
 				udpRequestsTotal.Inc()
-				if err := processStream("udp", bb.NewReader(), compressMethod, useLocalTimestamp, cp); err != nil {
+				if err := processStream("udp", bb.NewReader(), encoding, useLocalTimestamp, cp); err != nil {
 					logger.Errorf("syslog: cannot process UDP data from %s at %s: %s", remoteAddr, localAddr, err)
 				}
 			}
@@ -323,7 +321,7 @@ func serveUDP(ln net.PacketConn, tenantID logstorage.TenantID, compressMethod st
 	wg.Wait()
 }
 
-func serveTCP(ln net.Listener, tenantID logstorage.TenantID, compressMethod string, useLocalTimestamp bool, streamFields, ignoreFields []string, extraFields []logstorage.Field) {
+func serveTCP(ln net.Listener, tenantID logstorage.TenantID, encoding string, useLocalTimestamp bool, streamFields, ignoreFields []string, extraFields []logstorage.Field) {
 	var cm ingestserver.ConnsMap
 	cm.Init("syslog")
 
@@ -353,8 +351,8 @@ func serveTCP(ln net.Listener, tenantID logstorage.TenantID, compressMethod stri
 
 		wg.Add(1)
 		go func() {
-			cp := insertutils.GetCommonParamsForSyslog(tenantID, streamFields, ignoreFields, extraFields)
-			if err := processStream("tcp", c, compressMethod, useLocalTimestamp, cp); err != nil {
+			cp := insertutil.GetCommonParamsForSyslog(tenantID, streamFields, ignoreFields, extraFields)
+			if err := processStream("tcp", c, encoding, useLocalTimestamp, cp); err != nil {
 				logger.Errorf("syslog: cannot process TCP data at %q: %s", addr, err)
 			}
 
@@ -369,52 +367,29 @@ func serveTCP(ln net.Listener, tenantID logstorage.TenantID, compressMethod stri
 }
 
 // processStream parses a stream of syslog messages from r and ingests them into vlstorage.
-func processStream(protocol string, r io.Reader, compressMethod string, useLocalTimestamp bool, cp *insertutils.CommonParams) error {
+func processStream(protocol string, r io.Reader, encoding string, useLocalTimestamp bool, cp *insertutil.CommonParams) error {
 	if err := vlstorage.CanWriteData(); err != nil {
 		return err
 	}
 
-	lmp := cp.NewLogMessageProcessor("syslog_" + protocol)
-	err := processStreamInternal(r, compressMethod, useLocalTimestamp, lmp)
+	lmp := cp.NewLogMessageProcessor("syslog_"+protocol, true)
+	err := processStreamInternal(r, encoding, useLocalTimestamp, lmp)
 	lmp.MustClose()
 
 	return err
 }
 
-func processStreamInternal(r io.Reader, compressMethod string, useLocalTimestamp bool, lmp insertutils.LogMessageProcessor) error {
-	switch compressMethod {
-	case "", "none":
-	case "gzip":
-		zr, err := common.GetGzipReader(r)
-		if err != nil {
-			return fmt.Errorf("cannot read gzipped data: %w", err)
-		}
-		r = zr
-	case "deflate":
-		zr, err := common.GetZlibReader(r)
-		if err != nil {
-			return fmt.Errorf("cannot read deflated data: %w", err)
-		}
-		r = zr
-	default:
-		logger.Panicf("BUG: unsupported compressMethod=%q; supported values: none, gzip, deflate", compressMethod)
+func processStreamInternal(r io.Reader, encoding string, useLocalTimestamp bool, lmp insertutil.LogMessageProcessor) error {
+	reader, err := protoparserutil.GetUncompressedReader(r, encoding)
+	if err != nil {
+		return fmt.Errorf("cannot decode syslog data: %w", err)
 	}
+	defer protoparserutil.PutUncompressedReader(reader)
 
-	err := processUncompressedStream(r, useLocalTimestamp, lmp)
-
-	switch compressMethod {
-	case "gzip":
-		zr := r.(*gzip.Reader)
-		common.PutGzipReader(zr)
-	case "deflate":
-		zr := r.(io.ReadCloser)
-		common.PutZlibReader(zr)
-	}
-
-	return err
+	return processUncompressedStream(reader, useLocalTimestamp, lmp)
 }
 
-func processUncompressedStream(r io.Reader, useLocalTimestamp bool, lmp insertutils.LogMessageProcessor) error {
+func processUncompressedStream(r io.Reader, useLocalTimestamp bool, lmp insertutil.LogMessageProcessor) error {
 	wcr := writeconcurrencylimiter.GetReader(r)
 	defer writeconcurrencylimiter.PutReader(wcr)
 
@@ -499,7 +474,7 @@ again:
 			slr.err = fmt.Errorf("cannot parse message length from %q: %w", msgLenStr, err)
 			return false
 		}
-		if maxMsgLen := insertutils.MaxLineSizeBytes.IntN(); msgLen > uint64(maxMsgLen) {
+		if maxMsgLen := insertutil.MaxLineSizeBytes.IntN(); msgLen > uint64(maxMsgLen) {
 			slr.err = fmt.Errorf("cannot read message longer than %d bytes; msgLen=%d", maxMsgLen, msgLen)
 			return false
 		}
@@ -551,7 +526,7 @@ func putSyslogLineReader(slr *syslogLineReader) {
 
 var syslogLineReaderPool sync.Pool
 
-func processLine(line []byte, currentYear int, timezone *time.Location, useLocalTimestamp bool, lmp insertutils.LogMessageProcessor) error {
+func processLine(line []byte, currentYear int, timezone *time.Location, useLocalTimestamp bool, lmp insertutil.LogMessageProcessor) error {
 	p := logstorage.GetSyslogParser(currentYear, timezone)
 	lineStr := bytesutil.ToUnsafeString(line)
 	p.Parse(lineStr)
@@ -560,7 +535,7 @@ func processLine(line []byte, currentYear int, timezone *time.Location, useLocal
 	if useLocalTimestamp {
 		ts = time.Now().UnixNano()
 	} else {
-		nsecs, err := insertutils.ExtractTimestampFromFields("timestamp", p.Fields)
+		nsecs, err := insertutil.ExtractTimestampFromFields("timestamp", p.Fields)
 		if err != nil {
 			return fmt.Errorf("cannot get timestamp from syslog line %q: %w", line, err)
 		}

app/vlinsert/syslog/syslog_test.go

@@ -6,7 +6,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlinsert/insertutil"
 )
 
 func TestSyslogLineReader_Success(t *testing.T) {
@@ -84,7 +84,7 @@ func TestProcessStreamInternal_Success(t *testing.T) {
 		globalTimezone = time.UTC
 		globalCurrentYear.Store(int64(currentYear))
 
-		tlp := &insertutils.TestLogMessageProcessor{}
+		tlp := &insertutil.TestLogMessageProcessor{}
 		r := bytes.NewBufferString(data)
 		if err := processStreamInternal(r, "", false, tlp); err != nil {
 			t.Fatalf("unexpected error: %s", err)
@@ -113,7 +113,7 @@ func TestProcessStreamInternal_Failure(t *testing.T) {
 		MustInit()
 		defer MustStop()
 
-		tlp := &insertutils.TestLogMessageProcessor{}
+		tlp := &insertutil.TestLogMessageProcessor{}
 		r := bytes.NewBufferString(data)
 		if err := processStreamInternal(r, "", false, tlp); err == nil {
 			t.Fatalf("expecting non-nil error")

app/vlogscli/json_prettifier.go

@@ -154,7 +154,7 @@ func readNextJSONObject(d *json.Decoder) ([]logstorage.Field, error) {
 		}
 		value, ok := t.(string)
 		if !ok {
-			return nil, fmt.Errorf("unexpected token read for oject value: %v; want string", t)
+			return nil, fmt.Errorf("unexpected token read for object value: %v; want string", t)
 		}
 
 		fields = append(fields, logstorage.Field{

app/vlogsgenerator/main.go

@@ -31,7 +31,7 @@ var (
 	totalStreams  = flag.Int("totalStreams", 0, "The number of total log streams; if -totalStreams > -activeStreams, then some active streams are substituted with new streams "+
 		"during data generation")
 	logsPerStream     = flag.Int64("logsPerStream", 1_000, "The number of log entries to generate per each log stream. Log entries are evenly distributed between -start and -end")
-	constFieldsPerLog = flag.Int("constFieldsPerLog", 3, "The number of fields with constaint values to generate per each log entry; "+
+	constFieldsPerLog = flag.Int("constFieldsPerLog", 3, "The number of fields with constant values to generate per each log entry; "+
 		"see https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model")
 	varFieldsPerLog = flag.Int("varFieldsPerLog", 1, "The number of fields with variable values to generate per each log entry; "+
 		"see https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model")
@@ -177,7 +177,10 @@ func generateAndPushLogs(cfg *workerConfig, workerID int) {
 	sw := &statWriter{
 		w: pw,
 	}
-	bw := bufio.NewWriter(sw)
+
+	// The 1MB write buffer increases data ingestion performance by reducing the number of send() syscalls
+	bw := bufio.NewWriterSize(sw, 1024*1024)
+
 	doneCh := make(chan struct{})
 	go func() {
 		generateLogs(bw, workerID, cfg.activeStreams, cfg.totalStreams)

app/vlselect/internalselect/internalselect.go

@@ -0,0 +1,324 @@
+package internalselect
+
+import (
+	"context"
+	"flag"
+	"fmt"
+	"net/http"
+	"strconv"
+	"sync"
+	"time"
+
+	"github.com/VictoriaMetrics/metrics"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage/netselect"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/atomicutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding/zstd"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/netutil"
+)
+
+var disableSelect = flag.Bool("internalselect.disable", false, "Whether to disable /internal/select/* HTTP endpoints")
+
+// RequestHandler processes requests to /internal/select/*
+func RequestHandler(ctx context.Context, w http.ResponseWriter, r *http.Request) {
+	if *disableSelect {
+		httpserver.Errorf(w, r, "requests to /internal/select/* are disabled with -internalselect.disable command-line flag")
+		return
+	}
+
+	startTime := time.Now()
+
+	path := r.URL.Path
+	rh := requestHandlers[path]
+	if rh == nil {
+		httpserver.Errorf(w, r, "unsupported endpoint requested: %s", path)
+		return
+	}
+
+	metrics.GetOrCreateCounter(fmt.Sprintf(`vl_http_requests_total{path=%q}`, path)).Inc()
+	if err := rh(ctx, w, r); err != nil && !netutil.IsTrivialNetworkError(err) {
+		metrics.GetOrCreateCounter(fmt.Sprintf(`vl_http_request_errors_total{path=%q}`, path)).Inc()
+		httpserver.Errorf(w, r, "%s", err)
+		// The return is skipped intentionally in order to track the duration of failed queries.
+	}
+	metrics.GetOrCreateSummary(fmt.Sprintf(`vl_http_request_duration_seconds{path=%q}`, path)).UpdateDuration(startTime)
+}
+
+var requestHandlers = map[string]func(ctx context.Context, w http.ResponseWriter, r *http.Request) error{
+	"/internal/select/query":               processQueryRequest,
+	"/internal/select/field_names":         processFieldNamesRequest,
+	"/internal/select/field_values":        processFieldValuesRequest,
+	"/internal/select/stream_field_names":  processStreamFieldNamesRequest,
+	"/internal/select/stream_field_values": processStreamFieldValuesRequest,
+	"/internal/select/streams":             processStreamsRequest,
+	"/internal/select/stream_ids":          processStreamIDsRequest,
+}
+
+func processQueryRequest(ctx context.Context, w http.ResponseWriter, r *http.Request) error {
+	cp, err := getCommonParams(r, netselect.QueryProtocolVersion)
+	if err != nil {
+		return err
+	}
+
+	w.Header().Set("Content-Type", "application/octet-stream")
+
+	var wLock sync.Mutex
+	var dataLenBuf []byte
+
+	sendBuf := func(bb *bytesutil.ByteBuffer) error {
+		if len(bb.B) == 0 {
+			return nil
+		}
+
+		data := bb.B
+		if !cp.DisableCompression {
+			bufLen := len(bb.B)
+			bb.B = zstd.CompressLevel(bb.B, bb.B, 1)
+			data = bb.B[bufLen:]
+		}
+
+		wLock.Lock()
+		dataLenBuf = encoding.MarshalUint64(dataLenBuf[:0], uint64(len(data)))
+		_, err := w.Write(dataLenBuf)
+		if err == nil {
+			_, err = w.Write(data)
+		}
+		wLock.Unlock()
+
+		// Reset the sent buf
+		bb.Reset()
+
+		return err
+	}
+
+	var bufs atomicutil.Slice[bytesutil.ByteBuffer]
+
+	var errGlobalLock sync.Mutex
+	var errGlobal error
+
+	writeBlock := func(workerID uint, db *logstorage.DataBlock) {
+		if errGlobal != nil {
+			return
+		}
+
+		bb := bufs.Get(workerID)
+
+		bb.B = db.Marshal(bb.B)
+
+		if len(bb.B) < 1024*1024 {
+			// Fast path - the bb is too small to be sent to the client yet.
+			return
+		}
+
+		// Slow path - the bb must be sent to the client.
+		if err := sendBuf(bb); err != nil {
+			errGlobalLock.Lock()
+			if errGlobal != nil {
+				errGlobal = err
+			}
+			errGlobalLock.Unlock()
+		}
+	}
+
+	if err := vlstorage.RunQuery(ctx, cp.TenantIDs, cp.Query, writeBlock); err != nil {
+		return err
+	}
+	if errGlobal != nil {
+		return errGlobal
+	}
+
+	// Send the remaining data
+	for _, bb := range bufs.GetSlice() {
+		if err := sendBuf(bb); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func processFieldNamesRequest(ctx context.Context, w http.ResponseWriter, r *http.Request) error {
+	cp, err := getCommonParams(r, netselect.FieldNamesProtocolVersion)
+	if err != nil {
+		return err
+	}
+
+	fieldNames, err := vlstorage.GetFieldNames(ctx, cp.TenantIDs, cp.Query)
+	if err != nil {
+		return fmt.Errorf("cannot obtain field names: %w", err)
+	}
+
+	return writeValuesWithHits(w, fieldNames, cp.DisableCompression)
+}
+
+func processFieldValuesRequest(ctx context.Context, w http.ResponseWriter, r *http.Request) error {
+	cp, err := getCommonParams(r, netselect.FieldValuesProtocolVersion)
+	if err != nil {
+		return err
+	}
+
+	fieldName := r.FormValue("field")
+
+	limit, err := getInt64FromRequest(r, "limit")
+	if err != nil {
+		return err
+	}
+
+	fieldValues, err := vlstorage.GetFieldValues(ctx, cp.TenantIDs, cp.Query, fieldName, uint64(limit))
+	if err != nil {
+		return fmt.Errorf("cannot obtain field values: %w", err)
+	}
+
+	return writeValuesWithHits(w, fieldValues, cp.DisableCompression)
+}
+
+func processStreamFieldNamesRequest(ctx context.Context, w http.ResponseWriter, r *http.Request) error {
+	cp, err := getCommonParams(r, netselect.StreamFieldNamesProtocolVersion)
+	if err != nil {
+		return err
+	}
+
+	fieldNames, err := vlstorage.GetStreamFieldNames(ctx, cp.TenantIDs, cp.Query)
+	if err != nil {
+		return fmt.Errorf("cannot obtain stream field names: %w", err)
+	}
+
+	return writeValuesWithHits(w, fieldNames, cp.DisableCompression)
+}
+
+func processStreamFieldValuesRequest(ctx context.Context, w http.ResponseWriter, r *http.Request) error {
+	cp, err := getCommonParams(r, netselect.StreamFieldValuesProtocolVersion)
+	if err != nil {
+		return err
+	}
+
+	fieldName := r.FormValue("field")
+
+	limit, err := getInt64FromRequest(r, "limit")
+	if err != nil {
+		return err
+	}
+
+	fieldValues, err := vlstorage.GetStreamFieldValues(ctx, cp.TenantIDs, cp.Query, fieldName, uint64(limit))
+	if err != nil {
+		return fmt.Errorf("cannot obtain stream field values: %w", err)
+	}
+
+	return writeValuesWithHits(w, fieldValues, cp.DisableCompression)
+}
+
+func processStreamsRequest(ctx context.Context, w http.ResponseWriter, r *http.Request) error {
+	cp, err := getCommonParams(r, netselect.StreamsProtocolVersion)
+	if err != nil {
+		return err
+	}
+
+	limit, err := getInt64FromRequest(r, "limit")
+	if err != nil {
+		return err
+	}
+
+	streams, err := vlstorage.GetStreams(ctx, cp.TenantIDs, cp.Query, uint64(limit))
+	if err != nil {
+		return fmt.Errorf("cannot obtain streams: %w", err)
+	}
+
+	return writeValuesWithHits(w, streams, cp.DisableCompression)
+}
+
+func processStreamIDsRequest(ctx context.Context, w http.ResponseWriter, r *http.Request) error {
+	cp, err := getCommonParams(r, netselect.StreamIDsProtocolVersion)
+	if err != nil {
+		return err
+	}
+
+	limit, err := getInt64FromRequest(r, "limit")
+	if err != nil {
+		return err
+	}
+
+	streamIDs, err := vlstorage.GetStreamIDs(ctx, cp.TenantIDs, cp.Query, uint64(limit))
+	if err != nil {
+		return fmt.Errorf("cannot obtain streams: %w", err)
+	}
+
+	return writeValuesWithHits(w, streamIDs, cp.DisableCompression)
+}
+
+type commonParams struct {
+	TenantIDs []logstorage.TenantID
+	Query     *logstorage.Query
+
+	DisableCompression bool
+}
+
+func getCommonParams(r *http.Request, expectedProtocolVersion string) (*commonParams, error) {
+	version := r.FormValue("version")
+	if version != expectedProtocolVersion {
+		return nil, fmt.Errorf("unexpected version=%q; want %q", version, expectedProtocolVersion)
+	}
+
+	tenantIDsStr := r.FormValue("tenant_ids")
+	tenantIDs, err := logstorage.UnmarshalTenantIDs([]byte(tenantIDsStr))
+	if err != nil {
+		return nil, fmt.Errorf("cannot unmarshal tenant_ids=%q: %w", tenantIDsStr, err)
+	}
+
+	timestamp, err := getInt64FromRequest(r, "timestamp")
+	if err != nil {
+		return nil, err
+	}
+
+	qStr := r.FormValue("query")
+	q, err := logstorage.ParseQueryAtTimestamp(qStr, timestamp)
+	if err != nil {
+		return nil, fmt.Errorf("cannot unmarshal query=%q: %w", qStr, err)
+	}
+
+	s := r.FormValue("disable_compression")
+	disableCompression, err := strconv.ParseBool(s)
+	if err != nil {
+		return nil, fmt.Errorf("cannot parse disable_compression=%q: %w", s, err)
+	}
+
+	cp := &commonParams{
+		TenantIDs: tenantIDs,
+		Query:     q,
+
+		DisableCompression: disableCompression,
+	}
+	return cp, nil
+}
+
+func writeValuesWithHits(w http.ResponseWriter, vhs []logstorage.ValueWithHits, disableCompression bool) error {
+	var b []byte
+	for i := range vhs {
+		b = vhs[i].Marshal(b)
+	}
+
+	if !disableCompression {
+		b = zstd.CompressLevel(nil, b, 1)
+	}
+
+	w.Header().Set("Content-Type", "application/octet-stream")
+
+	if _, err := w.Write(b); err != nil {
+		return fmt.Errorf("cannot send response to the client: %w", err)
+	}
+
+	return nil
+}
+
+func getInt64FromRequest(r *http.Request, argName string) (int64, error) {
+	s := r.FormValue(argName)
+	n, err := strconv.ParseInt(s, 10, 64)
+	if err != nil {
+		return 0, fmt.Errorf("cannot parse %s=%q: %w", argName, s, err)
+	}
+	return n, nil
+}

app/vlselect/logsql/buffered_writer.go

@@ -1,47 +0,0 @@
-package logsql
-
-import (
-	"bufio"
-	"io"
-	"sync"
-)
-
-func getBufferedWriter(w io.Writer) *bufferedWriter {
-	v := bufferedWriterPool.Get()
-	if v == nil {
-		return &bufferedWriter{
-			bw: bufio.NewWriter(w),
-		}
-	}
-	bw := v.(*bufferedWriter)
-	bw.bw.Reset(w)
-	return bw
-}
-
-func putBufferedWriter(bw *bufferedWriter) {
-	bw.reset()
-	bufferedWriterPool.Put(bw)
-}
-
-var bufferedWriterPool sync.Pool
-
-type bufferedWriter struct {
-	mu sync.Mutex
-	bw *bufio.Writer
-}
-
-func (bw *bufferedWriter) reset() {
-	// nothing to do
-}
-
-func (bw *bufferedWriter) WriteIgnoreErrors(p []byte) {
-	bw.mu.Lock()
-	_, _ = bw.bw.Write(p)
-	bw.mu.Unlock()
-}
-
-func (bw *bufferedWriter) FlushIgnoreErrors() {
-	bw.mu.Lock()
-	_ = bw.bw.Flush()
-	bw.mu.Unlock()
-}

app/vlselect/logsql/logsql.go

@@ -3,6 +3,7 @@ package logsql
 import (
 	"context"
 	"fmt"
+	"io"
 	"math"
 	"net/http"
 	"regexp"
@@ -11,15 +12,17 @@ import (
 	"strconv"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/VictoriaMetrics/metrics"
 	"github.com/valyala/fastjson"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/atomicutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeutil"
@@ -35,32 +38,35 @@ func ProcessFacetsRequest(ctx context.Context, w http.ResponseWriter, r *http.Re
 		return
 	}
 
-	limit, err := httputils.GetInt(r, "limit")
+	limit, err := httputil.GetInt(r, "limit")
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
 	}
-	maxValuesPerField, err := httputils.GetInt(r, "max_values_per_field")
+	maxValuesPerField, err := httputil.GetInt(r, "max_values_per_field")
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
 	}
-	maxValueLen, err := httputils.GetInt(r, "max_value_len")
+	maxValueLen, err := httputil.GetInt(r, "max_value_len")
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
 	}
-	keepConstFields := httputils.GetBool(r, "keep_const_fields")
+	keepConstFields := httputil.GetBool(r, "keep_const_fields")
 
 	q.DropAllPipes()
 	q.AddFacetsPipe(limit, maxValuesPerField, maxValueLen, keepConstFields)
 
 	var mLock sync.Mutex
 	m := make(map[string][]facetEntry)
-	writeBlock := func(_ uint, _ []int64, columns []logstorage.BlockColumn) {
-		if len(columns) == 0 || len(columns[0].Values) == 0 {
+	writeBlock := func(_ uint, db *logstorage.DataBlock) {
+		rowsCount := db.RowsCount()
+		if rowsCount == 0 {
 			return
 		}
+
+		columns := db.Columns
 		if len(columns) != 3 {
 			logger.Panicf("BUG: expecting 3 columns; got %d columns", len(columns))
 		}
@@ -141,7 +147,7 @@ func ProcessHitsRequest(ctx context.Context, w http.ResponseWriter, r *http.Requ
 	fields := r.Form["field"]
 
 	// Obtain limit on the number of top fields entries.
-	fieldsLimit, err := httputils.GetInt(r, "fields_limit")
+	fieldsLimit, err := httputil.GetInt(r, "fields_limit")
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
@@ -156,17 +162,19 @@ func ProcessHitsRequest(ctx context.Context, w http.ResponseWriter, r *http.Requ
 
 	var mLock sync.Mutex
 	m := make(map[string]*hitsSeries)
-	writeBlock := func(_ uint, timestamps []int64, columns []logstorage.BlockColumn) {
-		if len(columns) == 0 || len(columns[0].Values) == 0 {
+	writeBlock := func(_ uint, db *logstorage.DataBlock) {
+		rowsCount := db.RowsCount()
+		if rowsCount == 0 {
 			return
 		}
 
+		columns := db.Columns
 		timestampValues := columns[0].Values
 		hitsValues := columns[len(columns)-1].Values
 		columns = columns[1 : len(columns)-1]
 
 		bb := blockResultPool.Get()
-		for i := range timestamps {
+		for i := 0; i < rowsCount; i++ {
 			timestampStr := strings.Clone(timestampValues[i])
 			hitsStr := strings.Clone(hitsValues[i])
 			hits, err := strconv.ParseUint(hitsStr, 10, 64)
@@ -205,6 +213,8 @@ func ProcessHitsRequest(ctx context.Context, w http.ResponseWriter, r *http.Requ
 	WriteHitsSeries(w, m)
 }
 
+var blockResultPool bytesutil.ByteBufferPool
+
 func getTopHitsSeries(m map[string]*hitsSeries, fieldsLimit int) map[string]*hitsSeries {
 	if fieldsLimit <= 0 || fieldsLimit >= len(m) {
 		return m
@@ -310,7 +320,7 @@ func ProcessFieldValuesRequest(ctx context.Context, w http.ResponseWriter, r *ht
 	}
 
 	// Parse limit query arg
-	limit, err := httputils.GetInt(r, "limit")
+	limit, err := httputil.GetInt(r, "limit")
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
@@ -370,7 +380,7 @@ func ProcessStreamFieldValuesRequest(ctx context.Context, w http.ResponseWriter,
 	}
 
 	// Parse limit query arg
-	limit, err := httputils.GetInt(r, "limit")
+	limit, err := httputil.GetInt(r, "limit")
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
@@ -401,7 +411,7 @@ func ProcessStreamIDsRequest(ctx context.Context, w http.ResponseWriter, r *http
 	}
 
 	// Parse limit query arg
-	limit, err := httputils.GetInt(r, "limit")
+	limit, err := httputil.GetInt(r, "limit")
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
@@ -432,7 +442,7 @@ func ProcessStreamsRequest(ctx context.Context, w http.ResponseWriter, r *http.R
 	}
 
 	// Parse limit query arg
-	limit, err := httputils.GetInt(r, "limit")
+	limit, err := httputil.GetInt(r, "limit")
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
@@ -470,21 +480,21 @@ func ProcessLiveTailRequest(ctx context.Context, w http.ResponseWriter, r *http.
 		return
 	}
 
-	refreshIntervalMsecs, err := httputils.GetDuration(r, "refresh_interval", 1000)
+	refreshIntervalMsecs, err := httputil.GetDuration(r, "refresh_interval", 1000)
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
 	}
 	refreshInterval := time.Millisecond * time.Duration(refreshIntervalMsecs)
 
-	startOffsetMsecs, err := httputils.GetDuration(r, "start_offset", 5*1000)
+	startOffsetMsecs, err := httputil.GetDuration(r, "start_offset", 5*1000)
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
 	}
 	startOffset := startOffsetMsecs * 1e6
 
-	offsetMsecs, err := httputils.GetDuration(r, "offset", 1000)
+	offsetMsecs, err := httputil.GetDuration(r, "offset", 1000)
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
@@ -536,7 +546,7 @@ var liveTailRequests = metrics.NewCounter(`vl_live_tailing_requests`)
 const tailOffsetNsecs = 5e9
 
 type logRow struct {
-	timestamp int64
+	timestamp string
 	fields    []logstorage.Field
 }
 
@@ -552,7 +562,7 @@ type tailProcessor struct {
 	mu sync.Mutex
 
 	perStreamRows  map[string][]logRow
-	lastTimestamps map[string]int64
+	lastTimestamps map[string]string
 
 	err error
 }
@@ -562,12 +572,12 @@ func newTailProcessor(cancel func()) *tailProcessor {
 		cancel: cancel,
 
 		perStreamRows:  make(map[string][]logRow),
-		lastTimestamps: make(map[string]int64),
+		lastTimestamps: make(map[string]string),
 	}
 }
 
-func (tp *tailProcessor) writeBlock(_ uint, timestamps []int64, columns []logstorage.BlockColumn) {
-	if len(timestamps) == 0 {
+func (tp *tailProcessor) writeBlock(_ uint, db *logstorage.DataBlock) {
+	if db.RowsCount() == 0 {
 		return
 	}
 
@@ -579,14 +589,8 @@ func (tp *tailProcessor) writeBlock(_ uint, timestamps []int64, columns []logsto
 	}
 
 	// Make sure columns contain _time field, since it is needed for proper tail work.
-	hasTime := false
-	for _, c := range columns {
-		if c.Name == "_time" {
-			hasTime = true
-			break
-		}
-	}
-	if !hasTime {
+	timestamps, ok := db.GetTimestamps()
+	if !ok {
 		tp.err = fmt.Errorf("missing _time field")
 		tp.cancel()
 		return
@@ -595,8 +599,8 @@ func (tp *tailProcessor) writeBlock(_ uint, timestamps []int64, columns []logsto
 	// Copy block rows to tp.perStreamRows
 	for i, timestamp := range timestamps {
 		streamID := ""
-		fields := make([]logstorage.Field, len(columns))
-		for j, c := range columns {
+		fields := make([]logstorage.Field, len(db.Columns))
+		for j, c := range db.Columns {
 			name := strings.Clone(c.Name)
 			value := strings.Clone(c.Values[i])
 
@@ -688,12 +692,15 @@ func ProcessStatsQueryRangeRequest(ctx context.Context, w http.ResponseWriter, r
 	m := make(map[string]*statsSeries)
 	var mLock sync.Mutex
 
-	writeBlock := func(_ uint, timestamps []int64, columns []logstorage.BlockColumn) {
+	writeBlock := func(_ uint, db *logstorage.DataBlock) {
+		rowsCount := db.RowsCount()
+
+		columns := db.Columns
 		clonedColumnNames := make([]string, len(columns))
 		for i, c := range columns {
 			clonedColumnNames[i] = strings.Clone(c.Name)
 		}
-		for i := range timestamps {
+		for i := 0; i < rowsCount; i++ {
 			// Do not move q.GetTimestamp() outside writeBlock, since ts
 			// must be initialized to query timestamp for every processed log row.
 			// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8312
@@ -802,12 +809,14 @@ func ProcessStatsQueryRequest(ctx context.Context, w http.ResponseWriter, r *htt
 	var rowsLock sync.Mutex
 
 	timestamp := q.GetTimestamp()
-	writeBlock := func(_ uint, timestamps []int64, columns []logstorage.BlockColumn) {
+	writeBlock := func(_ uint, db *logstorage.DataBlock) {
+		rowsCount := db.RowsCount()
+		columns := db.Columns
 		clonedColumnNames := make([]string, len(columns))
 		for i, c := range columns {
 			clonedColumnNames[i] = strings.Clone(c.Name)
 		}
-		for i := range timestamps {
+		for i := 0; i < rowsCount; i++ {
 			labels := make([]logstorage.Field, 0, len(byFields))
 			for j, c := range columns {
 				if slices.Contains(byFields, c.Name) {
@@ -863,17 +872,27 @@ func ProcessQueryRequest(ctx context.Context, w http.ResponseWriter, r *http.Req
 	}
 
 	// Parse limit query arg
-	limit, err := httputils.GetInt(r, "limit")
+	limit, err := httputil.GetInt(r, "limit")
 	if err != nil {
 		httpserver.Errorf(w, r, "%s", err)
 		return
 	}
 
-	bw := getBufferedWriter(w)
+	sw := &syncWriter{
+		w: w,
+	}
+
+	var bwShards atomicutil.Slice[bufferedWriter]
+	bwShards.Init = func(shard *bufferedWriter) {
+		shard.sw = sw
+	}
 	defer func() {
-		bw.FlushIgnoreErrors()
-		putBufferedWriter(bw)
+		shards := bwShards.GetSlice()
+		for _, shard := range shards {
+			shard.FlushIgnoreErrors()
+		}
 	}()
+
 	w.Header().Set("Content-Type", "application/stream+json")
 
 	if limit > 0 {
@@ -883,32 +902,34 @@ func ProcessQueryRequest(ctx context.Context, w http.ResponseWriter, r *http.Req
 				httpserver.Errorf(w, r, "%s", err)
 				return
 			}
-			bb := blockResultPool.Get()
-			b := bb.B
+			bw := bwShards.Get(0)
 			for i := range rows {
-				b = logstorage.MarshalFieldsToJSON(b[:0], rows[i].fields)
-				b = append(b, '\n')
-				bw.WriteIgnoreErrors(b)
+				bw.buf = logstorage.MarshalFieldsToJSON(bw.buf, rows[i].fields)
+				bw.buf = append(bw.buf, '\n')
+				if len(bw.buf) > 16*1024 {
+					bw.FlushIgnoreErrors()
+				}
 			}
-			bb.B = b
-			blockResultPool.Put(bb)
 			return
 		}
 
 		q.AddPipeLimit(uint64(limit))
 	}
 
-	writeBlock := func(_ uint, timestamps []int64, columns []logstorage.BlockColumn) {
-		if len(columns) == 0 || len(columns[0].Values) == 0 {
+	writeBlock := func(workerID uint, db *logstorage.DataBlock) {
+		rowsCount := db.RowsCount()
+		if rowsCount == 0 {
 			return
 		}
+		columns := db.Columns
 
-		bb := blockResultPool.Get()
-		for i := range timestamps {
-			WriteJSONRow(bb, columns, i)
+		bw := bwShards.Get(workerID)
+		for i := 0; i < rowsCount; i++ {
+			WriteJSONRow(bw, columns, i)
+			if len(bw.buf) > 16*1024 {
+				bw.FlushIgnoreErrors()
+			}
 		}
-		bw.WriteIgnoreErrors(bb.B)
-		blockResultPool.Put(bb)
 	}
 
 	if err := vlstorage.RunQuery(ctx, tenantIDs, q, writeBlock); err != nil {
@@ -917,14 +938,37 @@ func ProcessQueryRequest(ctx context.Context, w http.ResponseWriter, r *http.Req
 	}
 }
 
-var blockResultPool bytesutil.ByteBufferPool
-
-type row struct {
-	timestamp int64
-	fields    []logstorage.Field
+type syncWriter struct {
+	mu sync.Mutex
+	w  io.Writer
 }
 
-func getLastNQueryResults(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, limit int) ([]row, error) {
+func (sw *syncWriter) Write(p []byte) (int, error) {
+	sw.mu.Lock()
+	n, err := sw.w.Write(p)
+	sw.mu.Unlock()
+	return n, err
+}
+
+type bufferedWriter struct {
+	buf []byte
+	sw  *syncWriter
+}
+
+func (bw *bufferedWriter) Write(p []byte) (int, error) {
+	bw.buf = append(bw.buf, p...)
+
+	// Do not send bw.buf to bw.sw here, since the data at bw.buf may be incomplete (it must end with '\n')
+
+	return len(p), nil
+}
+
+func (bw *bufferedWriter) FlushIgnoreErrors() {
+	_, _ = bw.sw.Write(bw.buf)
+	bw.buf = bw.buf[:0]
+}
+
+func getLastNQueryResults(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, limit int) ([]logRow, error) {
 	limitUpper := 2 * limit
 	q.AddPipeLimit(uint64(limitUpper))
 
@@ -993,7 +1037,7 @@ func getLastNQueryResults(ctx context.Context, tenantIDs []logstorage.TenantID,
 	}
 }
 
-func getLastNRows(rows []row, limit int) []row {
+func getLastNRows(rows []logRow, limit int) []logRow {
 	sort.Slice(rows, func(i, j int) bool {
 		return rows[i].timestamp < rows[j].timestamp
 	})
@@ -1003,18 +1047,31 @@ func getLastNRows(rows []row, limit int) []row {
 	return rows
 }
 
-func getQueryResultsWithLimit(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, limit int) ([]row, error) {
+func getQueryResultsWithLimit(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, limit int) ([]logRow, error) {
 	ctxWithCancel, cancel := context.WithCancel(ctx)
 	defer cancel()
 
-	var rows []row
+	var missingTimeColumn atomic.Bool
+	var rows []logRow
 	var rowsLock sync.Mutex
-	writeBlock := func(_ uint, timestamps []int64, columns []logstorage.BlockColumn) {
+	writeBlock := func(_ uint, db *logstorage.DataBlock) {
+		if missingTimeColumn.Load() {
+			return
+		}
+
+		columns := db.Columns
 		clonedColumnNames := make([]string, len(columns))
 		for i, c := range columns {
 			clonedColumnNames[i] = strings.Clone(c.Name)
 		}
 
+		timestamps, ok := db.GetTimestamps()
+		if !ok {
+			missingTimeColumn.Store(true)
+			cancel()
+			return
+		}
+
 		for i, timestamp := range timestamps {
 			fields := make([]logstorage.Field, len(columns))
 			for j := range columns {
@@ -1024,7 +1081,7 @@ func getQueryResultsWithLimit(ctx context.Context, tenantIDs []logstorage.Tenant
 			}
 
 			rowsLock.Lock()
-			rows = append(rows, row{
+			rows = append(rows, logRow{
 				timestamp: timestamp,
 				fields:    fields,
 			})
@@ -1035,11 +1092,13 @@ func getQueryResultsWithLimit(ctx context.Context, tenantIDs []logstorage.Tenant
 			cancel()
 		}
 	}
-	if err := vlstorage.RunQuery(ctxWithCancel, tenantIDs, q, writeBlock); err != nil {
-		return nil, err
+	err := vlstorage.RunQuery(ctxWithCancel, tenantIDs, q, writeBlock)
+
+	if missingTimeColumn.Load() {
+		return nil, fmt.Errorf("missing _time column in the result for the query [%s]", q)
 	}
 
-	return rows, nil
+	return rows, err
 }
 
 func parseCommonArgs(r *http.Request) (*logstorage.Query, []logstorage.TenantID, error) {
@@ -1098,20 +1157,22 @@ func parseCommonArgs(r *http.Request) (*logstorage.Query, []logstorage.TenantID,
 	}
 
 	// Parse optional extra_filters
-	extraFiltersStr := r.FormValue("extra_filters")
-	extraFilters, err := parseExtraFilters(extraFiltersStr)
-	if err != nil {
-		return nil, nil, err
+	for _, extraFiltersStr := range r.Form["extra_filters"] {
+		extraFilters, err := parseExtraFilters(extraFiltersStr)
+		if err != nil {
+			return nil, nil, err
+		}
+		q.AddExtraFilters(extraFilters)
 	}
-	q.AddExtraFilters(extraFilters)
 
 	// Parse optional extra_stream_filters
-	extraStreamFiltersStr := r.FormValue("extra_stream_filters")
-	extraStreamFilters, err := parseExtraStreamFilters(extraStreamFiltersStr)
-	if err != nil {
-		return nil, nil, err
+	for _, extraStreamFiltersStr := range r.Form["extra_stream_filters"] {
+		extraStreamFilters, err := parseExtraStreamFilters(extraStreamFiltersStr)
+		if err != nil {
+			return nil, nil, err
+		}
+		q.AddExtraFilters(extraStreamFilters)
 	}
-	q.AddExtraFilters(extraStreamFilters)
 
 	return q, tenantIDs, nil
 }

app/vlselect/main.go

@@ -9,10 +9,11 @@ import (
 	"strings"
 	"time"
 
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlselect/internalselect"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlselect/logsql"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -71,7 +72,8 @@ var vmuiFileServer = http.FileServer(http.FS(vmuiFiles))
 // RequestHandler handles select requests for VictoriaLogs
 func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 	path := r.URL.Path
-	if !strings.HasPrefix(path, "/select/") {
+
+	if !strings.HasPrefix(path, "/select/") && !strings.HasPrefix(path, "/internal/select/") {
 		// Skip requests, which do not start with /select/, since these aren't our requests.
 		return false
 	}
@@ -98,25 +100,75 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		return true
 	}
 
+	ctx := r.Context()
+	if path == "/select/logsql/tail" {
+		logsqlTailRequests.Inc()
+		// Process live tailing request without timeout, since it is OK to run live tailing requests for very long time.
+		// Also do not apply concurrency limit to tail requests, since these limits are intended for non-tail requests.
+		logsql.ProcessLiveTailRequest(ctx, w, r)
+		return true
+	}
+
 	// Limit the number of concurrent queries, which can consume big amounts of CPU time.
 	startTime := time.Now()
-	ctx := r.Context()
 	d := getMaxQueryDuration(r)
 	ctxWithTimeout, cancel := context.WithTimeout(ctx, d)
 	defer cancel()
 
-	stopCh := ctxWithTimeout.Done()
+	if !incRequestConcurrency(ctxWithTimeout, w, r) {
+		return true
+	}
+	defer decRequestConcurrency()
+
+	if strings.HasPrefix(path, "/internal/select/") {
+		// Process internal request from vlselect without timeout (e.g. use ctx instead of ctxWithTimeout),
+		// since the timeout must be controlled by the vlselect.
+		internalselect.RequestHandler(ctx, w, r)
+		return true
+	}
+
+	ok := processSelectRequest(ctxWithTimeout, w, r, path)
+	if !ok {
+		return false
+	}
+
+	logRequestErrorIfNeeded(ctxWithTimeout, w, r, startTime)
+	return true
+}
+
+func logRequestErrorIfNeeded(ctx context.Context, w http.ResponseWriter, r *http.Request, startTime time.Time) {
+	err := ctx.Err()
+	switch err {
+	case nil:
+		// nothing to do
+	case context.Canceled:
+		// do not log canceled requests, since they are expected and legal.
+	case context.DeadlineExceeded:
+		err = &httpserver.ErrorWithStatusCode{
+			Err: fmt.Errorf("the request couldn't be executed in %.3f seconds; possible solutions: "+
+				"to increase -search.maxQueryDuration=%s; to pass bigger value to 'timeout' query arg", time.Since(startTime).Seconds(), maxQueryDuration),
+			StatusCode: http.StatusServiceUnavailable,
+		}
+		httpserver.Errorf(w, r, "%s", err)
+	default:
+		httpserver.Errorf(w, r, "unexpected error: %s", err)
+	}
+}
+
+func incRequestConcurrency(ctx context.Context, w http.ResponseWriter, r *http.Request) bool {
+	startTime := time.Now()
+	stopCh := ctx.Done()
 	select {
 	case concurrencyLimitCh <- struct{}{}:
-		defer func() { <-concurrencyLimitCh }()
+		return true
 	default:
 		// Sleep for a while until giving up. This should resolve short bursts in requests.
 		concurrencyLimitReached.Inc()
 		select {
 		case concurrencyLimitCh <- struct{}{}:
-			defer func() { <-concurrencyLimitCh }()
+			return true
 		case <-stopCh:
-			switch ctxWithTimeout.Err() {
+			switch ctx.Err() {
 			case context.Canceled:
 				remoteAddr := httpserver.GetQuotedRemoteAddr(r)
 				requestURI := httpserver.GetRequestURI(r)
@@ -129,49 +181,18 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 						"are executed. Possible solutions: to reduce query load; to add more compute resources to the server; "+
 						"to increase -search.maxQueueDuration=%s; to increase -search.maxQueryDuration=%s; to increase -search.maxConcurrentRequests; "+
 						"to pass bigger value to 'timeout' query arg",
-						d.Seconds(), *maxConcurrentRequests, maxQueueDuration, maxQueryDuration),
+						time.Since(startTime).Seconds(), *maxConcurrentRequests, maxQueueDuration, maxQueryDuration),
 					StatusCode: http.StatusServiceUnavailable,
 				}
 				httpserver.Errorf(w, r, "%s", err)
 			}
-			return true
+			return false
 		}
 	}
+}
 
-	if path == "/select/logsql/tail" {
-		logsqlTailRequests.Inc()
-		// Process live tailing request without timeout (e.g. use ctx instead of ctxWithTimeout),
-		// since it is OK to run live tailing requests for very long time.
-		logsql.ProcessLiveTailRequest(ctx, w, r)
-		return true
-	}
-
-	ok := processSelectRequest(ctxWithTimeout, w, r, path)
-	if !ok {
-		return false
-	}
-
-	err := ctxWithTimeout.Err()
-	switch err {
-	case nil:
-		// nothing to do
-	case context.Canceled:
-		remoteAddr := httpserver.GetQuotedRemoteAddr(r)
-		requestURI := httpserver.GetRequestURI(r)
-		logger.Infof("client has canceled the request after %.3f seconds: remoteAddr=%s, requestURI: %q",
-			time.Since(startTime).Seconds(), remoteAddr, requestURI)
-	case context.DeadlineExceeded:
-		err = &httpserver.ErrorWithStatusCode{
-			Err: fmt.Errorf("the request couldn't be executed in %.3f seconds; possible solutions: "+
-				"to increase -search.maxQueryDuration=%s; to pass bigger value to 'timeout' query arg", d.Seconds(), maxQueryDuration),
-			StatusCode: http.StatusServiceUnavailable,
-		}
-		httpserver.Errorf(w, r, "%s", err)
-	default:
-		httpserver.Errorf(w, r, "unexpected error: %s", err)
-	}
-
-	return true
+func decRequestConcurrency() {
+	<-concurrencyLimitCh
 }
 
 func processSelectRequest(ctx context.Context, w http.ResponseWriter, r *http.Request, path string) bool {
@@ -240,7 +261,7 @@ func processSelectRequest(ctx context.Context, w http.ResponseWriter, r *http.Re
 
 // getMaxQueryDuration returns the maximum duration for query from r.
 func getMaxQueryDuration(r *http.Request) time.Duration {
-	dms, err := httputils.GetDuration(r, "timeout", 0)
+	dms, err := httputil.GetDuration(r, "timeout", 0)
 	if err != nil {
 		dms = 0
 	}

app/vlselect/vmui/assets/MetricsQL-jNgFe_2G.md

@@ -5,9 +5,13 @@ menu:
   docs:
     parent: 'victoriametrics'
     weight: 23
+tags:
+  - metrics
 aliases:
 - /ExtendedPromQL.html
 - /MetricsQL.html
+- /metricsql/index.html
+- /metricsql/
 ---
 [VictoriaMetrics](https://github.com/VictoriaMetrics/VictoriaMetrics) implements MetricsQL -
 query language inspired by [PromQL](https://prometheus.io/docs/prometheus/latest/querying/basics/).
@@ -2349,4 +2353,4 @@ VictoriaMetrics performs the following implicit conversions for incoming queries
   is passed to [rollup function](#rollup-functions), then a [subquery](#subqueries) with `1i` lookbehind window and `1i` step is automatically formed.
   For example, `rate(sum(up))` is automatically converted to `rate((sum(default_rollup(up)))[1i:1i])`.
   This behavior can be disabled or logged via `-search.disableImplicitConversion` and `-search.logImplicitConversion` command-line flags
-  starting from [`v1.101.0` release](https://docs.victoriametrics.com/changelog/).
+  starting from [`v1.102.0-rc2` release](https://docs.victoriametrics.com/changelog/changelog_2024/#v11020-rc2).

app/vlselect/vmui/index.html

@@ -35,10 +35,10 @@
   <meta property="og:title" content="UI for VictoriaLogs">
   <meta property="og:url" content="https://victoriametrics.com/products/victorialogs/">
   <meta property="og:description" content="Explore your log data with VictoriaLogs UI">
-  <script type="module" crossorigin src="./assets/index-DuTUAk-m.js"></script>
-  <link rel="modulepreload" crossorigin href="./assets/vendor-DojlIpLz.js">
+  <script type="module" crossorigin src="./assets/index-Wi3UrT0H.js"></script>
+  <link rel="modulepreload" crossorigin href="./assets/vendor-C-vZmbyg.js">
   <link rel="stylesheet" crossorigin href="./assets/vendor-D1GxaB_c.css">
-  <link rel="stylesheet" crossorigin href="./assets/index-CEiptoJw.css">
+  <link rel="stylesheet" crossorigin href="./assets/index-Brup_hCI.css">
 </head>
 <body>
 <noscript>You need to enable JavaScript to run this app.</noscript>

app/vlstorage/main.go

@@ -10,11 +10,14 @@ import (
 
 	"github.com/VictoriaMetrics/metrics"
 
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage/netinsert"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage/netselect"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 )
 
 var (
@@ -38,15 +41,59 @@ var (
 	minFreeDiskSpaceBytes = flagutil.NewBytes("storage.minFreeDiskSpaceBytes", 10e6, "The minimum free disk space at -storageDataPath after which "+
 		"the storage stops accepting new data")
 
-	forceMergeAuthKey = flagutil.NewPassword("forceMergeAuthKey", "authKey, which must be passed in query string to /internal/force_merge pages. It overrides -httpAuth.*")
+	forceMergeAuthKey = flagutil.NewPassword("forceMergeAuthKey", "authKey, which must be passed in query string to /internal/force_merge . It overrides -httpAuth.* . "+
+		"See https://docs.victoriametrics.com/victorialogs/#forced-merge")
+	forceFlushAuthKey = flagutil.NewPassword("forceFlushAuthKey", "authKey, which must be passed in query string to /internal/force_flush . It overrides -httpAuth.* . "+
+		"See https://docs.victoriametrics.com/victorialogs/#forced-flush")
+
+	storageNodeAddrs = flagutil.NewArrayString("storageNode", "Comma-separated list of TCP addresses for storage nodes to route the ingested logs to and to send select queries to. "+
+		"If the list is empty, then the ingested logs are stored and queried locally from -storageDataPath")
+	insertConcurrency        = flag.Int("insert.concurrency", 2, "The average number of concurrent data ingestion requests, which can be sent to every -storageNode")
+	insertDisableCompression = flag.Bool("insert.disableCompression", false, "Whether to disable compression when sending the ingested data to -storageNode nodes. "+
+		"Disabled compression reduces CPU usage at the cost of higher network usage")
+	selectDisableCompression = flag.Bool("select.disableCompression", false, "Whether to disable compression for select query responses received from -storageNode nodes. "+
+		"Disabled compression reduces CPU usage at the cost of higher network usage")
+
+	storageNodeUsername     = flagutil.NewArrayString("storageNode.username", "Optional basic auth username to use for the corresponding -storageNode")
+	storageNodePassword     = flagutil.NewArrayString("storageNode.password", "Optional basic auth password to use for the corresponding -storageNode")
+	storageNodePasswordFile = flagutil.NewArrayString("storageNode.passwordFile", "Optional path to basic auth password to use for the corresponding -storageNode. "+
+		"The file is re-read every second")
+	storageNodeBearerToken     = flagutil.NewArrayString("storageNode.bearerToken", "Optional bearer auth token to use for the corresponding -storageNode")
+	storageNodeBearerTokenFile = flagutil.NewArrayString("storageNode.bearerTokenFile", "Optional path to bearer token file to use for the corresponding -storageNode. "+
+		"The token is re-read from the file every second")
+
+	storageNodeTLS = flagutil.NewArrayBool("storageNode.tls", "Whether to use TLS (HTTPS) protocol for communicating with the corresponding -storageNode. "+
+		"By default communication is performed via HTTP")
+	storageNodeTLSCAFile = flagutil.NewArrayString("storageNode.tlsCAFile", "Optional path to TLS CA file to use for verifying connections to the corresponding -storageNode. "+
+		"By default, system CA is used")
+	storageNodeTLSCertFile = flagutil.NewArrayString("storageNode.tlsCertFile", "Optional path to client-side TLS certificate file to use when connecting "+
+		"to the corresponding -storageNode")
+	storageNodeTLSKeyFile    = flagutil.NewArrayString("storageNode.tlsKeyFile", "Optional path to client-side TLS certificate key to use when connecting to the corresponding -storageNode")
+	storageNodeTLSServerName = flagutil.NewArrayString("storageNode.tlsServerName", "Optional TLS server name to use for connections to the corresponding -storageNode. "+
+		"By default, the server name from -storageNode is used")
+	storageNodeTLSInsecureSkipVerify = flagutil.NewArrayBool("storageNode.tlsInsecureSkipVerify", "Whether to skip tls verification when connecting to the corresponding -storageNode")
 )
 
+var localStorage *logstorage.Storage
+var localStorageMetrics *metrics.Set
+
+var netstorageInsert *netinsert.Storage
+var netstorageSelect *netselect.Storage
+
 // Init initializes vlstorage.
 //
 // Stop must be called when vlstorage is no longer needed
 func Init() {
-	if strg != nil {
-		logger.Panicf("BUG: Init() has been already called")
+	if len(*storageNodeAddrs) == 0 {
+		initLocalStorage()
+	} else {
+		initNetworkStorage()
+	}
+}
+
+func initLocalStorage() {
+	if localStorage != nil {
+		logger.Panicf("BUG: initLocalStorage() has been already called")
 	}
 
 	if retentionPeriod.Duration() < 24*time.Hour {
@@ -63,60 +110,157 @@ func Init() {
 	}
 	logger.Infof("opening storage at -storageDataPath=%s", *storageDataPath)
 	startTime := time.Now()
-	strg = logstorage.MustOpenStorage(*storageDataPath, cfg)
+	localStorage = logstorage.MustOpenStorage(*storageDataPath, cfg)
 
 	var ss logstorage.StorageStats
-	strg.UpdateStats(&ss)
+	localStorage.UpdateStats(&ss)
 	logger.Infof("successfully opened storage in %.3f seconds; smallParts: %d; bigParts: %d; smallPartBlocks: %d; bigPartBlocks: %d; smallPartRows: %d; bigPartRows: %d; "+
 		"smallPartSize: %d bytes; bigPartSize: %d bytes",
 		time.Since(startTime).Seconds(), ss.SmallParts, ss.BigParts, ss.SmallPartBlocks, ss.BigPartBlocks, ss.SmallPartRowsCount, ss.BigPartRowsCount,
 		ss.CompressedSmallPartSize, ss.CompressedBigPartSize)
 
-	// register storage metrics
-	storageMetrics = metrics.NewSet()
-	storageMetrics.RegisterMetricsWriter(func(w io.Writer) {
-		writeStorageMetrics(w, strg)
+	// register local storage metrics
+	localStorageMetrics = metrics.NewSet()
+	localStorageMetrics.RegisterMetricsWriter(func(w io.Writer) {
+		writeStorageMetrics(w, localStorage)
 	})
-	metrics.RegisterSet(storageMetrics)
+	metrics.RegisterSet(localStorageMetrics)
+}
+
+func initNetworkStorage() {
+	if netstorageInsert != nil || netstorageSelect != nil {
+		logger.Panicf("BUG: initNetworkStorage() has been already called")
+	}
+
+	authCfgs := make([]*promauth.Config, len(*storageNodeAddrs))
+	isTLSs := make([]bool, len(*storageNodeAddrs))
+	for i := range authCfgs {
+		authCfgs[i] = newAuthConfigForStorageNode(i)
+		isTLSs[i] = storageNodeTLS.GetOptionalArg(i)
+	}
+
+	logger.Infof("starting insert service for nodes %s", *storageNodeAddrs)
+	netstorageInsert = netinsert.NewStorage(*storageNodeAddrs, authCfgs, isTLSs, *insertConcurrency, *insertDisableCompression)
+
+	logger.Infof("initializing select service for nodes %s", *storageNodeAddrs)
+	netstorageSelect = netselect.NewStorage(*storageNodeAddrs, authCfgs, isTLSs, *selectDisableCompression)
+
+	logger.Infof("initialized all the network services")
+}
+
+func newAuthConfigForStorageNode(argIdx int) *promauth.Config {
+	username := storageNodeUsername.GetOptionalArg(argIdx)
+	password := storageNodePassword.GetOptionalArg(argIdx)
+	passwordFile := storageNodePasswordFile.GetOptionalArg(argIdx)
+	var basicAuthCfg *promauth.BasicAuthConfig
+	if username != "" || password != "" || passwordFile != "" {
+		basicAuthCfg = &promauth.BasicAuthConfig{
+			Username:     username,
+			Password:     promauth.NewSecret(password),
+			PasswordFile: passwordFile,
+		}
+	}
+
+	token := storageNodeBearerToken.GetOptionalArg(argIdx)
+	tokenFile := storageNodeBearerTokenFile.GetOptionalArg(argIdx)
+
+	tlsCfg := &promauth.TLSConfig{
+		CAFile:             storageNodeTLSCAFile.GetOptionalArg(argIdx),
+		CertFile:           storageNodeTLSCertFile.GetOptionalArg(argIdx),
+		KeyFile:            storageNodeTLSKeyFile.GetOptionalArg(argIdx),
+		ServerName:         storageNodeTLSServerName.GetOptionalArg(argIdx),
+		InsecureSkipVerify: storageNodeTLSInsecureSkipVerify.GetOptionalArg(argIdx),
+	}
+
+	opts := &promauth.Options{
+		BasicAuth:       basicAuthCfg,
+		BearerToken:     token,
+		BearerTokenFile: tokenFile,
+		TLSConfig:       tlsCfg,
+	}
+	ac, err := opts.NewConfig()
+	if err != nil {
+		logger.Panicf("FATAL: cannot populate auth config for storage node #%d: %s", argIdx, err)
+	}
+
+	return ac
 }
 
 // Stop stops vlstorage.
 func Stop() {
-	metrics.UnregisterSet(storageMetrics, true)
-	storageMetrics = nil
+	if localStorage != nil {
+		metrics.UnregisterSet(localStorageMetrics, true)
+		localStorageMetrics = nil
 
-	strg.MustClose()
-	strg = nil
+		localStorage.MustClose()
+		localStorage = nil
+	} else {
+		netstorageInsert.MustStop()
+		netstorageInsert = nil
+
+		netstorageSelect.MustStop()
+		netstorageSelect = nil
+	}
 }
 
 // RequestHandler is a storage request handler.
 func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 	path := r.URL.Path
-	if path == "/internal/force_merge" {
-		if !httpserver.CheckAuthFlag(w, r, forceMergeAuthKey) {
-			return true
-		}
-		// Run force merge in background
-		partitionNamePrefix := r.FormValue("partition_prefix")
-		go func() {
-			activeForceMerges.Inc()
-			defer activeForceMerges.Dec()
-			logger.Infof("forced merge for partition_prefix=%q has been started", partitionNamePrefix)
-			startTime := time.Now()
-			strg.MustForceMerge(partitionNamePrefix)
-			logger.Infof("forced merge for partition_prefix=%q has been successfully finished in %.3f seconds", partitionNamePrefix, time.Since(startTime).Seconds())
-		}()
-		return true
+	switch path {
+	case "/internal/force_merge":
+		return processForceMerge(w, r)
+	case "/internal/force_flush":
+		return processForceFlush(w, r)
 	}
 	return false
 }
 
-var strg *logstorage.Storage
-var storageMetrics *metrics.Set
+func processForceMerge(w http.ResponseWriter, r *http.Request) bool {
+	if localStorage == nil {
+		// Force merge isn't supported by non-local storage
+		return false
+	}
 
-// CanWriteData returns non-nil error if it cannot write data to vlstorage.
+	if !httpserver.CheckAuthFlag(w, r, forceMergeAuthKey) {
+		return true
+	}
+
+	// Run force merge in background
+	partitionNamePrefix := r.FormValue("partition_prefix")
+	go func() {
+		activeForceMerges.Inc()
+		defer activeForceMerges.Dec()
+		logger.Infof("forced merge for partition_prefix=%q has been started", partitionNamePrefix)
+		startTime := time.Now()
+		localStorage.MustForceMerge(partitionNamePrefix)
+		logger.Infof("forced merge for partition_prefix=%q has been successfully finished in %.3f seconds", partitionNamePrefix, time.Since(startTime).Seconds())
+	}()
+	return true
+}
+
+func processForceFlush(w http.ResponseWriter, r *http.Request) bool {
+	if localStorage == nil {
+		// Force merge isn't supported by non-local storage
+		return false
+	}
+
+	if !httpserver.CheckAuthFlag(w, r, forceFlushAuthKey) {
+		return true
+	}
+
+	logger.Infof("flushing storage to make pending data available for reading")
+	localStorage.DebugFlush()
+	return true
+}
+
+// CanWriteData returns non-nil error if it cannot write data to vlstorage
 func CanWriteData() error {
-	if strg.IsReadOnly() {
+	if localStorage == nil {
+		// The data can be always written in non-local mode.
+		return nil
+	}
+
+	if localStorage.IsReadOnly() {
 		return &httpserver.ErrorWithStatusCode{
 			Err: fmt.Errorf("cannot add rows into storage in read-only mode; the storage can be in read-only mode "+
 				"because of lack of free disk space at -storageDataPath=%s", *storageDataPath),
@@ -130,50 +274,77 @@ func CanWriteData() error {
 //
 // It is advised to call CanWriteData() before calling MustAddRows()
 func MustAddRows(lr *logstorage.LogRows) {
-	strg.MustAddRows(lr)
+	if localStorage != nil {
+		// Store lr in the local storage.
+		localStorage.MustAddRows(lr)
+	} else {
+		// Store lr across the remote storage nodes.
+		lr.ForEachRow(netstorageInsert.AddRow)
+	}
 }
 
 // RunQuery runs the given q and calls writeBlock for the returned data blocks
-func RunQuery(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, writeBlock logstorage.WriteBlockFunc) error {
-	return strg.RunQuery(ctx, tenantIDs, q, writeBlock)
+func RunQuery(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, writeBlock logstorage.WriteDataBlockFunc) error {
+	if localStorage != nil {
+		return localStorage.RunQuery(ctx, tenantIDs, q, writeBlock)
+	}
+	return netstorageSelect.RunQuery(ctx, tenantIDs, q, writeBlock)
 }
 
 // GetFieldNames executes q and returns field names seen in results.
 func GetFieldNames(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query) ([]logstorage.ValueWithHits, error) {
-	return strg.GetFieldNames(ctx, tenantIDs, q)
+	if localStorage != nil {
+		return localStorage.GetFieldNames(ctx, tenantIDs, q)
+	}
+	return netstorageSelect.GetFieldNames(ctx, tenantIDs, q)
 }
 
 // GetFieldValues executes q and returns unique values for the fieldName seen in results.
 //
 // If limit > 0, then up to limit unique values are returned.
 func GetFieldValues(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, fieldName string, limit uint64) ([]logstorage.ValueWithHits, error) {
-	return strg.GetFieldValues(ctx, tenantIDs, q, fieldName, limit)
+	if localStorage != nil {
+		return localStorage.GetFieldValues(ctx, tenantIDs, q, fieldName, limit)
+	}
+	return netstorageSelect.GetFieldValues(ctx, tenantIDs, q, fieldName, limit)
 }
 
 // GetStreamFieldNames executes q and returns stream field names seen in results.
 func GetStreamFieldNames(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query) ([]logstorage.ValueWithHits, error) {
-	return strg.GetStreamFieldNames(ctx, tenantIDs, q)
+	if localStorage != nil {
+		return localStorage.GetStreamFieldNames(ctx, tenantIDs, q)
+	}
+	return netstorageSelect.GetStreamFieldNames(ctx, tenantIDs, q)
 }
 
 // GetStreamFieldValues executes q and returns stream field values for the given fieldName seen in results.
 //
 // If limit > 0, then up to limit unique stream field values are returned.
 func GetStreamFieldValues(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, fieldName string, limit uint64) ([]logstorage.ValueWithHits, error) {
-	return strg.GetStreamFieldValues(ctx, tenantIDs, q, fieldName, limit)
+	if localStorage != nil {
+		return localStorage.GetStreamFieldValues(ctx, tenantIDs, q, fieldName, limit)
+	}
+	return netstorageSelect.GetStreamFieldValues(ctx, tenantIDs, q, fieldName, limit)
 }
 
 // GetStreams executes q and returns streams seen in query results.
 //
 // If limit > 0, then up to limit unique streams are returned.
 func GetStreams(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, limit uint64) ([]logstorage.ValueWithHits, error) {
-	return strg.GetStreams(ctx, tenantIDs, q, limit)
+	if localStorage != nil {
+		return localStorage.GetStreams(ctx, tenantIDs, q, limit)
+	}
+	return netstorageSelect.GetStreams(ctx, tenantIDs, q, limit)
 }
 
 // GetStreamIDs executes q and returns streamIDs seen in query results.
 //
 // If limit > 0, then up to limit unique streamIDs are returned.
 func GetStreamIDs(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, limit uint64) ([]logstorage.ValueWithHits, error) {
-	return strg.GetStreamIDs(ctx, tenantIDs, q, limit)
+	if localStorage != nil {
+		return localStorage.GetStreamIDs(ctx, tenantIDs, q, limit)
+	}
+	return netstorageSelect.GetStreamIDs(ctx, tenantIDs, q, limit)
 }
 
 func writeStorageMetrics(w io.Writer, strg *logstorage.Storage) {

app/vlstorage/netinsert/netinsert.go

@@ -0,0 +1,369 @@
+package netinsert
+
+import (
+	"errors"
+	"fmt"
+	"io"
+	"net/http"
+	"net/url"
+	"sync"
+	"sync/atomic"
+	"time"
+
+	"github.com/valyala/fastrand"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/contextutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding/zstd"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timerpool"
+)
+
+// the maximum size of a single data block sent to storage node.
+const maxInsertBlockSize = 2 * 1024 * 1024
+
+// ProtocolVersion is the version of the data ingestion protocol.
+//
+// It must be changed every time the data encoding at /internal/insert HTTP endpoint is changed.
+const ProtocolVersion = "v1"
+
+// Storage is a network storage for sending data to remote storage nodes in the cluster.
+type Storage struct {
+	sns []*storageNode
+
+	disableCompression bool
+
+	srt *streamRowsTracker
+
+	pendingDataBuffers chan *bytesutil.ByteBuffer
+
+	stopCh chan struct{}
+	wg     sync.WaitGroup
+}
+
+type storageNode struct {
+	// scheme is http or https scheme to communicate with addr
+	scheme string
+
+	// addr is TCP address of storage node to send the ingested data to
+	addr string
+
+	// s is a storage, which holds the given storageNode
+	s *Storage
+
+	// c is an http client used for sending data blocks to addr.
+	c *http.Client
+
+	// ac is auth config used for setting request headers such as Authorization and Host.
+	ac *promauth.Config
+
+	// pendingData contains pending data, which must be sent to the storage node at the addr.
+	pendingDataMu        sync.Mutex
+	pendingData          *bytesutil.ByteBuffer
+	pendingDataLastFlush time.Time
+
+	// the unix timestamp until the storageNode is disabled for data writing.
+	disabledUntil atomic.Uint64
+}
+
+func newStorageNode(s *Storage, addr string, ac *promauth.Config, isTLS bool) *storageNode {
+	tr := httputil.NewTransport(false, "vlinsert_backend")
+	tr.TLSHandshakeTimeout = 20 * time.Second
+	tr.DisableCompression = true
+
+	scheme := "http"
+	if isTLS {
+		scheme = "https"
+	}
+
+	sn := &storageNode{
+		scheme: scheme,
+		addr:   addr,
+		s:      s,
+		c: &http.Client{
+			Transport: ac.NewRoundTripper(tr),
+		},
+		ac: ac,
+
+		pendingData: &bytesutil.ByteBuffer{},
+	}
+
+	s.wg.Add(1)
+	go func() {
+		defer s.wg.Done()
+		sn.backgroundFlusher()
+	}()
+
+	return sn
+}
+
+func (sn *storageNode) backgroundFlusher() {
+	t := time.NewTicker(time.Second)
+	defer t.Stop()
+
+	for {
+		select {
+		case <-sn.s.stopCh:
+			return
+		case <-t.C:
+			sn.flushPendingData()
+		}
+	}
+}
+
+func (sn *storageNode) flushPendingData() {
+	sn.pendingDataMu.Lock()
+	if time.Since(sn.pendingDataLastFlush) < time.Second {
+		// nothing to flush
+		sn.pendingDataMu.Unlock()
+		return
+	}
+
+	pendingData := sn.grabPendingDataForFlushLocked()
+	sn.pendingDataMu.Unlock()
+
+	sn.mustSendInsertRequest(pendingData)
+}
+
+func (sn *storageNode) addRow(r *logstorage.InsertRow) {
+	bb := bbPool.Get()
+	b := bb.B
+
+	b = r.Marshal(b)
+
+	if len(b) > maxInsertBlockSize {
+		logger.Warnf("skipping too long log entry, since its length exceeds %d bytes; the actual log entry length is %d bytes; log entry contents: %s", maxInsertBlockSize, len(b), b)
+		return
+	}
+
+	var pendingData *bytesutil.ByteBuffer
+	sn.pendingDataMu.Lock()
+	if sn.pendingData.Len()+len(b) > maxInsertBlockSize {
+		pendingData = sn.grabPendingDataForFlushLocked()
+	}
+	sn.pendingData.MustWrite(b)
+	sn.pendingDataMu.Unlock()
+
+	bb.B = b
+	bbPool.Put(bb)
+
+	if pendingData != nil {
+		sn.mustSendInsertRequest(pendingData)
+	}
+}
+
+var bbPool bytesutil.ByteBufferPool
+
+func (sn *storageNode) grabPendingDataForFlushLocked() *bytesutil.ByteBuffer {
+	sn.pendingDataLastFlush = time.Now()
+	pendingData := sn.pendingData
+	sn.pendingData = <-sn.s.pendingDataBuffers
+
+	return pendingData
+}
+
+func (sn *storageNode) mustSendInsertRequest(pendingData *bytesutil.ByteBuffer) {
+	defer func() {
+		pendingData.Reset()
+		sn.s.pendingDataBuffers <- pendingData
+	}()
+
+	err := sn.sendInsertRequest(pendingData)
+	if err == nil {
+		return
+	}
+
+	if !errors.Is(err, errTemporarilyDisabled) {
+		logger.Warnf("%s; re-routing the data block to the remaining nodes", err)
+	}
+	for !sn.s.sendInsertRequestToAnyNode(pendingData) {
+		logger.Errorf("cannot send pending data to all storage nodes, since all of them are unavailable; re-trying to send the data in a second")
+
+		t := timerpool.Get(time.Second)
+		select {
+		case <-sn.s.stopCh:
+			timerpool.Put(t)
+			logger.Errorf("dropping %d bytes of data, since there are no available storage nodes", pendingData.Len())
+			return
+		case <-t.C:
+			timerpool.Put(t)
+		}
+	}
+}
+
+func (sn *storageNode) sendInsertRequest(pendingData *bytesutil.ByteBuffer) error {
+	dataLen := pendingData.Len()
+	if dataLen == 0 {
+		// Nothing to send.
+		return nil
+	}
+
+	if sn.disabledUntil.Load() > fasttime.UnixTimestamp() {
+		return errTemporarilyDisabled
+	}
+
+	ctx, cancel := contextutil.NewStopChanContext(sn.s.stopCh)
+	defer cancel()
+
+	var body io.Reader
+	if !sn.s.disableCompression {
+		bb := zstdBufPool.Get()
+		defer zstdBufPool.Put(bb)
+
+		bb.B = zstd.CompressLevel(bb.B[:0], pendingData.B, 1)
+		body = bb.NewReader()
+	} else {
+		body = pendingData.NewReader()
+	}
+
+	reqURL := sn.getRequestURL("/internal/insert")
+	req, err := http.NewRequestWithContext(ctx, "POST", reqURL, body)
+	if err != nil {
+		logger.Panicf("BUG: unexpected error when creating an http request: %s", err)
+	}
+	req.Header.Set("Content-Type", "application/octet-stream")
+	if !sn.s.disableCompression {
+		req.Header.Set("Content-Encoding", "zstd")
+	}
+	if err := sn.ac.SetHeaders(req, true); err != nil {
+		return fmt.Errorf("cannot set auth headers for %q: %w", reqURL, err)
+	}
+
+	resp, err := sn.c.Do(req)
+	if err != nil {
+		// Disable sn for data writing for 10 seconds.
+		sn.disabledUntil.Store(fasttime.UnixTimestamp() + 10)
+
+		return fmt.Errorf("cannot send data block with the length %d to %q: %s", pendingData.Len(), reqURL, err)
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode/100 == 2 {
+		return nil
+	}
+
+	respBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		respBody = []byte(fmt.Sprintf("%s", err))
+	}
+
+	// Disable sn for data writing for 10 seconds.
+	sn.disabledUntil.Store(fasttime.UnixTimestamp() + 10)
+
+	return fmt.Errorf("unexpected status code returned when sending data block to %q: %d; want 2xx; response body: %q", reqURL, resp.StatusCode, respBody)
+}
+
+func (sn *storageNode) getRequestURL(path string) string {
+	return fmt.Sprintf("%s://%s%s?version=%s", sn.scheme, sn.addr, path, url.QueryEscape(ProtocolVersion))
+}
+
+var zstdBufPool bytesutil.ByteBufferPool
+
+// NewStorage returns new Storage for the given addrs with the given authCfgs.
+//
+// The concurrency is the average number of concurrent connections per every addr.
+//
+// If disableCompression is set, then the data is sent uncompressed to the remote storage.
+//
+// Call MustStop on the returned storage when it is no longer needed.
+func NewStorage(addrs []string, authCfgs []*promauth.Config, isTLSs []bool, concurrency int, disableCompression bool) *Storage {
+	pendingDataBuffers := make(chan *bytesutil.ByteBuffer, concurrency*len(addrs))
+	for i := 0; i < cap(pendingDataBuffers); i++ {
+		pendingDataBuffers <- &bytesutil.ByteBuffer{}
+	}
+
+	s := &Storage{
+		disableCompression: disableCompression,
+		pendingDataBuffers: pendingDataBuffers,
+		stopCh:             make(chan struct{}),
+	}
+
+	sns := make([]*storageNode, len(addrs))
+	for i, addr := range addrs {
+		sns[i] = newStorageNode(s, addr, authCfgs[i], isTLSs[i])
+	}
+	s.sns = sns
+
+	s.srt = newStreamRowsTracker(len(sns))
+
+	return s
+}
+
+// MustStop stops the s.
+func (s *Storage) MustStop() {
+	close(s.stopCh)
+	s.wg.Wait()
+	s.sns = nil
+}
+
+// AddRow adds the given log row into s.
+func (s *Storage) AddRow(streamHash uint64, r *logstorage.InsertRow) {
+	idx := s.srt.getNodeIdx(streamHash)
+	sn := s.sns[idx]
+	sn.addRow(r)
+}
+
+func (s *Storage) sendInsertRequestToAnyNode(pendingData *bytesutil.ByteBuffer) bool {
+	startIdx := int(fastrand.Uint32n(uint32(len(s.sns))))
+	for i := range s.sns {
+		idx := (startIdx + i) % len(s.sns)
+		sn := s.sns[idx]
+		err := sn.sendInsertRequest(pendingData)
+		if err == nil {
+			return true
+		}
+		if !errors.Is(err, errTemporarilyDisabled) {
+			logger.Warnf("cannot send pending data to the storage node %q: %s; trying to send it to another storage node", sn.addr, err)
+		}
+	}
+	return false
+}
+
+var errTemporarilyDisabled = fmt.Errorf("writing to the node is temporarily disabled")
+
+type streamRowsTracker struct {
+	mu sync.Mutex
+
+	nodesCount    int64
+	rowsPerStream map[uint64]uint64
+}
+
+func newStreamRowsTracker(nodesCount int) *streamRowsTracker {
+	return &streamRowsTracker{
+		nodesCount:    int64(nodesCount),
+		rowsPerStream: make(map[uint64]uint64),
+	}
+}
+
+func (srt *streamRowsTracker) getNodeIdx(streamHash uint64) uint64 {
+	if srt.nodesCount == 1 {
+		// Fast path for a single node.
+		return 0
+	}
+
+	srt.mu.Lock()
+	defer srt.mu.Unlock()
+
+	streamRows := srt.rowsPerStream[streamHash] + 1
+	srt.rowsPerStream[streamHash] = streamRows
+
+	if streamRows <= 1000 {
+		// Write the initial rows for the stream to a single storage node for better locality.
+		// This should work great for log streams containing small number of logs, since will be distributed
+		// evenly among available storage nodes because they have different streamHash.
+		return streamHash % uint64(srt.nodesCount)
+	}
+
+	// The log stream contains more than 1000 rows. Distribute them among storage nodes at random
+	// in order to improve query performance over this stream (the data for the log stream
+	// can be processed in parallel on all the storage nodes).
+	//
+	// The random distribution is preferred over round-robin distribution in order to avoid possible
+	// dependency between the order of the ingested logs and the number of storage nodes,
+	// which may lead to non-uniform distribution of logs among storage nodes.
+	return uint64(fastrand.Uint32n(uint32(srt.nodesCount)))
+}

app/vlstorage/netinsert/netinsert_test.go

@@ -0,0 +1,57 @@
+package netinsert
+
+import (
+	"fmt"
+	"math"
+	"math/rand"
+	"testing"
+
+	"github.com/cespare/xxhash/v2"
+)
+
+func TestStreamRowsTracker(t *testing.T) {
+	f := func(rowsCount, streamsCount, nodesCount int) {
+		t.Helper()
+
+		// generate stream hashes
+		streamHashes := make([]uint64, streamsCount)
+		for i := range streamHashes {
+			streamHashes[i] = xxhash.Sum64([]byte(fmt.Sprintf("stream %d.", i)))
+		}
+
+		srt := newStreamRowsTracker(nodesCount)
+
+		rng := rand.New(rand.NewSource(0))
+		rowsPerNode := make([]uint64, nodesCount)
+		for i := 0; i < rowsCount; i++ {
+			streamIdx := rng.Intn(streamsCount)
+			h := streamHashes[streamIdx]
+			nodeIdx := srt.getNodeIdx(h)
+			rowsPerNode[nodeIdx]++
+		}
+
+		// Verify that rows are uniformly distributed among nodes.
+		expectedRowsPerNode := float64(rowsCount) / float64(nodesCount)
+		for nodeIdx, nodeRows := range rowsPerNode {
+			if math.Abs(float64(nodeRows)-expectedRowsPerNode)/expectedRowsPerNode > 0.15 {
+				t.Fatalf("non-uniform distribution of rows among nodes; node %d has %d rows, while it must have %v rows; rowsPerNode=%d",
+					nodeIdx, nodeRows, expectedRowsPerNode, rowsPerNode)
+			}
+		}
+	}
+
+	rowsCount := 10000
+	streamsCount := 9
+	nodesCount := 2
+	f(rowsCount, streamsCount, nodesCount)
+
+	rowsCount = 10000
+	streamsCount = 100
+	nodesCount = 2
+	f(rowsCount, streamsCount, nodesCount)
+
+	rowsCount = 100000
+	streamsCount = 1000
+	nodesCount = 9
+	f(rowsCount, streamsCount, nodesCount)
+}

app/vlstorage/netselect/netselect.go

@@ -0,0 +1,469 @@
+package netselect
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"math"
+	"net/http"
+	"net/url"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/contextutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding/zstd"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/slicesutil"
+)
+
+const (
+	// FieldNamesProtocolVersion is the version of the protocol used for /internal/select/field_names HTTP endpoint.
+	//
+	// It must be updated every time the protocol changes.
+	FieldNamesProtocolVersion = "v1"
+
+	// FieldValuesProtocolVersion is the version of the protocol used for /internal/select/field_values HTTP endpoint.
+	//
+	// It must be updated every time the protocol changes.
+	FieldValuesProtocolVersion = "v1"
+
+	// StreamFieldNamesProtocolVersion is the version of the protocol used for /internal/select/stream_field_names HTTP endpoint.
+	//
+	// It must be updated every time the protocol changes.
+	StreamFieldNamesProtocolVersion = "v1"
+
+	// StreamFieldValuesProtocolVersion is the version of the protocol used for /internal/select/stream_field_values HTTP endpoint.
+	//
+	// It must be updated every time the protocol changes.
+	StreamFieldValuesProtocolVersion = "v1"
+
+	// StreamsProtocolVersion is the version of the protocol used for /internal/select/streams HTTP endpoint.
+	//
+	// It must be updated every time the protocol changes.
+	StreamsProtocolVersion = "v1"
+
+	// StreamIDsProtocolVersion is the version of the protocol used for /internal/select/stream_ids HTTP endpoint.
+	//
+	// It must be updated every time the protocol changes.
+	StreamIDsProtocolVersion = "v1"
+
+	// QueryProtocolVersion is the version of the protocol used for /internal/select/query HTTP endpoint.
+	//
+	// It must be updated every time the protocol changes.
+	QueryProtocolVersion = "v1"
+)
+
+// Storage is a network storage for querying remote storage nodes in the cluster.
+type Storage struct {
+	sns []*storageNode
+
+	disableCompression bool
+}
+
+type storageNode struct {
+	// scheme is http or https scheme to communicate with addr
+	scheme string
+
+	// addr is TCP address of the storage node to query
+	addr string
+
+	// s is a storage, which holds the given storageNode
+	s *Storage
+
+	// c is an http client used for querying storage node at addr.
+	c *http.Client
+
+	// ac is auth config used for setting request headers such as Authorization and Host.
+	ac *promauth.Config
+}
+
+func newStorageNode(s *Storage, addr string, ac *promauth.Config, isTLS bool) *storageNode {
+	tr := httputil.NewTransport(false, "vlselect_backend")
+	tr.TLSHandshakeTimeout = 20 * time.Second
+	tr.DisableCompression = true
+
+	scheme := "http"
+	if isTLS {
+		scheme = "https"
+	}
+
+	sn := &storageNode{
+		scheme: scheme,
+		addr:   addr,
+		s:      s,
+		c: &http.Client{
+			Transport: ac.NewRoundTripper(tr),
+		},
+		ac: ac,
+	}
+	return sn
+}
+
+func (sn *storageNode) runQuery(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, processBlock func(db *logstorage.DataBlock)) error {
+	args := sn.getCommonArgs(QueryProtocolVersion, tenantIDs, q)
+
+	reqURL := sn.getRequestURL("/internal/select/query", args)
+	req, err := http.NewRequestWithContext(ctx, "GET", reqURL, nil)
+	if err != nil {
+		logger.Panicf("BUG: unexpected error when creating a request: %s", err)
+	}
+	if err := sn.ac.SetHeaders(req, true); err != nil {
+		return fmt.Errorf("cannot set auth headers for %q: %w", reqURL, err)
+	}
+
+	// send the request to the storage node
+	resp, err := sn.c.Do(req)
+	if err != nil {
+		return err
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode != http.StatusOK {
+		responseBody, err := io.ReadAll(resp.Body)
+		if err != nil {
+			responseBody = []byte(err.Error())
+		}
+		return fmt.Errorf("unexpected status code for the request to %q: %d; want %d; response: %q", reqURL, resp.StatusCode, http.StatusOK, responseBody)
+	}
+
+	// read the response
+	var dataLenBuf [8]byte
+	var buf []byte
+	var db logstorage.DataBlock
+	var valuesBuf []string
+	for {
+		if _, err := io.ReadFull(resp.Body, dataLenBuf[:]); err != nil {
+			if errors.Is(err, io.EOF) {
+				// The end of response stream
+				return nil
+			}
+			return fmt.Errorf("cannot read block size from %q: %w", reqURL, err)
+		}
+		blockLen := encoding.UnmarshalUint64(dataLenBuf[:])
+		if blockLen > math.MaxInt {
+			return fmt.Errorf("too big data block: %d bytes; mustn't exceed %v bytes", blockLen, math.MaxInt)
+		}
+
+		buf = slicesutil.SetLength(buf, int(blockLen))
+		if _, err := io.ReadFull(resp.Body, buf); err != nil {
+			return fmt.Errorf("cannot read block with size of %d bytes from %q: %w", blockLen, reqURL, err)
+		}
+
+		src := buf
+		if !sn.s.disableCompression {
+			bufLen := len(buf)
+			var err error
+			buf, err = zstd.Decompress(buf, buf)
+			if err != nil {
+				return fmt.Errorf("cannot decompress data block: %w", err)
+			}
+			src = buf[bufLen:]
+		}
+
+		for len(src) > 0 {
+			tail, vb, err := db.UnmarshalInplace(src, valuesBuf[:0])
+			if err != nil {
+				return fmt.Errorf("cannot unmarshal data block received from %q: %w", reqURL, err)
+			}
+			valuesBuf = vb
+			src = tail
+
+			processBlock(&db)
+
+			clear(valuesBuf)
+		}
+	}
+}
+
+func (sn *storageNode) getFieldNames(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query) ([]logstorage.ValueWithHits, error) {
+	args := sn.getCommonArgs(FieldNamesProtocolVersion, tenantIDs, q)
+
+	return sn.getValuesWithHits(ctx, "/internal/select/field_names", args)
+}
+
+func (sn *storageNode) getFieldValues(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, fieldName string, limit uint64) ([]logstorage.ValueWithHits, error) {
+	args := sn.getCommonArgs(FieldValuesProtocolVersion, tenantIDs, q)
+	args.Set("field", fieldName)
+	args.Set("limit", fmt.Sprintf("%d", limit))
+
+	return sn.getValuesWithHits(ctx, "/internal/select/field_values", args)
+}
+
+func (sn *storageNode) getStreamFieldNames(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query) ([]logstorage.ValueWithHits, error) {
+	args := sn.getCommonArgs(StreamFieldNamesProtocolVersion, tenantIDs, q)
+
+	return sn.getValuesWithHits(ctx, "/internal/select/stream_field_names", args)
+}
+
+func (sn *storageNode) getStreamFieldValues(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, fieldName string, limit uint64) ([]logstorage.ValueWithHits, error) {
+	args := sn.getCommonArgs(StreamFieldValuesProtocolVersion, tenantIDs, q)
+	args.Set("field", fieldName)
+	args.Set("limit", fmt.Sprintf("%d", limit))
+
+	return sn.getValuesWithHits(ctx, "/internal/select/stream_field_values", args)
+}
+
+func (sn *storageNode) getStreams(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, limit uint64) ([]logstorage.ValueWithHits, error) {
+	args := sn.getCommonArgs(StreamsProtocolVersion, tenantIDs, q)
+	args.Set("limit", fmt.Sprintf("%d", limit))
+
+	return sn.getValuesWithHits(ctx, "/internal/select/streams", args)
+}
+
+func (sn *storageNode) getStreamIDs(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, limit uint64) ([]logstorage.ValueWithHits, error) {
+	args := sn.getCommonArgs(StreamIDsProtocolVersion, tenantIDs, q)
+	args.Set("limit", fmt.Sprintf("%d", limit))
+
+	return sn.getValuesWithHits(ctx, "/internal/select/stream_ids", args)
+}
+
+func (sn *storageNode) getCommonArgs(version string, tenantIDs []logstorage.TenantID, q *logstorage.Query) url.Values {
+	args := url.Values{}
+	args.Set("version", version)
+	args.Set("tenant_ids", string(logstorage.MarshalTenantIDs(nil, tenantIDs)))
+	args.Set("query", q.String())
+	args.Set("timestamp", fmt.Sprintf("%d", q.GetTimestamp()))
+	args.Set("disable_compression", fmt.Sprintf("%v", sn.s.disableCompression))
+	return args
+}
+
+func (sn *storageNode) getValuesWithHits(ctx context.Context, path string, args url.Values) ([]logstorage.ValueWithHits, error) {
+	data, err := sn.executeRequestAt(ctx, path, args)
+	if err != nil {
+		return nil, err
+	}
+	return unmarshalValuesWithHits(data)
+}
+
+func (sn *storageNode) executeRequestAt(ctx context.Context, path string, args url.Values) ([]byte, error) {
+	reqURL := sn.getRequestURL(path, args)
+	req, err := http.NewRequestWithContext(ctx, "GET", reqURL, nil)
+	if err != nil {
+		logger.Panicf("BUG: unexpected error when creating a request: %s", err)
+	}
+
+	// send the request to the storage node
+	resp, err := sn.c.Do(req)
+	if err != nil {
+		return nil, err
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode != http.StatusOK {
+		responseBody, err := io.ReadAll(resp.Body)
+		if err != nil {
+			responseBody = []byte(err.Error())
+		}
+		return nil, fmt.Errorf("unexpected status code for the request to %q: %d; want %d; response: %q", reqURL, resp.StatusCode, http.StatusOK, responseBody)
+	}
+
+	// read the response
+	var bb bytesutil.ByteBuffer
+	if _, err := bb.ReadFrom(resp.Body); err != nil {
+		return nil, fmt.Errorf("cannot read response from %q: %w", reqURL, err)
+	}
+
+	if sn.s.disableCompression {
+		return bb.B, nil
+	}
+
+	bbLen := len(bb.B)
+	bb.B, err = zstd.Decompress(bb.B, bb.B)
+	if err != nil {
+		return nil, err
+	}
+	return bb.B[bbLen:], nil
+}
+
+func (sn *storageNode) getRequestURL(path string, args url.Values) string {
+	return fmt.Sprintf("%s://%s%s?%s", sn.scheme, sn.addr, path, args.Encode())
+}
+
+// NewStorage returns new Storage for the given addrs and the given authCfgs.
+//
+// If disableCompression is set, then uncompressed responses are received from storage nodes.
+//
+// Call MustStop on the returned storage when it is no longer needed.
+func NewStorage(addrs []string, authCfgs []*promauth.Config, isTLSs []bool, disableCompression bool) *Storage {
+	s := &Storage{
+		disableCompression: disableCompression,
+	}
+
+	sns := make([]*storageNode, len(addrs))
+	for i, addr := range addrs {
+		sns[i] = newStorageNode(s, addr, authCfgs[i], isTLSs[i])
+	}
+	s.sns = sns
+
+	return s
+}
+
+// MustStop stops the s.
+func (s *Storage) MustStop() {
+	s.sns = nil
+}
+
+// RunQuery runs the given q and calls writeBlock for the returned data blocks
+func (s *Storage) RunQuery(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, writeBlock logstorage.WriteDataBlockFunc) error {
+	nqr, err := logstorage.NewNetQueryRunner(ctx, tenantIDs, q, s.RunQuery, writeBlock)
+	if err != nil {
+		return err
+	}
+
+	search := func(stopCh <-chan struct{}, q *logstorage.Query, writeBlock logstorage.WriteDataBlockFunc) error {
+		return s.runQuery(stopCh, tenantIDs, q, writeBlock)
+	}
+
+	concurrency := q.GetConcurrency()
+	return nqr.Run(ctx, concurrency, search)
+}
+
+func (s *Storage) runQuery(stopCh <-chan struct{}, tenantIDs []logstorage.TenantID, q *logstorage.Query, writeBlock logstorage.WriteDataBlockFunc) error {
+	ctxWithCancel, cancel := contextutil.NewStopChanContext(stopCh)
+	defer cancel()
+
+	errs := make([]error, len(s.sns))
+
+	var wg sync.WaitGroup
+	for i := range s.sns {
+		wg.Add(1)
+		go func(nodeIdx int) {
+			defer wg.Done()
+			sn := s.sns[nodeIdx]
+			err := sn.runQuery(ctxWithCancel, tenantIDs, q, func(db *logstorage.DataBlock) {
+				writeBlock(uint(nodeIdx), db)
+			})
+			if err != nil {
+				// Cancel the remaining parallel queries
+				cancel()
+			}
+
+			errs[nodeIdx] = err
+		}(i)
+	}
+	wg.Wait()
+
+	return getFirstNonCancelError(errs)
+}
+
+// GetFieldNames executes q and returns field names seen in results.
+func (s *Storage) GetFieldNames(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query) ([]logstorage.ValueWithHits, error) {
+	return s.getValuesWithHits(ctx, 0, false, func(ctx context.Context, sn *storageNode) ([]logstorage.ValueWithHits, error) {
+		return sn.getFieldNames(ctx, tenantIDs, q)
+	})
+}
+
+// GetFieldValues executes q and returns unique values for the fieldName seen in results.
+//
+// If limit > 0, then up to limit unique values are returned.
+func (s *Storage) GetFieldValues(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, fieldName string, limit uint64) ([]logstorage.ValueWithHits, error) {
+	return s.getValuesWithHits(ctx, limit, true, func(ctx context.Context, sn *storageNode) ([]logstorage.ValueWithHits, error) {
+		return sn.getFieldValues(ctx, tenantIDs, q, fieldName, limit)
+	})
+}
+
+// GetStreamFieldNames executes q and returns stream field names seen in results.
+func (s *Storage) GetStreamFieldNames(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query) ([]logstorage.ValueWithHits, error) {
+	return s.getValuesWithHits(ctx, 0, false, func(ctx context.Context, sn *storageNode) ([]logstorage.ValueWithHits, error) {
+		return sn.getStreamFieldNames(ctx, tenantIDs, q)
+	})
+}
+
+// GetStreamFieldValues executes q and returns stream field values for the given fieldName seen in results.
+//
+// If limit > 0, then up to limit unique stream field values are returned.
+func (s *Storage) GetStreamFieldValues(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, fieldName string, limit uint64) ([]logstorage.ValueWithHits, error) {
+	return s.getValuesWithHits(ctx, limit, true, func(ctx context.Context, sn *storageNode) ([]logstorage.ValueWithHits, error) {
+		return sn.getStreamFieldValues(ctx, tenantIDs, q, fieldName, limit)
+	})
+}
+
+// GetStreams executes q and returns streams seen in query results.
+//
+// If limit > 0, then up to limit unique streams are returned.
+func (s *Storage) GetStreams(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, limit uint64) ([]logstorage.ValueWithHits, error) {
+	return s.getValuesWithHits(ctx, limit, true, func(ctx context.Context, sn *storageNode) ([]logstorage.ValueWithHits, error) {
+		return sn.getStreams(ctx, tenantIDs, q, limit)
+	})
+}
+
+// GetStreamIDs executes q and returns streamIDs seen in query results.
+//
+// If limit > 0, then up to limit unique streamIDs are returned.
+func (s *Storage) GetStreamIDs(ctx context.Context, tenantIDs []logstorage.TenantID, q *logstorage.Query, limit uint64) ([]logstorage.ValueWithHits, error) {
+	return s.getValuesWithHits(ctx, limit, true, func(ctx context.Context, sn *storageNode) ([]logstorage.ValueWithHits, error) {
+		return sn.getStreamIDs(ctx, tenantIDs, q, limit)
+	})
+}
+
+func (s *Storage) getValuesWithHits(ctx context.Context, limit uint64, resetHitsOnLimitExceeded bool,
+	callback func(ctx context.Context, sn *storageNode) ([]logstorage.ValueWithHits, error)) ([]logstorage.ValueWithHits, error) {
+
+	ctxWithCancel, cancel := context.WithCancel(ctx)
+	defer cancel()
+
+	results := make([][]logstorage.ValueWithHits, len(s.sns))
+	errs := make([]error, len(s.sns))
+
+	var wg sync.WaitGroup
+	for i := range s.sns {
+		wg.Add(1)
+		go func(nodeIdx int) {
+			defer wg.Done()
+
+			sn := s.sns[nodeIdx]
+			vhs, err := callback(ctxWithCancel, sn)
+			results[nodeIdx] = vhs
+			errs[nodeIdx] = err
+
+			if err != nil {
+				// Cancel the remaining parallel requests
+				cancel()
+			}
+		}(i)
+	}
+	wg.Wait()
+
+	if err := getFirstNonCancelError(errs); err != nil {
+		return nil, err
+	}
+
+	vhs := logstorage.MergeValuesWithHits(results, limit, resetHitsOnLimitExceeded)
+
+	return vhs, nil
+}
+
+func getFirstNonCancelError(errs []error) error {
+	for _, err := range errs {
+		if err != nil && !errors.Is(err, context.Canceled) {
+			return err
+		}
+	}
+	return nil
+}
+
+func unmarshalValuesWithHits(src []byte) ([]logstorage.ValueWithHits, error) {
+	var vhs []logstorage.ValueWithHits
+	for len(src) > 0 {
+		var vh logstorage.ValueWithHits
+		tail, err := vh.UnmarshalInplace(src)
+		if err != nil {
+			return nil, fmt.Errorf("cannot unmarshal ValueWithHits #%d: %w", len(vhs), err)
+		}
+		src = tail
+
+		// Clone vh.Value, since it points to src.
+		vh.Value = strings.Clone(vh.Value)
+
+		vhs = append(vhs, vh)
+	}
+
+	return vhs, nil
+}

app/vmagent/README.md

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

app/vmagent/csvimport/request_handler.go

@@ -7,9 +7,9 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/remotewrite"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
-	parser "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/csvimport"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/csvimport"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/csvimport/stream"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -22,16 +22,16 @@ var (
 
 // InsertHandler processes csv data from req.
 func InsertHandler(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
-	return stream.Parse(req, func(rows []parser.Row) error {
+	return stream.Parse(req, func(rows []csvimport.Row) error {
 		return insertRows(at, rows, extraLabels)
 	})
 }
 
-func insertRows(at *auth.Token, rows []parser.Row, extraLabels []prompbmarshal.Label) error {
+func insertRows(at *auth.Token, rows []csvimport.Row, extraLabels []prompbmarshal.Label) error {
 	ctx := common.GetPushCtx()
 	defer common.PutPushCtx(ctx)
 

app/vmagent/datadogsketches/request_handler.go

@@ -7,10 +7,10 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/remotewrite"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogsketches"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogsketches/stream"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -23,7 +23,7 @@ var (
 
 // InsertHandlerForHTTP processes remote write for DataDog POST /api/beta/sketches request.
 func InsertHandlerForHTTP(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
@@ -56,7 +56,7 @@ func insertRows(at *auth.Token, sketches []*datadogsketches.Sketch, extraLabels
 				})
 			}
 			for _, tag := range sketch.Tags {
-				name, value := datadogutils.SplitTag(tag)
+				name, value := datadogutil.SplitTag(tag)
 				if name == "host" {
 					name = "exported_host"
 				}

app/vmagent/datadogv1/request_handler.go

@@ -7,10 +7,10 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/remotewrite"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogv1"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogv1/stream"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -23,7 +23,7 @@ var (
 
 // InsertHandlerForHTTP processes remote write for DataDog POST /api/v1/series request.
 func InsertHandlerForHTTP(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
@@ -62,7 +62,7 @@ func insertRows(at *auth.Token, series []datadogv1.Series, extraLabels []prompbm
 			})
 		}
 		for _, tag := range ss.Tags {
-			name, value := datadogutils.SplitTag(tag)
+			name, value := datadogutil.SplitTag(tag)
 			if name == "host" {
 				name = "exported_host"
 			}

app/vmagent/datadogv2/request_handler.go

@@ -7,10 +7,10 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/remotewrite"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogv2"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/datadogv2/stream"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -25,7 +25,7 @@ var (
 //
 // See https://docs.datadoghq.com/api/latest/metrics/#submit-metrics
 func InsertHandlerForHTTP(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
@@ -65,7 +65,7 @@ func insertRows(at *auth.Token, series []datadogv2.Series, extraLabels []prompbm
 			})
 		}
 		for _, tag := range ss.Tags {
-			name, value := datadogutils.SplitTag(tag)
+			name, value := datadogutil.SplitTag(tag)
 			if name == "host" {
 				name = "exported_host"
 			}

app/vmagent/graphite/request_handler.go

@@ -21,7 +21,7 @@ var (
 //
 // See https://graphite.readthedocs.io/en/latest/feeding-carbon.html#the-plaintext-protocol
 func InsertHandler(r io.Reader) error {
-	return stream.Parse(r, false, func(rows []parser.Row) error {
+	return stream.Parse(r, "", func(rows []parser.Row) error {
 		return insertRows(nil, rows)
 	})
 }

app/vmagent/influx/request_handler.go

@@ -12,9 +12,9 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
-	parser "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/influx"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/influx"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/influx/stream"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -35,8 +35,8 @@ var (
 // InsertHandlerForReader processes remote write for influx line protocol.
 //
 // See https://github.com/influxdata/telegraf/tree/master/plugins/inputs/socket_listener/
-func InsertHandlerForReader(at *auth.Token, r io.Reader, isGzipped bool) error {
-	return stream.Parse(r, true, isGzipped, "", "", func(db string, rows []parser.Row) error {
+func InsertHandlerForReader(at *auth.Token, r io.Reader, encoding string) error {
+	return stream.Parse(r, encoding, true, "", "", func(db string, rows []influx.Row) error {
 		return insertRows(at, db, rows, nil)
 	})
 }
@@ -45,22 +45,22 @@ func InsertHandlerForReader(at *auth.Token, r io.Reader, isGzipped bool) error {
 //
 // See https://github.com/influxdata/influxdb/blob/4cbdc197b8117fee648d62e2e5be75c6575352f0/tsdb/README.md
 func InsertHandlerForHTTP(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
-	isGzipped := req.Header.Get("Content-Encoding") == "gzip"
-	isStreamMode := req.Header.Get("Stream-Mode") == "1"
 	q := req.URL.Query()
 	precision := q.Get("precision")
 	// Read db tag from https://docs.influxdata.com/influxdb/v1.7/tools/api/#write-http-endpoint
 	db := q.Get("db")
-	return stream.Parse(req.Body, isStreamMode, isGzipped, precision, db, func(db string, rows []parser.Row) error {
+	encoding := req.Header.Get("Content-Encoding")
+	isStreamMode := req.Header.Get("Stream-Mode") == "1"
+	return stream.Parse(req.Body, encoding, isStreamMode, precision, db, func(db string, rows []influx.Row) error {
 		return insertRows(at, db, rows, extraLabels)
 	})
 }
 
-func insertRows(at *auth.Token, db string, rows []parser.Row, extraLabels []prompbmarshal.Label) error {
+func insertRows(at *auth.Token, db string, rows []influx.Row, extraLabels []prompbmarshal.Label) error {
 	ctx := getPushCtx()
 	defer putPushCtx(ctx)
 

app/vmagent/main.go

@@ -34,7 +34,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/envflag"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/influxutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/influxutil"
 	graphiteserver "github.com/VictoriaMetrics/VictoriaMetrics/lib/ingestserver/graphite"
 	influxserver "github.com/VictoriaMetrics/VictoriaMetrics/lib/ingestserver/influx"
 	opentsdbserver "github.com/VictoriaMetrics/VictoriaMetrics/lib/ingestserver/opentsdb"
@@ -42,8 +42,8 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/procutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentelemetry/firehose"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/pushmetrics"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/stringsutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeserieslimits"
@@ -105,7 +105,7 @@ func main() {
 	// Some workloads may need increased GOGC values. Then such values can be set via GOGC environment variable.
 	// It is recommended increasing GOGC if go_memstats_gc_cpu_fraction metric exposed at /metrics page
 	// exceeds 0.05 for extended periods of time.
-	cgroup.SetGOGC(30)
+	cgroup.SetGOGC(50)
 
 	// Write flags and help message to stdout, since it is easier to grep or pipe.
 	flag.CommandLine.SetOutput(os.Stdout)
@@ -145,10 +145,10 @@ func main() {
 	startTime := time.Now()
 	remotewrite.StartIngestionRateLimiter()
 	remotewrite.Init()
-	common.StartUnmarshalWorkers()
+	protoparserutil.StartUnmarshalWorkers()
 	if len(*influxListenAddr) > 0 {
 		influxServer = influxserver.MustStart(*influxListenAddr, *influxUseProxyProtocol, func(r io.Reader) error {
-			return influx.InsertHandlerForReader(nil, r, false)
+			return influx.InsertHandlerForReader(nil, r, "")
 		})
 	}
 	if len(*graphiteListenAddr) > 0 {
@@ -195,7 +195,7 @@ func main() {
 	if len(*opentsdbHTTPListenAddr) > 0 {
 		opentsdbhttpServer.MustStop()
 	}
-	common.StopUnmarshalWorkers()
+	protoparserutil.StopUnmarshalWorkers()
 	remotewrite.Stop()
 
 	logger.Infof("successfully stopped vmagent in %.3f seconds", time.Since(startTime).Seconds())
@@ -282,7 +282,7 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 	}
 	switch path {
 	case "/prometheus/api/v1/write", "/api/v1/write", "/api/v1/push", "/prometheus/api/v1/push":
-		if common.HandleVMProtoServerHandshake(w, r) {
+		if protoparserutil.HandleVMProtoServerHandshake(w, r) {
 			return true
 		}
 		prometheusWriteRequests.Inc()
@@ -331,11 +331,11 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 		return true
 	case "/influx/query", "/query":
 		influxQueryRequests.Inc()
-		influxutils.WriteDatabaseNames(w)
+		influxutil.WriteDatabaseNames(w)
 		return true
 	case "/influx/health":
 		influxHealthRequests.Inc()
-		influxutils.WriteHealthCheckResponse(w)
+		influxutil.WriteHealthCheckResponse(w)
 		return true
 	case "/opentelemetry/api/v1/push", "/opentelemetry/v1/metrics":
 		opentelemetryPushRequests.Inc()
@@ -443,8 +443,10 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 	case "/prometheus/api/v1/targets", "/api/v1/targets":
 		promscrapeAPIV1TargetsRequests.Inc()
 		w.Header().Set("Content-Type", "application/json")
+		// https://prometheus.io/docs/prometheus/latest/querying/api/#targets
 		state := r.FormValue("state")
-		promscrape.WriteAPIV1Targets(w, state)
+		scrapePool := r.FormValue("scrapePool")
+		promscrape.WriteAPIV1Targets(w, state, scrapePool)
 		return true
 	case "/prometheus/target_response", "/target_response":
 		promscrapeTargetResponseRequests.Inc()
@@ -587,11 +589,11 @@ func processMultitenantRequest(w http.ResponseWriter, r *http.Request, path stri
 		return true
 	case "influx/query":
 		influxQueryRequests.Inc()
-		influxutils.WriteDatabaseNames(w)
+		influxutil.WriteDatabaseNames(w)
 		return true
 	case "influx/health":
 		influxHealthRequests.Inc()
-		influxutils.WriteHealthCheckResponse(w)
+		influxutil.WriteHealthCheckResponse(w)
 		return true
 	case "opentelemetry/api/v1/push", "opentelemetry/v1/metrics":
 		opentelemetryPushRequests.Inc()

app/vmagent/native/request_handler.go

@@ -9,8 +9,8 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/native/stream"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -25,12 +25,12 @@ var (
 //
 // See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/6
 func InsertHandler(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
-	isGzip := req.Header.Get("Content-Encoding") == "gzip"
-	return stream.Parse(req.Body, isGzip, func(block *stream.Block) error {
+	encoding := req.Header.Get("Content-Encoding")
+	return stream.Parse(req.Body, encoding, func(block *stream.Block) error {
 		return insertRows(at, block, extraLabels)
 	})
 }

app/vmagent/newrelic/request_handler.go

@@ -10,9 +10,9 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/newrelic"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/newrelic/stream"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 )
 
@@ -24,13 +24,12 @@ var (
 
 // InsertHandlerForHTTP processes remote write for NewRelic POST /infra/v2/metrics/events/bulk request.
 func InsertHandlerForHTTP(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
-	ce := req.Header.Get("Content-Encoding")
-	isGzip := ce == "gzip"
-	return stream.Parse(req.Body, isGzip, func(rows []newrelic.Row) error {
+	encoding := req.Header.Get("Content-Encoding")
+	return stream.Parse(req.Body, encoding, func(rows []newrelic.Row) error {
 		return insertRows(at, rows, extraLabels)
 	})
 }

app/vmagent/opentelemetry/request_handler.go

@@ -8,9 +8,9 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/remotewrite"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentelemetry/firehose"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentelemetry/stream"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -23,11 +23,11 @@ var (
 
 // InsertHandler processes opentelemetry metrics.
 func InsertHandler(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
-	isGzipped := req.Header.Get("Content-Encoding") == "gzip"
+	encoding := req.Header.Get("Content-Encoding")
 	var processBody func([]byte) ([]byte, error)
 	if req.Header.Get("Content-Type") == "application/json" {
 		if req.Header.Get("X-Amz-Firehose-Protocol-Version") != "" {
@@ -36,7 +36,7 @@ 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, isGzipped, processBody, func(tss []prompbmarshal.TimeSeries) error {
+	return stream.ParseStream(req.Body, encoding, processBody, func(tss []prompbmarshal.TimeSeries) error {
 		return insertRows(at, tss, extraLabels)
 	})
 }

app/vmagent/opentsdbhttp/request_handler.go

@@ -7,9 +7,9 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/remotewrite"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
-	parser "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentsdbhttp"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentsdbhttp"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentsdbhttp/stream"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/metrics"
 )
 
@@ -21,16 +21,16 @@ var (
 // InsertHandler processes HTTP OpenTSDB put requests.
 // See http://opentsdb.net/docs/build/html/api_http/put.html
 func InsertHandler(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
-	return stream.Parse(req, func(rows []parser.Row) error {
+	return stream.Parse(req, func(rows []opentsdbhttp.Row) error {
 		return insertRows(at, rows, extraLabels)
 	})
 }
 
-func insertRows(at *auth.Token, rows []parser.Row, extraLabels []prompbmarshal.Label) error {
+func insertRows(at *auth.Token, rows []opentsdbhttp.Row, extraLabels []prompbmarshal.Label) error {
 	ctx := common.GetPushCtx()
 	defer common.PutPushCtx(ctx)
 

app/vmagent/prometheusimport/request_handler.go

@@ -8,9 +8,9 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
-	parser "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/prometheus"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/prometheus"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/prometheus/stream"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -23,23 +23,23 @@ var (
 
 // InsertHandler processes `/api/v1/import/prometheus` request.
 func InsertHandler(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
-	defaultTimestamp, err := parserCommon.GetTimestamp(req)
+	defaultTimestamp, err := protoparserutil.GetTimestamp(req)
 	if err != nil {
 		return err
 	}
-	isGzipped := req.Header.Get("Content-Encoding") == "gzip"
-	return stream.Parse(req.Body, defaultTimestamp, isGzipped, true, func(rows []parser.Row) error {
+	encoding := req.Header.Get("Content-Encoding")
+	return stream.Parse(req.Body, defaultTimestamp, encoding, true, func(rows []prometheus.Row) error {
 		return insertRows(at, rows, extraLabels)
 	}, func(s string) {
 		httpserver.LogError(req, s)
 	})
 }
 
-func insertRows(at *auth.Token, rows []parser.Row, extraLabels []prompbmarshal.Label) error {
+func insertRows(at *auth.Token, rows []prometheus.Row, extraLabels []prompbmarshal.Label) error {
 	ctx := common.GetPushCtx()
 	defer common.PutPushCtx(ctx)
 

app/vmagent/prometheusimport/request_handler_test.go

@@ -12,7 +12,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmagent/remotewrite"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 )
 
 var (
@@ -26,7 +26,7 @@ func TestInsertHandler(t *testing.T) {
 	req := httptest.NewRequest(http.MethodPost, "/insert/0/api/v1/import/prometheus", bytes.NewBufferString(`{"foo":"bar"}
 go_memstats_alloc_bytes_total 1`))
 	if err := InsertHandler(nil, req); err != nil {
-		t.Fatalf("unxepected error %s", err)
+		t.Fatalf("unexpected error %s", err)
 	}
 	expectedMsg := "cannot unmarshal Prometheus line"
 	if !strings.Contains(testOutput.String(), expectedMsg) {
@@ -44,14 +44,14 @@ func setUp() {
 		log.Fatalf("unable to set %q with value %q, err: %v", remoteWriteFlag, srv.URL, err)
 	}
 	logger.Init()
-	common.StartUnmarshalWorkers()
+	protoparserutil.StartUnmarshalWorkers()
 	remotewrite.Init()
 	testOutput = &bytes.Buffer{}
 	logger.SetOutputForTests(testOutput)
 }
 
 func tearDown() {
-	common.StopUnmarshalWorkers()
+	protoparserutil.StopUnmarshalWorkers()
 	srv.Close()
 	logger.ResetOutputForTest()
 	tmpDataDir := flag.Lookup("remoteWrite.tmpDataPath").Value.String()

app/vmagent/promremotewrite/request_handler.go

@@ -8,8 +8,8 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/auth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/promremotewrite/stream"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 	"github.com/VictoriaMetrics/metrics"
 )
@@ -22,7 +22,7 @@ var (
 
 // InsertHandler processes remote write for prometheus.
 func InsertHandler(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}

app/vmagent/remotewrite/client.go

@@ -10,19 +10,22 @@ import (
 	"strconv"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/awsapi"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding/zstd"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/netutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/persistentqueue"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/ratelimiter"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timerpool"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeutil"
 	"github.com/VictoriaMetrics/metrics"
+	"github.com/golang/snappy"
 )
 
 var (
@@ -87,7 +90,8 @@ type client struct {
 	remoteWriteURL string
 
 	// Whether to use VictoriaMetrics remote write protocol for sending the data to remoteWriteURL
-	useVMProto bool
+	useVMProto          atomic.Bool
+	canDowngradeVMProto atomic.Bool
 
 	fq *persistentqueue.FastQueue
 	hc *http.Client
@@ -124,14 +128,14 @@ func newHTTPClient(argIdx int, remoteWriteURL, sanitizedURL string, fq *persiste
 	if err != nil {
 		logger.Fatalf("cannot initialize AWS Config for -remoteWrite.url=%q: %s", remoteWriteURL, err)
 	}
-	tr := &http.Transport{
-		DialContext:         netutil.NewStatDialFunc("vmagent_remotewrite"),
-		TLSHandshakeTimeout: tlsHandshakeTimeout.GetOptionalArg(argIdx),
-		MaxConnsPerHost:     2 * concurrency,
-		MaxIdleConnsPerHost: 2 * concurrency,
-		IdleConnTimeout:     time.Minute,
-		WriteBufferSize:     64 * 1024,
-	}
+
+	tr := httputil.NewTransport(false, "vmagent_remotewrite")
+	tr.TLSHandshakeTimeout = tlsHandshakeTimeout.GetOptionalArg(argIdx)
+	tr.MaxConnsPerHost = 2 * concurrency
+	tr.MaxIdleConnsPerHost = 2 * concurrency
+	tr.IdleConnTimeout = time.Minute
+	tr.WriteBufferSize = 64 * 1024
+
 	pURL := proxyURL.GetOptionalArg(argIdx)
 	if len(pURL) > 0 {
 		if !strings.Contains(pURL, "://") {
@@ -166,17 +170,11 @@ func newHTTPClient(argIdx int, remoteWriteURL, sanitizedURL string, fq *persiste
 		logger.Fatalf("-remoteWrite.useVMProto and -remoteWrite.usePromProto cannot be set simultaneously for -remoteWrite.url=%s", sanitizedURL)
 	}
 	if !useVMProto && !usePromProto {
-		// Auto-detect whether the remote storage supports VictoriaMetrics remote write protocol.
-		doRequest := func(url string) (*http.Response, error) {
-			return c.doRequest(url, nil)
-		}
-		useVMProto = common.HandleVMProtoClientHandshake(c.remoteWriteURL, doRequest)
-		if !useVMProto {
-			logger.Infof("the remote storage at %q doesn't support VictoriaMetrics remote write protocol. Switching to Prometheus remote write protocol. "+
-				"See https://docs.victoriametrics.com/vmagent/#victoriametrics-remote-write-protocol", sanitizedURL)
-		}
+		// The VM protocol could be downgraded later at runtime if unsupported media type response status is received.
+		useVMProto = true
+		c.canDowngradeVMProto.Store(true)
 	}
-	c.useVMProto = useVMProto
+	c.useVMProto.Store(useVMProto)
 
 	return c
 }
@@ -384,7 +382,7 @@ func (c *client) newRequest(url string, body []byte) (*http.Request, error) {
 	h := req.Header
 	h.Set("User-Agent", "vmagent")
 	h.Set("Content-Type", "application/x-protobuf")
-	if c.useVMProto {
+	if encoding.IsZstd(body) {
 		h.Set("Content-Encoding", "zstd")
 		h.Set("X-VictoriaMetrics-Remote-Write-Version", "1")
 	} else {
@@ -420,7 +418,7 @@ again:
 		if retryDuration > maxRetryDuration {
 			retryDuration = maxRetryDuration
 		}
-		logger.Warnf("couldn't send a block with size %d bytes to %q: %s; re-sending the block in %.3f seconds",
+		remoteWriteRetryLogger.Warnf("couldn't send a block with size %d bytes to %q: %s; re-sending the block in %.3f seconds",
 			len(block), c.sanitizedURL, err, retryDuration.Seconds())
 		t := timerpool.Get(retryDuration)
 		select {
@@ -433,6 +431,7 @@ again:
 		c.retriesCount.Inc()
 		goto again
 	}
+
 	statusCode := resp.StatusCode
 	if statusCode/100 == 2 {
 		_ = resp.Body.Close()
@@ -441,24 +440,46 @@ again:
 		c.blocksSent.Inc()
 		return true
 	}
+
 	metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_requests_total{url=%q, status_code="%d"}`, c.sanitizedURL, statusCode)).Inc()
-	if statusCode == 409 || statusCode == 400 {
-		body, err := io.ReadAll(resp.Body)
-		_ = resp.Body.Close()
-		if err != nil {
-			remoteWriteRejectedLogger.Errorf("sending a block with size %d bytes to %q was rejected (skipping the block): status code %d; "+
-				"failed to read response body: %s",
-				len(block), c.sanitizedURL, statusCode, err)
-		} else {
-			remoteWriteRejectedLogger.Errorf("sending a block with size %d bytes to %q was rejected (skipping the block): status code %d; response body: %s",
-				len(block), c.sanitizedURL, statusCode, string(body))
-		}
-		// Just drop block on 409 and 400 status codes like Prometheus does.
+	if statusCode == 409 {
+		logBlockRejected(block, c.sanitizedURL, resp)
+
+		// Just drop block on 409 status code like Prometheus does.
 		// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/873
 		// and https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1149
 		_ = resp.Body.Close()
 		c.packetsDropped.Inc()
 		return true
+		// - Remote Write v1 specification implicitly expects a `400 Bad Request` when the encoding is not supported.
+		// - Remote Write v2 specification explicitly specifies a `415 Unsupported Media Type` for unsupported encodings.
+		// - Real-world implementations of v1 use both 400 and 415 status codes.
+		// See more in research: https://github.com/VictoriaMetrics/VictoriaMetrics/pull/8462#issuecomment-2786918054
+	} else if statusCode == 415 || statusCode == 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/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/vmagent/#victoriametrics-remote-write-protocol", c.sanitizedURL)
+
+			block = mustRepackBlockFromZstdToSnappy(block)
+
+			c.retriesCount.Inc()
+			_ = resp.Body.Close()
+			goto again
+		}
+
+		// Just drop snappy blocks on 400 or 415 status codes like Prometheus does.
+		// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/873
+		// and https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1149
+		logBlockRejected(block, c.sanitizedURL, resp)
+		_ = resp.Body.Close()
+		c.packetsDropped.Inc()
+		return true
 	}
 
 	// Unexpected status code returned
@@ -488,6 +509,7 @@ again:
 }
 
 var remoteWriteRejectedLogger = logger.WithThrottler("remoteWriteRejected", 5*time.Second)
+var remoteWriteRetryLogger = logger.WithThrottler("remoteWriteRetry", 5*time.Second)
 
 // getRetryDuration returns retry duration.
 // retryAfterDuration has the highest priority.
@@ -510,6 +532,28 @@ func getRetryDuration(retryAfterDuration, retryDuration, maxRetryDuration time.D
 	return retryDuration
 }
 
+func mustRepackBlockFromZstdToSnappy(zstdBlock []byte) []byte {
+	plainBlock := make([]byte, 0, len(zstdBlock)*2)
+	plainBlock, err := zstd.Decompress(plainBlock, zstdBlock)
+	if err != nil {
+		logger.Panicf("FATAL: cannot re-pack block with size %d bytes from Zstd to Snappy: %s", len(zstdBlock), err)
+	}
+
+	return snappy.Encode(nil, plainBlock)
+}
+
+func logBlockRejected(block []byte, sanitizedURL string, resp *http.Response) {
+	body, err := io.ReadAll(resp.Body)
+	if err != nil {
+		remoteWriteRejectedLogger.Errorf("sending a block with size %d bytes to %q was rejected (skipping the block): status code %d; "+
+			"failed to read response body: %s",
+			len(block), sanitizedURL, resp.StatusCode, err)
+	} else {
+		remoteWriteRejectedLogger.Errorf("sending a block with size %d bytes to %q was rejected (skipping the block): status code %d; response body: %s",
+			len(block), sanitizedURL, resp.StatusCode, string(body))
+	}
+}
+
 // parseRetryAfterHeader parses `Retry-After` value retrieved from HTTP response header.
 // retryAfterString should be in either HTTP-date or a number of seconds.
 // It will return time.Duration(0) if `retryAfterString` does not follow RFC 7231.

app/vmagent/remotewrite/client_test.go

@@ -5,6 +5,9 @@ import (
 	"net/http"
 	"testing"
 	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
+	"github.com/golang/snappy"
 )
 
 func TestCalculateRetryDuration(t *testing.T) {
@@ -97,3 +100,19 @@ func helper(d time.Duration) time.Duration {
 
 	return d + dv
 }
+
+func TestRepackBlockFromZstdToSnappy(t *testing.T) {
+	expectedPlainBlock := []byte(`foobar`)
+
+	zstdBlock := encoding.CompressZSTDLevel(nil, expectedPlainBlock, 1)
+	snappyBlock := mustRepackBlockFromZstdToSnappy(zstdBlock)
+
+	actualPlainBlock, err := snappy.Decode(nil, snappyBlock)
+	if err != nil {
+		t.Fatalf("unexpected error: %s", err)
+	}
+
+	if string(actualPlainBlock) != string(expectedPlainBlock) {
+		t.Fatalf("unexpected plain block; got %q; want %q", actualPlainBlock, expectedPlainBlock)
+	}
+}

app/vmagent/remotewrite/pendingseries.go

@@ -16,6 +16,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/persistentqueue"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/slicesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeutil"
 	"github.com/VictoriaMetrics/metrics"
 	"github.com/golang/snappy"
@@ -39,7 +40,7 @@ type pendingSeries struct {
 	periodicFlusherWG sync.WaitGroup
 }
 
-func newPendingSeries(fq *persistentqueue.FastQueue, isVMRemoteWrite bool, significantFigures, roundDigits int) *pendingSeries {
+func newPendingSeries(fq *persistentqueue.FastQueue, isVMRemoteWrite *atomic.Bool, significantFigures, roundDigits int) *pendingSeries {
 	var ps pendingSeries
 	ps.wr.fq = fq
 	ps.wr.isVMRemoteWrite = isVMRemoteWrite
@@ -99,7 +100,7 @@ type writeRequest struct {
 	fq *persistentqueue.FastQueue
 
 	// Whether to encode the write request with VictoriaMetrics remote write protocol.
-	isVMRemoteWrite bool
+	isVMRemoteWrite *atomic.Bool
 
 	// How many significant figures must be left before sending the writeRequest to fq.
 	significantFigures int
@@ -118,7 +119,7 @@ type writeRequest struct {
 }
 
 func (wr *writeRequest) reset() {
-	// Do not reset lastFlushTime, fq, isVMRemoteWrite, significantFigures and roundDigits, since they are re-used.
+	// Do not reset lastFlushTime, fq, isVMRemoteWrite, significantFigures and roundDigits, since they are reused.
 
 	wr.wr.Timeseries = nil
 
@@ -137,7 +138,7 @@ 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
-	if !tryPushWriteRequest(&wr.wr, wr.mustWriteBlock, wr.isVMRemoteWrite) {
+	if !tryPushWriteRequest(&wr.wr, wr.mustWriteBlock, wr.isVMRemoteWrite.Load()) {
 		logger.Panicf("BUG: final flush must always return true")
 	}
 	wr.reset()
@@ -151,7 +152,7 @@ func (wr *writeRequest) mustWriteBlock(block []byte) bool {
 func (wr *writeRequest) tryFlush() bool {
 	wr.wr.Timeseries = wr.tss
 	wr.lastFlushTime.Store(fasttime.UnixTimestamp())
-	if !tryPushWriteRequest(&wr.wr, wr.fq.TryWriteBlock, wr.isVMRemoteWrite) {
+	if !tryPushWriteRequest(&wr.wr, wr.fq.TryWriteBlock, wr.isVMRemoteWrite.Load()) {
 		return false
 	}
 	wr.reset()
@@ -197,28 +198,43 @@ func (wr *writeRequest) tryPush(src []prompbmarshal.TimeSeries) bool {
 }
 
 func (wr *writeRequest) copyTimeSeries(dst, src *prompbmarshal.TimeSeries) {
-	labelsDst := wr.labels
+	labelsSrc := src.Labels
+
+	// Pre-allocate memory for labels.
 	labelsLen := len(wr.labels)
-	samplesDst := wr.samples
-	buf := wr.buf
-	for i := range src.Labels {
-		labelsDst = append(labelsDst, prompbmarshal.Label{})
-		dstLabel := &labelsDst[len(labelsDst)-1]
-		srcLabel := &src.Labels[i]
+	wr.labels = slicesutil.SetLength(wr.labels, labelsLen+len(labelsSrc))
+	labelsDst := wr.labels[labelsLen:]
 
-		buf = append(buf, srcLabel.Name...)
-		dstLabel.Name = bytesutil.ToUnsafeString(buf[len(buf)-len(srcLabel.Name):])
-		buf = append(buf, srcLabel.Value...)
-		dstLabel.Value = bytesutil.ToUnsafeString(buf[len(buf)-len(srcLabel.Value):])
+	// Pre-allocate memory for byte slice needed for storing label names and values.
+	neededBufLen := 0
+	for i := range labelsSrc {
+		label := &labelsSrc[i]
+		neededBufLen += len(label.Name) + len(label.Value)
 	}
-	dst.Labels = labelsDst[labelsLen:]
+	bufLen := len(wr.buf)
+	wr.buf = slicesutil.SetLength(wr.buf, bufLen+neededBufLen)
+	buf := wr.buf[:bufLen]
 
-	samplesDst = append(samplesDst, src.Samples...)
-	dst.Samples = samplesDst[len(samplesDst)-len(src.Samples):]
+	// Copy labels
+	for i := range labelsSrc {
+		dstLabel := &labelsDst[i]
+		srcLabel := &labelsSrc[i]
 
-	wr.samples = samplesDst
-	wr.labels = labelsDst
+		bufLen := len(buf)
+		buf = append(buf, srcLabel.Name...)
+		dstLabel.Name = bytesutil.ToUnsafeString(buf[bufLen:])
+
+		bufLen = len(buf)
+		buf = append(buf, srcLabel.Value...)
+		dstLabel.Value = bytesutil.ToUnsafeString(buf[bufLen:])
+	}
 	wr.buf = buf
+	dst.Labels = labelsDst
+
+	// Copy samples
+	samplesLen := len(wr.samples)
+	wr.samples = append(wr.samples, src.Samples...)
+	dst.Samples = wr.samples[samplesLen:]
 }
 
 // marshalConcurrency limits the maximum number of concurrent workers, which marshal and compress WriteRequest.

app/vmagent/remotewrite/relabel_test.go

@@ -6,7 +6,7 @@ import (
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 )
 
 func TestApplyRelabeling(t *testing.T) {
@@ -63,7 +63,7 @@ func TestAppendExtraLabels(t *testing.T) {
 func parseSeries(data string) []prompbmarshal.TimeSeries {
 	var tss []prompbmarshal.TimeSeries
 	tss = append(tss, prompbmarshal.TimeSeries{
-		Labels: promutils.MustNewLabelsFromString(data).GetLabels(),
+		Labels: promutil.MustNewLabelsFromString(data).GetLabels(),
 	})
 	return tss
 }

app/vmagent/remotewrite/remotewrite.go

@@ -15,6 +15,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bloomfilter"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/consistenthash"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
@@ -25,7 +26,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/procutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/ratelimiter"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/streamaggr"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeserieslimits"
@@ -43,7 +44,7 @@ var (
 		"according to https://docs.victoriametrics.com/#how-to-import-time-series-data ."+
 		"See https://docs.victoriametrics.com/vmagent/#multitenancy for details")
 
-	shardByURL = flag.Bool("remoteWrite.shardByURL", false, "Whether to shard outgoing series across all the remote storage systems enumerated via -remoteWrite.url . "+
+	shardByURL = flag.Bool("remoteWrite.shardByURL", false, "Whether to shard outgoing series across all the remote storage systems enumerated via -remoteWrite.url. "+
 		"By default the data is replicated across all the -remoteWrite.url . See https://docs.victoriametrics.com/vmagent/#sharding-among-remote-storages . "+
 		"See also -remoteWrite.shardByURLReplicas")
 	shardByURLReplicas = flag.Int("remoteWrite.shardByURLReplicas", 1, "How many copies of data to make among remote storage systems enumerated via -remoteWrite.url "+
@@ -54,7 +55,6 @@ var (
 	shardByURLIgnoreLabels = flagutil.NewArrayString("remoteWrite.shardByURL.ignoreLabels", "Optional list of labels, which must be ignored when sharding outgoing samples "+
 		"among remote storage systems if -remoteWrite.shardByURL command-line flag is set. By default all the labels are used for sharding in order to gain "+
 		"even distribution of series over the specified -remoteWrite.url systems. See also -remoteWrite.shardByURL.labels")
-
 	tmpDataPath = flag.String("remoteWrite.tmpDataPath", "vmagent-remotewrite-data", "Path to directory for storing pending data, which isn't sent to the configured -remoteWrite.url . "+
 		"See also -remoteWrite.maxDiskUsagePerURL and -remoteWrite.disableOnDiskQueue")
 	keepDanglingQueues = flag.Bool("remoteWrite.keepDanglingQueues", false, "Keep persistent queues contents at -remoteWrite.tmpDataPath in case there are no matching -remoteWrite.url. "+
@@ -96,7 +96,9 @@ var (
 
 var (
 	// rwctxsGlobal contains statically populated entries when -remoteWrite.url is specified.
-	rwctxsGlobal []*remoteWriteCtx
+	rwctxsGlobal              []*remoteWriteCtx
+	rwctxsGlobalIdx           []int
+	rwctxConsistentHashGlobal *consistenthash.ConsistentHash
 
 	// ErrQueueFullHTTPRetry must be returned when TryPush() returns false.
 	ErrQueueFullHTTPRetry = &httpserver.ErrorWithStatusCode{
@@ -205,7 +207,7 @@ func Init() {
 
 	initStreamAggrConfigGlobal()
 
-	rwctxsGlobal = newRemoteWriteCtxs(*remoteWriteURLs)
+	initRemoteWriteCtxs(*remoteWriteURLs)
 
 	disableOnDiskQueues := []bool(*disableOnDiskQueue)
 	disableOnDiskQueueAny = slices.Contains(disableOnDiskQueues, true)
@@ -290,7 +292,7 @@ var (
 	relabelConfigTimestamp    = metrics.NewCounter(`vmagent_relabel_config_last_reload_success_timestamp_seconds`)
 )
 
-func newRemoteWriteCtxs(urls []string) []*remoteWriteCtx {
+func initRemoteWriteCtxs(urls []string) {
 	if len(urls) == 0 {
 		logger.Panicf("BUG: urls must be non-empty")
 	}
@@ -306,6 +308,7 @@ func newRemoteWriteCtxs(urls []string) []*remoteWriteCtx {
 		maxInmemoryBlocks = 2
 	}
 	rwctxs := make([]*remoteWriteCtx, len(urls))
+	rwctxIdx := make([]int, len(urls))
 	for i, remoteWriteURLRaw := range urls {
 		remoteWriteURL, err := url.Parse(remoteWriteURLRaw)
 		if err != nil {
@@ -316,8 +319,19 @@ func newRemoteWriteCtxs(urls []string) []*remoteWriteCtx {
 			sanitizedURL = fmt.Sprintf("%d:%s", i+1, remoteWriteURL)
 		}
 		rwctxs[i] = newRemoteWriteCtx(i, remoteWriteURL, maxInmemoryBlocks, sanitizedURL)
+		rwctxIdx[i] = i
 	}
-	return rwctxs
+
+	if *shardByURL {
+		consistentHashNodes := make([]string, 0, len(urls))
+		for i, url := range urls {
+			consistentHashNodes = append(consistentHashNodes, fmt.Sprintf("%d:%s", i+1, url))
+		}
+		rwctxConsistentHashGlobal = consistenthash.NewConsistentHash(consistentHashNodes, 0)
+	}
+
+	rwctxsGlobal = rwctxs
+	rwctxsGlobalIdx = rwctxIdx
 }
 
 var (
@@ -501,6 +515,10 @@ func tryPush(at *auth.Token, wr *prompbmarshal.WriteRequest, forceDropSamplesOnF
 	return true
 }
 
+// getEligibleRemoteWriteCtxs checks whether writes to configured remote storage systems are blocked and
+// returns only the unblocked rwctx.
+//
+// calculateHealthyRwctxIdx will rely on the order of rwctx to be in ascending order.
 func getEligibleRemoteWriteCtxs(tss []prompbmarshal.TimeSeries, forceDropSamplesOnFailure bool) ([]*remoteWriteCtx, bool) {
 	if !disableOnDiskQueueAny {
 		return rwctxsGlobal, true
@@ -517,6 +535,12 @@ func getEligibleRemoteWriteCtxs(tss []prompbmarshal.TimeSeries, forceDropSamples
 				return nil, false
 			}
 			rowsCount := getRowsCount(tss)
+			if *shardByURL {
+				// Todo: When shardByURL is enabled, the following metrics won't be 100% accurate. Because vmagent don't know
+				// which rwctx should data be pushed to yet. Let's consider the hashing algorithm fair and will distribute
+				// data to all rwctxs evenly.
+				rowsCount = rowsCount / len(rwctxsGlobal)
+			}
 			rwctx.rowsDroppedOnPushFailure.Add(rowsCount)
 		}
 	}
@@ -528,6 +552,7 @@ func pushToRemoteStoragesTrackDropped(tss []prompbmarshal.TimeSeries) {
 	if len(rwctxs) == 0 {
 		return
 	}
+
 	if !tryPushBlockToRemoteStorages(rwctxs, tss, true) {
 		logger.Panicf("BUG: tryPushBlockToRemoteStorages() must return true when forceDropSamplesOnFailure=true")
 	}
@@ -578,42 +603,7 @@ func tryShardingBlockAmongRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []pr
 	defer putTSSShards(x)
 
 	shards := x.shards
-	tmpLabels := promutils.GetLabels()
-	for _, ts := range tssBlock {
-		hashLabels := ts.Labels
-		if len(shardByURLLabelsMap) > 0 {
-			hashLabels = tmpLabels.Labels[:0]
-			for _, label := range ts.Labels {
-				if _, ok := shardByURLLabelsMap[label.Name]; ok {
-					hashLabels = append(hashLabels, label)
-				}
-			}
-			tmpLabels.Labels = hashLabels
-		} else if len(shardByURLIgnoreLabelsMap) > 0 {
-			hashLabels = tmpLabels.Labels[:0]
-			for _, label := range ts.Labels {
-				if _, ok := shardByURLIgnoreLabelsMap[label.Name]; !ok {
-					hashLabels = append(hashLabels, label)
-				}
-			}
-			tmpLabels.Labels = hashLabels
-		}
-		h := getLabelsHash(hashLabels)
-		idx := h % uint64(len(shards))
-		i := 0
-		for {
-			shards[idx] = append(shards[idx], ts)
-			i++
-			if i >= replicas {
-				break
-			}
-			idx++
-			if idx >= uint64(len(shards)) {
-				idx = 0
-			}
-		}
-	}
-	promutils.PutLabels(tmpLabels)
+	shardAmountRemoteWriteCtx(tssBlock, shards, rwctxs, replicas)
 
 	// Push sharded samples to remote storage systems in parallel in order to reduce
 	// the time needed for sending the data to multiple remote storage systems.
@@ -636,6 +626,86 @@ func tryShardingBlockAmongRemoteStorages(rwctxs []*remoteWriteCtx, tssBlock []pr
 	return !anyPushFailed.Load()
 }
 
+// calculateHealthyRwctxIdx returns the index of healthyRwctxs in rwctxsGlobal.
+// It relies on the order of rwctx in healthyRwctxs, which is appended by getEligibleRemoteWriteCtxs.
+func calculateHealthyRwctxIdx(healthyRwctxs []*remoteWriteCtx) ([]int, []int) {
+	// fast path: all rwctxs are healthy.
+	if len(healthyRwctxs) == len(rwctxsGlobal) {
+		return rwctxsGlobalIdx, nil
+	}
+
+	unhealthyIdx := make([]int, 0, len(rwctxsGlobal))
+	healthyIdx := make([]int, 0, len(rwctxsGlobal))
+
+	var i int
+	for j := range rwctxsGlobal {
+		if i < len(healthyRwctxs) && rwctxsGlobal[j].idx == healthyRwctxs[i].idx {
+			healthyIdx = append(healthyIdx, j)
+			i++
+		} else {
+			unhealthyIdx = append(unhealthyIdx, j)
+		}
+	}
+
+	return healthyIdx, unhealthyIdx
+}
+
+// shardAmountRemoteWriteCtx distribute time series to shards by consistent hashing.
+func shardAmountRemoteWriteCtx(tssBlock []prompbmarshal.TimeSeries, shards [][]prompbmarshal.TimeSeries, rwctxs []*remoteWriteCtx, replicas int) {
+	tmpLabels := promutil.GetLabels()
+	defer promutil.PutLabels(tmpLabels)
+
+	healthyIdx, unhealthyIdx := calculateHealthyRwctxIdx(rwctxs)
+
+	// shardsIdxMap is a map to find which the shard idx by rwctxs idx.
+	// rwctxConsistentHashGlobal will tell which the rwctxs idx a time series should be written to.
+	// And this time series should be appended to the shards by correct shard idx.
+	shardsIdxMap := make(map[int]int, len(healthyIdx))
+	for idx, rwctxsIdx := range healthyIdx {
+		shardsIdxMap[rwctxsIdx] = idx
+	}
+
+	for _, ts := range tssBlock {
+		hashLabels := ts.Labels
+		if len(shardByURLLabelsMap) > 0 {
+			hashLabels = tmpLabels.Labels[:0]
+			for _, label := range ts.Labels {
+				if _, ok := shardByURLLabelsMap[label.Name]; ok {
+					hashLabels = append(hashLabels, label)
+				}
+			}
+			tmpLabels.Labels = hashLabels
+		} else if len(shardByURLIgnoreLabelsMap) > 0 {
+			hashLabels = tmpLabels.Labels[:0]
+			for _, label := range ts.Labels {
+				if _, ok := shardByURLIgnoreLabelsMap[label.Name]; !ok {
+					hashLabels = append(hashLabels, label)
+				}
+			}
+			tmpLabels.Labels = hashLabels
+		}
+		h := getLabelsHash(hashLabels)
+
+		// Get the rwctxIdx through consistent hashing and then map it to the index in shards.
+		// The rwctxIdx is not always equal to the shardIdx, for example, when some rwctx are not available.
+		rwctxIdx := rwctxConsistentHashGlobal.GetNodeIdx(h, unhealthyIdx)
+		shardIdx := shardsIdxMap[rwctxIdx]
+
+		replicated := 0
+		for {
+			shards[shardIdx] = append(shards[shardIdx], ts)
+			replicated++
+			if replicated >= replicas {
+				break
+			}
+			shardIdx++
+			if shardIdx >= len(shards) {
+				shardIdx = 0
+			}
+		}
+	}
+}
+
 type tssShards struct {
 	shards [][]prompbmarshal.TimeSeries
 }
@@ -807,7 +877,7 @@ func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, maxInmemoryBlocks in
 	}
 	pss := make([]*pendingSeries, pssLen)
 	for i := range pss {
-		pss[i] = newPendingSeries(fq, c.useVMProto, sf, rd)
+		pss[i] = newPendingSeries(fq, &c.useVMProto, sf, rd)
 	}
 
 	rwctx := &remoteWriteCtx{

app/vmagent/remotewrite/remotewrite_test.go

@@ -4,13 +4,17 @@ import (
 	"fmt"
 	"math"
 	"reflect"
+	"strconv"
+	"sync/atomic"
 	"testing"
 	"time"
 
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/consistenthash"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/prometheus"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/streamaggr"
+
 	"github.com/VictoriaMetrics/metrics"
 )
 
@@ -68,7 +72,9 @@ func TestRemoteWriteContext_TryPush_ImmutableTimeseries(t *testing.T) {
 		allRelabelConfigs.Store(rcs)
 
 		pss := make([]*pendingSeries, 1)
-		pss[0] = newPendingSeries(nil, true, 0, 100)
+		isVMProto := &atomic.Bool{}
+		isVMProto.Store(true)
+		pss[0] = newPendingSeries(nil, isVMProto, 0, 100)
 		rwctx := &remoteWriteCtx{
 			idx:                    0,
 			streamAggrKeepInput:    keepInput,
@@ -171,3 +177,173 @@ metric{env="dev"} 15
 metric{env="bar"} 25
 `)
 }
+
+func TestShardAmountRemoteWriteCtx(t *testing.T) {
+	// 1. distribute 100000 series to n nodes.
+	// 2. remove the last node from healthy list.
+	// 3. distribute the same 10000 series to (n-1) node again.
+	// 4. check active time series change rate:
+	// change rate must < (3/total nodes). e.g. +30% if 10 you have 10 nodes.
+
+	f := func(remoteWriteCount int, healthyIdx []int, replicas int) {
+		t.Helper()
+		defer func() {
+			rwctxsGlobal = nil
+			rwctxsGlobalIdx = nil
+			rwctxConsistentHashGlobal = nil
+		}()
+
+		rwctxsGlobal = make([]*remoteWriteCtx, remoteWriteCount)
+		rwctxsGlobalIdx = make([]int, remoteWriteCount)
+		rwctxs := make([]*remoteWriteCtx, 0, len(healthyIdx))
+
+		for i := range remoteWriteCount {
+			rwCtx := &remoteWriteCtx{
+				idx: i,
+			}
+			rwctxsGlobalIdx[i] = i
+
+			if i >= len(healthyIdx) {
+				rwctxsGlobal[i] = rwCtx
+				continue
+			}
+			hIdx := healthyIdx[i]
+			if hIdx != i {
+				rwctxs = append(rwctxs, &remoteWriteCtx{
+					idx: hIdx,
+				})
+			} else {
+				rwctxs = append(rwctxs, rwCtx)
+			}
+			rwctxsGlobal[i] = rwCtx
+		}
+
+		seriesCount := 100000
+		// build 1000000 series
+		tssBlock := make([]prompbmarshal.TimeSeries, 0, seriesCount)
+		for i := 0; i < seriesCount; i++ {
+			tssBlock = append(tssBlock, prompbmarshal.TimeSeries{
+				Labels: []prompbmarshal.Label{
+					{
+						Name:  "label",
+						Value: strconv.Itoa(i),
+					},
+				},
+				Samples: []prompbmarshal.Sample{
+					{
+						Timestamp: 0,
+						Value:     0,
+					},
+				},
+			})
+		}
+
+		// build consistent hash for x remote write context
+		// build active time series set
+		nodes := make([]string, 0, remoteWriteCount)
+		activeTimeSeriesByNodes := make([]map[string]struct{}, remoteWriteCount)
+		for i := 0; i < remoteWriteCount; i++ {
+			nodes = append(nodes, fmt.Sprintf("node%d", i))
+			activeTimeSeriesByNodes[i] = make(map[string]struct{})
+		}
+		rwctxConsistentHashGlobal = consistenthash.NewConsistentHash(nodes, 0)
+
+		// create shards
+		x := getTSSShards(len(rwctxs))
+		shards := x.shards
+
+		// execute
+		shardAmountRemoteWriteCtx(tssBlock, shards, rwctxs, replicas)
+
+		for i, nodeIdx := range healthyIdx {
+			for _, ts := range shards[i] {
+				// add it to node[nodeIdx]'s active time series
+				activeTimeSeriesByNodes[nodeIdx][prompbmarshal.LabelsToString(ts.Labels)] = struct{}{}
+			}
+		}
+
+		totalActiveTimeSeries := 0
+		for _, activeTimeSeries := range activeTimeSeriesByNodes {
+			totalActiveTimeSeries += len(activeTimeSeries)
+		}
+		avgActiveTimeSeries1 := totalActiveTimeSeries / remoteWriteCount
+		putTSSShards(x)
+
+		// removed last node
+		rwctxs = rwctxs[:len(rwctxs)-1]
+		healthyIdx = healthyIdx[:len(healthyIdx)-1]
+
+		x = getTSSShards(len(rwctxs))
+		shards = x.shards
+
+		// execute
+		shardAmountRemoteWriteCtx(tssBlock, shards, rwctxs, replicas)
+		for i, nodeIdx := range healthyIdx {
+			for _, ts := range shards[i] {
+				// add it to node[nodeIdx]'s active time series
+				activeTimeSeriesByNodes[nodeIdx][prompbmarshal.LabelsToString(ts.Labels)] = struct{}{}
+			}
+		}
+
+		totalActiveTimeSeries = 0
+		for _, activeTimeSeries := range activeTimeSeriesByNodes {
+			totalActiveTimeSeries += len(activeTimeSeries)
+		}
+		avgActiveTimeSeries2 := totalActiveTimeSeries / remoteWriteCount
+
+		changed := math.Abs(float64(avgActiveTimeSeries2-avgActiveTimeSeries1) / float64(avgActiveTimeSeries1))
+		threshold := 3 / float64(remoteWriteCount)
+
+		if changed >= threshold {
+			t.Fatalf("average active time series before: %d, after: %d, changed: %.2f. threshold: %.2f", avgActiveTimeSeries1, avgActiveTimeSeries2, changed, threshold)
+		}
+
+	}
+
+	f(5, []int{0, 1, 2, 3, 4}, 1)
+
+	f(5, []int{0, 1, 2, 3, 4}, 2)
+
+	f(10, []int{0, 1, 2, 3, 4, 5, 6, 7, 9}, 1)
+
+	f(10, []int{0, 1, 2, 3, 4, 5, 6, 7, 9}, 3)
+}
+
+func TestCalculateHealthyRwctxIdx(t *testing.T) {
+	f := func(total int, healthyIdx []int, unhealthyIdx []int) {
+		t.Helper()
+
+		healthyMap := make(map[int]bool)
+		for _, idx := range healthyIdx {
+			healthyMap[idx] = true
+		}
+		rwctxsGlobal = make([]*remoteWriteCtx, total)
+		rwctxsGlobalIdx = make([]int, total)
+		rwctxs := make([]*remoteWriteCtx, 0, len(healthyIdx))
+		for i := range rwctxsGlobal {
+			rwctx := &remoteWriteCtx{idx: i}
+			rwctxsGlobal[i] = rwctx
+			if healthyMap[i] {
+				rwctxs = append(rwctxs, rwctx)
+			}
+			rwctxsGlobalIdx[i] = i
+		}
+
+		gotHealthyIdx, gotUnhealthyIdx := calculateHealthyRwctxIdx(rwctxs)
+		if !reflect.DeepEqual(healthyIdx, gotHealthyIdx) {
+			t.Errorf("calculateHealthyRwctxIdx want healthyIdx = %v, got %v", healthyIdx, gotHealthyIdx)
+		}
+		if !reflect.DeepEqual(unhealthyIdx, gotUnhealthyIdx) {
+			t.Errorf("calculateHealthyRwctxIdx want unhealthyIdx = %v, got %v", unhealthyIdx, gotUnhealthyIdx)
+		}
+	}
+
+	f(5, []int{0, 1, 2, 3, 4}, nil)
+	f(5, []int{0, 1, 2, 4}, []int{3})
+	f(5, []int{2, 4}, []int{0, 1, 3})
+	f(5, []int{0, 2, 4}, []int{1, 3})
+	f(5, []int{}, []int{0, 1, 2, 3, 4})
+	f(5, []int{4}, []int{0, 1, 2, 3})
+	f(1, []int{0}, nil)
+	f(1, []int{}, []int{0})
+}

app/vmagent/remotewrite/streamaggr.go

@@ -56,7 +56,7 @@ var (
 		"See https://docs.victoriametrics.com/stream-aggregation/#ignoring-old-samples")
 	streamAggrIgnoreFirstIntervals = flagutil.NewArrayInt("remoteWrite.streamAggr.ignoreFirstIntervals", 0, "Number of aggregation intervals to skip after the start "+
 		"for the corresponding -remoteWrite.streamAggr.config at the corresponding -remoteWrite.url. Increase this value if "+
-		"you observe incorrect aggregation results after vmagent restarts. It could be caused by receiving bufferred delayed data from clients pushing data into the vmagent. "+
+		"you observe incorrect aggregation results after vmagent restarts. It could be caused by receiving buffered delayed data from clients pushing data into the vmagent. "+
 		"See https://docs.victoriametrics.com/stream-aggregation/#ignore-aggregation-intervals-on-start")
 	streamAggrDropInputLabels = flagutil.NewArrayString("remoteWrite.streamAggr.dropInputLabels", "An optional list of labels to drop from samples "+
 		"before stream de-duplication and aggregation with -remoteWrite.streamAggr.config and -remoteWrite.streamAggr.dedupInterval at the corresponding -remoteWrite.url. "+
@@ -131,7 +131,7 @@ func reloadStreamAggrConfigGlobal() {
 func initStreamAggrConfigGlobal() {
 	sas, err := newStreamAggrConfigGlobal()
 	if err != nil {
-		logger.Fatalf("cannot initialize gloabl stream aggregators: %s", err)
+		logger.Fatalf("cannot initialize global stream aggregators: %s", err)
 	}
 	if sas != nil {
 		filePath := sas.FilePath()

app/vmagent/vmimport/request_handler.go

@@ -9,8 +9,8 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	parserCommon "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/common"
-	parser "github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/vmimport"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/vmimport"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/vmimport/stream"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/tenantmetrics"
 	"github.com/VictoriaMetrics/metrics"
@@ -26,17 +26,17 @@ var (
 //
 // See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/6
 func InsertHandler(at *auth.Token, req *http.Request) error {
-	extraLabels, err := parserCommon.GetExtraLabels(req)
+	extraLabels, err := protoparserutil.GetExtraLabels(req)
 	if err != nil {
 		return err
 	}
-	isGzipped := req.Header.Get("Content-Encoding") == "gzip"
-	return stream.Parse(req.Body, isGzipped, func(rows []parser.Row) error {
+	encoding := req.Header.Get("Content-Encoding")
+	return stream.Parse(req.Body, encoding, func(rows []vmimport.Row) error {
 		return insertRows(at, rows, extraLabels)
 	})
 }
 
-func insertRows(at *auth.Token, rows []parser.Row, extraLabels []prompbmarshal.Label) error {
+func insertRows(at *auth.Token, rows []vmimport.Row, extraLabels []prompbmarshal.Label) error {
 	ctx := common.GetPushCtx()
 	defer common.PutPushCtx(ctx)
 

app/vmalert-tool/README.md

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

app/vmalert-tool/deployment/Dockerfile

@@ -0,0 +1,8 @@
+ARG base_image=non-existing
+FROM $base_image
+
+EXPOSE 8880
+
+ENTRYPOINT ["/vmalert-tool-prod"]
+ARG src_binary=non-existing
+COPY $src_binary ./vmalert-tool-prod

app/vmalert-tool/unittest/alerting.go

@@ -1,15 +1,15 @@
 package unittest
 
 import (
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 )
 
 // alertTestCase holds alert_rule_test cases defined in test file
 type alertTestCase struct {
-	EvalTime  *promutils.Duration `yaml:"eval_time"`
-	GroupName string              `yaml:"groupname"`
-	Alertname string              `yaml:"alertname"`
-	ExpAlerts []expAlert          `yaml:"exp_alerts"`
+	EvalTime  *promutil.Duration `yaml:"eval_time"`
+	GroupName string             `yaml:"groupname"`
+	Alertname string             `yaml:"alertname"`
+	ExpAlerts []expAlert         `yaml:"exp_alerts"`
 }
 
 // expAlert holds exp_alerts defined in test file

app/vmalert-tool/unittest/input.go

@@ -14,7 +14,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/decimal"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 	"github.com/VictoriaMetrics/metricsql"
 )
 
@@ -41,7 +41,7 @@ func httpWrite(address string, r io.Reader) {
 }
 
 // writeInputSeries send input series to vmstorage and flush them
-func writeInputSeries(input []series, interval *promutils.Duration, startStamp time.Time, dst string) error {
+func writeInputSeries(input []series, interval *promutil.Duration, startStamp time.Time, dst string) error {
 	r := testutil.WriteRequest{}
 	var err error
 	r.Timeseries, err = parseInputSeries(input, interval, startStamp)
@@ -56,7 +56,7 @@ func writeInputSeries(input []series, interval *promutils.Duration, startStamp t
 	return nil
 }
 
-func parseInputSeries(input []series, interval *promutils.Duration, startStamp time.Time) ([]testutil.TimeSeries, error) {
+func parseInputSeries(input []series, interval *promutil.Duration, startStamp time.Time) ([]testutil.TimeSeries, error) {
 	var res []testutil.TimeSeries
 	for _, data := range input {
 		expr, err := metricsql.Parse(data.Series)

app/vmalert-tool/unittest/input_test.go

@@ -5,7 +5,7 @@ import (
 	"time"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/decimal"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 )
 
 func TestParseInputValue_Failure(t *testing.T) {
@@ -70,7 +70,7 @@ func TestParseInputValue_Success(t *testing.T) {
 func TestParseInputSeries_Success(t *testing.T) {
 	f := func(input []series) {
 		t.Helper()
-		var interval promutils.Duration
+		var interval promutil.Duration
 		_, err := parseInputSeries(input, &interval, time.Now())
 		if err != nil {
 			t.Fatalf("expect to see no error: %v", err)
@@ -86,7 +86,7 @@ func TestParseInputSeries_Success(t *testing.T) {
 func TestParseInputSeries_Fail(t *testing.T) {
 	f := func(input []series) {
 		t.Helper()
-		var interval promutils.Duration
+		var interval promutil.Duration
 		_, err := parseInputSeries(input, &interval, time.Now())
 		if err == nil {
 			t.Fatalf("expect to see error: %v", err)

app/vmalert-tool/unittest/recording.go

@@ -10,15 +10,15 @@ import (
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/datasource"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 	"github.com/VictoriaMetrics/metricsql"
 )
 
 // metricsqlTestCase holds metricsql_expr_test cases defined in test file
 type metricsqlTestCase struct {
-	Expr       string              `yaml:"expr"`
-	EvalTime   *promutils.Duration `yaml:"eval_time"`
-	ExpSamples []expSample         `yaml:"exp_samples"`
+	Expr       string             `yaml:"expr"`
+	EvalTime   *promutil.Duration `yaml:"eval_time"`
+	ExpSamples []expSample        `yaml:"exp_samples"`
 }
 
 type expSample struct {
@@ -95,7 +95,7 @@ Outer:
 	return
 }
 
-func durationToTime(pd *promutils.Duration) time.Time {
+func durationToTime(pd *promutil.Duration) time.Time {
 	if pd == nil {
 		return time.Time{}
 	}

app/vmalert-tool/unittest/unittest.go

@@ -36,7 +36,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 	"github.com/VictoriaMetrics/metrics"
 )
 
@@ -182,7 +182,7 @@ func ruleUnitTest(filename string, content []byte, externalLabels map[string]str
 
 	if unitTestInp.EvaluationInterval.Duration() == 0 {
 		fmt.Println("evaluation_interval set to 1m by default")
-		unitTestInp.EvaluationInterval = &promutils.Duration{D: 1 * time.Minute}
+		unitTestInp.EvaluationInterval = &promutil.Duration{D: 1 * time.Minute}
 	}
 
 	groupOrderMap := make(map[string]int)
@@ -312,7 +312,7 @@ func (tg *testGroup) test(evalInterval time.Duration, groupOrderMap map[string]i
 	defer tearDown()
 
 	if tg.Interval == nil {
-		tg.Interval = promutils.NewDuration(evalInterval)
+		tg.Interval = promutil.NewDuration(evalInterval)
 	}
 	err := writeInputSeries(tg.InputSeries, tg.Interval, testStartTime, fmt.Sprintf("http://127.0.0.1:%s/api/v1/write", httpListenAddr))
 	if err != nil {
@@ -472,15 +472,15 @@ func (tg *testGroup) test(evalInterval time.Duration, groupOrderMap map[string]i
 
 // unitTestFile holds the contents of a single unit test file
 type unitTestFile struct {
-	RuleFiles          []string            `yaml:"rule_files"`
-	EvaluationInterval *promutils.Duration `yaml:"evaluation_interval"`
-	GroupEvalOrder     []string            `yaml:"group_eval_order"`
-	Tests              []testGroup         `yaml:"tests"`
+	RuleFiles          []string           `yaml:"rule_files"`
+	EvaluationInterval *promutil.Duration `yaml:"evaluation_interval"`
+	GroupEvalOrder     []string           `yaml:"group_eval_order"`
+	Tests              []testGroup        `yaml:"tests"`
 }
 
 // testGroup is a group of input series and test cases associated with it
 type testGroup struct {
-	Interval           *promutils.Duration `yaml:"interval"`
+	Interval           *promutil.Duration  `yaml:"interval"`
 	InputSeries        []series            `yaml:"input_series"`
 	AlertRuleTests     []alertTestCase     `yaml:"alert_rule_test"`
 	MetricsqlExprTests []metricsqlTestCase `yaml:"metricsql_expr_test"`

app/vmalert/README.md

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

app/vmalert/config/config.go

@@ -2,7 +2,6 @@ package config
 
 import (
 	"bytes"
-	"crypto/md5"
 	"flag"
 	"fmt"
 	"hash/fnv"
@@ -11,11 +10,12 @@ import (
 	"sort"
 	"strings"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/config/log"
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/utils"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/envtemplate"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
 	"gopkg.in/yaml.v2"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/config/log"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/vmalertutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/envtemplate"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 )
 
 var (
@@ -27,15 +27,15 @@ var (
 type Group struct {
 	Type       Type `yaml:"type,omitempty"`
 	File       string
-	Name       string              `yaml:"name"`
-	Interval   *promutils.Duration `yaml:"interval,omitempty"`
-	EvalOffset *promutils.Duration `yaml:"eval_offset,omitempty"`
+	Name       string             `yaml:"name"`
+	Interval   *promutil.Duration `yaml:"interval,omitempty"`
+	EvalOffset *promutil.Duration `yaml:"eval_offset,omitempty"`
 	// EvalDelay will adjust the `time` parameter of rule evaluation requests to compensate intentional query delay from datasource.
 	// see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5155
-	EvalDelay   *promutils.Duration `yaml:"eval_delay,omitempty"`
-	Limit       int                 `yaml:"limit,omitempty"`
-	Rules       []Rule              `yaml:"rules"`
-	Concurrency int                 `yaml:"concurrency"`
+	EvalDelay   *promutil.Duration `yaml:"eval_delay,omitempty"`
+	Limit       int                `yaml:"limit,omitempty"`
+	Rules       []Rule             `yaml:"rules"`
+	Concurrency int                `yaml:"concurrency"`
 	// Labels is a set of label value pairs, that will be added to every rule.
 	// It has priority over the external labels.
 	Labels map[string]string `yaml:"labels"`
@@ -67,7 +67,7 @@ func (g *Group) UnmarshalYAML(unmarshal func(any) error) error {
 	if g.Type.Get() == "" {
 		g.Type = NewRawType(*defaultRuleType)
 	}
-	h := md5.New()
+	h := fnv.New64a()
 	h.Write(b)
 	g.Checksum = fmt.Sprintf("%x", h.Sum(nil))
 	return nil
@@ -88,6 +88,9 @@ func (g *Group) Validate(validateTplFn ValidateTplFn, validateExpressions bool)
 	if g.EvalOffset.Duration() > g.Interval.Duration() {
 		return fmt.Errorf("eval_offset should be smaller than interval; now eval_offset: %v, interval: %v", g.EvalOffset.Duration(), g.Interval.Duration())
 	}
+	if g.EvalOffset != nil && g.EvalDelay != nil {
+		return fmt.Errorf("eval_offset cannot be used with eval_delay")
+	}
 	if g.Limit < 0 {
 		return fmt.Errorf("invalid limit %d, shouldn't be less than 0", g.Limit)
 	}
@@ -132,15 +135,15 @@ func (g *Group) Validate(validateTplFn ValidateTplFn, validateExpressions bool)
 // recording rule or alerting rule.
 type Rule struct {
 	ID     uint64
-	Record string              `yaml:"record,omitempty"`
-	Alert  string              `yaml:"alert,omitempty"`
-	Expr   string              `yaml:"expr"`
-	For    *promutils.Duration `yaml:"for,omitempty"`
+	Record string             `yaml:"record,omitempty"`
+	Alert  string             `yaml:"alert,omitempty"`
+	Expr   string             `yaml:"expr"`
+	For    *promutil.Duration `yaml:"for,omitempty"`
 	// Alert will continue firing for this long even when the alerting expression no longer has results.
-	KeepFiringFor *promutils.Duration `yaml:"keep_firing_for,omitempty"`
-	Labels        map[string]string   `yaml:"labels,omitempty"`
-	Annotations   map[string]string   `yaml:"annotations,omitempty"`
-	Debug         bool                `yaml:"debug,omitempty"`
+	KeepFiringFor *promutil.Duration `yaml:"keep_firing_for,omitempty"`
+	Labels        map[string]string  `yaml:"labels,omitempty"`
+	Annotations   map[string]string  `yaml:"annotations,omitempty"`
+	Debug         bool               `yaml:"debug,omitempty"`
 	// UpdateEntriesLimit defines max number of rule's state updates stored in memory.
 	// Overrides `-rule.updateEntriesLimit`.
 	UpdateEntriesLimit *int `yaml:"update_entries_limit,omitempty"`
@@ -262,7 +265,7 @@ func Parse(pathPatterns []string, validateTplFn ValidateTplFn, validateExpressio
 }
 
 func parse(files map[string][]byte, validateTplFn ValidateTplFn, validateExpressions bool) ([]Group, error) {
-	errGroup := new(utils.ErrGroup)
+	errGroup := new(vmalertutil.ErrGroup)
 	var groups []Group
 	for file, data := range files {
 		uniqueGroups := map[string]struct{}{}

app/vmalert/config/config_test.go

@@ -11,7 +11,7 @@ import (
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/notifier"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/templates"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 	"gopkg.in/yaml.v2"
 )
 
@@ -162,7 +162,7 @@ func TestGroupValidate_Failure(t *testing.T) {
 		Rules: []Rule{
 			{
 				Expr: "sum(up == 0 ) by (host)",
-				For:  promutils.NewDuration(10 * time.Millisecond),
+				For:  promutil.NewDuration(10 * time.Millisecond),
 			},
 			{
 				Expr: "sumSeries(time('foo.bar',10))",
@@ -172,13 +172,13 @@ func TestGroupValidate_Failure(t *testing.T) {
 
 	f(&Group{
 		Name:     "negative interval",
-		Interval: promutils.NewDuration(-1),
+		Interval: promutil.NewDuration(-1),
 	}, false, "interval shouldn't be lower than 0")
 
 	f(&Group{
 		Name:       "wrong eval_offset",
-		Interval:   promutils.NewDuration(time.Minute),
-		EvalOffset: promutils.NewDuration(2 * time.Minute),
+		Interval:   promutil.NewDuration(time.Minute),
+		EvalOffset: promutil.NewDuration(2 * time.Minute),
 	}, false, "eval_offset should be smaller than interval")
 
 	f(&Group{
@@ -309,7 +309,7 @@ func TestGroupValidate_Failure(t *testing.T) {
 				Record: "r1",
 				ID:     1,
 				Expr:   "sumSeries(time('foo.bar',10))",
-				For:    promutils.NewDuration(10 * time.Millisecond),
+				For:    promutil.NewDuration(10 * time.Millisecond),
 			},
 			{
 				Record: "r2",
@@ -326,7 +326,7 @@ func TestGroupValidate_Failure(t *testing.T) {
 			{
 				Record: "r1",
 				Expr:   "sum(up == 0 ) by (host)",
-				For:    promutils.NewDuration(10 * time.Millisecond),
+				For:    promutil.NewDuration(10 * time.Millisecond),
 			},
 		},
 	}, true, "bad LogsQL expr")
@@ -338,7 +338,7 @@ func TestGroupValidate_Failure(t *testing.T) {
 			{
 				Record: "r1",
 				Expr:   "* | stats by (path) count()",
-				For:    promutils.NewDuration(10 * time.Millisecond),
+				For:    promutil.NewDuration(10 * time.Millisecond),
 			},
 		},
 	}, true, "bad prometheus expr")
@@ -488,7 +488,7 @@ func TestHashRule_Equal(t *testing.T) {
 	f(Rule{Alert: "record", Expr: "up == 1"}, Rule{Alert: "record", Expr: "up == 1"})
 
 	f(Rule{
-		Alert: "alert", Expr: "up == 1", For: promutils.NewDuration(time.Minute), KeepFiringFor: promutils.NewDuration(time.Minute),
+		Alert: "alert", Expr: "up == 1", For: promutil.NewDuration(time.Minute), KeepFiringFor: promutil.NewDuration(time.Minute),
 	}, Rule{Alert: "alert", Expr: "up == 1"})
 }
 

app/vmalert/config/testdata/dir/rules0-bad.rules

@@ -9,7 +9,7 @@ groups:
         annotations:
           summary: "{{ }}"
           description: "{{$labels}}"
-      - alert: UnkownAnnotationsFunction
+      - alert: UnknownAnnotationsFunction
         for: 5m
         expr: vm_rows > 0
         labels:

app/vmalert/config/testdata/dir/rules2-bad.rules

@@ -1,7 +1,7 @@
 groups:
   - name: group
     rules:
-      - alert: UnkownLabelFunction
+      - alert: UnknownLabelFunction
         for: 5m
         expr: vm_rows > 0
         labels:

app/vmalert/datasource/client_prom.go

@@ -35,6 +35,8 @@ type promResponse struct {
 	Stats struct {
 		SeriesFetched *string `json:"seriesFetched,omitempty"`
 	} `json:"stats,omitempty"`
+	// IsPartial supported by VictoriaMetrics
+	IsPartial *bool `json:"isPartial,omitempty"`
 }
 
 // see https://prometheus.io/docs/prometheus/latest/querying/api/#instant-queries
@@ -209,7 +211,7 @@ func parsePrometheusResponse(req *http.Request, resp *http.Response) (res Result
 	if err != nil {
 		return res, err
 	}
-	res = Result{Data: ms}
+	res = Result{Data: ms, IsPartial: r.IsPartial}
 	if r.Stats.SeriesFetched != nil {
 		intV, err := strconv.Atoi(*r.Stats.SeriesFetched)
 		if err != nil {

app/vmalert/datasource/client_test.go

@@ -12,7 +12,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/utils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/vmalertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
 )
@@ -72,12 +72,14 @@ func TestVMInstantQuery(t *testing.T) {
 			w.Write([]byte(`{"status":"success","data":{"resultType":"scalar","result":[1583786142, "1"]}}`))
 		case 7:
 			w.Write([]byte(`{"status":"success","data":{"resultType":"scalar","result":[1583786142, "1"]},"stats":{"seriesFetched": "42"}}`))
+		case 8:
+			w.Write([]byte(`{"status":"success", "isPartial":true, "data":{"resultType":"scalar","result":[1583786142, "1"]}}`))
 		}
 	})
 	mux.HandleFunc("/render", func(w http.ResponseWriter, _ *http.Request) {
 		c++
 		switch c {
-		case 8:
+		case 9:
 			w.Write([]byte(`[{"target":"constantLine(10)","tags":{"name":"constantLine(10)"},"datapoints":[[10,1611758343],[10,1611758373],[10,1611758403]]}]`))
 		}
 	})
@@ -100,9 +102,9 @@ func TestVMInstantQuery(t *testing.T) {
 			t.Fatalf("failed to parse 'time' query param %q: %s", timeParam, err)
 		}
 		switch c {
-		case 9:
-			w.Write([]byte("[]"))
 		case 10:
+			w.Write([]byte("[]"))
+		case 11:
 			w.Write([]byte(`{"status":"success","data":{"resultType":"vector","result":[{"metric":{"__name__":"total","foo":"bar"},"value":[1583786142,"13763"]},{"metric":{"__name__":"total","foo":"baz"},"value":[1583786140,"2000"]}]}}`))
 		}
 	})
@@ -203,10 +205,18 @@ func TestVMInstantQuery(t *testing.T) {
 			*res.SeriesFetched)
 	}
 
+	res, _, err = pq.Query(ctx, vmQuery, ts) // 8
+	if err != nil {
+		t.Fatalf("unexpected %s", err)
+	}
+	if res.IsPartial != nil && !*res.IsPartial {
+		t.Fatalf("unexpected metric isPartial want %+v", true)
+	}
+
 	// test graphite
 	gq := s.BuildWithParams(QuerierParams{DataSourceType: string(datasourceGraphite)})
 
-	res, _, err = gq.Query(ctx, queryRender, ts) // 8 - graphite
+	res, _, err = gq.Query(ctx, queryRender, ts) // 9 - graphite
 	if err != nil {
 		t.Fatalf("unexpected %s", err)
 	}
@@ -226,9 +236,9 @@ func TestVMInstantQuery(t *testing.T) {
 	vlogs := datasourceVLogs
 	pq = s.BuildWithParams(QuerierParams{DataSourceType: string(vlogs), EvaluationInterval: 15 * time.Second})
 
-	expErr(vlogsQuery, "error parsing response") // 9
+	expErr(vlogsQuery, "error parsing response") // 10
 
-	res, _, err = pq.Query(ctx, vlogsQuery, ts) // 10
+	res, _, err = pq.Query(ctx, vlogsQuery, ts) // 11
 	if err != nil {
 		t.Fatalf("unexpected %s", err)
 	}
@@ -753,7 +763,7 @@ func TestHeaders(t *testing.T) {
 
 	// basic auth
 	f(func() *Client {
-		cfg, err := utils.AuthConfig(utils.WithBasicAuth("foo", "bar", ""))
+		cfg, err := vmalertutil.AuthConfig(vmalertutil.WithBasicAuth("foo", "bar", ""))
 		if err != nil {
 			t.Fatalf("Error get auth config: %s", err)
 		}
@@ -766,7 +776,7 @@ func TestHeaders(t *testing.T) {
 
 	// bearer auth
 	f(func() *Client {
-		cfg, err := utils.AuthConfig(utils.WithBearer("foo", ""))
+		cfg, err := vmalertutil.AuthConfig(vmalertutil.WithBearer("foo", ""))
 		if err != nil {
 			t.Fatalf("Error get auth config: %s", err)
 		}
@@ -798,7 +808,7 @@ func TestHeaders(t *testing.T) {
 
 	// custom header overrides basic auth
 	f(func() *Client {
-		cfg, err := utils.AuthConfig(utils.WithBasicAuth("foo", "bar", ""))
+		cfg, err := vmalertutil.AuthConfig(vmalertutil.WithBasicAuth("foo", "bar", ""))
 		if err != nil {
 			t.Fatalf("Error get auth config: %s", err)
 		}

app/vmalert/datasource/datasource.go

@@ -34,6 +34,9 @@ type Result struct {
 	// If nil, then this feature is not supported by the datasource.
 	// SeriesFetched is supported by VictoriaMetrics since v1.90.
 	SeriesFetched *int
+	// IsPartial is used by VictoriaMetrics to indicate
+	// whether response data is partial.
+	IsPartial *bool
 }
 
 // QuerierBuilder builds Querier with given params.

app/vmalert/datasource/faker.go

@@ -10,8 +10,9 @@ import (
 // FakeQuerier is a mock querier that return predefined results and error message
 type FakeQuerier struct {
 	sync.Mutex
-	metrics []Metric
-	err     error
+	metrics   []Metric
+	err       error
+	isPartial *bool
 }
 
 // SetErr sets query error message
@@ -21,11 +22,19 @@ func (fq *FakeQuerier) SetErr(err error) {
 	fq.Unlock()
 }
 
+// SetPartialResponse marks query response as partial
+func (fq *FakeQuerier) SetPartialResponse(partial bool) {
+	fq.Lock()
+	fq.isPartial = &partial
+	fq.Unlock()
+}
+
 // Reset reset querier's error message and results
 func (fq *FakeQuerier) Reset() {
 	fq.Lock()
 	fq.err = nil
 	fq.metrics = fq.metrics[:0]
+	fq.isPartial = nil
 	fq.Unlock()
 }
 
@@ -57,7 +66,7 @@ func (fq *FakeQuerier) Query(_ context.Context, _ string, _ time.Time) (Result,
 	cp := make([]Metric, len(fq.metrics))
 	copy(cp, fq.metrics)
 	req, _ := http.NewRequest(http.MethodPost, "foo.com", nil)
-	return Result{Data: cp}, req, nil
+	return Result{Data: cp, IsPartial: fq.isPartial}, req, nil
 }
 
 // FakeQuerierWithRegistry can store different results for different query expr

app/vmalert/datasource/init.go

@@ -8,10 +8,10 @@ import (
 	"strings"
 	"time"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/utils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/vmalertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputils"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/netutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 )
 
 var (
@@ -79,15 +79,13 @@ type Param struct {
 // Provided extraParams will be added as GET params for
 // each request.
 func Init(extraParams url.Values) (QuerierBuilder, error) {
-	if *addr == "" {
-		return nil, fmt.Errorf("datasource.url is empty")
+	if err := httputil.CheckURL(*addr); err != nil {
+		return nil, fmt.Errorf("invalid -datasource.url: %w", err)
 	}
-
-	tr, err := httputils.Transport(*addr, *tlsCertFile, *tlsKeyFile, *tlsCAFile, *tlsServerName, *tlsInsecureSkipVerify)
+	tr, err := promauth.NewTLSTransport(*tlsCertFile, *tlsKeyFile, *tlsCAFile, *tlsServerName, *tlsInsecureSkipVerify, "vmalert_datasource")
 	if err != nil {
 		return nil, fmt.Errorf("failed to create transport for -datasource.url=%q: %w", *addr, err)
 	}
-	tr.DialContext = netutil.NewStatDialFunc("vmalert_datasource")
 	tr.DisableKeepAlives = *disableKeepAlive
 	tr.MaxIdleConnsPerHost = *maxIdleConnections
 	if tr.MaxIdleConns != 0 && tr.MaxIdleConns < tr.MaxIdleConnsPerHost {
@@ -106,11 +104,11 @@ func Init(extraParams url.Values) (QuerierBuilder, error) {
 	if err != nil {
 		return nil, fmt.Errorf("cannot parse JSON for -datasource.oauth2.endpointParams=%s: %w", *oauth2EndpointParams, err)
 	}
-	authCfg, err := utils.AuthConfig(
-		utils.WithBasicAuth(*basicAuthUsername, *basicAuthPassword, *basicAuthPasswordFile),
-		utils.WithBearer(*bearerToken, *bearerTokenFile),
-		utils.WithOAuth(*oauth2ClientID, *oauth2ClientSecret, *oauth2ClientSecretFile, *oauth2TokenURL, *oauth2Scopes, endpointParams),
-		utils.WithHeaders(*headers))
+	authCfg, err := vmalertutil.AuthConfig(
+		vmalertutil.WithBasicAuth(*basicAuthUsername, *basicAuthPassword, *basicAuthPasswordFile),
+		vmalertutil.WithBearer(*bearerToken, *bearerTokenFile),
+		vmalertutil.WithOAuth(*oauth2ClientID, *oauth2ClientSecret, *oauth2ClientSecretFile, *oauth2TokenURL, *oauth2Scopes, endpointParams),
+		vmalertutil.WithHeaders(*headers))
 	if err != nil {
 		return nil, fmt.Errorf("failed to configure auth: %w", err)
 	}

app/vmalert/main_test.go

@@ -169,7 +169,6 @@ groups:
 	checkCfg(nil)
 	groupsLen = lenLocked(m)
 	if groupsLen != 2 {
-		fmt.Println(m.groups)
 		t.Fatalf("expected to have exactly 2 groups loaded; got %d", groupsLen)
 	}
 

app/vmalert/manager.go

@@ -83,7 +83,8 @@ func (m *manager) close() {
 
 func (m *manager) startGroup(ctx context.Context, g *rule.Group, restore bool) error {
 	m.wg.Add(1)
-	id := g.ID()
+	id := g.GetID()
+	g.Init()
 	go func() {
 		defer m.wg.Done()
 		if restore {
@@ -112,7 +113,7 @@ func (m *manager) update(ctx context.Context, groupsCfg []config.Group, restore
 			}
 		}
 		ng := rule.NewGroup(cfg, m.querierBuilder, *evaluationInterval, m.labels)
-		groupsRegistry[ng.ID()] = ng
+		groupsRegistry[ng.GetID()] = ng
 	}
 
 	if rrPresent && m.rw == nil {
@@ -130,17 +131,17 @@ func (m *manager) update(ctx context.Context, groupsCfg []config.Group, restore
 
 	m.groupsMu.Lock()
 	for _, og := range m.groups {
-		ng, ok := groupsRegistry[og.ID()]
+		ng, ok := groupsRegistry[og.GetID()]
 		if !ok {
 			// old group is not present in new list,
 			// so must be stopped and deleted
 			og.Close()
-			delete(m.groups, og.ID())
+			delete(m.groups, og.GetID())
 			og = nil
 			continue
 		}
-		delete(groupsRegistry, ng.ID())
-		if og.Checksum != ng.Checksum {
+		delete(groupsRegistry, ng.GetID())
+		if og.GetCheckSum() != ng.GetCheckSum() {
 			toUpdate = append(toUpdate, updateItem{old: og, new: ng})
 		}
 	}

app/vmalert/manager_test.go

@@ -144,7 +144,7 @@ func TestManagerUpdate_Success(t *testing.T) {
 		}
 
 		for _, wantG := range groupsExpected {
-			gotG, ok := m.groups[wantG.ID()]
+			gotG, ok := m.groups[wantG.CreateID()]
 			if !ok {
 				t.Fatalf("expected to have group %q", wantG.Name)
 			}
@@ -258,7 +258,7 @@ func compareGroups(t *testing.T, a, b *rule.Group) {
 	}
 	for i, r := range a.Rules {
 		got, want := r, b.Rules[i]
-		if a.ID() != b.ID() {
+		if a.CreateID() != b.CreateID() {
 			t.Fatalf("expected to have rule %q; got %q", want.ID(), got.ID())
 		}
 		if err := rule.CompareRules(t, want, got); err != nil {

app/vmalert/notifier/alert.go

@@ -9,7 +9,7 @@ import (
 	"time"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/templates"
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/utils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/vmalertutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
 )
@@ -141,7 +141,7 @@ func ValidateTemplates(annotations map[string]string) error {
 func templateAnnotations(annotations map[string]string, data AlertTplData, tmpl *textTpl.Template, execute bool) (map[string]string, error) {
 	var builder strings.Builder
 	var buf bytes.Buffer
-	eg := new(utils.ErrGroup)
+	eg := new(vmalertutil.ErrGroup)
 	r := make(map[string]string, len(annotations))
 	tData := tplData{data, externalLabels, externalURL}
 	header := strings.Join(tplHeaders, "")

app/vmalert/notifier/alertmanager.go

@@ -12,8 +12,8 @@ import (
 
 	"github.com/VictoriaMetrics/metrics"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/utils"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/vmalertutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
@@ -37,8 +37,9 @@ type AlertManager struct {
 type notifierMetrics struct {
 	set *metrics.Set
 
-	alertsSent       *metrics.Counter
-	alertsSendErrors *metrics.Counter
+	alertsSent         *metrics.Counter
+	alertsSendErrors   *metrics.Counter
+	alertsSendDuration *metrics.Histogram
 }
 
 func newNotifierMetrics(addr string) *notifierMetrics {
@@ -46,9 +47,10 @@ func newNotifierMetrics(addr string) *notifierMetrics {
 	metrics.RegisterSet(set)
 
 	return &notifierMetrics{
-		set:              set,
-		alertsSent:       set.GetOrCreateCounter(fmt.Sprintf("vmalert_alerts_sent_total{addr=%q}", addr)),
-		alertsSendErrors: set.GetOrCreateCounter(fmt.Sprintf("vmalert_alerts_send_errors_total{addr=%q}", addr)),
+		set:                set,
+		alertsSent:         set.NewCounter(fmt.Sprintf("vmalert_alerts_sent_total{addr=%q}", addr)),
+		alertsSendErrors:   set.NewCounter(fmt.Sprintf("vmalert_alerts_send_errors_total{addr=%q}", addr)),
+		alertsSendDuration: set.NewHistogram(fmt.Sprintf("vmalert_alerts_send_duration_seconds{addr=%q}", addr)),
 	}
 }
 
@@ -72,7 +74,9 @@ func (am AlertManager) Addr() string {
 // Send an alert or resolve message
 func (am *AlertManager) Send(ctx context.Context, alerts []Alert, headers map[string]string) error {
 	am.metrics.alertsSent.Add(len(alerts))
+	startTime := time.Now()
 	err := am.send(ctx, alerts, headers)
+	am.metrics.alertsSendDuration.UpdateDuration(startTime)
 	if err != nil {
 		am.metrics.alertsSendErrors.Add(len(alerts))
 	}
@@ -148,11 +152,16 @@ const alertManagerPath = "/api/v2/alerts"
 func NewAlertManager(alertManagerURL string, fn AlertURLGenerator, authCfg promauth.HTTPClientConfig,
 	relabelCfg *promrelabel.ParsedConfigs, timeout time.Duration,
 ) (*AlertManager, error) {
+
+	if err := httputil.CheckURL(alertManagerURL); err != nil {
+		return nil, fmt.Errorf("invalid alertmanager URL: %w", err)
+	}
+
 	tls := &promauth.TLSConfig{}
 	if authCfg.TLSConfig != nil {
 		tls = authCfg.TLSConfig
 	}
-	tr, err := httputils.Transport(alertManagerURL, tls.CertFile, tls.KeyFile, tls.CAFile, tls.ServerName, tls.InsecureSkipVerify)
+	tr, err := promauth.NewTLSTransport(tls.CertFile, tls.KeyFile, tls.CAFile, tls.ServerName, tls.InsecureSkipVerify, "vmalert_notifier")
 	if err != nil {
 		return nil, fmt.Errorf("failed to create transport for alertmanager URL=%q: %w", alertManagerURL, err)
 
@@ -167,11 +176,11 @@ func NewAlertManager(alertManagerURL string, fn AlertURLGenerator, authCfg proma
 		oauth = authCfg.OAuth2
 	}
 
-	aCfg, err := utils.AuthConfig(
-		utils.WithBasicAuth(ba.Username, ba.Password.String(), ba.PasswordFile),
-		utils.WithBearer(authCfg.BearerToken.String(), authCfg.BearerTokenFile),
-		utils.WithOAuth(oauth.ClientID, oauth.ClientSecret.String(), oauth.ClientSecretFile, oauth.TokenURL, strings.Join(oauth.Scopes, ";"), oauth.EndpointParams),
-		utils.WithHeaders(strings.Join(authCfg.Headers, "^^")),
+	aCfg, err := vmalertutil.AuthConfig(
+		vmalertutil.WithBasicAuth(ba.Username, ba.Password.String(), ba.PasswordFile),
+		vmalertutil.WithBearer(authCfg.BearerToken.String(), authCfg.BearerTokenFile),
+		vmalertutil.WithOAuth(oauth.ClientID, oauth.ClientSecret.String(), oauth.ClientSecretFile, oauth.TokenURL, strings.Join(oauth.Scopes, ";"), oauth.EndpointParams),
+		vmalertutil.WithHeaders(strings.Join(authCfg.Headers, "^^")),
 	)
 	if err != nil {
 		return nil, fmt.Errorf("failed to configure auth: %w", err)
@@ -189,8 +198,10 @@ func NewAlertManager(alertManagerURL string, fn AlertURLGenerator, authCfg proma
 		argFunc:        fn,
 		authCfg:        aCfg,
 		relabelConfigs: relabelCfg,
-		client:         &http.Client{Transport: tr},
-		timeout:        timeout,
-		metrics:        newNotifierMetrics(alertManagerURL),
+		client: &http.Client{
+			Transport: tr,
+		},
+		timeout: timeout,
+		metrics: newNotifierMetrics(alertManagerURL),
 	}, nil
 }

app/vmalert/notifier/config.go

@@ -1,8 +1,8 @@
 package notifier
 
 import (
-	"crypto/md5"
 	"fmt"
+	"hash/fnv"
 	"net/url"
 	"os"
 	"path"
@@ -16,7 +16,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promrelabel"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/consul"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/dns"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 )
 
 // Config contains list of supported configuration settings
@@ -44,7 +44,7 @@ type Config struct {
 	// AlertRelabelConfigs contains list of relabeling rules alert labels
 	AlertRelabelConfigs []promrelabel.RelabelConfig `yaml:"alert_relabel_configs,omitempty"`
 	// The timeout used when sending alerts.
-	Timeout *promutils.Duration `yaml:"timeout,omitempty"`
+	Timeout *promutil.Duration `yaml:"timeout,omitempty"`
 
 	// Checksum stores the hash of yaml definition for the config.
 	// May be used to detect any changes to the config file.
@@ -82,7 +82,7 @@ func (cfg *Config) UnmarshalYAML(unmarshal func(any) error) error {
 		cfg.Scheme = "http"
 	}
 	if cfg.Timeout.Duration() == 0 {
-		cfg.Timeout = promutils.NewDuration(time.Second * 10)
+		cfg.Timeout = promutil.NewDuration(time.Second * 10)
 	}
 	rCfg, err := promrelabel.ParseRelabelConfigs(cfg.RelabelConfigs)
 	if err != nil {
@@ -99,7 +99,7 @@ func (cfg *Config) UnmarshalYAML(unmarshal func(any) error) error {
 	if err != nil {
 		return fmt.Errorf("failed to marshal configuration for checksum: %w", err)
 	}
-	h := md5.New()
+	h := fnv.New64a()
 	h.Write(b)
 	cfg.Checksum = fmt.Sprintf("%x", h.Sum(nil))
 	return nil
@@ -130,7 +130,7 @@ func parseConfig(path string) (*Config, error) {
 	return cfg, nil
 }
 
-func parseLabels(target string, metaLabels *promutils.Labels, cfg *Config) (string, *promutils.Labels, error) {
+func parseLabels(target string, metaLabels *promutil.Labels, cfg *Config) (string, *promutil.Labels, error) {
 	labels := mergeLabels(target, metaLabels, cfg)
 	labels.Labels = cfg.parsedRelabelConfigs.Apply(labels.Labels, 0)
 	labels.RemoveMetaLabels()
@@ -176,9 +176,9 @@ func addMissingPort(scheme, target string) string {
 	return target
 }
 
-func mergeLabels(target string, metaLabels *promutils.Labels, cfg *Config) *promutils.Labels {
+func mergeLabels(target string, metaLabels *promutil.Labels, cfg *Config) *promutil.Labels {
 	// See https://prometheus.io/docs/prometheus/latest/configuration/configuration/#relabel_config
-	m := promutils.NewLabels(3 + metaLabels.Len())
+	m := promutil.NewLabels(3 + metaLabels.Len())
 	address := target
 	scheme := cfg.Scheme
 	alertsPath := path.Join("/", cfg.PathPrefix, alertManagerPath)

app/vmalert/notifier/config_watcher.go

@@ -9,7 +9,7 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/consul"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discovery/dns"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 )
 
 // configWatcher supports dynamic reload of Notifier objects
@@ -157,7 +157,7 @@ func targetsFromLabels(labelsFn getLabels, cfg *Config, genFn AlertURLGenerator)
 	return targets, errors
 }
 
-type getLabels func() ([]*promutils.Labels, error)
+type getLabels func() ([]*promutil.Labels, error)
 
 func (cw *configWatcher) start() error {
 	if len(cw.cfg.StaticConfigs) > 0 {
@@ -183,8 +183,8 @@ func (cw *configWatcher) start() error {
 	}
 
 	if len(cw.cfg.ConsulSDConfigs) > 0 {
-		err := cw.add(TargetConsul, *consul.SDCheckInterval, func() ([]*promutils.Labels, error) {
-			var labels []*promutils.Labels
+		err := cw.add(TargetConsul, *consul.SDCheckInterval, func() ([]*promutil.Labels, error) {
+			var labels []*promutil.Labels
 			for i := range cw.cfg.ConsulSDConfigs {
 				sdc := &cw.cfg.ConsulSDConfigs[i]
 				targetLabels, err := sdc.GetLabels(cw.cfg.baseDir)
@@ -201,8 +201,8 @@ func (cw *configWatcher) start() error {
 	}
 
 	if len(cw.cfg.DNSSDConfigs) > 0 {
-		err := cw.add(TargetDNS, *dns.SDCheckInterval, func() ([]*promutils.Labels, error) {
-			var labels []*promutils.Labels
+		err := cw.add(TargetDNS, *dns.SDCheckInterval, func() ([]*promutil.Labels, error) {
+			var labels []*promutil.Labels
 			for i := range cw.cfg.DNSSDConfigs {
 				sdc := &cw.cfg.DNSSDConfigs[i]
 				targetLabels, err := sdc.GetLabels(cw.cfg.baseDir)

app/vmalert/notifier/init.go

@@ -9,7 +9,7 @@ import (
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promauth"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 )
 
 var (
@@ -102,9 +102,9 @@ func Init(gen AlertURLGenerator, extLabels map[string]string, extURL string) (fu
 		if len(*addrs) > 0 || *configPath != "" {
 			return nil, fmt.Errorf("only one of -notifier.blackhole, -notifier.url and -notifier.config flags must be specified")
 		}
-
+		notifier := newBlackHoleNotifier()
 		staticNotifiersFn = func() []Notifier {
-			return []Notifier{newBlackHoleNotifier()}
+			return []Notifier{notifier}
 		}
 		return staticNotifiersFn, nil
 	}
@@ -189,7 +189,7 @@ func notifiersFromFlags(gen AlertURLGenerator) ([]Notifier, error) {
 // list of labels added during discovery.
 type Target struct {
 	Notifier
-	Labels *promutils.Labels
+	Labels *promutil.Labels
 }
 
 // TargetType defines how the Target was discovered