extract vmsingle-specific code into vmstorageSingleNode type

Signed-off-by: Artem Fetishev <rtm@victoriametrics.com>

Makefile

@@ -485,8 +485,8 @@ apptest-legacy: victoria-metrics-race vmbackup-race vmrestore-race
 		curl --output-dir /tmp -LO $${URL}/$${VMSINGLE} && tar xzf /tmp/$${VMSINGLE} -C $${DIR} && \
 		curl --output-dir /tmp -LO $${URL}/$${VMCLUSTER} && tar xzf /tmp/$${VMCLUSTER} -C $${DIR} \
 	); \
-	VM_LEGACY_VMSINGLE_PATH=$${DIR}/victoria-metrics-prod \
-	VM_LEGACY_VMSTORAGE_PATH=$${DIR}/vmstorage-prod \
+	VMSINGLE_V1_132_0_PATH=$${DIR}/victoria-metrics-prod \
+	VMSTORAGE_V1_132_0_PATH=$${DIR}/vmstorage-prod \
 	go test ./apptest/tests -run="^TestLegacySingle.*"
 
 benchmark:

app/victoria-metrics/main.go

@@ -22,7 +22,6 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/procutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/pushmetrics"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
 )
 
 var (
@@ -30,23 +29,26 @@ var (
 	useProxyProtocol = flagutil.NewArrayBool("httpListenAddr.useProxyProtocol", "Whether to use proxy protocol for connections accepted at the corresponding -httpListenAddr . "+
 		"See https://www.haproxy.org/download/1.8/doc/proxy-protocol.txt . "+
 		"With enabled proxy protocol http server cannot serve regular /metrics endpoint. Use -pushmetrics.url for metrics pushing")
-	minScrapeInterval = flag.Duration("dedup.minScrapeInterval", 0, "Leave only the last sample in every time series per each discrete interval "+
-		"equal to -dedup.minScrapeInterval > 0. See also -streamAggr.dedupInterval and https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#deduplication")
 	dryRun = flag.Bool("dryRun", false, "Whether to check config files without running VictoriaMetrics. The following config files are checked: "+
 		"-promscrape.config, -relabelConfig and -streamAggr.config. Unknown config entries aren't allowed in -promscrape.config by default. "+
 		"This can be changed with -promscrape.config.strictParse=false command-line flag")
-	inmemoryDataFlushInterval = flag.Duration("inmemoryDataFlushInterval", 5*time.Second, "The interval for guaranteed saving of in-memory data to disk. "+
-		"The saved data survives unclean shutdowns such as OOM crash, hardware reset, SIGKILL, etc. "+
-		"Bigger intervals may help increase the lifetime of flash storage with limited write cycles (e.g. Raspberry PI). "+
-		"Smaller intervals increase disk IO load. Minimum supported value is 1s")
 	maxIngestionRate = flag.Int("maxIngestionRate", 0, "The maximum number of samples vmsingle can receive per second. Data ingestion is paused when the limit is exceeded. "+
 		"By default there are no limits on samples ingestion rate.")
-	finalDedupScheduleInterval = flag.Duration("storage.finalDedupScheduleCheckInterval", time.Hour, "The interval for checking when final deduplication process should be started."+
-		"Storage unconditionally adds 25% jitter to the interval value on each check evaluation."+
-		" Changing the interval to the bigger values may delay downsampling, deduplication for historical data."+
-		" See also https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#deduplication")
+	maxConcurrentRequests = flag.Int("search.maxConcurrentRequests", getDefaultMaxConcurrentRequests(), "The maximum number of concurrent search requests. "+
+		"It shouldn't be high, since a single request can saturate all the CPU cores, while many concurrently executed requests may require high amounts of memory. "+
+		"See also -search.maxQueueDuration and -search.maxMemoryPerQuery")
+	maxQueueDuration = flag.Duration("search.maxQueueDuration", 10*time.Second, "The maximum time the request waits for execution when -search.maxConcurrentRequests "+
+		"limit is reached; see also -search.maxQueryDuration")
 )
 
+func getDefaultMaxConcurrentRequests() int {
+	// A single request can saturate all the CPU cores, so there is no sense
+	// in allowing higher number of concurrent requests - they will just contend
+	// for unavailable CPU time.
+	n := min(cgroup.AvailableCPUs()*2, 16)
+	return n
+}
+
 func main() {
 	// VictoriaMetrics is optimized for reduced memory allocations,
 	// so it can run with the reduced GOGC in order to reduce the used memory,
@@ -87,14 +89,8 @@ func main() {
 	}
 	logger.Infof("starting VictoriaMetrics at %q...", listenAddrs)
 	startTime := time.Now()
-	storage.SetDedupInterval(*minScrapeInterval)
-	storage.SetDataFlushInterval(*inmemoryDataFlushInterval)
-	if *finalDedupScheduleInterval < time.Hour {
-		logger.Fatalf("-dedup.finalDedupScheduleCheckInterval cannot be smaller than 1 hour; got %s", *finalDedupScheduleInterval)
-	}
-	storage.SetFinalDedupScheduleInterval(*finalDedupScheduleInterval)
-	vmstorage.Init(promql.ResetRollupResultCacheIfNeeded)
-	vmselect.Init()
+	vmstorage.Init(*maxConcurrentRequests, promql.ResetRollupResultCacheIfNeeded)
+	vmselect.Init(*maxConcurrentRequests, *maxQueueDuration)
 	vminsertcommon.StartIngestionRateLimiter(*maxIngestionRate)
 	vminsert.Init()
 

app/victoria-metrics/self_scraper.go

@@ -93,7 +93,7 @@ func selfScraper(scrapeInterval time.Duration) {
 				mr.Value = r.Value
 			}
 		}
-		if err := vmstorage.AddRows(mrs); err != nil {
+		if err := vmstorage.VMInsertAPI.WriteRows(mrs); err != nil {
 			logger.Errorf("cannot store self-scraped metrics: %s", err)
 		}
 		if len(metadataRows.Rows) > 0 {
@@ -105,7 +105,7 @@ func selfScraper(scrapeInterval time.Duration) {
 					Type:             mm.Type,
 				})
 			}
-			if err := vmstorage.AddMetadataRows(mms); err != nil {
+			if err := vmstorage.VMInsertAPI.WriteMetadata(mms); err != nil {
 				logger.Errorf("cannot store self-scraped metrics metadata: %s", err)
 			}
 		}

app/vmagent/main.go

@@ -118,6 +118,7 @@ func main() {
 	remotewrite.InitSecretFlags()
 	buildinfo.Init()
 	logger.Init()
+	opentelemetry.Init()
 	timeserieslimits.Init(*maxLabelsPerTimeseries, *maxLabelNameLen, *maxLabelValueLen)
 
 	if promscrape.IsDryRun() {

app/vmagent/opentelemetry/request_handler.go

@@ -25,6 +25,11 @@ var (
 	rowsPerInsert          = metrics.NewHistogram(`vmagent_rows_per_insert{type="opentelemetry"}`)
 )
 
+// Init must be called after flag.Parse and before using the opentelemetry package.
+func Init() {
+	stream.InitDecodeOptions()
+}
+
 // InsertHandlerForReader processes metrics from given reader.
 func InsertHandlerForReader(at *auth.Token, r io.Reader, encoding string) error {
 	return stream.ParseStream(r, encoding, nil, func(tss []prompb.TimeSeries, mms []prompb.MetricMetadata) error {

app/vmalert-tool/unittest/input.go

@@ -52,7 +52,7 @@ func writeInputSeries(input []series, interval *promutil.Duration, startStamp ti
 	data := testutil.Compress(r)
 	// write input series to vm
 	httpWrite(dst, bytes.NewBuffer(data))
-	vmstorage.Storage.DebugFlush()
+	vmstorage.DebugFlush()
 	return nil
 }
 

app/vmalert-tool/unittest/unittest.go

@@ -108,7 +108,9 @@ func UnitTest(files []string, disableGroupLabel bool, externalLabels []string, e
 	storagePath = tmpFolder
 	processFlags()
 	vminsert.Init()
-	vmselect.Init()
+	const maxConcurrentRequests = 4
+	maxQueueDuration := 5 * time.Second
+	vmselect.Init(maxConcurrentRequests, maxQueueDuration)
 	// storagePath will be created again when closing vmselect, so remove it again.
 	defer fs.MustRemoveDir(storagePath)
 	defer vminsert.Stop()
@@ -279,7 +281,8 @@ func processFlags() {
 }
 
 func setUp() {
-	vmstorage.Init(promql.ResetRollupResultCacheIfNeeded)
+	const maxConcurrentRequests = 4
+	vmstorage.Init(maxConcurrentRequests, promql.ResetRollupResultCacheIfNeeded)
 	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
 	defer cancel()
 	readyCheckFunc := func() bool {
@@ -384,7 +387,7 @@ func (tg *testGroup) test(evalInterval time.Duration, groupOrderMap map[string]i
 				}
 			}
 			// flush series after each group evaluation
-			vmstorage.Storage.DebugFlush()
+			vmstorage.DebugFlush()
 		}
 
 		// check alert_rule_test case at every eval time

app/vmalert/config/config.go

@@ -113,15 +113,15 @@ func (g *Group) Validate(validateTplFn ValidateTplFn, validateExpressions bool)
 			// because correct types must be inherited after unmarshalling.
 			exprValidator := g.Type.ValidateExpr
 			if err := exprValidator(r.Expr); err != nil {
-				return fmt.Errorf("invalid expression for rule  %q: %w", ruleName, err)
+				return fmt.Errorf("invalid expression for rule %q: %w", ruleName, err)
 			}
 		}
 		if validateTplFn != nil {
 			if err := validateTplFn(r.Annotations); err != nil {
-				return fmt.Errorf("invalid annotations for rule  %q: %w", ruleName, err)
+				return fmt.Errorf("invalid annotations for rule %q: %w", ruleName, err)
 			}
 			if err := validateTplFn(r.Labels); err != nil {
-				return fmt.Errorf("invalid labels for rule  %q: %w", ruleName, err)
+				return fmt.Errorf("invalid labels for rule %q: %w", ruleName, err)
 			}
 		}
 	}

app/vmalert/config/config_test.go

@@ -121,7 +121,7 @@ func TestParse_Failure(t *testing.T) {
 	f([]string{"testdata/dir/rules2-bad.rules"}, "function \"unknown\" not defined")
 	f([]string{"testdata/dir/rules3-bad.rules"}, "either `record` or `alert` must be set")
 	f([]string{"testdata/dir/rules4-bad.rules"}, "either `record` or `alert` must be set")
-	f([]string{"testdata/rules/rules1-bad.rules"}, "bad graphite expr")
+	f([]string{"testdata/rules/rules1-bad.rules"}, "bad GraphiteQL expr")
 	f([]string{"testdata/rules/vlog-rules0-bad.rules"}, "bad LogsQL expr")
 	f([]string{"testdata/dir/rules6-bad.rules"}, "missing ':' in header")
 	f([]string{"testdata/rules/rules-multi-doc-bad.rules"}, "unknown fields")
@@ -283,7 +283,7 @@ func TestGroupValidate_Failure(t *testing.T) {
 				Expr:   "up | 0",
 			},
 		},
-	}, true, "bad prometheus expr")
+	}, true, "bad MetricsQL expr")
 
 	f(&Group{
 		Name: "test graphite expr",
@@ -293,7 +293,7 @@ func TestGroupValidate_Failure(t *testing.T) {
 				"description": "some-description",
 			}},
 		},
-	}, true, "bad graphite expr")
+	}, true, "bad GraphiteQL expr")
 
 	f(&Group{
 		Name: "test vlogs expr",
@@ -327,7 +327,7 @@ func TestGroupValidate_Failure(t *testing.T) {
 				Expr:   "sum(up == 0 ) by (host)",
 			},
 		},
-	}, true, "bad graphite expr")
+	}, true, "bad GraphiteQL expr")
 
 	f(&Group{
 		Name: "test vlogs with prometheus exp",
@@ -351,7 +351,7 @@ func TestGroupValidate_Failure(t *testing.T) {
 				For:    promutil.NewDuration(10 * time.Millisecond),
 			},
 		},
-	}, true, "bad prometheus expr")
+	}, true, "bad MetricsQL expr")
 }
 
 func TestGroupValidate_Success(t *testing.T) {

app/vmalert/config/types.go

@@ -66,11 +66,11 @@ func (t *Type) ValidateExpr(expr string) error {
 	switch t.String() {
 	case "graphite":
 		if _, err := graphiteql.Parse(expr); err != nil {
-			return fmt.Errorf("bad graphite expr: %q, err: %w", expr, err)
+			return fmt.Errorf("bad GraphiteQL expr: %q, err: %w", expr, err)
 		}
 	case "prometheus":
 		if _, err := metricsql.Parse(expr); err != nil {
-			return fmt.Errorf("bad prometheus expr: %q, err: %w", expr, err)
+			return fmt.Errorf("bad MetricsQL expr: %q, err: %w", expr, err)
 		}
 	case "vlogs":
 		q, err := logstorage.ParseStatsQuery(expr, 0)

app/vminsert/common/insert_ctx.go

@@ -184,7 +184,7 @@ func (ctx *InsertCtx) WriteMetadata(mmpbs []prompb.MetricMetadata) error {
 	}
 	ctx.mms = mms
 
-	err := vmstorage.AddMetadataRows(mms)
+	err := vmstorage.VMInsertAPI.WriteMetadata(mms)
 	if err != nil {
 		return &httpserver.ErrorWithStatusCode{
 			Err:        fmt.Errorf("cannot store metrics metadata: %w", err),
@@ -209,7 +209,7 @@ func (ctx *InsertCtx) WritePromMetadata(mmps []prometheus.Metadata) error {
 	}
 	ctx.mms = mms
 
-	err := vmstorage.AddMetadataRows(mms)
+	err := vmstorage.VMInsertAPI.WriteMetadata(mms)
 	if err != nil {
 		return &httpserver.ErrorWithStatusCode{
 			Err:        fmt.Errorf("cannot store prometheus metrics metadata: %w", err),
@@ -278,7 +278,7 @@ func (ctx *InsertCtx) FlushBufs() error {
 	// since the number of concurrent FlushBufs() calls should be already limited via writeconcurrencylimiter
 	// used at every stream.Parse() call under lib/protoparser/*
 
-	err := vmstorage.AddRows(ctx.mrs)
+	err := vmstorage.VMInsertAPI.WriteRows(ctx.mrs)
 	ctx.Reset(0)
 	if err == nil {
 		return nil

app/vminsert/common/streamaggr.go

@@ -283,7 +283,7 @@ func pushAggregateSeries(tss []prompb.TimeSeries) {
 	}
 	// There is no need in limiting the number of concurrent calls to vmstorage.AddRows() here,
 	// since the number of concurrent pushAggregateSeries() calls should be already limited by lib/streamaggr.
-	if err := vmstorage.AddRows(ctx.mrs); err != nil {
+	if err := vmstorage.VMInsertAPI.WriteRows(ctx.mrs); err != nil {
 		logger.Errorf("cannot flush aggregate series: %s", err)
 	}
 }

app/vminsert/main.go

@@ -89,6 +89,7 @@ var staticServer = http.FileServer(http.FS(staticFiles))
 func Init() {
 	relabel.Init()
 	common.InitStreamAggr()
+	opentelemetry.Init()
 	protoparserutil.StartUnmarshalWorkers()
 	if len(*graphiteListenAddr) > 0 {
 		graphiteServer = graphiteserver.MustStart(*graphiteListenAddr, *graphiteUseProxyProtocol, graphite.InsertHandler)

app/vminsert/opentelemetry/request_handler.go

@@ -20,6 +20,11 @@ var (
 	metadataInserted = metrics.NewCounter(`vm_metadata_rows_inserted_total{type="opentelemetry"}`)
 )
 
+// Init must be called after flag.Parse and before using the opentelemetry package.
+func Init() {
+	stream.InitDecodeOptions()
+}
+
 // InsertHandler processes opentelemetry metrics.
 func InsertHandler(req *http.Request) error {
 	extraLabels, err := protoparserutil.GetExtraLabels(req)

app/vmselect/graphite/metrics_api.go

@@ -1,7 +1,6 @@
 package graphite
 
 import (
-	"flag"
 	"fmt"
 	"math"
 	"net/http"
@@ -21,8 +20,6 @@ import (
 	"github.com/VictoriaMetrics/metricsql"
 )
 
-var maxTagValueSuffixes = flag.Int("search.maxTagValueSuffixesPerSearch", 100e3, "The maximum number of tag value suffixes returned from /metrics/find")
-
 // MetricsFindHandler implements /metrics/find handler.
 //
 // See https://graphite-api.readthedocs.io/en/latest/api.html#metrics-find
@@ -225,7 +222,7 @@ func metricsFind(tr storage.TimeRange, label, qHead, qTail string, delimiter byt
 	n := strings.IndexAny(qTail, "*{[")
 	if n < 0 {
 		query := qHead + qTail
-		suffixes, err := netstorage.TagValueSuffixes(nil, tr, label, query, delimiter, *maxTagValueSuffixes, deadline)
+		suffixes, err := netstorage.TagValueSuffixes(nil, tr, label, query, delimiter, 0, deadline)
 		if err != nil {
 			return nil, err
 		}
@@ -245,7 +242,7 @@ func metricsFind(tr storage.TimeRange, label, qHead, qTail string, delimiter byt
 	}
 	if n == len(qTail)-1 && strings.HasSuffix(qTail, "*") {
 		query := qHead + qTail[:len(qTail)-1]
-		suffixes, err := netstorage.TagValueSuffixes(nil, tr, label, query, delimiter, *maxTagValueSuffixes, deadline)
+		suffixes, err := netstorage.TagValueSuffixes(nil, tr, label, query, delimiter, 0, deadline)
 		if err != nil {
 			return nil, err
 		}

app/vmselect/graphite/tags_api.go

@@ -138,7 +138,9 @@ func registerMetrics(startTime time.Time, w http.ResponseWriter, r *http.Request
 		mr.MetricNameRaw = storage.MarshalMetricNameRaw(mr.MetricNameRaw[:0], labels)
 		mr.Timestamp = ct
 	}
-	vmstorage.RegisterMetricNames(nil, mrs)
+	if err := vmstorage.VMSelectAPI.RegisterMetricNames(nil, mrs, 0); err != nil {
+		return err
+	}
 
 	// Return response
 	contentType := "text/plain; charset=utf-8"

app/vmselect/main.go

@@ -21,7 +21,6 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/stats"
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/buildinfo"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
@@ -36,12 +35,6 @@ var (
 	deleteAuthKey                = flagutil.NewPassword("deleteAuthKey", "authKey for metrics' deletion via /api/v1/admin/tsdb/delete_series and /tags/delSeries. It could be passed via authKey query arg. It overrides -httpAuth.*")
 	metricNamesStatsResetAuthKey = flagutil.NewPassword("metricNamesStatsResetAuthKey", "authKey for resetting metric names usage cache via /api/v1/admin/status/metric_names_stats/reset. It overrides -httpAuth.*. "+
 		"See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#track-ingested-metrics-usage")
-
-	maxConcurrentRequests = flag.Int("search.maxConcurrentRequests", getDefaultMaxConcurrentRequests(), "The maximum number of concurrent search requests. "+
-		"It shouldn't be high, since a single request can saturate all the CPU cores, while many concurrently executed requests may require high amounts of memory. "+
-		"See also -search.maxQueueDuration and -search.maxMemoryPerQuery")
-	maxQueueDuration = flag.Duration("search.maxQueueDuration", 10*time.Second, "The maximum time the request waits for execution when -search.maxConcurrentRequests "+
-		"limit is reached; see also -search.maxQueryDuration")
 	resetCacheAuthKey    = flagutil.NewPassword("search.resetCacheAuthKey", "Optional authKey for resetting rollup cache via /internal/resetRollupResultCache call. It could be passed via authKey query arg. It overrides -httpAuth.*")
 	logSlowQueryDuration = flag.Duration("search.logSlowQueryDuration", 5*time.Second, "Log queries with execution time exceeding this value. Zero disables slow query logging. "+
 		"See also -search.logQueryMemoryUsage")
@@ -50,27 +43,22 @@ var (
 
 var slowQueries = metrics.NewCounter(`vm_slow_queries_total`)
 
-func getDefaultMaxConcurrentRequests() int {
-	// A single request can saturate all the CPU cores, so there is no sense
-	// in allowing higher number of concurrent requests - they will just contend
-	// for unavailable CPU time.
-	n := min(cgroup.AvailableCPUs()*2, 16)
-	return n
-}
-
 // Init initializes vmselect
-func Init() {
-	tmpDirPath := *vmstorage.DataPath + "/tmp"
+func Init(maxConcurrentRequests int, maxQueueDuration time.Duration) {
+	tmpDirPath := vmstorage.DataPath() + "/tmp"
 	fs.MustRemoveDirContents(tmpDirPath)
 	netstorage.InitTmpBlocksDir(tmpDirPath)
-	promql.InitRollupResultCache(*vmstorage.DataPath + "/cache/rollupResult")
-	prometheus.InitMaxUniqueTimeseries(*maxConcurrentRequests)
+	promql.InitRollupResultCache(vmstorage.DataPath() + "/cache/rollupResult")
 
-	concurrencyLimitCh = make(chan struct{}, *maxConcurrentRequests)
+	concurrencyLimitCh = make(chan struct{}, maxConcurrentRequests)
 	initVMUIConfig()
 	initVMAlertProxy()
 
 	flagutil.RegisterSecretFlag("vmalert.proxyURL")
+
+	RequestHandler = func(w http.ResponseWriter, r *http.Request) bool {
+		return requestHandler(w, r, maxConcurrentRequests, maxQueueDuration)
+	}
 }
 
 // Stop stops vmselect
@@ -90,9 +78,6 @@ var (
 	_ = metrics.NewGauge(`vm_concurrent_select_current`, func() float64 {
 		return float64(len(concurrencyLimitCh))
 	})
-	_ = metrics.NewGauge(`vm_search_max_unique_timeseries`, func() float64 {
-		return float64(prometheus.GetMaxUniqueTimeSeries())
-	})
 )
 
 //go:embed vmui
@@ -100,8 +85,10 @@ var vmuiFiles embed.FS
 
 var vmuiFileServer = http.FileServer(http.FS(vmuiFiles))
 
-// RequestHandler handles remote read API requests
-func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
+var RequestHandler func(w http.ResponseWriter, r *http.Request) bool
+
+// requestHandler handles remote read API requests
+func requestHandler(w http.ResponseWriter, r *http.Request, maxConcurrentRequests int, maxQueueDuration time.Duration) bool {
 	path := strings.ReplaceAll(r.URL.Path, "//", "/")
 
 	// Strip /prometheus and /graphite prefixes in order to provide path compatibility with cluster version
@@ -131,12 +118,12 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 	default:
 		// Sleep for a while until giving up. This should resolve short bursts in requests.
 		concurrencyLimitReached.Inc()
-		d := min(searchutil.GetMaxQueryDuration(r), *maxQueueDuration)
+		d := min(searchutil.GetMaxQueryDuration(r), maxQueueDuration)
 		t := timerpool.Get(d)
 		select {
 		case concurrencyLimitCh <- struct{}{}:
 			timerpool.Put(t)
-			qt.Printf("wait in queue because -search.maxConcurrentRequests=%d concurrent requests are executed", *maxConcurrentRequests)
+			qt.Printf("wait in queue because -search.maxConcurrentRequests=%d concurrent requests are executed", maxConcurrentRequests)
 			defer func() { <-concurrencyLimitCh }()
 		case <-r.Context().Done():
 			timerpool.Put(t)
@@ -152,7 +139,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 				Err: fmt.Errorf("couldn't start executing the request in %.3f seconds, since -search.maxConcurrentRequests=%d concurrent requests "+
 					"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; to increase -search.maxConcurrentRequests",
-					d.Seconds(), *maxConcurrentRequests, maxQueueDuration),
+					d.Seconds(), maxConcurrentRequests, maxQueueDuration),
 				StatusCode: http.StatusTooManyRequests,
 			}
 			w.Header().Add("Retry-After", "10")

app/vmselect/netstorage/netstorage.go

@@ -27,10 +27,6 @@ import (
 )
 
 var (
-	maxTagKeysPerSearch = flag.Int("search.maxTagKeys", 100e3, "The maximum number of tag keys returned from /api/v1/labels . "+
-		"See also -search.maxLabelsAPISeries and -search.maxLabelsAPIDuration")
-	maxTagValuesPerSearch = flag.Int("search.maxTagValues", 100e3, "The maximum number of tag values returned from /api/v1/label/<label_name>/values . "+
-		"See also -search.maxLabelsAPISeries and -search.maxLabelsAPIDuration")
 	maxSamplesPerSeries = flag.Int("search.maxSamplesPerSeries", 30e6, "The maximum number of raw samples a single query can scan per each time series. This option allows limiting memory usage")
 	maxSamplesPerQuery  = flag.Int("search.maxSamplesPerQuery", 1e9, "The maximum number of raw samples a single query can process across all time series. "+
 		"This protects from heavy queries, which select unexpectedly high number of raw samples. See also -search.maxSamplesPerSeries")
@@ -80,7 +76,7 @@ func (rss *Results) Cancel() {
 }
 
 func (rss *Results) mustClose() {
-	putStorageSearch(rss.sr)
+	vmstorage.PutSearch(rss.sr)
 	rss.sr = nil
 	putTmpBlocksFile(rss.tbf)
 	rss.tbf = nil
@@ -758,12 +754,7 @@ var sbhPool sync.Pool
 func DeleteSeries(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline searchutil.Deadline) (int, error) {
 	qt = qt.NewChild("delete series: %s", sq)
 	defer qt.Done()
-	tr := sq.GetTimeRange()
-	tfss, err := setupTfss(qt, tr, sq.TagFilterss, sq.MaxMetrics, deadline)
-	if err != nil {
-		return 0, err
-	}
-	return vmstorage.DeleteSeries(qt, tfss, sq.MaxMetrics)
+	return vmstorage.VMSelectAPI.DeleteSeries(qt, sq, deadline.Deadline())
 }
 
 // LabelNames returns label names matching the given sq until the given deadline.
@@ -773,15 +764,7 @@ func LabelNames(qt *querytracer.Tracer, sq *storage.SearchQuery, maxLabelNames i
 	if deadline.Exceeded() {
 		return nil, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
 	}
-	if maxLabelNames > *maxTagKeysPerSearch || maxLabelNames <= 0 {
-		maxLabelNames = *maxTagKeysPerSearch
-	}
-	tr := sq.GetTimeRange()
-	tfss, err := setupTfss(qt, tr, sq.TagFilterss, sq.MaxMetrics, deadline)
-	if err != nil {
-		return nil, err
-	}
-	labels, err := vmstorage.SearchLabelNames(qt, tfss, tr, maxLabelNames, sq.MaxMetrics, deadline.Deadline())
+	labels, err := vmstorage.VMSelectAPI.LabelNames(qt, sq, maxLabelNames, deadline.Deadline())
 	if err != nil {
 		return nil, fmt.Errorf("error during labels search on time range: %w", err)
 	}
@@ -841,15 +824,7 @@ func LabelValues(qt *querytracer.Tracer, labelName string, sq *storage.SearchQue
 	if deadline.Exceeded() {
 		return nil, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
 	}
-	if maxLabelValues > *maxTagValuesPerSearch || maxLabelValues <= 0 {
-		maxLabelValues = *maxTagValuesPerSearch
-	}
-	tr := sq.GetTimeRange()
-	tfss, err := setupTfss(qt, tr, sq.TagFilterss, sq.MaxMetrics, deadline)
-	if err != nil {
-		return nil, err
-	}
-	labelValues, err := vmstorage.SearchLabelValues(qt, labelName, tfss, tr, maxLabelValues, sq.MaxMetrics, deadline.Deadline())
+	labelValues, err := vmstorage.VMSelectAPI.LabelValues(qt, sq, labelName, maxLabelValues, deadline.Deadline())
 	if err != nil {
 		return nil, fmt.Errorf("error during label values search on time range for labelName=%q: %w", labelName, err)
 	}
@@ -864,7 +839,10 @@ func GetMetricsMetadata(qt *querytracer.Tracer, limit int, metricName string) ([
 	qt = qt.NewChild("get metrics metadata: limit=%d, metric_name=%q", limit, metricName)
 	defer qt.Done()
 
-	metadata := vmstorage.Storage.GetMetadataRows(qt, limit, metricName)
+	metadata, err := vmstorage.VMSelectAPI.GetMetadataRecords(qt, nil, limit, metricName, 0)
+	if err != nil {
+		return nil, err
+	}
 
 	sort.Slice(metadata, func(i, j int) bool {
 		return string(metadata[i].MetricFamilyName) < string(metadata[j].MetricFamilyName)
@@ -912,16 +890,11 @@ func TagValueSuffixes(qt *querytracer.Tracer, tr storage.TimeRange, tagKey, tagV
 	if deadline.Exceeded() {
 		return nil, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
 	}
-	suffixes, err := vmstorage.SearchTagValueSuffixes(qt, tr, tagKey, tagValuePrefix, delimiter, maxSuffixes, deadline.Deadline())
+	suffixes, err := vmstorage.VMSelectAPI.TagValueSuffixes(qt, 0, 0, tr, tagKey, tagValuePrefix, delimiter, maxSuffixes, deadline.Deadline())
 	if err != nil {
 		return nil, fmt.Errorf("error during search for suffixes for tagKey=%q, tagValuePrefix=%q, delimiter=%c on time range %s: %w",
 			tagKey, tagValuePrefix, delimiter, tr.String(), err)
 	}
-	if len(suffixes) >= maxSuffixes {
-		return nil, fmt.Errorf("more than -search.maxTagValueSuffixesPerSearch=%d tag value suffixes found for tagKey=%q, tagValuePrefix=%q, delimiter=%c on time range %s; "+
-			"either narrow down the query or increase -search.maxTagValueSuffixesPerSearch command-line flag value",
-			maxSuffixes, tagKey, tagValuePrefix, delimiter, tr.String())
-	}
 	return suffixes, nil
 }
 
@@ -934,13 +907,7 @@ func TSDBStatus(qt *querytracer.Tracer, sq *storage.SearchQuery, focusLabel stri
 	if deadline.Exceeded() {
 		return nil, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
 	}
-	tr := sq.GetTimeRange()
-	tfss, err := setupTfss(qt, tr, sq.TagFilterss, sq.MaxMetrics, deadline)
-	if err != nil {
-		return nil, err
-	}
-	date := uint64(tr.MinTimestamp) / (3600 * 24 * 1000)
-	status, err := vmstorage.GetTSDBStatus(qt, tfss, date, focusLabel, topN, sq.MaxMetrics, deadline.Deadline())
+	status, err := vmstorage.VMSelectAPI.TSDBStatus(qt, sq, focusLabel, topN, deadline.Deadline())
 	if err != nil {
 		return nil, fmt.Errorf("error during tsdb status request: %w", err)
 	}
@@ -954,28 +921,13 @@ func SeriesCount(qt *querytracer.Tracer, deadline searchutil.Deadline) (uint64,
 	if deadline.Exceeded() {
 		return 0, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
 	}
-	n, err := vmstorage.GetSeriesCount(deadline.Deadline())
+	n, err := vmstorage.VMSelectAPI.SeriesCount(qt, 0, 0, deadline.Deadline())
 	if err != nil {
 		return 0, fmt.Errorf("error during series count request: %w", err)
 	}
 	return n, nil
 }
 
-func getStorageSearch() *storage.Search {
-	v := ssPool.Get()
-	if v == nil {
-		return &storage.Search{}
-	}
-	return v.(*storage.Search)
-}
-
-func putStorageSearch(sr *storage.Search) {
-	sr.MustClose()
-	ssPool.Put(sr)
-}
-
-var ssPool sync.Pool
-
 // ExportBlocks searches for time series matching sq and calls f for each found block.
 //
 // f is called in parallel from multiple goroutines.
@@ -989,18 +941,12 @@ func ExportBlocks(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline sear
 	if deadline.Exceeded() {
 		return fmt.Errorf("timeout exceeded before starting data export: %s", deadline.String())
 	}
-	tr := sq.GetTimeRange()
-	tfss, err := setupTfss(qt, tr, sq.TagFilterss, sq.MaxMetrics, deadline)
+
+	sr, _, err := vmstorage.GetSearch(qt, sq, deadline.Deadline())
 	if err != nil {
 		return err
 	}
-
-	vmstorage.WG.Add(1)
-	defer vmstorage.WG.Done()
-
-	sr := getStorageSearch()
-	defer putStorageSearch(sr)
-	sr.Init(qt, vmstorage.Storage, tfss, tr, sq.MaxMetrics, deadline.Deadline())
+	defer vmstorage.PutSearch(sr)
 
 	// Start workers that call f in parallel on available CPU cores.
 	workCh := make(chan *exportWork, gomaxprocs*8)
@@ -1010,6 +956,7 @@ func ExportBlocks(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline sear
 		mustStop      atomic.Bool
 	)
 	var wg sync.WaitGroup
+	tr := sq.GetTimeRange()
 	for workerID := range gomaxprocs {
 		wg.Go(func() {
 			for xw := range workCh {
@@ -1093,14 +1040,7 @@ func SearchMetricNames(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline
 		return nil, fmt.Errorf("timeout exceeded before starting to search metric names: %s", deadline.String())
 	}
 
-	// Setup search.
-	tr := sq.GetTimeRange()
-	tfss, err := setupTfss(qt, tr, sq.TagFilterss, sq.MaxMetrics, deadline)
-	if err != nil {
-		return nil, err
-	}
-
-	metricNames, err := vmstorage.SearchMetricNames(qt, tfss, tr, sq.MaxMetrics, deadline.Deadline())
+	metricNames, err := vmstorage.VMSelectAPI.SearchMetricNames(qt, sq, deadline.Deadline())
 	if err != nil {
 		return nil, fmt.Errorf("cannot find metric names: %w", err)
 	}
@@ -1119,18 +1059,11 @@ func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadlin
 		return nil, fmt.Errorf("timeout exceeded before starting the query processing: %s", deadline.String())
 	}
 
-	// Setup search.
-	tr := sq.GetTimeRange()
-	tfss, err := setupTfss(qt, tr, sq.TagFilterss, sq.MaxMetrics, deadline)
+	sr, maxSeriesCount, err := vmstorage.GetSearch(qt, sq, deadline.Deadline())
 	if err != nil {
 		return nil, err
 	}
 
-	vmstorage.WG.Add(1)
-	defer vmstorage.WG.Done()
-
-	sr := getStorageSearch()
-	maxSeriesCount := sr.Init(qt, vmstorage.Storage, tfss, tr, sq.MaxMetrics, deadline.Deadline())
 	type blockRefs struct {
 		brs []blockRef
 	}
@@ -1168,7 +1101,7 @@ func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadlin
 		blocksRead++
 		if deadline.Exceeded() {
 			putTmpBlocksFile(tbf)
-			putStorageSearch(sr)
+			vmstorage.PutSearch(sr)
 			return nil, fmt.Errorf("timeout exceeded while fetching data block #%d from storage: %s", blocksRead, deadline.String())
 		}
 		br := sr.MetricBlockRef.BlockRef
@@ -1180,7 +1113,7 @@ func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadlin
 		samples += br.RowsCount()
 		if *maxSamplesPerQuery > 0 && samples > *maxSamplesPerQuery {
 			putTmpBlocksFile(tbf)
-			putStorageSearch(sr)
+			vmstorage.PutSearch(sr)
 			return nil, fmt.Errorf("cannot select more than -search.maxSamplesPerQuery=%d samples; possible solutions: increase the -search.maxSamplesPerQuery; "+
 				"reduce time range for the query; use more specific label filters in order to select fewer series", *maxSamplesPerQuery)
 		}
@@ -1189,7 +1122,7 @@ func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadlin
 		addr, err := tbf.WriteBlockRefData(buf)
 		if err != nil {
 			putTmpBlocksFile(tbf)
-			putStorageSearch(sr)
+			vmstorage.PutSearch(sr)
 			return nil, fmt.Errorf("cannot write %d bytes to temporary file: %w", len(buf), err)
 		}
 
@@ -1247,7 +1180,7 @@ func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadlin
 
 	if err := sr.Error(); err != nil {
 		putTmpBlocksFile(tbf)
-		putStorageSearch(sr)
+		vmstorage.PutSearch(sr)
 		if errors.Is(err, storage.ErrDeadlineExceeded) {
 			return nil, fmt.Errorf("timeout exceeded during the query: %s", deadline.String())
 		}
@@ -1255,13 +1188,13 @@ func ProcessSearchQuery(qt *querytracer.Tracer, sq *storage.SearchQuery, deadlin
 	}
 	if err := tbf.Finalize(); err != nil {
 		putTmpBlocksFile(tbf)
-		putStorageSearch(sr)
+		vmstorage.PutSearch(sr)
 		return nil, fmt.Errorf("cannot finalize temporary file: %w", err)
 	}
 	qt.Printf("fetch unique series=%d, blocks=%d, samples=%d, bytes=%d", len(m), blocksRead, samples, tbf.Len())
 
 	var rss Results
-	rss.tr = tr
+	rss.tr = sq.GetTimeRange()
 	rss.deadline = deadline
 	pts := make([]packedTimeseries, len(orderedMetricNames))
 	for i, metricName := range orderedMetricNames {
@@ -1302,35 +1235,6 @@ func getBlockRefsEnd(a []blockRef) uintptr {
 	return uintptr(unsafe.Pointer(unsafe.SliceData(a))) + uintptr(len(a))*unsafe.Sizeof(blockRef{})
 }
 
-func setupTfss(qt *querytracer.Tracer, tr storage.TimeRange, tagFilterss [][]storage.TagFilter, maxMetrics int, deadline searchutil.Deadline) ([]*storage.TagFilters, error) {
-	tfss := make([]*storage.TagFilters, 0, len(tagFilterss))
-	for _, tagFilters := range tagFilterss {
-		tfs := storage.NewTagFilters()
-		for i := range tagFilters {
-			tf := &tagFilters[i]
-			if string(tf.Key) == "__graphite__" {
-				query := tf.Value
-				paths, err := vmstorage.SearchGraphitePaths(qt, tr, query, maxMetrics, deadline.Deadline())
-				if err != nil {
-					return nil, fmt.Errorf("error when searching for Graphite paths for query %q: %w", query, err)
-				}
-				if len(paths) >= maxMetrics {
-					return nil, fmt.Errorf("more than %d time series match Graphite query %q; "+
-						"either narrow down the query or increase the corresponding -search.max* command-line flag value; "+
-						"see https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#resource-usage-limits", maxMetrics, query)
-				}
-				tfs.AddGraphiteQuery(query, paths, tf.IsNegative)
-				continue
-			}
-			if err := tfs.Add(tf.Key, tf.Value, tf.IsNegative, tf.IsRegexp); err != nil {
-				return nil, fmt.Errorf("cannot parse tag filter %s: %w", tf, err)
-			}
-		}
-		tfss = append(tfss, tfs)
-	}
-	return tfss, nil
-}
-
 func applyGraphiteRegexpFilter(filter string, ss []string) ([]string, error) {
 	// Anchor filter regexp to the beginning of the string as Graphite does.
 	// See https://github.com/graphite-project/graphite-web/blob/3ad279df5cb90b211953e39161df416e54a84948/webapp/graphite/tags/localdatabase.py#L157
@@ -1357,13 +1261,12 @@ const maxFastAllocBlockSize = 32 * 1024
 func GetMetricNamesStats(qt *querytracer.Tracer, limit, le int, matchPattern string) (metricnamestats.StatsResult, error) {
 	qt = qt.NewChild("get metric names usage statistics with limit: %d, less or equal to: %d, match pattern=%q", limit, le, matchPattern)
 	defer qt.Done()
-	return vmstorage.GetMetricNamesStats(qt, limit, le, matchPattern)
+	return vmstorage.VMSelectAPI.GetMetricNamesUsageStats(qt, nil, limit, le, matchPattern, 0)
 }
 
 // ResetMetricNamesStats resets state of metric names usage
 func ResetMetricNamesStats(qt *querytracer.Tracer) error {
 	qt = qt.NewChild("reset metric names usage stats")
 	defer qt.Done()
-	vmstorage.ResetMetricNamesStats(qt)
-	return nil
+	return vmstorage.VMSelectAPI.ResetMetricNamesUsageStats(qt, 0)
 }

app/vmselect/prometheus/prometheus.go

@@ -28,8 +28,6 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputil"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/memory"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/netutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
@@ -50,9 +48,6 @@ var (
 		"If set to true, the query model becomes closer to InfluxDB data model. If set to true, then -search.maxLookback and -search.maxStalenessInterval are ignored")
 	maxStepForPointsAdjustment = flag.Duration("search.maxStepForPointsAdjustment", time.Minute, "The maximum step when /api/v1/query_range handler adjusts "+
 		"points with timestamps closer than -search.latencyOffset to the current time. The adjustment is needed because such points may contain incomplete data")
-
-	maxUniqueTimeseries = flag.Int("search.maxUniqueTimeseries", 0, "The maximum number of unique time series, which can be selected during /api/v1/query and /api/v1/query_range queries. This option allows limiting memory usage. "+
-		"When set to zero, the limit is automatically calculated based on -search.maxConcurrentRequests (inversely proportional) and memory available to the process (proportional).")
 	maxFederateSeries       = flag.Int("search.maxFederateSeries", 1e6, "The maximum number of time series, which can be returned from /federate. This option allows limiting memory usage")
 	maxExportSeries         = flag.Int("search.maxExportSeries", 10e6, "The maximum number of time series, which can be returned from /api/v1/export* APIs. This option allows limiting memory usage")
 	maxTSDBStatusSeries     = flag.Int("search.maxTSDBStatusSeries", 10e6, "The maximum number of time series, which can be processed during the call to /api/v1/status/tsdb. This option allows limiting memory usage")
@@ -853,7 +848,7 @@ func QueryHandler(qt *querytracer.Tracer, startTime time.Time, w http.ResponseWr
 		End:                 start,
 		Step:                step,
 		MaxPointsPerSeries:  *maxPointsPerTimeseries,
-		MaxSeries:           GetMaxUniqueTimeSeries(),
+		MaxSeries:           0, // let vmstorage use maxUniqueTimeseries by default
 		QuotedRemoteAddr:    httpserver.GetQuotedRemoteAddr(r),
 		Deadline:            deadline,
 		MayCache:            mayCache,
@@ -964,7 +959,7 @@ func queryRangeHandler(qt *querytracer.Tracer, startTime time.Time, w http.Respo
 		End:                 end,
 		Step:                step,
 		MaxPointsPerSeries:  *maxPointsPerTimeseries,
-		MaxSeries:           GetMaxUniqueTimeSeries(),
+		MaxSeries:           0, // let vmstorage use maxUniqueTimeseries by default
 		QuotedRemoteAddr:    httpserver.GetQuotedRemoteAddr(r),
 		Deadline:            deadline,
 		MayCache:            mayCache,
@@ -1300,43 +1295,6 @@ func (sw *scalableWriter) flush() error {
 	return sw.bw.Flush()
 }
 
-var (
-	maxUniqueTimeseriesValueOnce sync.Once
-	maxUniqueTimeseriesValue     int
-)
-
-// InitMaxUniqueTimeseries init the max metrics limit calculated by available resources.
-// The calculation is split into calculateMaxUniqueTimeSeriesForResource for unit testing.
-func InitMaxUniqueTimeseries(maxConcurrentRequests int) {
-	maxUniqueTimeseriesValueOnce.Do(func() {
-		maxUniqueTimeseriesValue = *maxUniqueTimeseries
-		if maxUniqueTimeseriesValue <= 0 {
-			maxUniqueTimeseriesValue = calculateMaxUniqueTimeSeriesForResource(maxConcurrentRequests, memory.Remaining())
-		}
-	})
-}
-
-// calculateMaxUniqueTimeSeriesForResource calculate the max metrics limit calculated by available resources.
-func calculateMaxUniqueTimeSeriesForResource(maxConcurrentRequests, remainingMemory int) int {
-	if maxConcurrentRequests <= 0 {
-		// This line should NOT be reached unless the user has set an incorrect `search.maxConcurrentRequests`.
-		// In such cases, fallback to unlimited.
-		logger.Warnf("limiting -search.maxUniqueTimeseries to %v because -search.maxConcurrentRequests=%d.", 2e9, maxConcurrentRequests)
-		return 2e9
-	}
-
-	// Calculate the max metrics limit for a single request in the worst-case concurrent scenario.
-	// The approximate size of 1 unique series that could occupy in the vmstorage is 200 bytes.
-	mts := remainingMemory / 200 / maxConcurrentRequests
-	logger.Infof("limiting -search.maxUniqueTimeseries to %d according to -search.maxConcurrentRequests=%d and remaining memory=%d bytes. To increase the limit, reduce -search.maxConcurrentRequests or increase memory available to the process.", mts, maxConcurrentRequests, remainingMemory)
-	return mts
-}
-
-// GetMaxUniqueTimeSeries returns the max metrics limit calculated by available resources.
-func GetMaxUniqueTimeSeries() int {
-	return maxUniqueTimeseriesValue
-}
-
 // copied from https://github.com/prometheus/common/blob/adea6285c1c7447fcb7bfdeb6abfc6eff893e0a7/model/metric.go#L483
 // it's not possible to use direct import due to increased binary size
 func unescapePrometheusLabelName(name string) string {

app/vmselect/prometheus/prometheus_test.go

@@ -4,7 +4,6 @@ import (
 	"math"
 	"net/http"
 	"reflect"
-	"runtime"
 	"testing"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/netstorage"
@@ -230,29 +229,3 @@ func TestGetLatencyOffsetMillisecondsFailure(t *testing.T) {
 	}
 	f("http://localhost?latency_offset=foobar")
 }
-
-func TestCalculateMaxMetricsLimitByResource(t *testing.T) {
-	f := func(maxConcurrentRequest, remainingMemory, expect int) {
-		t.Helper()
-		maxMetricsLimit := calculateMaxUniqueTimeSeriesForResource(maxConcurrentRequest, remainingMemory)
-		if maxMetricsLimit != expect {
-			t.Fatalf("unexpected max metrics limit: got %d, want %d", maxMetricsLimit, expect)
-		}
-	}
-
-	// Skip when GOARCH=386
-	if runtime.GOARCH != "386" {
-		// 8 CPU & 32 GiB
-		f(16, int(math.Round(32*1024*1024*1024*0.4)), 4294967)
-		// 4 CPU & 32 GiB
-		f(8, int(math.Round(32*1024*1024*1024*0.4)), 8589934)
-	}
-
-	// 2 CPU & 4 GiB
-	f(4, int(math.Round(4*1024*1024*1024*0.4)), 2147483)
-
-	// other edge cases
-	f(0, int(math.Round(4*1024*1024*1024*0.4)), 2e9)
-	f(4, 0, 0)
-
-}

app/vmstorage/main.go

@@ -1,7 +1,6 @@
 package vmstorage
 
 import (
-	"errors"
 	"flag"
 	"fmt"
 	"io"
@@ -9,12 +8,10 @@ import (
 	"net/http"
 	"strconv"
 	"strings"
-	"sync"
 	"time"
 
 	"github.com/VictoriaMetrics/metrics"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
@@ -23,14 +20,13 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/mergeset"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricnamestats"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricsmetadata"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/stringsutil"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/syncwg"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/vminsertapi"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/vmselectapi"
 )
 
 var (
+	storageDataPath = flag.String("storageDataPath", "victoria-metrics-data", "Path to storage data")
 	retentionPeriod = flagutil.NewRetentionDuration("retentionPeriod", "1M", "Data with timestamps outside the retentionPeriod is automatically deleted. The minimum retentionPeriod is 24h or 1d. "+
 		"See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#retention. See also -retentionFilter")
 	futureRetention = flagutil.NewRetentionDuration("futureRetention", "2d", "Data with timestamps bigger than now+futureRetention is automatically deleted. "+
@@ -38,14 +34,8 @@ var (
 	snapshotAuthKey   = flagutil.NewPassword("snapshotAuthKey", "authKey, which must be passed in query string to /snapshot* pages. It overrides -httpAuth.*")
 	forceMergeAuthKey = flagutil.NewPassword("forceMergeAuthKey", "authKey, which must be passed in query string to /internal/force_merge pages. It overrides -httpAuth.*")
 	forceFlushAuthKey = flagutil.NewPassword("forceFlushAuthKey", "authKey, which must be passed in query string to /internal/force_flush pages. It overrides -httpAuth.*")
-	snapshotsMaxAge   = flagutil.NewRetentionDuration("snapshotsMaxAge", "3d", "Automatically delete snapshots older than -snapshotsMaxAge if it is set to non-zero duration. Make sure that backup process has enough time to finish the backup before the corresponding snapshot is automatically deleted")
 	_                 = flag.Duration("snapshotCreateTimeout", 0, "Deprecated: this flag does nothing")
 
-	precisionBits = flag.Int("precisionBits", 64, "The number of precision bits to store per each value. Lower precision bits improves data compression at the cost of precision loss")
-
-	// DataPath is a path to storage data.
-	DataPath = flag.String("storageDataPath", "victoria-metrics-data", "Path to storage data")
-
 	_ = flag.Duration("finalMergeDelay", 0, "Deprecated: this flag does nothing")
 	_ = flag.Int("bigMergeConcurrency", 0, "Deprecated: this flag does nothing")
 	_ = flag.Int("smallMergeConcurrency", 0, "Deprecated: this flag does nothing")
@@ -53,6 +43,12 @@ var (
 	retentionTimezoneOffset = flag.Duration("retentionTimezoneOffset", 0, "The offset for performing indexdb rotation. "+
 		"If set to 0, then the indexdb rotation is performed at 4am UTC time per each -retentionPeriod. "+
 		"If set to 2h, then the indexdb rotation is performed at 4am EET time (the timezone with +2h offset)")
+	minScrapeInterval = flag.Duration("dedup.minScrapeInterval", 0, "Leave only the last sample in every time series per each discrete interval "+
+		"equal to -dedup.minScrapeInterval > 0. See also -streamAggr.dedupInterval and https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#deduplication")
+	inmemoryDataFlushInterval = flag.Duration("inmemoryDataFlushInterval", 5*time.Second, "The interval for guaranteed saving of in-memory data to disk. "+
+		"The saved data survives unclean shutdowns such as OOM crash, hardware reset, SIGKILL, etc. "+
+		"Bigger intervals may help increase the lifetime of flash storage with limited write cycles (e.g. Raspberry PI). "+
+		"Smaller intervals increase disk IO load. Minimum supported value is 1s")
 
 	logNewSeries = flag.Bool("logNewSeries", false, "Whether to log new series. This option is for debug purposes only. It can lead to performance issues "+
 		"when big number of new series are ingested into VictoriaMetrics")
@@ -70,6 +66,11 @@ var (
 
 	minFreeDiskSpaceBytes = flagutil.NewBytes("storage.minFreeDiskSpaceBytes", 100e6, "The minimum free disk space at -storageDataPath after which the storage stops accepting new data")
 
+	finalDedupScheduleInterval = flag.Duration("storage.finalDedupScheduleCheckInterval", time.Hour, "The interval for checking when final deduplication process should be started."+
+		"Storage unconditionally adds 25% jitter to the interval value on each check evaluation."+
+		" Changing the interval to the bigger values may delay downsampling, deduplication for historical data."+
+		" See also https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#deduplication")
+
 	cacheSizeStorageTSID = flagutil.NewBytes("storage.cacheSizeStorageTSID", 0, "Overrides max size for storage/tsid cache. "+
 		"See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#cache-tuning")
 	cacheSizeStorageMetricName = flagutil.NewBytes("storage.cacheSizeStorageMetricName", 0, "Overrides max size for storage/metricName cache. "+
@@ -103,17 +104,22 @@ var (
 		"If set to 0 or a negative value, defaults to 1% of allowed memory.")
 )
 
-// Init initializes vmstorage.
-func Init(resetCacheIfNeeded func(mrs []storage.MetricRow)) {
-	if err := encoding.CheckPrecisionBits(uint8(*precisionBits)); err != nil {
-		logger.Fatalf("invalid `-precisionBits`: %s", err)
-	}
+func DataPath() string {
+	return *storageDataPath
+}
 
-	resetResponseCacheIfNeeded = resetCacheIfNeeded
+// Init initializes vmstorage.
+func Init(maxConcurrentRequests int, resetCacheIfNeeded func(mrs []storage.MetricRow)) {
+	storage.SetDedupInterval(*minScrapeInterval)
+	storage.SetDataFlushInterval(*inmemoryDataFlushInterval)
 	storage.LegacySetRetentionTimezoneOffset(*retentionTimezoneOffset)
 	storage.SetFreeDiskSpaceLimit(minFreeDiskSpaceBytes.N)
 	storage.SetTSIDCacheSize(cacheSizeStorageTSID.IntN())
 	storage.SetTagFiltersCacheSize(cacheSizeIndexDBTagFilters.IntN())
+	if *finalDedupScheduleInterval < time.Hour {
+		logger.Fatalf("-storage.finalDedupScheduleCheckInterval cannot be smaller than 1 hour; got %s", *finalDedupScheduleInterval)
+	}
+	storage.SetFinalDedupScheduleInterval(*finalDedupScheduleInterval)
 	storage.SetMetricNamesStatsCacheSize(cacheSizeMetricNamesStats.IntN())
 	storage.SetMetricNameCacheSize(cacheSizeStorageMetricName.IntN())
 	storage.SetMetadataStorageSize(metadataStorageSize.IntN())
@@ -132,9 +138,9 @@ func Init(resetCacheIfNeeded func(mrs []storage.MetricRow)) {
 	if *idbPrefillStart > 23*time.Hour {
 		logger.Panicf("-storage.idbPrefillStart cannot exceed 23 hours; got %s", idbPrefillStart)
 	}
-	logger.Infof("opening storage at %q with -retentionPeriod=%s", *DataPath, retentionPeriod)
+	fs.RegisterPathFsMetrics(*storageDataPath)
+	logger.Infof("opening storage at %q with -retentionPeriod=%s", *storageDataPath, retentionPeriod)
 	startTime := time.Now()
-	WG = syncwg.WaitGroup{}
 	opts := storage.OpenOptions{
 		Retention:                   retentionPeriod.Duration(),
 		FutureRetention:             futureRetention.Duration(),
@@ -146,9 +152,8 @@ func Init(resetCacheIfNeeded func(mrs []storage.MetricRow)) {
 		IDBPrefillStart:             *idbPrefillStart,
 		LogNewSeries:                *logNewSeries,
 	}
-	strg := storage.MustOpenStorage(*DataPath, opts)
-	Storage = strg
-	initStaleSnapshotsRemover(strg)
+	strg := storage.MustOpenStorage(*storageDataPath, opts)
+	vmStorage = newVMStorageSingleNode(strg, maxConcurrentRequests, resetCacheIfNeeded)
 
 	var m storage.Metrics
 	strg.UpdateMetrics(&m)
@@ -158,152 +163,46 @@ func Init(resetCacheIfNeeded func(mrs []storage.MetricRow)) {
 	rowsCount := tm.SmallRowsCount + tm.BigRowsCount
 	sizeBytes := tm.SmallSizeBytes + tm.BigSizeBytes
 	logger.Infof("successfully opened storage %q in %.3f seconds; partsCount: %d; blocksCount: %d; rowsCount: %d; sizeBytes: %d",
-		*DataPath, time.Since(startTime).Seconds(), partsCount, blocksCount, rowsCount, sizeBytes)
+		*storageDataPath, time.Since(startTime).Seconds(), partsCount, blocksCount, rowsCount, sizeBytes)
 
 	// register storage metrics
 	storageMetrics = metrics.NewSet()
 	storageMetrics.RegisterMetricsWriter(func(w io.Writer) {
-		writeStorageMetrics(w, strg)
+		vmStorage.writeStorageMetrics(w)
 	})
 	metrics.RegisterSet(storageMetrics)
-	fs.RegisterPathFsMetrics(*DataPath)
+
+	VMInsertAPI = vmStorage
+	VMSelectAPI = vmStorage
+	GetSearch = func(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (*storage.Search, int, error) {
+		return vmStorage.GetSearch(qt, sq, deadline)
+	}
+	PutSearch = func(sr *storage.Search) {
+		vmStorage.PutSearch(sr)
+	}
+	RequestHandler = func(w http.ResponseWriter, r *http.Request) bool {
+		return vmStorage.requestHandler(w, r)
+	}
+	DebugFlush = func() {
+		vmStorage.s.DebugFlush()
+	}
 }
 
 var storageMetrics *metrics.Set
 
-// Storage is a storage.
-//
-// Every storage call must be wrapped into WG.Add(1) ... WG.Done()
-// for proper graceful shutdown when Stop is called.
-var Storage *storage.Storage
+var (
+	// vmStorageSingleNode is an instance of vmstorage used by vminsert and
+	// vmselect for writing and reading data.
+	vmStorage      *VMStorageSingleNode
+	VMInsertAPI    vminsertapi.API
+	VMSelectAPI    vmselectapi.API
+	GetSearch      func(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (*storage.Search, int, error)
+	PutSearch      func(sr *storage.Search)
+	RequestHandler func(w http.ResponseWriter, r *http.Request) bool
 
-// WG must be incremented before Storage call.
-//
-// Use syncwg instead of sync, since Add is called from concurrent goroutines.
-var WG syncwg.WaitGroup
-
-// resetResponseCacheIfNeeded is a callback for automatic resetting of response cache if needed.
-var resetResponseCacheIfNeeded func(mrs []storage.MetricRow)
-
-// AddRows adds mrs to the storage.
-//
-// The caller should limit the number of concurrent calls to AddRows() in order to limit memory usage.
-func AddRows(mrs []storage.MetricRow) error {
-	if Storage.IsReadOnly() {
-		return errReadOnly
-	}
-	resetResponseCacheIfNeeded(mrs)
-	WG.Add(1)
-	Storage.AddRows(mrs, uint8(*precisionBits))
-	WG.Done()
-	return nil
-}
-
-// AddMetadataRows adds mrs to the storage.
-//
-// The caller should limit the number of concurrent calls to AddMetadataRows() in order to limit memory usage.
-func AddMetadataRows(mms []metricsmetadata.Row) error {
-	if Storage.IsReadOnly() {
-		return errReadOnly
-	}
-	WG.Add(1)
-	Storage.AddMetadataRows(mms)
-	WG.Done()
-	return nil
-}
-
-var errReadOnly = errors.New("the storage is in read-only mode; check -storage.minFreeDiskSpaceBytes command-line flag value")
-
-// RegisterMetricNames registers all the metrics from mrs in the storage.
-func RegisterMetricNames(qt *querytracer.Tracer, mrs []storage.MetricRow) {
-	WG.Add(1)
-	Storage.RegisterMetricNames(qt, mrs)
-	WG.Done()
-}
-
-// DeleteSeries deletes series matching tfss.
-//
-// Returns the number of deleted series.
-func DeleteSeries(qt *querytracer.Tracer, tfss []*storage.TagFilters, maxMetrics int) (int, error) {
-	WG.Add(1)
-	n, err := Storage.DeleteSeries(qt, tfss, maxMetrics)
-	WG.Done()
-	return n, err
-}
-
-// GetMetricNamesStats returns metric names usage stats with give limit and lte predicate
-func GetMetricNamesStats(qt *querytracer.Tracer, limit, le int, matchPattern string) (metricnamestats.StatsResult, error) {
-	WG.Add(1)
-	r := Storage.GetMetricNamesStats(qt, limit, le, matchPattern)
-	WG.Done()
-	return r, nil
-}
-
-// ResetMetricNamesStats resets state for metric names usage tracker
-func ResetMetricNamesStats(qt *querytracer.Tracer) {
-	WG.Add(1)
-	Storage.ResetMetricNamesStats(qt)
-	WG.Done()
-}
-
-// SearchMetricNames returns metric names for the given tfss on the given tr.
-func SearchMetricNames(qt *querytracer.Tracer, tfss []*storage.TagFilters, tr storage.TimeRange, maxMetrics int, deadline uint64) ([]string, error) {
-	WG.Add(1)
-	metricNames, err := Storage.SearchMetricNames(qt, tfss, tr, maxMetrics, deadline)
-	WG.Done()
-	return metricNames, err
-}
-
-// SearchLabelNames searches for tag keys matching the given tfss on tr.
-func SearchLabelNames(qt *querytracer.Tracer, tfss []*storage.TagFilters, tr storage.TimeRange, maxTagKeys, maxMetrics int, deadline uint64) ([]string, error) {
-	WG.Add(1)
-	labelNames, err := Storage.SearchLabelNames(qt, tfss, tr, maxTagKeys, maxMetrics, deadline)
-	WG.Done()
-	return labelNames, err
-}
-
-// SearchLabelValues searches for label values for the given labelName, tfss and
-// tr.
-func SearchLabelValues(qt *querytracer.Tracer, labelName string, tfss []*storage.TagFilters, tr storage.TimeRange, maxLabelValues, maxMetrics int, deadline uint64) ([]string, error) {
-	WG.Add(1)
-	labelValues, err := Storage.SearchLabelValues(qt, labelName, tfss, tr, maxLabelValues, maxMetrics, deadline)
-	WG.Done()
-	return labelValues, err
-}
-
-// SearchTagValueSuffixes returns all the tag value suffixes for the given tagKey and tagValuePrefix on the given tr.
-//
-// This allows implementing https://graphite-api.readthedocs.io/en/latest/api.html#metrics-find or similar APIs.
-func SearchTagValueSuffixes(qt *querytracer.Tracer, tr storage.TimeRange, tagKey, tagValuePrefix string, delimiter byte, maxTagValueSuffixes int, deadline uint64) ([]string, error) {
-	WG.Add(1)
-	suffixes, err := Storage.SearchTagValueSuffixes(qt, tr, tagKey, tagValuePrefix, delimiter, maxTagValueSuffixes, deadline)
-	WG.Done()
-	return suffixes, err
-}
-
-// SearchGraphitePaths returns all the metric names matching the given Graphite query.
-func SearchGraphitePaths(qt *querytracer.Tracer, tr storage.TimeRange, query []byte, maxPaths int, deadline uint64) ([]string, error) {
-	WG.Add(1)
-	paths, err := Storage.SearchGraphitePaths(qt, tr, query, maxPaths, deadline)
-	WG.Done()
-	return paths, err
-}
-
-// GetTSDBStatus returns TSDB status for given filters on the given date.
-func GetTSDBStatus(qt *querytracer.Tracer, tfss []*storage.TagFilters, date uint64, focusLabel string, topN, maxMetrics int, deadline uint64) (*storage.TSDBStatus, error) {
-	WG.Add(1)
-	status, err := Storage.GetTSDBStatus(qt, tfss, date, focusLabel, topN, maxMetrics, deadline)
-	WG.Done()
-	return status, err
-}
-
-// GetSeriesCount returns the number of time series in the storage.
-func GetSeriesCount(deadline uint64) (uint64, error) {
-	WG.Add(1)
-	n, err := Storage.GetSeriesCount(deadline)
-	WG.Done()
-	return n, err
-}
+	// TODO(@rtm0): Remove this dependency from vmalert-tool unit tests.
+	DebugFlush func()
+)
 
 // Stop stops the vmstorage
 func Stop() {
@@ -311,19 +210,24 @@ func Stop() {
 	metrics.UnregisterSet(storageMetrics, true)
 	storageMetrics = nil
 
-	logger.Infof("gracefully closing the storage at %s", *DataPath)
+	logger.Infof("gracefully closing the storage at %s", *storageDataPath)
 	startTime := time.Now()
-	WG.WaitAndBlock()
-	stopStaleSnapshotsRemover()
-	Storage.MustClose()
+	vmStorage.Stop()
 	logger.Infof("successfully closed the storage in %.3f seconds", time.Since(startTime).Seconds())
 
 	fs.MustStopDirRemover()
-	logger.Infof("the storage has been stopped")
+	logger.Infof("the vmstorage has been stopped")
 }
 
-// RequestHandler is a storage request handler.
-func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
+func (api *VMStorageSingleNode) requestHandler(w http.ResponseWriter, r *http.Request) bool {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.requestHandler(w, r)
+}
+
+// requestHandler is a storage request handler.
+// TODO(@rtm0): Move to a separate file, request_handler.go
+func (api *VMStorage) requestHandler(w http.ResponseWriter, r *http.Request) bool {
 	path := r.URL.Path
 	if path == "/internal/force_merge" {
 		if !httpserver.CheckAuthFlag(w, r, forceMergeAuthKey) {
@@ -336,8 +240,9 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 			defer activeForceMerges.Dec()
 			logger.Infof("forced merge for partition_prefix=%q has been started", partitionNamePrefix)
 			startTime := time.Now()
-			if err := Storage.ForceMergePartitions(partitionNamePrefix); err != nil {
+			if err := api.s.ForceMergePartitions(partitionNamePrefix); err != nil {
 				logger.Errorf("error in forced merge for partition_prefix=%q: %s", partitionNamePrefix, err)
+				return
 			}
 			logger.Infof("forced merge for partition_prefix=%q has been successfully finished in %.3f seconds", partitionNamePrefix, time.Since(startTime).Seconds())
 		}()
@@ -348,9 +253,10 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 			return true
 		}
 		logger.Infof("flushing storage to make pending data available for reading")
-		Storage.DebugFlush()
+		api.s.DebugFlush()
 		return true
 	}
+
 	if path == "/internal/log_new_series" {
 		if !httpserver.CheckAuthFlag(w, r, logNewSeriesAuthKey) {
 			return true
@@ -367,7 +273,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		}
 		logger.Infof("enabling logging of new series for the next %s. This may increase resource usage during this period.", time.Duration(dealine)*time.Second)
 		endTime := fasttime.UnixTimestamp() + uint64(dealine)
-		Storage.SetLogNewSeriesUntil(endTime)
+		api.s.SetLogNewSeriesUntil(endTime)
 		fmt.Fprintf(w, `{"status":"success","data":{"logEndTime":%q}}`, time.Unix(int64(endTime), 0))
 		return true
 	}
@@ -389,13 +295,13 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 	case "/create":
 		snapshotsCreateTotal.Inc()
 		w.Header().Set("Content-Type", "application/json")
-		snapshotName := Storage.MustCreateSnapshot()
+		snapshotName := api.s.MustCreateSnapshot()
 
 		// Verify whether the client already closed the connection.
 		// In this case it is better to drop the created snapshot, since the client isn't interested in it.
 		if err := r.Context().Err(); err != nil {
 			logger.Infof("deleting already created snapshot at %s because the client canceled the request", snapshotName)
-			if err := deleteSnapshot(snapshotName); err != nil {
+			if err := api.deleteSnapshot(snapshotName); err != nil {
 				logger.Infof("cannot delete just created snapshot: %s", err)
 				return true
 			}
@@ -411,7 +317,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 	case "/list":
 		snapshotsListTotal.Inc()
 		w.Header().Set("Content-Type", "application/json")
-		snapshots := Storage.MustListSnapshots()
+		snapshots := api.s.MustListSnapshots()
 		fmt.Fprintf(w, `{"status":"ok","snapshots":[`)
 		if len(snapshots) > 0 {
 			for _, snapshot := range snapshots[:len(snapshots)-1] {
@@ -425,7 +331,7 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 		snapshotsDeleteTotal.Inc()
 		w.Header().Set("Content-Type", "application/json")
 		snapshotName := r.FormValue("snapshot")
-		if err := deleteSnapshot(snapshotName); err != nil {
+		if err := api.deleteSnapshot(snapshotName); err != nil {
 			jsonResponseError(w, err)
 			snapshotsDeleteErrorsTotal.Inc()
 			return true
@@ -435,9 +341,10 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 	case "/delete_all":
 		snapshotsDeleteAllTotal.Inc()
 		w.Header().Set("Content-Type", "application/json")
-		snapshots := Storage.MustListSnapshots()
+		snapshots := api.s.MustListSnapshots()
 		for _, snapshotName := range snapshots {
-			if err := Storage.DeleteSnapshot(snapshotName); err != nil {
+			// TODO(@rtm0): Use VMStorage.deleteSnapshot()?
+			if err := api.s.DeleteSnapshot(snapshotName); err != nil {
 				err = fmt.Errorf("cannot delete snapshot %q: %w", snapshotName, err)
 				jsonResponseError(w, err)
 				snapshotsDeleteAllErrorsTotal.Inc()
@@ -451,50 +358,6 @@ func RequestHandler(w http.ResponseWriter, r *http.Request) bool {
 	}
 }
 
-func deleteSnapshot(snapshotName string) error {
-	snapshots := Storage.MustListSnapshots()
-	for _, snName := range snapshots {
-		if snName == snapshotName {
-			if err := Storage.DeleteSnapshot(snName); err != nil {
-				return fmt.Errorf("cannot delete snapshot %q: %w", snName, err)
-			}
-			return nil
-		}
-	}
-	return fmt.Errorf("cannot find snapshot %q", snapshotName)
-}
-
-func initStaleSnapshotsRemover(strg *storage.Storage) {
-	staleSnapshotsRemoverCh = make(chan struct{})
-	if snapshotsMaxAge.Duration() <= 0 {
-		return
-	}
-	snapshotsMaxAgeDur := snapshotsMaxAge.Duration()
-	staleSnapshotsRemoverWG.Go(func() {
-		d := timeutil.AddJitterToDuration(time.Second * 11)
-		t := time.NewTicker(d)
-		defer t.Stop()
-		for {
-			select {
-			case <-staleSnapshotsRemoverCh:
-				return
-			case <-t.C:
-			}
-			strg.MustDeleteStaleSnapshots(snapshotsMaxAgeDur)
-		}
-	})
-}
-
-func stopStaleSnapshotsRemover() {
-	close(staleSnapshotsRemoverCh)
-	staleSnapshotsRemoverWG.Wait()
-}
-
-var (
-	staleSnapshotsRemoverCh chan struct{}
-	staleSnapshotsRemoverWG sync.WaitGroup
-)
-
 var (
 	activeForceMerges = metrics.NewCounter("vm_active_force_merges")
 
@@ -509,21 +372,23 @@ var (
 	snapshotsDeleteAllErrorsTotal = metrics.NewCounter(`vm_http_request_errors_total{path="/snapshot/delete_all"}`)
 )
 
-func writeStorageMetrics(w io.Writer, strg *storage.Storage) {
+// TODO(@rtm0): Move to metrics.go.
+func (api *VMStorage) writeStorageMetrics(w io.Writer) {
+	strg := api.s
 	var m storage.Metrics
 	strg.UpdateMetrics(&m)
 	tm := &m.TableMetrics
 	idbm := &m.TableMetrics.IndexDBMetrics
 
-	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_free_disk_space_bytes{path=%q}`, *DataPath), fs.MustGetFreeSpace(*DataPath))
-	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_free_disk_space_limit_bytes{path=%q}`, *DataPath), uint64(minFreeDiskSpaceBytes.N))
-	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_total_disk_space_bytes{path=%q}`, *DataPath), fs.MustGetTotalSpace(*DataPath))
+	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_free_disk_space_bytes{path=%q}`, *storageDataPath), fs.MustGetFreeSpace(*storageDataPath))
+	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_free_disk_space_limit_bytes{path=%q}`, *storageDataPath), uint64(minFreeDiskSpaceBytes.N))
+	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_total_disk_space_bytes{path=%q}`, *storageDataPath), fs.MustGetTotalSpace(*storageDataPath))
 
 	isReadOnly := 0
 	if strg.IsReadOnly() {
 		isReadOnly = 1
 	}
-	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_storage_is_read_only{path=%q}`, *DataPath), uint64(isReadOnly))
+	metrics.WriteGaugeUint64(w, fmt.Sprintf(`vm_storage_is_read_only{path=%q}`, *storageDataPath), uint64(isReadOnly))
 
 	metrics.WriteGaugeUint64(w, `vm_active_merges{type="storage/inmemory"}`, tm.ActiveInmemoryMerges)
 	metrics.WriteGaugeUint64(w, `vm_active_merges{type="storage/small"}`, tm.ActiveSmallMerges)
@@ -733,6 +598,8 @@ func writeStorageMetrics(w io.Writer, strg *storage.Storage) {
 	metrics.WriteGaugeUint64(w, `vm_downsampling_partitions_scheduled`, tm.ScheduledDownsamplingPartitions)
 	metrics.WriteGaugeUint64(w, `vm_downsampling_partitions_scheduled_size_bytes`, tm.ScheduledDownsamplingPartitionsSize)
 
+	metrics.WriteGaugeUint64(w, `vm_search_max_unique_timeseries`, uint64(api.maxUniqueTimeSeriesCalculated))
+
 	metrics.WriteGaugeUint64(w, `vm_metrics_metadata_storage_items`, m.MetadataStorageItemsCurrent)
 	metrics.WriteCounterUint64(w, `vm_metrics_metadata_storage_size_bytes`, m.MetadataStorageCurrentSizeBytes)
 	metrics.WriteCounterUint64(w, `vm_metrics_metadata_storage_max_size_bytes`, m.MetadataStorageMaxSizeBytes)

app/vmstorage/vmstorage.go

@@ -0,0 +1,396 @@
+package vmstorage
+
+import (
+	"flag"
+	"fmt"
+	"sync"
+	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/memory"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricnamestats"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricsmetadata"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/vmselectapi"
+)
+
+var (
+	precisionBits = flag.Int("precisionBits", 64, "The number of precision bits to store per each value. Lower precision bits improves data compression "+
+		"at the cost of precision loss")
+	maxUniqueTimeseries = flag.Int("search.maxUniqueTimeseries", 0, "The maximum number of unique time series, which can be scanned during every query. "+
+		"This allows protecting against heavy queries, which select unexpectedly high number of series. When set to zero, the limit is automatically calculated based on -search.maxConcurrentRequests (inversely proportional) and memory available to the process (proportional). See also -search.max* command-line flags at vmselect")
+	maxTagKeys = flag.Int("search.maxTagKeys", 100e3, "The maximum number of tag keys returned per search. "+
+		"See also -search.maxLabelsAPISeries and -search.maxLabelsAPIDuration")
+	maxTagValues = flag.Int("search.maxTagValues", 100e3, "The maximum number of tag values returned per search. "+
+		"See also -search.maxLabelsAPISeries and -search.maxLabelsAPIDuration")
+	maxTagValueSuffixesPerSearch = flag.Int("search.maxTagValueSuffixesPerSearch", 100e3, "The maximum number of tag value suffixes returned from /metrics/find")
+	snapshotsMaxAge              = flagutil.NewRetentionDuration("snapshotsMaxAge", "3d", "Automatically delete snapshots older than -snapshotsMaxAge if it is set to non-zero duration. Make sure that backup process has enough time to finish the backup before the corresponding snapshot is automatically deleted")
+)
+
+// newVMStorage creates a new instance of of VMStorage.
+//
+// The created VMStorage instance takes ownership of s.
+func newVMStorage(s *storage.Storage, maxConcurrentRequests int) *VMStorage {
+	if err := encoding.CheckPrecisionBits(uint8(*precisionBits)); err != nil {
+		logger.Fatalf("invalid -precisionBits: %d", err)
+	}
+
+	maxUniqueTimeseriesCalculated := *maxUniqueTimeseries
+	if maxUniqueTimeseriesCalculated <= 0 {
+		maxUniqueTimeseriesCalculated = calculateMaxUniqueTimeseries(maxConcurrentRequests, memory.Remaining())
+	}
+
+	vms := &VMStorage{
+		s:                             s,
+		maxUniqueTimeseries:           *maxUniqueTimeseries,
+		maxUniqueTimeSeriesCalculated: maxUniqueTimeseriesCalculated,
+		staleSnapshotsRemoverCh:       make(chan struct{}),
+	}
+	vms.initStaleSnapshotsRemover()
+	return vms
+}
+
+// calculateMaxUniqueTimeseries calculates the maxUniqueTimeseries based on the
+// available system resources.
+func calculateMaxUniqueTimeseries(maxConcurrentRequests, remainingMemory int) int {
+	if maxConcurrentRequests <= 0 {
+		// This line should NOT be reached unless the user has set an incorrect `search.maxConcurrentRequests`.
+		// In such cases, fallback to unlimited.
+		logger.Warnf("limiting -search.maxUniqueTimeseries to %v because -search.maxConcurrentRequests=%d.", 2e9, maxConcurrentRequests)
+		return 2e9
+	}
+
+	// Calculate the max metrics limit for a single request in the worst-case concurrent scenario.
+	// The approximate size of 1 unique series that could occupy in the vmstorage is 200 bytes.
+	mts := remainingMemory / 200 / maxConcurrentRequests
+	logger.Infof("limiting -search.maxUniqueTimeseries to %d according to -search.maxConcurrentRequests=%d and remaining memory=%d bytes. To increase the limit, reduce -search.maxConcurrentRequests or increase memory available to the process.", mts, maxConcurrentRequests, remainingMemory)
+	return mts
+}
+
+// VMStorage impelements vmselectapi.API and vminsertapi.API.
+type VMStorage struct {
+	s                             *storage.Storage
+	maxUniqueTimeseries           int
+	maxUniqueTimeSeriesCalculated int
+	staleSnapshotsRemoverCh       chan struct{}
+	staleSnapshotsRemoverWG       sync.WaitGroup
+}
+
+func (api *VMStorage) initStaleSnapshotsRemover() {
+	if snapshotsMaxAge.Duration() <= 0 {
+		return
+	}
+	snapshotsMaxAgeDuration := snapshotsMaxAge.Duration()
+	api.staleSnapshotsRemoverWG.Go(func() {
+		d := timeutil.AddJitterToDuration(time.Second * 11)
+		t := time.NewTicker(d)
+		defer t.Stop()
+		for {
+			select {
+			case <-api.staleSnapshotsRemoverCh:
+				return
+			case <-t.C:
+			}
+			api.s.MustDeleteStaleSnapshots(snapshotsMaxAgeDuration)
+		}
+	})
+}
+
+func (api *VMStorage) Stop() {
+	close(api.staleSnapshotsRemoverCh)
+	api.staleSnapshotsRemoverWG.Wait()
+	api.s.MustClose()
+}
+
+// WriteRows writes metric rows to the storage.
+//
+// The caller should limit the number of concurrent calls to WriteRows() in
+// order to limit memory usage.
+func (api *VMStorage) WriteRows(rows []storage.MetricRow) error {
+	api.s.AddRows(rows, uint8(*precisionBits))
+	return nil
+}
+
+// WriteMetadata writes metrics metadata to storage.
+//
+// The caller should limit the number of concurrent calls to WriteMetadata() in
+// order to limit memory usage.
+func (api *VMStorage) WriteMetadata(rows []metricsmetadata.Row) error {
+	api.s.AddMetadataRows(rows)
+	return nil
+}
+
+// IsReadOnly returns true is the storage is in read-only mode.
+func (api *VMStorage) IsReadOnly() bool {
+	return api.s.IsReadOnly()
+}
+
+func (api *VMStorage) InitSearch(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (vmselectapi.BlockIterator, error) {
+	tr := sq.GetTimeRange()
+	maxMetrics := api.getMaxMetrics(sq.MaxMetrics)
+	tfss, err := api.setupTfss(qt, sq, tr, maxMetrics, deadline)
+	if err != nil {
+		return nil, err
+	}
+	if len(tfss) == 0 {
+		return nil, fmt.Errorf("missing tag filters")
+	}
+	bi := getBlockIterator()
+	bi.sr.Init(qt, api.s, tfss, tr, maxMetrics, deadline)
+	if err := bi.sr.Error(); err != nil {
+		bi.MustClose()
+		return nil, err
+	}
+	return bi, nil
+}
+
+func (api *VMStorage) getMaxMetrics(searchQueryLimit int) int {
+	if searchQueryLimit <= 0 {
+		return api.maxUniqueTimeSeriesCalculated
+	}
+	// searchQueryLimit cannot exceed `-search.maxUniqueTimeseries`
+	if api.maxUniqueTimeseries != 0 && searchQueryLimit > api.maxUniqueTimeseries {
+		searchQueryLimit = api.maxUniqueTimeseries
+	}
+	return searchQueryLimit
+}
+
+// blockIterator implements vmselectapi.BlockIterator
+type blockIterator struct {
+	sr storage.Search
+	mb storage.MetricBlock
+}
+
+var blockIteratorsPool sync.Pool
+
+func (bi *blockIterator) MustClose() {
+	bi.sr.MustClose()
+	bi.mb.MetricName = nil
+	bi.mb.Block.Reset()
+	blockIteratorsPool.Put(bi)
+}
+
+func getBlockIterator() *blockIterator {
+	v := blockIteratorsPool.Get()
+	if v == nil {
+		v = &blockIterator{}
+	}
+	return v.(*blockIterator)
+}
+
+func (bi *blockIterator) NextBlock(dst []byte) ([]byte, bool) {
+	if !bi.sr.NextMetricBlock() {
+		return dst, false
+	}
+	mb := bi.mb
+	mb.MetricName = bi.sr.MetricBlockRef.MetricName
+	bi.sr.MetricBlockRef.BlockRef.MustReadBlock(&mb.Block)
+	dst = mb.Marshal(dst[:0])
+	return dst, true
+}
+
+func (bi *blockIterator) Error() error {
+	return bi.sr.Error()
+}
+
+// SearchMetricNames returns metric names for the given tfss on the given tr.
+func (api *VMStorage) SearchMetricNames(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) ([]string, error) {
+	tr := sq.GetTimeRange()
+	maxMetrics := sq.MaxMetrics
+	if maxMetrics <= 0 {
+		// fallback to maxUniqueTimeSeries if no limit is provided,
+		// see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/7857
+		maxMetrics = api.maxUniqueTimeSeriesCalculated
+	}
+	tfss, err := api.setupTfss(qt, sq, tr, maxMetrics, deadline)
+	if err != nil {
+		return nil, err
+	}
+	if len(tfss) == 0 {
+		return nil, fmt.Errorf("missing tag filters")
+	}
+	return api.s.SearchMetricNames(qt, tfss, tr, maxMetrics, deadline)
+}
+
+// SearchLabelValues searches for label values for the given labelName, tfss and
+// tr.
+func (api *VMStorage) LabelValues(qt *querytracer.Tracer, sq *storage.SearchQuery, labelName string, maxLabelValues int, deadline uint64) ([]string, error) {
+	tr := sq.GetTimeRange()
+	if maxLabelValues <= 0 || maxLabelValues > *maxTagValues {
+		maxLabelValues = *maxTagValues
+	}
+	maxMetrics := sq.MaxMetrics
+	if maxMetrics <= 0 {
+		// fallback to maxUniqueTimeSeries if no limit is provided,
+		// see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/7857
+		maxMetrics = api.maxUniqueTimeSeriesCalculated
+	}
+	tfss, err := api.setupTfss(qt, sq, tr, maxMetrics, deadline)
+	if err != nil {
+		return nil, err
+	}
+	return api.s.SearchLabelValues(qt, labelName, tfss, tr, maxLabelValues, maxMetrics, deadline)
+}
+
+// TagValueSuffixes returns all the tag value suffixes for the given tagKey and
+// tagValuePrefix on the given tr.
+//
+// This allows implementing
+// https://graphite-api.readthedocs.io/en/latest/api.html#metrics-find or
+// similar APIs.
+func (api *VMStorage) TagValueSuffixes(qt *querytracer.Tracer, _, _ uint32, tr storage.TimeRange, tagKey, tagValuePrefix string, delimiter byte,
+	maxSuffixes int, deadline uint64) ([]string, error) {
+	if maxSuffixes <= 0 || maxSuffixes > *maxTagValueSuffixesPerSearch {
+		maxSuffixes = *maxTagValueSuffixesPerSearch
+	}
+	suffixes, err := api.s.SearchTagValueSuffixes(qt, tr, tagKey, tagValuePrefix, delimiter, maxSuffixes, deadline)
+	if err != nil {
+		return nil, err
+	}
+	if len(suffixes) >= maxSuffixes {
+		return nil, fmt.Errorf("more than -search.maxTagValueSuffixesPerSearch=%d suffixes returned; "+
+			"either narrow down the search or increase -search.maxTagValueSuffixesPerSearch command-line flag value", maxSuffixes)
+	}
+	return suffixes, nil
+}
+
+// SearchLabelNames searches for tag keys matching the given tfss on tr.
+func (api *VMStorage) LabelNames(qt *querytracer.Tracer, sq *storage.SearchQuery, maxLabelNames int, deadline uint64) ([]string, error) {
+	tr := sq.GetTimeRange()
+	if maxLabelNames <= 0 || maxLabelNames > *maxTagKeys {
+		maxLabelNames = *maxTagKeys
+	}
+	maxMetrics := sq.MaxMetrics
+	if maxMetrics <= 0 {
+		// fallback to maxUniqueTimeSeries if no limit is provided,
+		// see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/7857
+		maxMetrics = api.maxUniqueTimeSeriesCalculated
+	}
+	tfss, err := api.setupTfss(qt, sq, tr, maxMetrics, deadline)
+	if err != nil {
+		return nil, err
+	}
+	return api.s.SearchLabelNames(qt, tfss, tr, maxLabelNames, maxMetrics, deadline)
+}
+
+func (api *VMStorage) SeriesCount(_ *querytracer.Tracer, _, _ uint32, deadline uint64) (uint64, error) {
+	return api.s.GetSeriesCount(deadline)
+}
+
+func (api *VMStorage) Tenants(_ *querytracer.Tracer, _ storage.TimeRange, _ uint64) ([]string, error) {
+	return nil, nil
+}
+
+// GetTSDBStatus returns TSDB status for given filters on the given date.
+func (api *VMStorage) TSDBStatus(qt *querytracer.Tracer, sq *storage.SearchQuery, focusLabel string, topN int, deadline uint64) (*storage.TSDBStatus, error) {
+	tr := sq.GetTimeRange()
+	maxMetrics := sq.MaxMetrics
+	if maxMetrics <= 0 {
+		// fallback to maxUniqueTimeSeries if no limit is provided,
+		// see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/7857
+		maxMetrics = api.maxUniqueTimeSeriesCalculated
+	}
+	tfss, err := api.setupTfss(qt, sq, tr, maxMetrics, deadline)
+	if err != nil {
+		return nil, err
+	}
+	date := uint64(sq.MinTimestamp) / (24 * 3600 * 1000)
+	return api.s.GetTSDBStatus(qt, tfss, date, focusLabel, topN, maxMetrics, deadline)
+}
+
+// DeleteSeries deletes series matching tfss.
+//
+// Returns the number of deleted series.
+func (api *VMStorage) DeleteSeries(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (int, error) {
+	// TODO(@rtm0): Return an error if the storage is in read-only mode?
+
+	tr := sq.GetTimeRange()
+	maxMetrics := sq.MaxMetrics
+	if maxMetrics <= 0 {
+		// fallback to maxUniqueTimeSeries if no limit is provided,
+		// see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/7857
+		maxMetrics = api.maxUniqueTimeSeriesCalculated
+	}
+	tfss, err := api.setupTfss(qt, sq, tr, maxMetrics, deadline)
+	if err != nil {
+		return 0, err
+	}
+	if len(tfss) == 0 {
+		return 0, fmt.Errorf("missing tag filters")
+	}
+	return api.s.DeleteSeries(qt, tfss, maxMetrics)
+}
+
+func (api *VMStorage) RegisterMetricNames(qt *querytracer.Tracer, mrs []storage.MetricRow, _ uint64) error {
+	// TODO(@rtm0): Return an error if the storage is in read-only mode?
+
+	api.s.RegisterMetricNames(qt, mrs)
+	return nil
+}
+
+// GetMetricNamesUsageStats returns metric name usage stats.
+func (api *VMStorage) GetMetricNamesUsageStats(qt *querytracer.Tracer, _ *storage.TenantToken, limit, le int, matchPattern string, _ uint64) (metricnamestats.StatsResult, error) {
+
+	return api.s.GetMetricNamesStats(qt, limit, le, matchPattern), nil
+}
+
+// ResetMetricNamesStats resets state for metric names usage tracker
+func (api *VMStorage) ResetMetricNamesUsageStats(qt *querytracer.Tracer, _ uint64) error {
+	api.s.ResetMetricNamesStats(qt)
+	return nil
+}
+
+func (api *VMStorage) setupTfss(qt *querytracer.Tracer, sq *storage.SearchQuery, tr storage.TimeRange, maxMetrics int, deadline uint64) ([]*storage.TagFilters, error) {
+	tfss := make([]*storage.TagFilters, 0, len(sq.TagFilterss))
+	for _, tagFilters := range sq.TagFilterss {
+		tfs := storage.NewTagFilters()
+		for i := range tagFilters {
+			tf := &tagFilters[i]
+			if string(tf.Key) == "__graphite__" {
+				query := tf.Value
+				qtChild := qt.NewChild("searching for series matching __graphite__=%q", query)
+				paths, err := api.s.SearchGraphitePaths(qtChild, tr, query, maxMetrics, deadline)
+				qtChild.Donef("found %d series", len(paths))
+				if err != nil {
+					return nil, fmt.Errorf("error when searching for Graphite paths for query %q: %w", query, err)
+				}
+				if len(paths) >= maxMetrics {
+					return nil, fmt.Errorf("more than %d time series match Graphite query %q; "+
+						"either narrow down the query or increase the corresponding -search.max* command-line flag value at vmselect nodes; "+
+						"see https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#resource-usage-limits", maxMetrics, query)
+				}
+				tfs.AddGraphiteQuery(query, paths, tf.IsNegative)
+				continue
+			}
+			if err := tfs.Add(tf.Key, tf.Value, tf.IsNegative, tf.IsRegexp); err != nil {
+				return nil, fmt.Errorf("cannot parse tag filter %s: %w", tf, err)
+			}
+		}
+		tfss = append(tfss, tfs)
+	}
+	return tfss, nil
+}
+
+func (api *VMStorage) GetMetadataRecords(qt *querytracer.Tracer, _ *storage.TenantToken, limit int, metricName string, _ uint64) ([]*metricsmetadata.Row, error) {
+	return api.s.GetMetadataRows(qt, limit, metricName), nil
+}
+
+// deleteSnapshot deletes a snapshot by its name.
+//
+// Callers must wrap the call with wg.Add(1)...wg.Done().
+func (api *VMStorage) deleteSnapshot(snapshotName string) error {
+	snapshots := api.s.MustListSnapshots()
+	for _, snName := range snapshots {
+		if snName == snapshotName {
+			if err := api.s.DeleteSnapshot(snName); err != nil {
+				return fmt.Errorf("cannot delete snapshot %q: %w", snName, err)
+			}
+			return nil
+		}
+	}
+	return fmt.Errorf("cannot find snapshot %q", snapshotName)
+}

app/vmstorage/vmstorage_single_node.go

@@ -0,0 +1,202 @@
+package vmstorage
+
+import (
+	"errors"
+	"sync"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricnamestats"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricsmetadata"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/syncwg"
+)
+
+// newVMStorageSingleNode creates a new instance of of VMStorage for vmsingle.
+func newVMStorageSingleNode(s *storage.Storage, maxConcurrentRequests int, resetCacheIfNeeded func(mrs []storage.MetricRow)) *VMStorageSingleNode {
+	vms := newVMStorage(s, maxConcurrentRequests)
+	return &VMStorageSingleNode{
+		VMStorage:          vms,
+		wg:                 syncwg.WaitGroup{},
+		resetCacheIfNeeded: resetCacheIfNeeded,
+	}
+}
+
+type VMStorageSingleNode struct {
+	*VMStorage
+
+	// wg is used to wrap every storage call into wg.Add(1) ... wg.Done()
+	// for proper graceful shutdown when Stop is called.
+	//
+	// Use syncwg instead of sync, since Add is called from concurrent
+	// goroutines.
+	wg syncwg.WaitGroup
+
+	// resetCacheIfNeeded is a callback for automatic resetting of response
+	// cache if needed.
+	resetCacheIfNeeded func(mrs []storage.MetricRow)
+}
+
+func (api *VMStorageSingleNode) Stop() {
+	api.wg.WaitAndBlock()
+	api.VMStorage.Stop()
+}
+
+// WriteRows writes metric rows to the storage.
+//
+// Returns an error if the storage is in read-only mode.
+//
+// The caller should limit the number of concurrent calls to WriteRows() in
+// order to limit memory usage.
+func (api *VMStorageSingleNode) WriteRows(rows []storage.MetricRow) error {
+	api.wg.Add(1)
+	defer api.wg.Done()
+
+	if api.s.IsReadOnly() {
+		return errReadOnly
+	}
+	api.resetCacheIfNeeded(rows)
+	return api.VMStorage.WriteRows(rows)
+}
+
+// WriteMetadata writes metrics metadata to storage.
+//
+// Returns an error if the storage is in read-only mode.
+//
+// The caller should limit the number of concurrent calls to WriteMetadata() in
+// order to limit memory usage.
+func (api *VMStorageSingleNode) WriteMetadata(rows []metricsmetadata.Row) error {
+	api.wg.Add(1)
+	defer api.wg.Done()
+
+	if api.s.IsReadOnly() {
+		return errReadOnly
+	}
+	return api.VMStorage.WriteMetadata(rows)
+}
+
+var errReadOnly = errors.New("the storage is in read-only mode; check -storage.minFreeDiskSpaceBytes command-line flag value")
+
+func (api *VMStorageSingleNode) IsReadOnly() bool {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.IsReadOnly()
+}
+
+// GetSearch sets up an instance of storage search and returns it to the caller
+// along with the max series count that the search can return.
+//
+// This method is not part of the vmselectapi.API and must only be used by
+// vmsingle HTTP handlers.
+//
+// Callers of this method must call PutSearch() once the search instance is not
+// needed anymore.
+func (api *VMStorageSingleNode) GetSearch(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (*storage.Search, int, error) {
+	api.wg.Add(1)
+
+	tr := sq.GetTimeRange()
+	maxMetrics := api.getMaxMetrics(sq.MaxMetrics)
+	tfss, err := api.setupTfss(qt, sq, tr, maxMetrics, deadline)
+	if err != nil {
+		return nil, 0, err
+	}
+
+	sr := getSearch()
+	maxSeriesCount := sr.Init(qt, api.s, tfss, tr, sq.MaxMetrics, deadline)
+	return sr, maxSeriesCount, nil
+}
+
+// PutSearch resets the search once it is not needed anymore and puts it aside
+// for future reuse.
+//
+// This method is not part of the vmselectapi.API and must only be used by
+// vmsingle HTTP handlers.
+//
+// The method must only be used on search instances that have been created with
+// GetSearch().
+func (api *VMStorageSingleNode) PutSearch(sr *storage.Search) {
+	api.wg.Done()
+	putSearch(sr)
+}
+
+func getSearch() *storage.Search {
+	v := ssPool.Get()
+	if v == nil {
+		return &storage.Search{}
+	}
+	return v.(*storage.Search)
+}
+
+func putSearch(sr *storage.Search) {
+	sr.MustClose()
+	ssPool.Put(sr)
+}
+
+var ssPool sync.Pool
+
+func (api *VMStorageSingleNode) SearchMetricNames(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) ([]string, error) {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.SearchMetricNames(qt, sq, deadline)
+}
+
+func (api *VMStorageSingleNode) LabelValues(qt *querytracer.Tracer, sq *storage.SearchQuery, labelName string, maxLabelValues int, deadline uint64) ([]string, error) {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.LabelValues(qt, sq, labelName, maxLabelValues, deadline)
+}
+
+func (api *VMStorageSingleNode) TagValueSuffixes(qt *querytracer.Tracer, accountID, projectID uint32, tr storage.TimeRange, tagKey, tagValuePrefix string, delimiter byte, maxSuffixes int, deadline uint64) ([]string, error) {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.TagValueSuffixes(qt, accountID, projectID, tr, tagKey, tagValuePrefix, delimiter, maxSuffixes, deadline)
+}
+
+func (api *VMStorageSingleNode) LabelNames(qt *querytracer.Tracer, sq *storage.SearchQuery, maxLabelNames int, deadline uint64) ([]string, error) {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.LabelNames(qt, sq, maxLabelNames, deadline)
+}
+
+func (api *VMStorageSingleNode) SeriesCount(qt *querytracer.Tracer, accountID, projectID uint32, deadline uint64) (uint64, error) {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.SeriesCount(qt, accountID, projectID, deadline)
+}
+
+func (api *VMStorageSingleNode) TSDBStatus(qt *querytracer.Tracer, sq *storage.SearchQuery, focusLabel string, topN int, deadline uint64) (*storage.TSDBStatus, error) {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.TSDBStatus(qt, sq, focusLabel, topN, deadline)
+}
+
+func (api *VMStorageSingleNode) DeleteSeries(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (int, error) {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	// TODO(@rtm0): Return an error if the storage is in read-only mode?
+	return api.VMStorage.DeleteSeries(qt, sq, deadline)
+}
+
+func (api *VMStorageSingleNode) RegisterMetricNames(qt *querytracer.Tracer, mrs []storage.MetricRow, deadline uint64) error {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	// TODO(@rtm0): Return an error if the storage is in read-only mode?
+	return api.VMStorage.RegisterMetricNames(qt, mrs, deadline)
+}
+
+func (api *VMStorageSingleNode) GetMetricNamesUsageStats(qt *querytracer.Tracer, tt *storage.TenantToken, limit, le int, matchPattern string, deadline uint64) (metricnamestats.StatsResult, error) {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.GetMetricNamesUsageStats(qt, tt, limit, le, matchPattern, deadline)
+}
+
+func (api *VMStorageSingleNode) ResetMetricNamesUsageStats(qt *querytracer.Tracer, deadline uint64) error {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.ResetMetricNamesUsageStats(qt, deadline)
+}
+
+func (api *VMStorageSingleNode) GetMetadataRecords(qt *querytracer.Tracer, tt *storage.TenantToken, limit int, metricName string, deadline uint64) ([]*metricsmetadata.Row, error) {
+	api.wg.Add(1)
+	defer api.wg.Done()
+	return api.VMStorage.GetMetadataRecords(qt, tt, limit, metricName, deadline)
+}

apptest/app.go

@@ -22,6 +22,7 @@ var (
 	vminsertAddrRE              = regexp.MustCompile(`accepting vminsert conns at (.*:\d{1,5})$`)
 	vminsertClusterNativeAddrRE = regexp.MustCompile(`started TCP clusternative server at "(.*:\d{1,5})"`)
 	vmselectAddrRE              = regexp.MustCompile(`accepting vmselect conns at (.*:\d{1,5})$`)
+	vmauthHttpListenAddrRE      = regexp.MustCompile(`pprof handlers are exposed at http://(.*:\d{1,5})/debug/pprof/`)
 )
 
 // app represents an instance of some VictoriaMetrics server (such as vmstorage,

apptest/testcase.go

@@ -88,19 +88,11 @@ func (tc *TestCase) MustStartDefaultVmsingle() *Vmsingle {
 }
 
 // MustStartVmsingle is a test helper function that starts an instance of
-// vmsingle located at ../../bin/victoria-metrics-race and fails the test if the app
-// fails to start.
+// vmsingle (latest version) and fails the test if the app fails to start.
 func (tc *TestCase) MustStartVmsingle(instance string, flags []string) *Vmsingle {
 	tc.t.Helper()
-	return tc.MustStartVmsingleAt(instance, "../../bin/victoria-metrics-race", flags)
-}
 
-// MustStartVmsingleAt is a test helper function that starts an instance of
-// vmsingle and fails the test if the app fails to start.
-func (tc *TestCase) MustStartVmsingleAt(instance, binary string, flags []string) *Vmsingle {
-	tc.t.Helper()
-
-	app, err := StartVmsingleAt(instance, binary, flags, tc.cli, tc.output)
+	app, err := StartVmsingle(instance, flags, tc.cli, tc.output)
 	if err != nil {
 		tc.t.Fatalf("Could not start %s: %v", instance, err)
 	}
@@ -109,19 +101,11 @@ func (tc *TestCase) MustStartVmsingleAt(instance, binary string, flags []string)
 }
 
 // MustStartVmstorage is a test helper function that starts an instance of
-// vmstorage located at ../../bin/vmstorage-race and fails the test if the app fails
-// to start.
+// vmstorage (latest version) and fails the test if the app fails to start.
 func (tc *TestCase) MustStartVmstorage(instance string, flags []string) *Vmstorage {
 	tc.t.Helper()
-	return tc.MustStartVmstorageAt(instance, "../../bin/vmstorage-race", flags)
-}
 
-// MustStartVmstorageAt is a test helper function that starts an instance of
-// vmstorage and fails the test if the app fails to start.
-func (tc *TestCase) MustStartVmstorageAt(instance string, binary string, flags []string) *Vmstorage {
-	tc.t.Helper()
-
-	app, err := StartVmstorageAt(instance, binary, flags, tc.cli, tc.output)
+	app, err := StartVmstorage(instance, flags, tc.cli, tc.output)
 	if err != nil {
 		tc.t.Fatalf("Could not start %s: %v", instance, err)
 	}
@@ -130,7 +114,7 @@ func (tc *TestCase) MustStartVmstorageAt(instance string, binary string, flags [
 }
 
 // MustStartVmselect is a test helper function that starts an instance of
-// vmselect and fails the test if the app fails to start.
+// vmselect (latest version) and fails the test if the app fails to start.
 func (tc *TestCase) MustStartVmselect(instance string, flags []string) *Vmselect {
 	tc.t.Helper()
 
@@ -290,10 +274,8 @@ func (tc *TestCase) MustStartDefaultCluster() *Vmcluster {
 // tests usually come paired with corresponding vmsingle tests.
 type ClusterOptions struct {
 	Vmstorage1Instance string
-	Vmstorage1Binary   string
 	Vmstorage1Flags    []string
 	Vmstorage2Instance string
-	Vmstorage2Binary   string
 	Vmstorage2Flags    []string
 	VminsertInstance   string
 	VminsertFlags      []string
@@ -305,15 +287,8 @@ type ClusterOptions struct {
 func (tc *TestCase) MustStartCluster(opts *ClusterOptions) *Vmcluster {
 	tc.t.Helper()
 
-	if opts.Vmstorage1Binary == "" {
-		opts.Vmstorage1Binary = "../../bin/vmstorage-race"
-	}
-	vmstorage1 := tc.MustStartVmstorageAt(opts.Vmstorage1Instance, opts.Vmstorage1Binary, opts.Vmstorage1Flags)
-
-	if opts.Vmstorage2Binary == "" {
-		opts.Vmstorage2Binary = "../../bin/vmstorage-race"
-	}
-	vmstorage2 := tc.MustStartVmstorageAt(opts.Vmstorage2Instance, opts.Vmstorage2Binary, opts.Vmstorage2Flags)
+	vmstorage1 := tc.MustStartVmstorage(opts.Vmstorage1Instance, opts.Vmstorage1Flags)
+	vmstorage2 := tc.MustStartVmstorage(opts.Vmstorage2Instance, opts.Vmstorage2Flags)
 
 	opts.VminsertFlags = append(opts.VminsertFlags, []string{
 		"-storageNode=" + vmstorage1.VminsertAddr() + "," + vmstorage2.VminsertAddr(),

apptest/testcase_legacy.go

@@ -0,0 +1,50 @@
+package apptest
+
+// MustStartVmsingle_v1_132_0 is a test helper function that starts an instance
+// of vmsingle-v1.132.0 (last version that uses legacy index) and fails the test
+// if the app fails to start.
+func (tc *TestCase) MustStartVmsingle_v1_132_0(instance string, flags []string) *Vmsingle {
+	tc.t.Helper()
+
+	app, err := StartVmsingle_v1_132_0(instance, flags, tc.cli, tc.output)
+	if err != nil {
+		tc.t.Fatalf("Could not start %s: %v", instance, err)
+	}
+	tc.addApp(instance, app)
+	return app
+}
+
+// MustStartVmstorage_v1_132_0 is a test helper function that starts an instance
+// of vmstorage-v1.132.0 (last version that uses legacy index)  and fails the
+// test if the app fails to start.
+func (tc *TestCase) MustStartVmstorage_v1_132_0(instance string, flags []string) *Vmstorage {
+	tc.t.Helper()
+
+	app, err := StartVmstorage_v1_132_0(instance, flags, tc.cli, tc.output)
+	if err != nil {
+		tc.t.Fatalf("Could not start %s: %v", instance, err)
+	}
+	tc.addApp(instance, app)
+	return app
+}
+
+// MustStartCluster_v1_132_0 starts a cluster with vmstorage-v1.132.0 with
+// custom flags.
+func (tc *TestCase) MustStartCluster_v1_132_0(opts *ClusterOptions) *Vmcluster {
+	tc.t.Helper()
+
+	vmstorage1 := tc.MustStartVmstorage_v1_132_0(opts.Vmstorage1Instance, opts.Vmstorage1Flags)
+	vmstorage2 := tc.MustStartVmstorage_v1_132_0(opts.Vmstorage2Instance, opts.Vmstorage2Flags)
+
+	opts.VminsertFlags = append(opts.VminsertFlags, []string{
+		"-storageNode=" + vmstorage1.VminsertAddr() + "," + vmstorage2.VminsertAddr(),
+	}...)
+	vminsert := tc.MustStartVminsert(opts.VminsertInstance, opts.VminsertFlags)
+
+	opts.VmselectFlags = append(opts.VmselectFlags, []string{
+		"-storageNode=" + vmstorage1.VmselectAddr() + "," + vmstorage2.VmselectAddr(),
+	}...)
+	vmselect := tc.MustStartVmselect(opts.VmselectInstance, opts.VmselectFlags)
+
+	return &Vmcluster{vminsert, vmselect, []*Vmstorage{vmstorage1, vmstorage2}}
+}

apptest/tests/legacy_indexdb_test.go

@@ -2,7 +2,6 @@ package tests
 
 import (
 	"fmt"
-	"os"
 	"path/filepath"
 	"slices"
 	"testing"
@@ -11,11 +10,6 @@ import (
 	at "github.com/VictoriaMetrics/VictoriaMetrics/apptest"
 )
 
-var (
-	legacyVmsinglePath  = os.Getenv("VM_LEGACY_VMSINGLE_PATH")
-	legacyVmstoragePath = os.Getenv("VM_LEGACY_VMSTORAGE_PATH")
-)
-
 type testLegacyDeleteSeriesOpts struct {
 	startLegacySUT func() at.PrometheusWriteQuerier
 	startNewSUT    func() at.PrometheusWriteQuerier
@@ -31,7 +25,7 @@ func TestLegacySingleDeleteSeries(t *testing.T) {
 
 	opts := testLegacyDeleteSeriesOpts{
 		startLegacySUT: func() at.PrometheusWriteQuerier {
-			return tc.MustStartVmsingleAt("vmsingle-legacy", legacyVmsinglePath, []string{
+			return tc.MustStartVmsingle_v1_132_0("vmsingle-legacy", []string{
 				"-storageDataPath=" + storageDataPath,
 				"-retentionPeriod=100y",
 				"-search.maxStalenessInterval=1m",
@@ -64,15 +58,13 @@ func TestLegacyClusterDeleteSeries(t *testing.T) {
 
 	opts := testLegacyDeleteSeriesOpts{
 		startLegacySUT: func() at.PrometheusWriteQuerier {
-			return tc.MustStartCluster(&at.ClusterOptions{
+			return tc.MustStartCluster_v1_132_0(&at.ClusterOptions{
 				Vmstorage1Instance: "vmstorage1-legacy",
-				Vmstorage1Binary:   legacyVmstoragePath,
 				Vmstorage1Flags: []string{
 					"-storageDataPath=" + storage1DataPath,
 					"-retentionPeriod=100y",
 				},
 				Vmstorage2Instance: "vmstorage2-legacy",
-				Vmstorage2Binary:   legacyVmstoragePath,
 				Vmstorage2Flags: []string{
 					"-storageDataPath=" + storage2DataPath,
 					"-retentionPeriod=100y",
@@ -255,7 +247,7 @@ func TestLegacySingleBackupRestore(t *testing.T) {
 
 	opts := testLegacyBackupRestoreOpts{
 		startLegacySUT: func() at.PrometheusWriteQuerier {
-			return tc.MustStartVmsingleAt("vmsingle-legacy", legacyVmsinglePath, []string{
+			return tc.MustStartVmsingle_v1_132_0("vmsingle-legacy", []string{
 				"-storageDataPath=" + storageDataPath,
 				"-retentionPeriod=100y",
 				"-search.disableCache=true",
@@ -298,15 +290,13 @@ func TestLegacyClusterBackupRestore(t *testing.T) {
 
 	opts := testLegacyBackupRestoreOpts{
 		startLegacySUT: func() at.PrometheusWriteQuerier {
-			return tc.MustStartCluster(&at.ClusterOptions{
+			return tc.MustStartCluster_v1_132_0(&at.ClusterOptions{
 				Vmstorage1Instance: "vmstorage1-legacy",
-				Vmstorage1Binary:   legacyVmstoragePath,
 				Vmstorage1Flags: []string{
 					"-storageDataPath=" + storage1DataPath,
 					"-retentionPeriod=100y",
 				},
 				Vmstorage2Instance: "vmstorage2-legacy",
-				Vmstorage2Binary:   legacyVmstoragePath,
 				Vmstorage2Flags: []string{
 					"-storageDataPath=" + storage2DataPath,
 					"-retentionPeriod=100y",
@@ -583,7 +573,7 @@ func TestLegacySingleDowngrade(t *testing.T) {
 
 	opts := testLegacyDowngradeOpts{
 		startLegacySUT: func() at.PrometheusWriteQuerier {
-			return tc.MustStartVmsingleAt("vmsingle-legacy", legacyVmsinglePath, []string{
+			return tc.MustStartVmsingle_v1_132_0("vmsingle-legacy", []string{
 				"-storageDataPath=" + storageDataPath,
 				"-retentionPeriod=100y",
 				"-search.disableCache=true",
@@ -618,15 +608,13 @@ func TestLegacyClusterDowngrade(t *testing.T) {
 
 	opts := testLegacyDowngradeOpts{
 		startLegacySUT: func() at.PrometheusWriteQuerier {
-			return tc.MustStartCluster(&at.ClusterOptions{
+			return tc.MustStartCluster_v1_132_0(&at.ClusterOptions{
 				Vmstorage1Instance: "vmstorage1-legacy",
-				Vmstorage1Binary:   legacyVmstoragePath,
 				Vmstorage1Flags: []string{
 					"-storageDataPath=" + storage1DataPath,
 					"-retentionPeriod=100y",
 				},
 				Vmstorage2Instance: "vmstorage2-legacy",
-				Vmstorage2Binary:   legacyVmstoragePath,
 				Vmstorage2Flags: []string{
 					"-storageDataPath=" + storage2DataPath,
 					"-retentionPeriod=100y",

apptest/vmagent.go

@@ -16,43 +16,63 @@ import (
 	"github.com/golang/snappy"
 )
 
-// Vmagent holds the state of a vmagent app and provides vmagent-specific functions
-type Vmagent struct {
-	*app
-	*metricsClient
-
-	httpListenAddr string
-
-	cli *Client
-}
-
-// StartVmagent starts an instance of vmagent with the given flags. It also
-// sets the default flags and populates the app instance state with runtime
-// values extracted from the application log (such as httpListenAddr)
+// StartVmagent starts the latest version of vmagent.
+//
+// The path to the binary can be provided via VMAGENT_PATH environment
+// variable. If the variable is not set, ../../bin/vmagent-race will be
+// used.
 func StartVmagent(instance string, flags []string, cli *Client, promScrapeConfigFilePath string, output io.Writer) (*Vmagent, error) {
-	extractREs := []*regexp.Regexp{
-		httpListenAddrRE,
+	binary := os.Getenv("VMAGENT_PATH")
+	if binary == "" {
+		binary = "../../bin/vmagent-race"
 	}
-
-	app, stderrExtracts, err := startApp(instance, "../../bin/vmagent-race", flags, &appOptions{
+	app, stderrExtracts, err := startApp(instance, binary, flags, &appOptions{
 		defaultFlags: map[string]string{
 			"-httpListenAddr":          "127.0.0.1:0",
 			"-promscrape.config":       promScrapeConfigFilePath,
 			"-remoteWrite.tmpDataPath": fmt.Sprintf("%s/%s-%d", os.TempDir(), instance, time.Now().UnixNano()),
 		},
-		extractREs: extractREs,
-		output:     output,
+		extractREs: []*regexp.Regexp{
+			httpListenAddrRE,
+		},
+		output: output,
 	})
 	if err != nil {
 		return nil, err
 	}
 
+	return newVmagent(app, cli, vmagentRuntimeValues{
+		httpListenAddr: stderrExtracts[0],
+	}), nil
+}
+
+type vmagentRuntimeValues struct {
+	httpListenAddr string
+}
+
+func newVmagent(app *app, cli *Client, rt vmagentRuntimeValues) *Vmagent {
 	return &Vmagent{
 		app:            app,
-		metricsClient:  newMetricsClient(cli, stderrExtracts[0]),
-		httpListenAddr: stderrExtracts[0],
 		cli:            cli,
-	}, nil
+		metricsClient:  newMetricsClient(cli, rt.httpListenAddr),
+		httpListenAddr: rt.httpListenAddr,
+	}
+}
+
+// Vmagent holds the state of a vmagent app and provides vmagent-specific
+// functions.
+type Vmagent struct {
+	*app
+	*metricsClient
+
+	cli            *Client
+	httpListenAddr string
+}
+
+// HTTPAddr returns the address at which the vmagent process is listening
+// for http connections.
+func (app *Vmagent) HTTPAddr() string {
+	return app.httpListenAddr
 }
 
 // APIV1ImportPrometheus is a test helper function that inserts a
@@ -203,12 +223,6 @@ func (app *Vmagent) PrometheusAPIV1Write(t *testing.T, wr prompb.WriteRequest, o
 	})
 }
 
-// HTTPAddr returns the address at which the vmagent process is listening
-// for http connections.
-func (app *Vmagent) HTTPAddr() string {
-	return app.httpListenAddr
-}
-
 // sendBlocking sends the data to vmstorage by executing `send` function and
 // waits until the data is actually sent.
 //

apptest/vmauth.go

@@ -2,6 +2,7 @@ package apptest
 
 import (
 	"io"
+	"os"
 	"regexp"
 	"syscall"
 	"testing"
@@ -10,7 +11,48 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
 )
 
-var httpBuilitinListenAddrRE = regexp.MustCompile(`pprof handlers are exposed at http://(.*:\d{1,5})/debug/pprof/`)
+// StartVmauth starts the latest version of vmauth.
+//
+// The path to the binary can be provided via VMAUTH_PATH environment
+// variable. If the variable is not set, ../../bin/vmauth-race will be
+// used.
+func StartVmauth(instance string, flags []string, cli *Client, configFilePath string, output io.Writer) (*Vmauth, error) {
+	binary := os.Getenv("VMAUTH_PATH")
+	if binary == "" {
+		binary = "../../bin/vmauth-race"
+	}
+	app, stderrExtracts, err := startApp(instance, binary, flags, &appOptions{
+		defaultFlags: map[string]string{
+			"-httpListenAddr": "127.0.0.1:0",
+			"-auth.config":    configFilePath,
+		},
+		extractREs: []*regexp.Regexp{
+			vmauthHttpListenAddrRE,
+		},
+		output: output,
+	})
+	if err != nil {
+		return nil, err
+	}
+
+	return newVmauth(app, cli, configFilePath, vmauthRuntimeValues{
+		httpListenAddr: stderrExtracts[0],
+	}), nil
+}
+
+type vmauthRuntimeValues struct {
+	httpListenAddr string
+}
+
+func newVmauth(app *app, cli *Client, configFilePath string, rt vmauthRuntimeValues) *Vmauth {
+	return &Vmauth{
+		app:            app,
+		metricsClient:  newMetricsClient(cli, rt.httpListenAddr),
+		httpListenAddr: rt.httpListenAddr,
+		configFilePath: configFilePath,
+		cli:            cli,
+	}
+}
 
 // Vmauth holds the state of a vmauth app and provides vmauth-specific
 // functions.
@@ -18,38 +60,14 @@ type Vmauth struct {
 	*app
 	*metricsClient
 
+	cli            *Client
 	httpListenAddr string
 	configFilePath string
-	cli            *Client
 }
 
-// StartVmauth starts an instance of vmauth with the given flags. It also
-// sets the default flags and populates the app instance state with runtime
-// values extracted from the application log (such as httpListenAddr)
-func StartVmauth(instance string, flags []string, cli *Client, configFilePath string, output io.Writer) (*Vmauth, error) {
-	extractREs := []*regexp.Regexp{
-		httpBuilitinListenAddrRE,
-	}
-
-	app, stderrExtracts, err := startApp(instance, "../../bin/vmauth-race", flags, &appOptions{
-		defaultFlags: map[string]string{
-			"-httpListenAddr": "127.0.0.1:0",
-			"-auth.config":    configFilePath,
-		},
-		extractREs: extractREs,
-		output:     output,
-	})
-	if err != nil {
-		return nil, err
-	}
-
-	return &Vmauth{
-		app:            app,
-		metricsClient:  newMetricsClient(cli, stderrExtracts[0]),
-		httpListenAddr: stderrExtracts[0],
-		configFilePath: configFilePath,
-		cli:            cli,
-	}, nil
+// GetHTTPListenAddr returns listen http addr
+func (app *Vmauth) GetHTTPListenAddr() string {
+	return app.httpListenAddr
 }
 
 // UpdateConfiguration updates the vmauth configuration file with the provided YAML content,
@@ -79,8 +97,3 @@ func (app *Vmauth) UpdateConfiguration(t *testing.T, configFileYAML string) {
 
 	t.Fatalf("config were not reloaded after SIGHUP signal; previous total: %d, current total: %d", prevTotal, currTotal)
 }
-
-// GetHTTPListenAddr returns listen http addr
-func (app *Vmauth) GetHTTPListenAddr() string {
-	return app.httpListenAddr
-}

apptest/vmbackup.go

@@ -1,15 +1,26 @@
 package apptest
 
-import "io"
+import (
+	"io"
+	"os"
+)
 
-// StartVmbackup starts an instance of vmbackup with the given flags and waits
-// until it exits.
+// StartVmbackup starts the latest version of vmbackup with the given flags and
+// waits until it exits.
+//
+// The path to the binary can be provided via VMBACKUP_PATH environment
+// variable. If the variable is not set, ../../bin/vmbackup-race will be
+// used.
 func StartVmbackup(instance, storageDataPath, snapshotCreateURL, dst string, output io.Writer) error {
+	binary := os.Getenv("VMBACKUP_PATH")
+	if binary == "" {
+		binary = "../../bin/vmbackup-race"
+	}
 	flags := []string{
 		"-storageDataPath=" + storageDataPath,
 		"-snapshot.createURL=" + snapshotCreateURL,
 		"-dst=" + dst,
 	}
-	_, _, err := startApp(instance, "../../bin/vmbackup-race", flags, &appOptions{wait: true, output: output})
+	_, _, err := startApp(instance, binary, flags, &appOptions{wait: true, output: output})
 	return err
 }

apptest/vmctl.go

@@ -1,9 +1,23 @@
 package apptest
 
-import "io"
+import (
+	"io"
+	"os"
+)
 
 // StartVmctl starts an instance of vmctl cli with the given flags
+
+// StartVmctl starts the latest version of vmctl with the given flags and
+// waits until it exits.
+//
+// The path to the binary can be provided via VMCTL_PATH environment
+// variable. If the variable is not set, ../../bin/vmctl-race will be
+// used.
 func StartVmctl(instance string, flags []string, output io.Writer) error {
-	_, _, err := startApp(instance, "../../bin/vmctl-race", flags, &appOptions{wait: true, output: output})
+	binary := os.Getenv("VMCTL_PATH")
+	if binary == "" {
+		binary = "../../bin/vmctl-race"
+	}
+	_, _, err := startApp(instance, binary, flags, &appOptions{wait: true, output: output})
 	return err
 }

apptest/vminsert.go

@@ -3,23 +3,13 @@ package apptest
 import (
 	"fmt"
 	"io"
+	"os"
 	"regexp"
 	"strings"
 	"testing"
 	"time"
 )
 
-// Vminsert holds the state of a vminsert app and provides vminsert-specific
-// functions.
-type Vminsert struct {
-	*app
-	*metricsClient
-	*vminsertClient
-
-	httpListenAddr          string
-	clusternativeListenAddr string
-}
-
 // storageNodes returns the storage node addresses passed to vminsert via
 // -storageNode command line flag.
 func storageNodes(flags []string) []string {
@@ -31,9 +21,11 @@ func storageNodes(flags []string) []string {
 	return nil
 }
 
-// StartVminsert starts an instance of vminsert with the given flags. It also
-// sets the default flags and populates the app instance state with runtime
-// values extracted from the application log (such as httpListenAddr)
+// StartVminsert starts the latest version of vminsert.
+//
+// The path to the binary can be provided via VMINSERT_PATH environment
+// variable. If the variable is not set, ../../bin/vminsert-race will be
+// used.
 func StartVminsert(instance string, flags []string, cli *Client, output io.Writer) (*Vminsert, error) {
 	extractREs := []*regexp.Regexp{
 		httpListenAddrRE,
@@ -48,11 +40,15 @@ func StartVminsert(instance string, flags []string, cli *Client, output io.Write
 		extractREs = append(extractREs, regexp.MustCompile(logRecord))
 	}
 
-	app, stderrExtracts, err := startApp(instance, "../../bin/vminsert-race", flags, &appOptions{
+	binary := os.Getenv("VMINSERT_PATH")
+	if binary == "" {
+		binary = "../../bin/vminsert-race"
+	}
+	app, stderrExtracts, err := startApp(instance, binary, flags, &appOptions{
 		defaultFlags: map[string]string{
 			"-httpListenAddr":                              "127.0.0.1:0",
 			"-clusternativeListenAddr":                     "127.0.0.1:0",
-			"-graphiteListenAddr":                          ":0",
+			"-graphiteListenAddr":                          "127.0.0.1:0",
 			"-opentsdbListenAddr":                          "127.0.0.1:0",
 			"-clusternative.vminsertConnsShutdownDuration": "1ms",
 		},
@@ -63,27 +59,56 @@ func StartVminsert(instance string, flags []string, cli *Client, output io.Write
 		return nil, err
 	}
 
-	metricsClient := newMetricsClient(cli, stderrExtracts[0])
-	return &Vminsert{
-		app:           app,
-		metricsClient: metricsClient,
-		vminsertClient: &vminsertClient{
-			vminsertCli: cli,
-			url: func(op, path string, opts QueryOpts) string {
-				return getClusterPath(stderrExtracts[0], op, path, opts)
-			},
-			openTSDBURL: func(op, path string, opts QueryOpts) string {
-				return getClusterPath(stderrExtracts[3], op, path, opts)
-			},
-			graphiteListenAddr: stderrExtracts[2],
-			sendBlocking: func(t *testing.T, numRecordsToSend int, send func()) {
-				t.Helper()
-				sendBlocking(t, metricsClient, numRecordsToSend, send)
-			},
-		},
+	return newVminsert(app, cli, vminsertRuntimeValues{
 		httpListenAddr:          stderrExtracts[0],
 		clusternativeListenAddr: stderrExtracts[1],
-	}, nil
+		graphiteListenAddr:      stderrExtracts[2],
+		openTSDBListenAddr:      stderrExtracts[3],
+	}), nil
+}
+
+type vminsertRuntimeValues struct {
+	httpListenAddr          string
+	clusternativeListenAddr string
+	graphiteListenAddr      string
+	openTSDBListenAddr      string
+}
+
+func newVminsert(app *app, cli *Client, rt vminsertRuntimeValues) *Vminsert {
+	metricsClient := newMetricsClient(cli, rt.httpListenAddr)
+	vminsertClient := &vminsertClient{
+		vminsertCli: cli,
+		url: func(op, path string, opts QueryOpts) string {
+			return getClusterPath(rt.httpListenAddr, op, path, opts)
+		},
+		openTSDBURL: func(op, path string, opts QueryOpts) string {
+			return getClusterPath(rt.openTSDBListenAddr, op, path, opts)
+		},
+		graphiteListenAddr: rt.graphiteListenAddr,
+		sendBlocking: func(t *testing.T, numRecordsToSend int, send func()) {
+			t.Helper()
+			sendBlocking(t, metricsClient, numRecordsToSend, send)
+		},
+	}
+
+	return &Vminsert{
+		app:                     app,
+		metricsClient:           metricsClient,
+		vminsertClient:          vminsertClient,
+		httpListenAddr:          rt.httpListenAddr,
+		clusternativeListenAddr: rt.clusternativeListenAddr,
+	}
+}
+
+// Vminsert holds the state of a vminsert app and provides vminsert-specific
+// functions.
+type Vminsert struct {
+	*app
+	*metricsClient
+	*vminsertClient
+
+	httpListenAddr          string
+	clusternativeListenAddr string
 }
 
 // ClusternativeListenAddr returns the address at which the vminsert process is

apptest/vmrestore.go

@@ -1,14 +1,25 @@
 package apptest
 
-import "io"
+import (
+	"io"
+	"os"
+)
 
-// StartVmrestore starts an instance of vmrestore with the given flags and waits
-// until it exits.
+// StartVmrestore starts the latest version of vmrestore with the given flags
+// and waits until it exits.
+//
+// The path to the binary can be provided via VMRESTORE_PATH environment
+// variable. If the variable is not set, ../../bin/vmrestore-race will be
+// used.
 func StartVmrestore(instance, src, storageDataPath string, output io.Writer) error {
+	binary := os.Getenv("VMRESTORE_PATH")
+	if binary == "" {
+		binary = "../../bin/vmrestore-race"
+	}
 	flags := []string{
 		"-src=" + src,
 		"-storageDataPath=" + storageDataPath,
 	}
-	_, _, err := startApp(instance, "../../bin/vmrestore-race", flags, &appOptions{wait: true, output: output})
+	_, _, err := startApp(instance, binary, flags, &appOptions{wait: true, output: output})
 	return err
 }

apptest/vmselect.go

@@ -3,26 +3,21 @@ package apptest
 import (
 	"fmt"
 	"io"
+	"os"
 	"regexp"
 )
 
-// Vmselect holds the state of a vmselect app and provides vmselect-specific
-// functions.
-type Vmselect struct {
-	*app
-	*metricsClient
-	*vmselectClient
-
-	httpListenAddr          string
-	clusternativeListenAddr string
-	cli                     *Client
-}
-
-// StartVmselect starts an instance of vmselect with the given flags. It also
-// sets the default flags and populates the app instance state with runtime
-// values extracted from the application log (such as httpListenAddr)
+// StartVmselect starts the latest version of vmselect.
+//
+// The path to the binary can be provided via VMSELECT_PATH environment
+// variable. If the variable is not set, ../../bin/vmselect-race will be
+// used.
 func StartVmselect(instance string, flags []string, cli *Client, output io.Writer) (*Vmselect, error) {
-	app, stderrExtracts, err := startApp(instance, "../../bin/vmselect-race", flags, &appOptions{
+	binary := os.Getenv("VMSELECT_PATH")
+	if binary == "" {
+		binary = "../../bin/vmselect-race"
+	}
+	app, stderrExtracts, err := startApp(instance, binary, flags, &appOptions{
 		defaultFlags: map[string]string{
 			"-httpListenAddr":          "127.0.0.1:0",
 			"-clusternativeListenAddr": "127.0.0.1:0",
@@ -37,21 +32,43 @@ func StartVmselect(instance string, flags []string, cli *Client, output io.Write
 		return nil, err
 	}
 
+	return newVmselect(app, cli, vmselectRuntimeValues{
+		httpListenAddr:          stderrExtracts[0],
+		clusternativeListenAddr: stderrExtracts[1],
+	}), nil
+}
+
+type vmselectRuntimeValues struct {
+	httpListenAddr          string
+	clusternativeListenAddr string
+}
+
+func newVmselect(app *app, cli *Client, rt vmselectRuntimeValues) *Vmselect {
 	return &Vmselect{
 		app:           app,
-		metricsClient: newMetricsClient(cli, stderrExtracts[0]),
+		metricsClient: newMetricsClient(cli, rt.httpListenAddr),
 		vmselectClient: &vmselectClient{
 			vmselectCli: cli,
 			url: func(op, path string, opts QueryOpts) string {
-				return getClusterPath(stderrExtracts[0], op, path, opts)
+				return getClusterPath(rt.httpListenAddr, op, path, opts)
 			},
-			metricNamesStatsResetURL: fmt.Sprintf("http://%s/admin/api/v1/admin/status/metric_names_stats/reset", stderrExtracts[0]),
-			tenantsURL:               fmt.Sprintf("http://%s/admin/tenants", stderrExtracts[0]),
+			metricNamesStatsResetURL: fmt.Sprintf("http://%s/admin/api/v1/admin/status/metric_names_stats/reset", rt.httpListenAddr),
+			tenantsURL:               fmt.Sprintf("http://%s/admin/tenants", rt.httpListenAddr),
 		},
-		httpListenAddr:          stderrExtracts[0],
-		clusternativeListenAddr: stderrExtracts[1],
-		cli:                     cli,
-	}, nil
+		httpListenAddr:          rt.httpListenAddr,
+		clusternativeListenAddr: rt.clusternativeListenAddr,
+	}
+}
+
+// Vmselect holds the state of a vmselect app and provides vmselect-specific
+// functions.
+type Vmselect struct {
+	*app
+	*metricsClient
+	*vmselectClient
+
+	httpListenAddr          string
+	clusternativeListenAddr string
 }
 
 // ClusternativeListenAddr returns the address at which the vmselect process is

apptest/vmsingle.go

@@ -9,28 +9,21 @@ import (
 	"time"
 )
 
-// Vmsingle holds the state of a vmsingle app and provides vmsingle-specific
-// functions.
-type Vmsingle struct {
-	*app
-	*metricsClient
-	*vmstorageClient
-	*vmselectClient
-	*vminsertClient
-
-	storageDataPath string
-	httpListenAddr  string
-}
-
-// StartVmsingleAt starts an instance of vmsingle with the given flags. It also
-// sets the default flags and populates the app instance state with runtime
-// values extracted from the application log (such as httpListenAddr).
-func StartVmsingleAt(instance, binary string, flags []string, cli *Client, output io.Writer) (*Vmsingle, error) {
+// StartVmsingle starts the latest version of vmsingle.
+//
+// The path to the binary can be provided via VMSINGLE_PATH environment
+// variable. If the variable is not set, ../../bin/victoria-metrics-race will be
+// used.
+func StartVmsingle(instance string, flags []string, cli *Client, output io.Writer) (*Vmsingle, error) {
+	binary := os.Getenv("VMSINGLE_PATH")
+	if binary == "" {
+		binary = "../../bin/victoria-metrics-race"
+	}
 	app, stderrExtracts, err := startApp(instance, binary, flags, &appOptions{
 		defaultFlags: map[string]string{
 			"-storageDataPath":    fmt.Sprintf("%s/%s-%d", os.TempDir(), instance, time.Now().UnixNano()),
 			"-httpListenAddr":     "127.0.0.1:0",
-			"-graphiteListenAddr": ":0",
+			"-graphiteListenAddr": "127.0.0.1:0",
 			"-opentsdbListenAddr": "127.0.0.1:0",
 		},
 		extractREs: []*regexp.Regexp{
@@ -45,38 +38,67 @@ func StartVmsingleAt(instance, binary string, flags []string, cli *Client, outpu
 		return nil, err
 	}
 
+	return newVmsingle(app, cli, vmsingleRuntimeValues{
+		storageDataPath:    stderrExtracts[0],
+		httpListenAddr:     stderrExtracts[1],
+		graphiteListenAddr: stderrExtracts[2],
+		openTSDBListenAddr: stderrExtracts[3],
+	}), nil
+}
+
+type vmsingleRuntimeValues struct {
+	storageDataPath    string
+	httpListenAddr     string
+	graphiteListenAddr string
+	openTSDBListenAddr string
+}
+
+func newVmsingle(app *app, cli *Client, rt vmsingleRuntimeValues) *Vmsingle {
 	return &Vmsingle{
 		app:           app,
-		metricsClient: newMetricsClient(cli, stderrExtracts[1]),
+		metricsClient: newMetricsClient(cli, rt.httpListenAddr),
 		vmstorageClient: &vmstorageClient{
 			vmstorageCli:   cli,
-			httpListenAddr: stderrExtracts[1],
+			httpListenAddr: rt.httpListenAddr,
 		},
 		vmselectClient: &vmselectClient{
 			vmselectCli: cli,
 			url: func(op, path string, opts QueryOpts) string {
-				return fmt.Sprintf("http://%s/%s", stderrExtracts[1], path)
+				return fmt.Sprintf("http://%s/%s", rt.httpListenAddr, path)
 			},
-			metricNamesStatsResetURL: fmt.Sprintf("http://%s/api/v1/admin/status/metric_names_stats/reset", stderrExtracts[1]),
+			metricNamesStatsResetURL: fmt.Sprintf("http://%s/api/v1/admin/status/metric_names_stats/reset", rt.httpListenAddr),
 			tenantsURL:               "vmsingle-does-not-serve-tenants",
 		},
 		vminsertClient: &vminsertClient{
 			vminsertCli: cli,
 			url: func(_, path string, _ QueryOpts) string {
-				return fmt.Sprintf("http://%s/%s", stderrExtracts[1], path)
+				return fmt.Sprintf("http://%s/%s", rt.httpListenAddr, path)
 			},
 			openTSDBURL: func(_, path string, _ QueryOpts) string {
-				return fmt.Sprintf("http://%s/%s", stderrExtracts[3], path)
+				return fmt.Sprintf("http://%s/%s", rt.openTSDBListenAddr, path)
 			},
-			graphiteListenAddr: stderrExtracts[2],
+			graphiteListenAddr: rt.graphiteListenAddr,
 			sendBlocking: func(t *testing.T, _ int, send func()) {
 				t.Helper()
 				send()
 			},
 		},
-		storageDataPath: stderrExtracts[0],
-		httpListenAddr:  stderrExtracts[1],
-	}, nil
+		storageDataPath: rt.storageDataPath,
+		httpListenAddr:  rt.httpListenAddr,
+	}
+}
+
+// Vmsingle holds the state of a vmsingle app and provides vmsingle-specific
+// functions.
+type Vmsingle struct {
+	*app
+	*metricsClient
+	*vmstorageClient
+	*vmselectClient
+	*vminsertClient
+
+	storageDataPath string
+	httpListenAddr  string
 }
 
 // HTTPAddr returns the address at which the vminsert process is

apptest/vmsingle_legacy.go

@@ -0,0 +1,43 @@
+package apptest
+
+import (
+	"fmt"
+	"io"
+	"os"
+	"regexp"
+	"time"
+)
+
+// StartVmsingle_v1_132_0 starts vmsingle-v1.132.0 (the last version that uses
+// legacy index).
+//
+// The path to the binary must be provided via VMSINGLE_V1_132_0_PATH
+// environment variable.
+func StartVmsingle_v1_132_0(instance string, flags []string, cli *Client, output io.Writer) (*Vmsingle, error) {
+	binary := os.Getenv("VMSINGLE_V1_132_0_PATH")
+	app, stderrExtracts, err := startApp(instance, binary, flags, &appOptions{
+		defaultFlags: map[string]string{
+			"-storageDataPath":    fmt.Sprintf("%s/%s-%d", os.TempDir(), instance, time.Now().UnixNano()),
+			"-httpListenAddr":     "127.0.0.1:0",
+			"-graphiteListenAddr": "127.0.0.1:0",
+			"-opentsdbListenAddr": "127.0.0.1:0",
+		},
+		extractREs: []*regexp.Regexp{
+			storageDataPathRE,
+			httpListenAddrRE,
+			graphiteListenAddrRE,
+			openTSDBListenAddrRE,
+		},
+		output: output,
+	})
+	if err != nil {
+		return nil, err
+	}
+
+	return newVmsingle(app, cli, vmsingleRuntimeValues{
+		storageDataPath:    stderrExtracts[0],
+		httpListenAddr:     stderrExtracts[1],
+		graphiteListenAddr: stderrExtracts[2],
+		openTSDBListenAddr: stderrExtracts[3],
+	}), nil
+}

apptest/vmstorage.go

@@ -8,23 +8,22 @@ import (
 	"time"
 )
 
-// Vmstorage holds the state of a vmstorage app and provides vmstorage-specific
-// functions.
-type Vmstorage struct {
-	*app
-	*metricsClient
-	*vmstorageClient
-
-	storageDataPath string
-	httpListenAddr  string
-	vminsertAddr    string
-	vmselectAddr    string
+// StartVmstorage starts the latest version of vmstorage.
+//
+// The path to the binary can be provided via VMSTORAGE_PATH environment
+// variable. If the variable is not set, ../../bin/vmstorage-race will be used.
+func StartVmstorage(instance string, flags []string, cli *Client, output io.Writer) (*Vmstorage, error) {
+	binary := os.Getenv("VMSTORAGE_PATH")
+	if binary == "" {
+		binary = "../../bin/vmstorage-race"
+	}
+	return startVmstorage(instance, binary, flags, cli, output)
 }
 
-// StartVmstorageAt starts an instance of vmstorage with the given flags. It also
+// startVmstorage starts an instance of vmstorage with the given flags. It also
 // sets the default flags and populates the app instance state with runtime
 // values extracted from the application log (such as httpListenAddr)
-func StartVmstorageAt(instance, binary string, flags []string, cli *Client, output io.Writer) (*Vmstorage, error) {
+func startVmstorage(instance, binary string, flags []string, cli *Client, output io.Writer) (*Vmstorage, error) {
 	app, stderrExtracts, err := startApp(instance, binary, flags, &appOptions{
 		defaultFlags: map[string]string{
 			"-storageDataPath": fmt.Sprintf("%s/%s-%d", os.TempDir(), instance, time.Now().UnixNano()),
@@ -44,18 +43,47 @@ func StartVmstorageAt(instance, binary string, flags []string, cli *Client, outp
 		return nil, err
 	}
 
-	return &Vmstorage{
-		app:           app,
-		metricsClient: newMetricsClient(cli, stderrExtracts[1]),
-		vmstorageClient: &vmstorageClient{
-			vmstorageCli:   cli,
-			httpListenAddr: stderrExtracts[1],
-		},
+	return newVmstorage(app, cli, vmstorageRuntimeValues{
 		storageDataPath: stderrExtracts[0],
 		httpListenAddr:  stderrExtracts[1],
 		vminsertAddr:    stderrExtracts[2],
 		vmselectAddr:    stderrExtracts[3],
-	}, nil
+	}), nil
+}
+
+type vmstorageRuntimeValues struct {
+	storageDataPath string
+	httpListenAddr  string
+	vminsertAddr    string
+	vmselectAddr    string
+}
+
+func newVmstorage(app *app, cli *Client, rt vmstorageRuntimeValues) *Vmstorage {
+	return &Vmstorage{
+		app:           app,
+		metricsClient: newMetricsClient(cli, rt.httpListenAddr),
+		vmstorageClient: &vmstorageClient{
+			vmstorageCli:   cli,
+			httpListenAddr: rt.httpListenAddr,
+		},
+		storageDataPath: rt.storageDataPath,
+		httpListenAddr:  rt.httpListenAddr,
+		vminsertAddr:    rt.vminsertAddr,
+		vmselectAddr:    rt.vmselectAddr,
+	}
+}
+
+// Vmstorage holds the state of a vmstorage app and provides vmstorage-specific
+// functions.
+type Vmstorage struct {
+	*app
+	*metricsClient
+	*vmstorageClient
+
+	storageDataPath string
+	httpListenAddr  string
+	vminsertAddr    string
+	vmselectAddr    string
 }
 
 // VminsertAddr returns the address at which the vmstorage process is listening

apptest/vmstorage_legacy.go

@@ -0,0 +1,16 @@
+package apptest
+
+import (
+	"io"
+	"os"
+)
+
+// StartVmstorage_v1_132_0 starts vmstorage-v1.132.0 (the last version that uses
+// legacy index).
+//
+// The path to the binary must be provided via VMSTORAGE_V1_132_0_PATH
+// environment variable.
+func StartVmstorage_v1_132_0(instance string, flags []string, cli *Client, output io.Writer) (*Vmstorage, error) {
+	binary := os.Getenv("VMSTORAGE_V1_132_0_PATH")
+	return startVmstorage(instance, binary, flags, cli, output)
+}

deployment/docker/compose-vm-cluster.yml

@@ -3,7 +3,7 @@ services:
   #  It scrapes targets defined in --promscrape.config
   #  And forward them to --remoteWrite.url
   vmagent:
-    image: victoriametrics/vmagent:v1.143.0
+    image: victoriametrics/vmagent:v1.144.0
     depends_on:
       - "vmauth"
     ports:
@@ -42,14 +42,14 @@ services:
   # vmstorage shards. Each shard receives 1/N of all metrics sent to vminserts,
   # where N is number of vmstorages (2 in this case).
   vmstorage-1:
-    image: victoriametrics/vmstorage:v1.143.0-cluster
+    image: victoriametrics/vmstorage:v1.144.0-cluster
     volumes:
       - strgdata-1:/storage
     command:
       - "--storageDataPath=/storage"
     restart: always
   vmstorage-2:
-    image: victoriametrics/vmstorage:v1.143.0-cluster
+    image: victoriametrics/vmstorage:v1.144.0-cluster
     volumes:
       - strgdata-2:/storage
     command:
@@ -59,7 +59,7 @@ services:
   # vminsert is ingestion frontend. It receives metrics pushed by vmagent,
   # pre-process them and distributes across configured vmstorage shards.
   vminsert-1:
-    image: victoriametrics/vminsert:v1.143.0-cluster
+    image: victoriametrics/vminsert:v1.144.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -68,7 +68,7 @@ services:
       - "--storageNode=vmstorage-2:8400"
     restart: always
   vminsert-2:
-    image: victoriametrics/vminsert:v1.143.0-cluster
+    image: victoriametrics/vminsert:v1.144.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -80,7 +80,7 @@ services:
   # vmselect is a query fronted. It serves read queries in MetricsQL or PromQL.
   # vmselect collects results from configured `--storageNode` shards.
   vmselect-1:
-    image: victoriametrics/vmselect:v1.143.0-cluster
+    image: victoriametrics/vmselect:v1.144.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -90,7 +90,7 @@ services:
       - "--vmalert.proxyURL=http://vmalert:8880"
     restart: always
   vmselect-2:
-    image: victoriametrics/vmselect:v1.143.0-cluster
+    image: victoriametrics/vmselect:v1.144.0-cluster
     depends_on:
       - "vmstorage-1"
       - "vmstorage-2"
@@ -105,7 +105,7 @@ services:
   # read requests from Grafana, vmui, vmalert among vmselects.
   # It can be used as an authentication proxy.
   vmauth:
-    image: victoriametrics/vmauth:v1.143.0
+    image: victoriametrics/vmauth:v1.144.0
     depends_on:
       - "vmselect-1"
       - "vmselect-2"
@@ -119,7 +119,7 @@ services:
 
   # vmalert executes alerting and recording rules
   vmalert:
-    image: victoriametrics/vmalert:v1.143.0
+    image: victoriametrics/vmalert:v1.144.0
     depends_on:
       - "vmauth"
     ports:

deployment/docker/compose-vm-single.yml

@@ -3,7 +3,7 @@ services:
   #  It scrapes targets defined in --promscrape.config
   #  And forward them to --remoteWrite.url
   vmagent:
-    image: victoriametrics/vmagent:v1.143.0
+    image: victoriametrics/vmagent:v1.144.0
     depends_on:
       - "victoriametrics"
     ports:
@@ -18,7 +18,7 @@ services:
   # VictoriaMetrics instance, a single process responsible for
   # storing metrics and serve read requests.
   victoriametrics:
-    image: victoriametrics/victoria-metrics:v1.143.0
+    image: victoriametrics/victoria-metrics:v1.144.0
     ports:
       - 8428:8428
       - 8089:8089
@@ -59,7 +59,7 @@ services:
 
   # vmalert executes alerting and recording rules
   vmalert:
-    image: victoriametrics/vmalert:v1.143.0
+    image: victoriametrics/vmalert:v1.144.0
     depends_on:
       - "victoriametrics"
       - "alertmanager"

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

@@ -1,6 +1,6 @@
 services:
   vmagent:
-    image: victoriametrics/vmagent:v1.143.0
+    image: victoriametrics/vmagent:v1.144.0
     depends_on:
       - "victoriametrics"
     ports:
@@ -14,7 +14,7 @@ services:
     restart: always
 
   victoriametrics:
-    image: victoriametrics/victoria-metrics:v1.143.0
+    image: victoriametrics/victoria-metrics:v1.144.0
     ports:
       - 8428:8428
     volumes:
@@ -40,7 +40,7 @@ services:
     restart: always
 
   vmalert:
-    image: victoriametrics/vmalert:v1.143.0
+    image: victoriametrics/vmalert:v1.144.0
     depends_on:
       - "victoriametrics"
     ports:

docs/ai-tools/README.md

@@ -88,6 +88,7 @@ These skills provide predefined workflows and capabilities such as:
 * Multi-signal investigations 
 * Cardinality optimization 
 * Unused metric detection
+* Stream aggregation configuration
 
 To install the available skills for AI agents, run:
 ```sh

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

@@ -240,23 +240,23 @@ vmagent will write data into VictoriaMetrics single-node and cluster (with tenan
 # compose.yaml
 services:
   vmsingle:
-    image: victoriametrics/victoria-metrics:v1.143.0
+    image: victoriametrics/victoria-metrics:v1.144.0
 
   vmstorage:
-    image: victoriametrics/vmstorage:v1.143.0-cluster
+    image: victoriametrics/vmstorage:v1.144.0-cluster
 
   vminsert:
-    image: victoriametrics/vminsert:v1.143.0-cluster
+    image: victoriametrics/vminsert:v1.144.0-cluster
     command:
       - -storageNode=vmstorage:8400
 
   vmselect:
-    image: victoriametrics/vmselect:v1.143.0-cluster
+    image: victoriametrics/vmselect:v1.144.0-cluster
     command:
       - -storageNode=vmstorage:8401
 
   vmagent:
-    image: victoriametrics/vmagent:v1.143.0
+    image: victoriametrics/vmagent:v1.144.0
     volumes:
       - ./scrape.yaml:/etc/vmagent/config.yaml
     command:
@@ -308,7 +308,7 @@ Now add the vmauth service to `compose.yaml`:
 # compose.yaml
 services:
   vmauth:
-    image: docker.io/victoriametrics/vmauth:v1.143.0
+    image: docker.io/victoriametrics/vmauth:v1.144.0
     ports:
       - 8427:8427
     volumes:

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

@@ -155,15 +155,15 @@ These services will store and query the metrics scraped by vmagent.
 # compose.yaml
 services:
   vmstorage:
-    image: victoriametrics/vmstorage:v1.143.0-cluster
+    image: victoriametrics/vmstorage:v1.144.0-cluster
 
   vminsert:
-    image: victoriametrics/vminsert:v1.143.0-cluster
+    image: victoriametrics/vminsert:v1.144.0-cluster
     command:
       - -storageNode=vmstorage:8400
 
   vmselect:
-    image: victoriametrics/vmselect:v1.143.0-cluster
+    image: victoriametrics/vmselect:v1.144.0-cluster
     command:
       - -storageNode=vmstorage:8401
     ports:
@@ -196,7 +196,7 @@ Add the vmauth service to `compose.yaml`:
 # compose.yaml
 services:
   vmauth:
-    image: victoriametrics/vmauth:v1.143.0-enterprise
+    image: victoriametrics/vmauth:v1.144.0-enterprise
     ports:
       - 8427:8427
     volumes:
@@ -251,7 +251,7 @@ Add the vmagent service to `compose.yaml` with OAuth2 configuration:
 # compose.yaml
 services:
   vmagent:
-    image: victoriametrics/vmagent:v1.143.0
+    image: victoriametrics/vmagent:v1.144.0
     volumes:
       - ./scrape.yaml:/etc/vmagent/config.yaml
     command:

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

@@ -107,7 +107,7 @@ The final piece is the Docker Compose file. This ties all the services together
 # compose.yml
 services:
   victoriametrics:
-    image: victoriametrics/victoria-metrics:v1.143.0
+    image: victoriametrics/victoria-metrics:v1.144.0
     command:
       - "--storageDataPath=/victoria-metrics-data"
       - "--selfScrapeInterval=10s"
@@ -128,7 +128,7 @@ services:
       - ./alertmanager.yml:/etc/alertmanager/alertmanager.yml:ro
 
   vmalert:
-    image: victoriametrics/vmalert:v1.143.0
+    image: victoriametrics/vmalert:v1.144.0
     depends_on:
       - victoriametrics
       - alertmanager

docs/playgrounds/README.md

@@ -150,8 +150,12 @@ You can experiment with your own data during the month‑long trial without depl
 are fast-booting Linux microVMs that run on a fleet of large bare-metal servers. You can start a playground right from your browser. 
 Once up and running, accessing a playground is no different from SSH-ing into a remote server rented from your favorite VPS or Cloud provider.
 
-Iximiuz Labs provides playgrounds for VictoriaMetrics software:
-- [VictoriaMetrics single node (on Ubuntu)](https://labs.iximiuz.com/playgrounds/victoriametrics-e2f9b613)
-- [VictoriaMetrics cluster (on Ubuntu)](https://labs.iximiuz.com/playgrounds/victoriametrics-cluster-8eacb19d)
-- [Getting Started with VictoriaMetrics on Kubernetes](https://labs.iximiuz.com/tutorials/victoriametrics-getting-started-kubernetes-0e9c0993)
-- [VictoriaMetrics Operator](https://labs.iximiuz.com/playgrounds/victoriametrics-kubernetes-9eebc258)
+Iximiuz Labs provides various [learning-by-doing resources for VictoriaMetrics](https://labs.iximiuz.com/v/victoriametrics):
+- Tutorial:
+  - [Getting Started with VictoriaMetrics on Kubernetes](https://labs.iximiuz.com/tutorials/victoriametrics-getting-started-kubernetes)
+- Playgrounds:
+  - [VictoriaMetrics single node](https://labs.iximiuz.com/playgrounds/victoriametrics)
+  - [VictoriaMetrics cluster](https://labs.iximiuz.com/playgrounds/victoriametrics-cluster)
+  - [VictoriaMetrics on Kubernetes](https://labs.iximiuz.com/playgrounds/victoriametrics-kubernetes)
+
+Iximiuz Labs requires a [free account](https://labs.iximiuz.com/signup) to access the materials.

docs/victoriametrics/Quick-Start.md

@@ -43,6 +43,9 @@ Just download VictoriaMetrics and follow [these instructions](https://docs.victo
 See [available integrations](https://docs.victoriametrics.com/victoriametrics/integrations/) with other systems like
 [Prometheus](https://docs.victoriametrics.com/victoriametrics/integrations/prometheus/) or [Grafana](https://docs.victoriametrics.com/victoriametrics/integrations/grafana/).
 
+> Want to see VictoriaMetrics in action, but without installing anything?
+> Try [Playgrounds](https://docs.victoriametrics.com/playgrounds/) - a list of publicly available playgrounds for VictoriaMetrics software.
+
 VictoriaMetrics is developed at a fast pace, so it is recommended to periodically check the [CHANGELOG](https://docs.victoriametrics.com/victoriametrics/changelog/)
 and perform [regular upgrades](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#how-to-upgrade-victoriametrics).
 
@@ -58,9 +61,9 @@ Download the newest available [VictoriaMetrics release](https://docs.victoriamet
 from [DockerHub](https://hub.docker.com/r/victoriametrics/victoria-metrics) or [Quay](https://quay.io/repository/victoriametrics/victoria-metrics?tab=tags):
 
 ```sh
-docker pull victoriametrics/victoria-metrics:v1.143.0
+docker pull victoriametrics/victoria-metrics:v1.144.0
 docker run -it --rm -v `pwd`/victoria-metrics-data:/victoria-metrics-data -p 8428:8428 \
- victoriametrics/victoria-metrics:v1.143.0 --selfScrapeInterval=5s -storageDataPath=victoria-metrics-data
+ victoriametrics/victoria-metrics:v1.144.0 --selfScrapeInterval=5s -storageDataPath=victoria-metrics-data
 ```
 
 _For Enterprise images, see [this link](https://docs.victoriametrics.com/victoriametrics/enterprise/#docker-images)._

docs/victoriametrics/changelog/CHANGELOG.md

@@ -26,6 +26,8 @@ See also [LTS releases](https://docs.victoriametrics.com/victoriametrics/lts-rel
 
 ## tip
 
+* FEATURE: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/), [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vminsert` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): add `-opentelemetry.promoteAllResourceAttributes` and `-opentelemetry.promoteScopeMetadata` command-line flags to allow managing label promotion for resource attributes and OTel scope metadata. See [OpenTelemetry](https://docs.victoriametrics.com/victoriametrics/integrations/opentelemetry/) docs and [#10931](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10931).
+
 ## [v1.144.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.144.0)
 
 Released at 2026-05-22
@@ -240,6 +242,28 @@ It enables back `Discovered targets` debug UI by default.
 * BUGFIX: `vmstorage` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): properly apply `extra_filters[]` filter when querying `vm_account_id` or `vm_project_id` labels via [multitenant](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#multitenancy) request for `/api/v1/label/…/values` API. Before, `extra_filters` was ignored. See [#10503](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10503).
 * BUGFIX: [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): revert the use of rollup result cache for [instant queries](https://docs.victoriametrics.com/keyConcepts.html#instant-query) that contain [`rate`](https://docs.victoriametrics.com/MetricsQL.html#rate) function with a lookbehind window larger than `-search.minWindowForInstantRollupOptimization`. The cache usage was removed since [v1.132.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.132.0). See [#10098](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10098#issuecomment-3895011084) for more details.
 
+## [v1.136.10](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.136.10)
+
+Released at 2026-05-22
+
+**v1.136.x is a line of [LTS releases](https://docs.victoriametrics.com/victoriametrics/lts-releases/). It contains important up-to-date bugfixes for [VictoriaMetrics enterprise](https://docs.victoriametrics.com/victoriametrics/enterprise/).
+All these fixes are also included in [the latest community release](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/latest).
+The v1.136.x line will be supported for at least 12 months since [v1.136.0](https://docs.victoriametrics.com/victoriametrics/changelog/#v11360) release**
+
+* BUGFIX: [stream aggregation](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/): stop emitting stale values for `quantiles(...)` outputs when a time series has no samples during the current aggregation interval. See [#10918](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10918). Thanks to @alexei38 for the contribution.
+* BUGFIX: [stream aggregation](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/): extend delay on aggregation windows flush by the biggest lag among pushed samples. Before, the delay was calculated as 95th percentile across samples, which could underrepresent outliers and reject them from aggregation as "too old". See [#10402](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10402).
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): fix a bug in [cardinality limiters](https://docs.victoriametrics.com/victoriametrics/vmagent/#cardinality-limiter) where series with different labels, like `{a="bc"}` and `{ab="c"}`, could be incorrectly treated as identical and dropped. See [#10937](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10937).
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): hide values passed to `-remoteWrite.headers` in startup logs, `/metrics`, and `/flags`, since they can contain sensitive HTTP headers such as `Authorization` and API keys.
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): hide values passed to `-remoteWrite.proxyURL` in startup logs, `/metrics`, and `/flags`, since they can contain sensitive credentials.
+* BUGFIX: [vmalert](https://docs.victoriametrics.com/victoriametrics/vmalert/): hide values passed to `-remoteWrite.headers`,`remoteRead.headers`, `datasource.headers` and `notifier.headers` in startup logs, `/metrics`, and `/flags`, since they can contain sensitive HTTP headers such as `Authorization` and API keys.
+* BUGFIX: `vminsert` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): properly establish [mtls](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#mtls-protection) connection between vmstorage and vminsert. Regression was introduced in v1.130.0 release for the enterprise version of vmstorage. See [#10972](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10972).
+* BUGFIX: [vmrestore](https://docs.victoriametrics.com/victoriametrics/vmrestore/): fix a bug where specifying `-storageDataPath` with a trailing slash could cause `vmrestore` to panic. See [#10823](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10823). Thanks to @utafrali for the contribution.
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): prevent unintentional rerouting of samples to other sharding targets when one of the `-remoteWrite.url` targets with `-remoteWrite.disableOnDiskQueue` becomes blocked. Previously this could break the sharding guarantee by sending samples to wrong targets instead of dropping or retrying them. See [#10507](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10507).
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): return error on startup if `-remoteWrite.disableOnDiskQueue` is not configured uniformly across all `-remoteWrite.url` targets when `-remoteWrite.shardByURL` is enabled. Either all targets must have it enabled or all must have it disabled. See [#10507](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10507).
+* BUGFIX: [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): hide values passed to `vmalert.proxyURL` in startup logs, `/metrics`, and `/flags`, since they can contain sensitive HTTP headers such as `Authorization` and API keys.
+* BUGFIX: [vmui](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmui): preserve exact series values in graph tooltips instead of rounding them by significant digits. See [#10952](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10952).
+* BUGFIX: [vmui](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmui): add missing `__timestamp__` and `__value__` columns to CSV exported from the table view on the Query tab. See [#10975](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10975).
+
 ## [v1.136.9](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.136.9)
 
 Released at 2026-05-08
@@ -561,6 +585,26 @@ See changes [here](https://docs.victoriametrics.com/victoriametrics/changelog/ch
 
 See changes [here](https://docs.victoriametrics.com/victoriametrics/changelog/changelog_2025/#v11230)
 
+## [v1.122.23](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.122.23)
+
+Released at 2026-05-22
+
+**v1.122.x is a line of [LTS releases](https://docs.victoriametrics.com/victoriametrics/lts-releases/). It contains important up-to-date bugfixes for [VictoriaMetrics enterprise](https://docs.victoriametrics.com/victoriametrics/enterprise/).
+All these fixes are also included in [the latest community release](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/latest).
+The v1.122.x line will be supported for at least 12 months since [v1.122.0](https://docs.victoriametrics.com/victoriametrics/changelog/#v11220) release**
+
+* BUGFIX: [stream aggregation](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/): stop emitting stale values for `quantiles(...)` outputs when a time series has no samples during the current aggregation interval. See [#10918](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10918). Thanks to @alexei38 for the contribution.
+* BUGFIX: [stream aggregation](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/): extend delay on aggregation windows flush by the biggest lag among pushed samples. Before, the delay was calculated as 95th percentile across samples, which could underrepresent outliers and reject them from aggregation as "too old". See [#10402](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10402).
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): fix a bug in [cardinality limiters](https://docs.victoriametrics.com/victoriametrics/vmagent/#cardinality-limiter) where series with different labels, like `{a="bc"}` and `{ab="c"}`, could be incorrectly treated as identical and dropped. See [#10937](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10937).
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): hide values passed to `-remoteWrite.headers` in startup logs, `/metrics`, and `/flags`, since they can contain sensitive HTTP headers such as `Authorization` and API keys.
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): hide values passed to `-remoteWrite.proxyURL` in startup logs, `/metrics`, and `/flags`, since they can contain sensitive credentials.
+* BUGFIX: [vmalert](https://docs.victoriametrics.com/victoriametrics/vmalert/): hide values passed to `-remoteWrite.headers`,`remoteRead.headers`, `datasource.headers` and `notifier.headers` in startup logs, `/metrics`, and `/flags`, since they can contain sensitive HTTP headers such as `Authorization` and API keys.
+* BUGFIX: [vmrestore](https://docs.victoriametrics.com/victoriametrics/vmrestore/): fix a bug where specifying `-storageDataPath` with a trailing slash could cause `vmrestore` to panic. See [#10823](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10823). Thanks to @utafrali for the contribution.
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): prevent unintentional rerouting of samples to other sharding targets when one of the `-remoteWrite.url` targets with `-remoteWrite.disableOnDiskQueue` becomes blocked. Previously this could break the sharding guarantee by sending samples to wrong targets instead of dropping or retrying them. See [#10507](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10507).
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): return error on startup if `-remoteWrite.disableOnDiskQueue` is not configured uniformly across all `-remoteWrite.url` targets when `-remoteWrite.shardByURL` is enabled. Either all targets must have it enabled or all must have it disabled. See [#10507](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10507).
+* BUGFIX: [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): hide values passed to `vmalert.proxyURL` in startup logs, `/metrics`, and `/flags`, since they can contain sensitive HTTP headers such as `Authorization` and API keys.
+* BUGFIX: [vmui](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmui): preserve exact series values in graph tooltips instead of rounding them by significant digits. See [#10952](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10952).
+
 ## [v1.122.22](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.122.22)
 
 Released at 2026-05-08

docs/victoriametrics/enterprise.md

@@ -121,7 +121,7 @@ It is allowed to run Enterprise components in [cases listed here](https://docs.v
 Binary releases of Enterprise components are available at [the releases page for VictoriaMetrics](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/latest),
 [the releases page for VictoriaLogs](https://github.com/VictoriaMetrics/VictoriaLogs/releases/latest)
 and [the releases page for VictoriaTraces](https://github.com/VictoriaMetrics/VictoriaTraces/releases/latest).
-Enterprise binaries and packages have `enterprise` suffix in their names. For example, `victoria-metrics-linux-amd64-v1.143.0-enterprise.tar.gz`.
+Enterprise binaries and packages have `enterprise` suffix in their names. For example, `victoria-metrics-linux-amd64-v1.144.0-enterprise.tar.gz`.
 
 In order to run binary release of Enterprise component, please download the `*-enterprise.tar.gz` archive for your OS and architecture
 from the corresponding releases page and unpack it. Then run the unpacked binary.
@@ -139,8 +139,8 @@ For example, the following command runs VictoriaMetrics Enterprise binary with t
 obtained at [this page](https://victoriametrics.com/products/enterprise/trial/):
 
 ```sh
-wget https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/v1.143.0/victoria-metrics-linux-amd64-v1.143.0-enterprise.tar.gz
-tar -xzf victoria-metrics-linux-amd64-v1.143.0-enterprise.tar.gz
+wget https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/v1.144.0/victoria-metrics-linux-amd64-v1.144.0-enterprise.tar.gz
+tar -xzf victoria-metrics-linux-amd64-v1.144.0-enterprise.tar.gz
 ./victoria-metrics-prod -license=BASE64_ENCODED_LICENSE_KEY
 ```
 
@@ -155,7 +155,7 @@ Alternatively, VictoriaMetrics Enterprise license can be stored in the file and
 It is allowed to run Enterprise components in [cases listed here](https://docs.victoriametrics.com/victoriametrics/enterprise/#valid-cases-for-victoriametrics-enterprise).
 
 Docker images for Enterprise components are available at [VictoriaMetrics Docker Hub](https://hub.docker.com/u/victoriametrics) and [VictoriaMetrics Quay](https://quay.io/organization/victoriametrics).
-Enterprise docker images have `enterprise` suffix in their names. For example, `victoriametrics/victoria-metrics:v1.143.0-enterprise`.
+Enterprise docker images have `enterprise` suffix in their names. For example, `victoriametrics/victoria-metrics:v1.144.0-enterprise`.
 
 In order to run Docker image of VictoriaMetrics Enterprise component, it is required to provide the license key via the command-line
 flag as described in the [binary-releases](https://docs.victoriametrics.com/victoriametrics/enterprise/#binary-releases) section.
@@ -165,13 +165,13 @@ Enterprise license key can be obtained at [this page](https://victoriametrics.co
 For example, the following command runs VictoriaMetrics Enterprise Docker image with the specified license key:
 
 ```sh
-docker run --name=victoria-metrics victoriametrics/victoria-metrics:v1.143.0-enterprise -license=BASE64_ENCODED_LICENSE_KEY
+docker run --name=victoria-metrics victoriametrics/victoria-metrics:v1.144.0-enterprise -license=BASE64_ENCODED_LICENSE_KEY
 ```
 
 Alternatively, the license code can be stored in the file and then referred via `-licenseFile` command-line flag:
 
 ```sh
-docker run --name=victoria-metrics -v /vm-license:/vm-license  victoriametrics/victoria-metrics:v1.143.0-enterprise -licenseFile=/path/to/vm-license
+docker run --name=victoria-metrics -v /vm-license:/vm-license  victoriametrics/victoria-metrics:v1.144.0-enterprise -licenseFile=/path/to/vm-license
 ```
 
 Example docker-compose configuration:
@@ -181,7 +181,7 @@ version: "3.5"
 services:
   victoriametrics:
     container_name: victoriametrics
-    image: victoriametrics/victoria-metrics:v1.143.0
+    image: victoriametrics/victoria-metrics:v1.144.0
     ports:
       - 8428:8428
     volumes:
@@ -213,7 +213,7 @@ is used to provide the license key in plain-text:
 ```yaml
 server:
   image:
-    tag: v1.143.0-enterprise
+    tag: v1.144.0-enterprise
 
 license:
   key: {BASE64_ENCODED_LICENSE_KEY}
@@ -224,7 +224,7 @@ In order to provide the license key via existing secret, the following values fi
 ```yaml
 server:
   image:
-    tag: v1.143.0-enterprise
+    tag: v1.144.0-enterprise
 
 license:
   secret:
@@ -274,7 +274,7 @@ spec:
   license:
     key: {BASE64_ENCODED_LICENSE_KEY}
   image:
-    tag: v1.143.0-enterprise
+    tag: v1.144.0-enterprise
 ```
 
 In order to provide the license key via an existing secret, the following custom resource is used:
@@ -291,7 +291,7 @@ spec:
       name: vm-license
       key: license
   image:
-    tag: v1.143.0-enterprise
+    tag: v1.144.0-enterprise
 ```
 
 Example secret with license key:
@@ -342,7 +342,7 @@ Builds are available for amd64 and arm64 architectures.
 
 Example archive:
 
-`victoria-metrics-linux-amd64-v1.143.0-enterprise.tar.gz`
+`victoria-metrics-linux-amd64-v1.144.0-enterprise.tar.gz`
 
 Includes:
 
@@ -351,7 +351,7 @@ Includes:
 
 Example Docker image:
 
-`victoriametrics/victoria-metrics:v1.143.0-enterprise-fips` – uses the FIPS-compatible binary and based on `scratch` image.
+`victoriametrics/victoria-metrics:v1.144.0-enterprise-fips` – uses the FIPS-compatible binary and based on `scratch` image.
 
 ## What Happens to Licensed Components When a License Expires
 

docs/victoriametrics/integrations/grafana/_index.md

@@ -59,6 +59,10 @@ Once connected, you can build graphs and dashboards using [PromQL](https://prome
 _Creating a datasource may require [specific permissions](https://grafana.com/docs/grafana/latest/administration/data-source-management/).
 If you don't see an option to create a data source - try contacting system administrator._
 
+If you run [vmalert](https://docs.victoriametrics.com/victoriametrics/vmalert/) and want to see its rules in [Grafana Alerting UI](https://grafana.com/docs/grafana/latest/alerting/),
+then set configure `-vmalert.proxyURL` on VictoriaMetrics [single-node](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmalert)
+or [vmselect in cluster version](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#vmalert).
+
 ## Multi-tenant access with vmauth and OIDC
 
 [vmauth](https://docs.victoriametrics.com/victoriametrics/vmauth/) can proxy Grafana datasource requests and enforce

docs/victoriametrics/integrations/opentelemetry.md

@@ -28,9 +28,17 @@ The following label sanitization options can be enabled:
 
 > These flags can be applied on vmagent, vminsert or VictoriaMetrics single-node.
 
+## Instrumentation Scope
+
+By default, VictoriaMetrics promotes [OTel scope metadata](https://opentelemetry.io/docs/specs/otel/common/instrumentation-scope/) to metric labels. This behavior can be disabled via `-opentelemetry.promoteScopeMetadata`.
+
 ## Resource Attributes
 
 By default, VictoriaMetrics promotes all [OpenTelemetry resource](https://opentelemetry.io/docs/specs/otel/resource/data-model/) attributes to labels and attaches them to all ingested OTLP metrics.
+The following attribute promotion options can be configured:
+* `-opentelemetry.promoteAllResourceAttributes` - promotes all resource attributes to labels, except for the ones configured with `-opentelemetry.ignoreResourceAttributes`.
+* `-opentelemetry.promoteResourceAttributes` - promotes specific list of resource attributes to labels. It cannot be configured simultaneously with `-opentelemetry.promoteAllResourceAttributes`.
+* `-opentelemetry.ignoreResourceAttributes` - controls which resource attributes to ignore, can only be set when `-opentelemetry.promoteAllResourceAttributes` is true.
 
 ## Exponential histograms
 

docs/victoriametrics/scrape_config_examples.md

@@ -35,8 +35,8 @@ scrape_configs:
 After you created the `scrape.yaml` file, download and unpack [single-node VictoriaMetrics](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) to the same directory:
 
 ```sh
-wget https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/v1.143.0/victoria-metrics-linux-amd64-v1.143.0.tar.gz
-tar xzf victoria-metrics-linux-amd64-v1.143.0.tar.gz
+wget https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/v1.144.0/victoria-metrics-linux-amd64-v1.144.0.tar.gz
+tar xzf victoria-metrics-linux-amd64-v1.144.0.tar.gz
 ```
 
 Then start VictoriaMetrics and instruct it to scrape targets defined in `scrape.yaml` and save scraped metrics
@@ -150,8 +150,8 @@ Then start [single-node VictoriaMetrics](https://docs.victoriametrics.com/victor
 
 ```yaml
 # Download and unpack single-node VictoriaMetrics
-wget https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/v1.143.0/victoria-metrics-linux-amd64-v1.143.0.tar.gz
-tar xzf victoria-metrics-linux-amd64-v1.143.0.tar.gz
+wget https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/v1.144.0/victoria-metrics-linux-amd64-v1.144.0.tar.gz
+tar xzf victoria-metrics-linux-amd64-v1.144.0.tar.gz
 
 # Run single-node VictoriaMetrics with the given scrape.yaml
 ./victoria-metrics-prod -promscrape.config=scrape.yaml

docs/victoriametrics/stream-aggregation/README.md

@@ -26,6 +26,7 @@ Stream aggregation has the following features:
   and/or scraped from [Prometheus-compatible targets](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#how-to-scrape-prometheus-exporters-such-as-node-exporter)
 - It can filter out raw samples matched by aggregation rules, so raw data will never reach the remote destination. See `-streamAggr.keepInput` and `-streamAggr.dropInput` in [aggregation config](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/);
 - It allows building [flexible processing pipelines](#routing);
+- It is [horizontally scalable](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#scaling-aggregation-horizontally).
 
 # Limitations
 
@@ -598,6 +599,47 @@ Below is an example of an `aggr.yaml` configuration that drops the `replica` and
   keep_metric_names: true
 ```
 
+## Scaling aggregation horizontally
+
+Aggregation output is only correct when all contributing samples are processed by the same aggregator instance.
+
+To scale the aggregation horizontally, always shard the input samples in a deterministic way. This can be achieved by
+building a two layer topology of vmagents where the first layer is responsible for sharding, and the second layer is responsible for aggregating:
+```mermaid
+flowchart LR
+    V1[vmagent-shard-1] -- requests_total{env=test, pod=foo} --> SV1[vmagent-aggr-1]
+    V1[vmagent-shard-1] -- requests_total{env=prod, pod=bar} --> SV2[vmagent-aggr-1]
+    V2[vmagent-shard-2] -- requests_total{env=prod, pod=baz} --> SV2[vmagent-aggr-2]
+    SV1 -- requests_total:5m_without_pod_total{env=test} --> x(( ))
+    SV2 -- requests_total:5m_without_pod_total{env=prod} --> y(( ))
+style x fill:none,stroke:none
+style y fill:none,stroke:none
+```
+
+The sharding layer of vmagents can be configured via the `-remoteWrite.shardByURL.labels` or `-remoteWrite.shardByURL.ignoreLabels`
+command line flags. See how to [shard data across remote write destinations](https://docs.victoriametrics.com/victoriametrics/vmagent/#sharding-among-remote-storages) for more details.
+
+The following requirements must be met for sharded aggregation to work correctly:
+- All sharding vmagents should have the same deterministic sharding configuration.
+- The sharding configuration must align with the `by` and `without` lists:
+  - Labels listed in `by` setting should be a subset of shard's routing key `-remoteWrite.shardByURL.labels`. 
+    With `-remoteWrite.shardByURL.labels=env,job` aggregator's `by` should include `by: env`, `by: job` or both: `by: [env, job]`.
+    This makes sure that all the samples for the same `env` and `job` are aggregated together and produce the complete output.
+  - Labels listed in `without` setting should be a superset of shard's routing key `--remoteWrite.shardByURL.ignoreLabels`.
+    With `-remoteWrite.shardByURL.ignoreLabels=env,job` aggegator's `without` should include at least both labels `without: [env,job]`.
+    This makes sure that `requests_total{env=test, job=foo}` and `requests_total{env=prod, job=foo}` are routed to the same aggregator
+    and are aggregated together. See also [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/5938#issuecomment-2018470324).
+- Aggregating vmagents should not produce collisions: the aggregation output should be unique across all the sharded agents.
+  For example, `requests_total:5m_without_env_pod_total` produced by both `vmagent-aggr-1` and `vmagent-aggr-2` will collide
+  unless they have labels uniquely identifying them. These labels should be either preserved during sharding and aggregation config,
+  or enforced on the output via `-remoteWrite.label` - see [these docs](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#cluster-mode) for more details.
+
+> Never shard histograms by `le` (or `vmrange` in case of VM histograms) label. A histogram is a logical group of series differing
+only in the bucket label. All of those buckets must land on the same aggregator at the same time so it can produce a
+coherent bucket set. See more about [aggregating histograms](https://docs.victoriametrics.com/stream-aggregation/#aggregating-histograms).
+
+See also [why you shouldn't put an aggregator behind a load balancer](https://docs.victoriametrics.com/stream-aggregation/#put-aggregator-behind-load-balancer).
+
 # Troubleshooting
 
 - [Unexpected spikes for `total` or `increase` outputs](#staleness).

docs/victoriametrics/victoria_metrics_common_flags.md

@@ -37,7 +37,7 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/
      Flag value can be read from the given file when using -deleteAuthKey=file:///abs/path/to/file or -deleteAuthKey=file://./relative/path/to/file.
      Flag value can be read from the given http/https url when using -deleteAuthKey=http://host/path or -deleteAuthKey=https://host/path
   -denyQueriesOutsideRetention
-     Whether to deny queries outside the configured -retentionPeriod. When set, then /api/v1/query_range would return '503 Service Unavailable' error for queries with 'from' value outside -retentionPeriod. This may be useful when multiple data sources with distinct retentions are hidden behind query-tee
+     Whether to deny queries outside the configured -retentionPeriod and -futureRetention. When set, then /api/v1/query_range will return an error for queries with 'from' value outside -retentionPeriod or 'to' value beyond -futureRetention. This may be useful when multiple data sources with distinct retentions are hidden behind query-tee
   -denyQueryTracing
      Whether to disable the ability to trace queries. See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#query-tracing
   -disablePerDayIndex
@@ -198,7 +198,7 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/
   -maxLabelsPerTimeseries int
      The maximum number of labels per time series to be accepted. Series with superfluous labels are ignored. In this case the vm_rows_ignored_total{reason="too_many_labels"} metric at /metrics page is incremented (default 40)
   -memory.allowedBytes size
-     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage
+     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage. The process may behave unexpectedly if this flag is set too small (e.g., 1 byte).
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 0)
   -memory.allowedPercent float
      Allowed percent of system memory VictoriaMetrics caches may occupy. See also -memory.allowedBytes. Too low a value may increase cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache which will result in higher disk IO usage (default 60)
@@ -217,9 +217,19 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 67108864)
   -opentelemetry.convertMetricNamesToPrometheus
      Whether to convert only metric names into Prometheus-compatible format for the metrics ingested via OpenTelemetry protocol; see https://docs.victoriametrics.com/victoriametrics/integrations/opentelemetry/
+  -opentelemetry.ignoreResourceAttributes array
+     Control which resource attributes to ignore, can only be set when 'opentelemetry.promoteAllResourceAttributes' is true.
+  -opentelemetry.labelNameUnderscoreSanitization
+     Whether to enable prepending of 'key' to labels starting with '_' when -opentelemetry.usePrometheusNaming is enabled. Reserved labels starting with '__' are not modified. See https://docs.victoriametrics.com/victoriametrics/integrations/opentelemetry/ (default true)
   -opentelemetry.maxRequestSize size
      The maximum size in bytes of a single OpenTelemetry request
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 67108864)
+  -opentelemetry.promoteAllResourceAttributes
+     Whether to promote all resource attributes to labels, except for the ones configured with 'opentelemetry.ignoreResourceAttributes'.
+  -opentelemetry.promoteResourceAttributes array
+     Promote specific list of resource attributes to labels.
+  -opentelemetry.promoteScopeMetadata
+     Whether to promote OTel scope metadata (i.e. name, version, schema URL, and attributes) to metric labels.
   -opentelemetry.usePrometheusNaming
      Whether to convert metric names and labels into Prometheus-compatible format for the metrics ingested via OpenTelemetry protocol; see https://docs.victoriametrics.com/victoriametrics/integrations/opentelemetry/
   -opentsdbHTTPListenAddr string

docs/victoriametrics/vmagent.md

@@ -182,15 +182,15 @@ among remote storage systems specified in `-remoteWrite.url`.
 > For example, if you set `-remoteWrite.url=srv+foo` and it's resolved to three addresses (`192.168.1.1`, `192.168.1.2`, `192.168.1.3`),
 > vmagent will only choose **one** randomly every time it (re-)creates the connection. In contrast, specifying the addresses manually (`-remoteWrite.url=192.168.1.1 -remoteWrite.url=192.168.1.2 -remoteWrite.url=192.168.1.3`) will shard samples across all three URLs.
 
-Sometimes, it may be necessary to use only a particular set of labels for sharding. For example, it may be necessary to route all the metrics with the same `instance` label
-to the same `-remoteWrite.url`. In this case, you can specify a comma-separated list of these labels in the `-remoteWrite.shardByURL.labels`
-command-line flag. For example, `-remoteWrite.shardByURL.labels=instance,__name__` would shard metrics with the same name and `instance`
-label to the same `-remoteWrite.url`.
+Use `-remoteWrite.shardByURL.labels` to route metrics among `-remoteWrite.url` based on their label values. 
+For example, `-remoteWrite.shardByURL.labels=instance,__name__` would shard metrics with the same name and `instance`
+label to the same `-remoteWrite.url`. This command-line flag allows specifying a comma-separated list of labels.
 
-Sometimes, it may be necessary to ignore some labels when sharding samples across multiple `-remoteWrite.url` backends.
-For example, if all the [raw samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) with the same set of labels
-except for the labels `instance` and `pod` must be routed to the same backend. In this case the list of ignored labels must be passed to
-`-remoteWrite.shardByURL.ignoreLabels` command-line flag: `-remoteWrite.shardByURL.ignoreLabels=instance,pod`.
+Alternatively, you can use `-remoteWrite.shardByURL.ignoreLabels` to route metrics among `-remoteWrite.url` based on their label values, excluding the specified labels.
+For example, `-remoteWrite.shardByURL.ignoreLabels=pod` would shard metrics `metric{pod="foo"}` and `metric{pod="bar"}` to the same `-remoteWrite.url`
+by ignoring the `pod` label. This command-line flag allows specifying a comma-separated list of labels.
+
+> Command-line flags `-remoteWrite.shardByURL.labels` and `-remoteWrite.shardByURL.ignoreLabels` are mutually exclusive.
 
 See also [how to scrape a large number of targets](#scraping-big-number-of-targets).
 

docs/victoriametrics/vmagent_common_flags.md

@@ -169,7 +169,7 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/vmagent/ .
   -maxLabelsPerTimeseries int
      The maximum number of labels per time series to be accepted. Series with superfluous labels are ignored. In this case the vm_rows_ignored_total{reason="too_many_labels"} metric at /metrics page is incremented
   -memory.allowedBytes size
-     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage
+     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage. The process may behave unexpectedly if this flag is set too small (e.g., 1 byte).
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 0)
   -memory.allowedPercent float
      Allowed percent of system memory VictoriaMetrics caches may occupy. See also -memory.allowedBytes. Too low a value may increase cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache which will result in higher disk IO usage (default 60)
@@ -184,9 +184,19 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/vmagent/ .
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 67108864)
   -opentelemetry.convertMetricNamesToPrometheus
      Whether to convert only metric names into Prometheus-compatible format for the metrics ingested via OpenTelemetry protocol; see https://docs.victoriametrics.com/victoriametrics/integrations/opentelemetry/
+  -opentelemetry.ignoreResourceAttributes array
+     Control which resource attributes to ignore, can only be set when 'opentelemetry.promoteAllResourceAttributes' is true.
+  -opentelemetry.labelNameUnderscoreSanitization
+     Whether to enable prepending of 'key' to labels starting with '_' when -opentelemetry.usePrometheusNaming is enabled. Reserved labels starting with '__' are not modified. See https://docs.victoriametrics.com/victoriametrics/integrations/opentelemetry/ (default true)
   -opentelemetry.maxRequestSize size
      The maximum size in bytes of a single OpenTelemetry request
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 67108864)
+  -opentelemetry.promoteAllResourceAttributes
+     Whether to promote all resource attributes to labels, except for the ones configured with 'opentelemetry.ignoreResourceAttributes'.
+  -opentelemetry.promoteResourceAttributes array
+     Promote specific list of resource attributes to labels.
+  -opentelemetry.promoteScopeMetadata
+     Whether to promote OTel scope metadata (i.e. name, version, schema URL, and attributes) to metric labels.
   -opentelemetry.usePrometheusNaming
      Whether to convert metric names and labels into Prometheus-compatible format for the metrics ingested via OpenTelemetry protocol; see https://docs.victoriametrics.com/victoriametrics/integrations/opentelemetry/
   -opentsdbHTTPListenAddr string

docs/victoriametrics/vmalert.md

@@ -21,23 +21,9 @@ Recording rules results are persisted via remote write protocols and require `-r
 `vmalert` is heavily inspired by [Prometheus](https://prometheus.io/docs/alerting/latest/overview/)
 implementation and aims to be compatible with its syntax.
 
-Configure `-vmalert.proxyURL` on VictoriaMetrics [single-node](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmalert)
-or [vmselect in cluster version](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#vmalert)
-to proxy requests to `vmalert`. Proxying is needed for the following cases:
-
-* to proxy requests from [Grafana Alerting UI](https://grafana.com/docs/grafana/latest/alerting/);
-* to access `vmalert`'s UI through [vmui](https://docs.victoriametrics.com/victoriametrics/#vmui).
-
-[VictoriaMetrics Cloud](https://console.victoriametrics.cloud/signUp?utm_source=website&utm_campaign=docs_vm_vmalert_intro)
-provides out-of-the-box alerting functionality based on `vmalert`. This service simplifies the setup
-and management of alerting and recording rules as well as the integration with Alertmanager. For more details,
-please refer to the [VictoriaMetrics Cloud documentation](https://docs.victoriametrics.com/victoriametrics-cloud/alertmanager-setup-for-deployment/).
-
 ## Features
 
-* Integration with [VictoriaMetrics](https://github.com/VictoriaMetrics/VictoriaMetrics) and [MetricsQL](https://docs.victoriametrics.com/victoriametrics/metricsql/);
-* Integration with [VictoriaLogs](https://docs.victoriametrics.com/victorialogs/) and [LogsQL](https://docs.victoriametrics.com/victorialogs/logsql/). See [this doc](https://docs.victoriametrics.com/victorialogs/vmalert/);
-* Integration with [VictoriaTraces](https://docs.victoriametrics.com/victoriatraces/) which also uses [LogsQL](https://docs.victoriametrics.com/victorialogs/logsql/). See [this doc](https://docs.victoriametrics.com/victoriatraces/vmalert/);
+* Integration with VictoriaMetrics, VictoriaLogs, VictoriaTraces, Graphite and Prometheus compatible storages. See [Integrations](https://docs.victoriametrics.com/victoriametrics/vmalert/#integrations) for details;
 * Prometheus [alerting rules definition format](https://prometheus.io/docs/prometheus/latest/configuration/alerting_rules/#defining-alerting-rules)
   support;
 * Integration with [Alertmanager](https://github.com/prometheus/alertmanager) starting from [Alertmanager v0.16.0-alpha](https://github.com/prometheus/alertmanager/releases/tag/v0.16.0-alpha.0);
@@ -62,8 +48,8 @@ please refer to the [VictoriaMetrics Cloud documentation](https://docs.victoriam
 
 To start using `vmalert` you will need the following things:
 
-* list of rules - PromQL/MetricsQL expressions to execute;
-* datasource address - reachable endpoint with [Prometheus HTTP API](https://prometheus.io/docs/prometheus/latest/querying/api/#http-api) support for running queries against;
+* list of rules - PromQL/MetricsQL/LogsQL/GraphiteQL expressions to execute;
+* datasource address - a storage that [vmalert integrates with](https://docs.victoriametrics.com/victoriametrics/vmalert/#integrations) for executing queries;
 * notifier address [optional] - reachable [Alert Manager](https://github.com/prometheus/alertmanager) instance for processing,
   aggregating alerts, and sending notifications. Please note, notifier address also supports Consul and DNS Service Discovery via
   [config file](https://docs.victoriametrics.com/victoriametrics/vmalert/#notifier-configuration-file).
@@ -73,7 +59,7 @@ To start using `vmalert` you will need the following things:
 * remote read address [optional] - MetricsQL compatible datasource to restore alerts state from.
 
 You can use the existing [docker-compose environment](https://github.com/VictoriaMetrics/VictoriaMetrics/tree/master/deployment/docker#victoriametrics-single-server)
-as example. It already contains vmalert configured with list of alerting rules and integrated with Alert Manager and VictoriaMetrics.
+as an example. It already contains vmalert configured with the list of alerting rules and integrated with Alert Manager and VictoriaMetrics.
 
 Alternatively, build `vmalert` from sources:
 
@@ -87,7 +73,7 @@ Then run `vmalert`:
 
 ```sh
 ./bin/vmalert -rule=alert.rules \            # Path to the file with rules configuration. Supports wildcard and HTTP URL (S3/GCS are available in Enterprise).
-    -datasource.url=http://localhost:8428 \  # Prometheus HTTP API compatible datasource
+    -datasource.url=http://localhost:8428 \  # VictoriaMetrics URL to query for rules evaluation. See other available Integrations above.
     -notifier.url=http://localhost:9093 \    # AlertManager URL (required if alerting rules are used)
     -notifier.url=http://127.0.0.1:9093 \    # AlertManager replica URL
     -remoteWrite.url=http://localhost:8428 \ # Remote write compatible storage to persist rules and alerts state info (required if recording rules are used)
@@ -107,7 +93,7 @@ See also [stream aggregation](https://docs.victoriametrics.com/victoriametrics/s
 
 See the full list of configuration flags in [configuration](#configuration) section.
 
-If you run multiple `vmalert` services on the same datastore or AlertManager and need to distinguish the results or alerts, 
+If you run multiple `vmalert` services for the same datasource or AlertManager and need to distinguish the results or alerts, 
 specify different `-external.label` command-line flags to indicate which `vmalert` generated them. 
 If rule result metrics have label that conflict with `-external.label`, `vmalert` will automatically rename
 it with prefix `exported_`.
@@ -116,6 +102,7 @@ Configuration for [recording](https://prometheus.io/docs/prometheus/latest/confi
 and [alerting](https://prometheus.io/docs/prometheus/latest/configuration/alerting_rules/) rules is very
 similar to Prometheus rules and configured using YAML. Configuration examples may be found
 in [testdata](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmalert/config/testdata) folder.
+
 Every `rule` belongs to a `group` and every configuration file may contain arbitrary number of groups:
 
 ```yaml
@@ -123,13 +110,7 @@ groups:
   [ - <rule_group> ]
 ```
 
-> Explore how to integrate `vmalert` with [VictoriaMetrics Anomaly Detection](https://docs.victoriametrics.com/anomaly-detection/) in the following [guide](https://docs.victoriametrics.com/anomaly-detection/guides/guide-vmanomaly-vmalert/).
-
-> For users of [VictoriaMetrics Cloud](https://console.victoriametrics.cloud/signUp?utm_source=website&utm_campaign=docs_vm_vmalert_config),
-> many of the configuration steps (including highly available setup of `vmalert` for cluster deployments) are handled automatically.
-> Please, refer to the [VictoriaMetrics Cloud documentation](https://docs.victoriametrics.com/victoriametrics-cloud/alertmanager-setup-for-deployment/) for more details.
-
-### Groups
+## Groups
 
 Each group has the following attributes:
 
@@ -230,9 +211,11 @@ rules:
 
 ### Rules
 
-Every rule contains `expr` field for [PromQL](https://prometheus.io/docs/prometheus/latest/querying/basics/)
-or [MetricsQL](https://docs.victoriametrics.com/victoriametrics/metricsql/) expression. `vmalert` will execute the configured
-expression and then act according to the Rule type.
+Every rule contains an `expr` field for the expression to evaluate against the configured datasource.
+Depending on `group.type` value or `-rule.defaultRuleType` cmd-line flag expression can be one of the following types:
+- `prometheus` (default) - [PromQL](https://prometheus.io/docs/prometheus/latest/querying/basics/) or [MetricsQL](https://docs.victoriametrics.com/victoriametrics/metricsql/) expression.
+- `vlogs` - [VictoriaLogs](https://docs.victoriametrics.com/victorialogs/vmalert/) expression.
+- `graphite` - [Graphite](https://graphite.readthedocs.io/en/stable/render_api.html) expression.
 
 There are two types of Rules:
 
@@ -244,8 +227,7 @@ There are two types of Rules:
   `-remoteWrite.url`. Recording rules are used to precompute frequently needed or computationally
   expensive expressions and save their result as a new set of time series ([Prometheus recording rules docs](https://prometheus.io/docs/prometheus/latest/configuration/recording_rules/)).
 
-`vmalert` forbids defining duplicates - rules with the same combination of name, expression, and labels
-within one group.
+> `vmalert` forbids defining duplicates - rules with the same combination of name, expression and labels within one group.
 
 #### Alerting rules
 
@@ -256,8 +238,8 @@ The syntax for alerting rule is the following:
 alert: <string>
 
 # The expression to evaluate. The expression language depends on the type value.
-# By default, PromQL/MetricsQL expression is used. If group.type="graphite", then the expression
-# must contain valid Graphite expression.
+# By default, PromQL/MetricsQL expression is used. Other available types are "graphite" and "vlogs".
+# See https://docs.victoriametrics.com/victoriametrics/vmalert/#integrations
 expr: <string>
 
 # Alerts are considered firing once they have been returned for this long.
@@ -303,7 +285,44 @@ annotations:
   [ <labelname>: <tmpl_string> ]
 ```
 
-#### Templating
+#### Recording rules
+
+The syntax for recording rules is the following:
+
+```yaml
+# The name of the time series to output to. Must be a valid metric name.
+record: <string>
+
+# The expression to evaluate. The expression language depends on the type value.
+# By default, PromQL/MetricsQL expression is used. Other available types are "graphite" and "vlogs".
+# See https://docs.victoriametrics.com/victoriametrics/vmalert/#integrations
+expr: <string>
+
+# Labels to add or overwrite labels from other external label sources, such as group labels, before storing the result.
+#
+# In case of conflicts, original labels are kept with prefix `exported_`.
+# As a special case, specifying a label with an empty string value removes the label from the result if it exists 
+# in the original query result; otherwise, it is ignored.
+#
+# Labels do not support templating in https://docs.victoriametrics.com/victoriametrics/vmalert/#templating due to cardinality concerns. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8171.
+labels:
+  [ <labelname>: <labelvalue> ]
+
+# Whether to print debug information into logs.
+# Information includes requests sent to the datasource.
+# information - it will be printed to logs.
+# Logs are printed with INFO level, so make sure that -loggerLevel=INFO to see the output.
+[ debug: <bool> | default = false ]
+
+# Defines the number of rule updates entries stored in memory
+# and available for view on rule Details page.
+# Overrides `rule.updateEntriesLimit` value for this specific rule.
+[ update_entries_limit: <integer> | default 0 ]
+```
+
+For recording rules to work `-remoteWrite.url` must be specified.
+
+## Templating
 
 It is allowed to use [Go templating](https://golang.org/pkg/text/template/) in annotations and labels(with limited support) to format data, iterate over
 or execute expressions.
@@ -325,7 +344,7 @@ The following variables are available in templating:
 
 Additionally, `vmalert` provides some extra templating functions listed in [template functions](#template-functions) and [reusable templates](#reusable-templates).
 
-#### Template functions
+### Template functions
 
 `vmalert` provides the following template functions, which can be used during [templating](#templating):
 
@@ -349,7 +368,7 @@ Additionally, `vmalert` provides some extra templating functions listed in [temp
 * `parseDurationTime` - parses the input string into [time.Duration](https://pkg.go.dev/time#Duration).
 * `pathEscape` - escapes the input string, so it can be safely put inside path part of URL.
 * `pathPrefix` - returns the path part of the `-external.url` command-line flag.
-* `query` - executes the [MetricsQL](https://docs.victoriametrics.com/victoriametrics/metricsql/) query against `-datasource.url` and returns the query result.
+* `query` - executes query against `-datasource.url` and returns the query result.
   For example, `{{ query "sort_desc(process_resident_memory_bytes)" | first | value }}` executes the `sort_desc(process_resident_memory_bytes)`
   query at `-datasource.url` and returns the first result.
 * `queryEscape` - escapes the input string, so it can be safely put inside [query arg](https://en.wikipedia.org/wiki/Percent-encoding) part of URL.
@@ -369,7 +388,7 @@ Additionally, `vmalert` provides some extra templating functions listed in [temp
 * `toUpper` - converts all the chars in the input string to uppercase.
 * `value` - returns the numeric value from the input query result.
 
-#### Reusable templates
+### Reusable templates
 
 Like in Alertmanager you can define [reusable templates](https://prometheus.io/docs/prometheus/latest/configuration/template_examples/#defining-reusable-templates)
 to share same templates across annotations. Just define the templates in a file and
@@ -407,44 +426,8 @@ groups:
 The `-rule.templates` flag supports wildcards so multiple files with templates can be loaded.
 The content of `-rule.templates` can be also [hot reloaded](#hot-config-reload).
 
-#### Recording rules
 
-The syntax for recording rules is following:
-
-```yaml
-# The name of the time series to output to. Must be a valid metric name.
-record: <string>
-
-# The expression to evaluate. The expression language depends on the type value.
-# By default, MetricsQL expression is used. If group.type="graphite", then the expression
-# must contain valid Graphite expression.
-expr: <string>
-
-# Labels to add or overwrite labels from other external label sources, such as group labels, before storing the result.
-#
-# In case of conflicts, original labels are kept with prefix `exported_`.
-# As a special case, specifying a label with an empty string value removes the label from the result if it exists 
-# in the original query result; otherwise, it is ignored.
-#
-# Labels do not support templating in https://docs.victoriametrics.com/victoriametrics/vmalert/#templating due to cardinality concerns. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8171.
-labels:
-  [ <labelname>: <labelvalue> ]
-
-# Whether to print debug information into logs.
-# Information includes requests sent to the datasource.
-# information - it will be printed to logs.
-# Logs are printed with INFO level, so make sure that -loggerLevel=INFO to see the output.
-[ debug: <bool> | default = false ]
-
-# Defines the number of rule's updates entries stored in memory
-# and available for view on rule's Details page.
-# Overrides `rule.updateEntriesLimit` value for this specific rule.
-[ update_entries_limit: <integer> | default 0 ]
-```
-
-For recording rules to work `-remoteWrite.url` must be specified.
-
-### Alerts state on restarts
+## Alerts state on restarts
 
 `vmalert` holds alerts state in the memory. Restart of the `vmalert` process will reset the state of all active alerts
 in the memory. To prevent `vmalert` from losing the state on restarts configure it to persist the state
@@ -465,7 +448,7 @@ in configured `-remoteRead.url`, weren't updated in the last `1h` (controlled by
 or received state doesn't match current `vmalert` rules configuration. `vmalert` marks successfully restored rules
 with `restored` label in [web UI](#web).
 
-### Link to alert source
+## Link to alert source
 
 Alerting notifications sent by vmalert always contain a `source` link. By default, the link format
 is the following `http://<vmalert-addr>/vmalert/alert?group_id=<group_id>&alert_id=<alert_id>`. On click, it opens
@@ -501,7 +484,9 @@ In addition to `source` link, some extra links could be added to alert's [annota
 field. See [how we use them](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/9751ea10983d42068487624849cac7ad6fd7e1d8/deployment/docker/rules/alerts-cluster.yml#L44)
 to link alerting rule and the corresponding panel on Grafana dashboard.
 
-### Multitenancy
+## Multitenancy
+
+> See how to use [multitenancy in rules for VictoriaLogs](https://docs.victoriametrics.com/victorialogs/vmalert/#how-to-use-multitenancy-in-rules).
 
 There are the following approaches exist for alerting and recording rules across
 [multiple tenants](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#multitenancy):
@@ -559,7 +544,7 @@ The enterprise version of vmalert is available in `vmutils-*-enterprise.tar.gz`
 at [release page](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/latest) and in `*-enterprise`
 tags at [Docker Hub](https://hub.docker.com/r/victoriametrics/vmalert/tags) and [Quay](https://quay.io/repository/victoriametrics/vmalert?tab=tags).
 
-### Reading rules from object storage
+## Reading rules from object storage
 
 [Enterprise version](https://docs.victoriametrics.com/victoriametrics/enterprise/) of `vmalert` may read alerting and recording rules
 from object storage:
@@ -578,7 +563,7 @@ The following [command-line flags](#flags) can be used for fine-tuning access to
 * `-s3.customEndpoint` - custom S3 endpoint for use with S3-compatible storages (e.g. MinIO). S3 is used if not set.
 * `-s3.forcePathStyle` - prefixing endpoint with bucket name when set false, true by default.
 
-### Topology examples
+## Topology examples
 
 The following sections are showing how `vmalert` may be used and configured
 for different scenarios.
@@ -589,7 +574,7 @@ Please note, not all flags in examples are required:
   you have recording rules or want to store [alerts state](#alerts-state-on-restarts) on `vmalert` restarts;
 * `-notifier.url` is optional and is needed only if you have alerting rules.
 
-#### Single-node VictoriaMetrics
+### Single-node VictoriaMetrics
 
 The simplest configuration where one single-node VM server is used for
 rules execution, storing recording rules results and alerts state.
@@ -607,7 +592,7 @@ rules execution, storing recording rules results and alerts state.
 ![vmalert single](vmalert_single.webp)
 {width="500"}
 
-#### Cluster VictoriaMetrics
+### Cluster VictoriaMetrics
 
 In [cluster mode](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/)
 VictoriaMetrics has separate components for writing and reading path:
@@ -630,7 +615,7 @@ Cluster mode could have multiple `vminsert` and `vmselect` components.
 In case when you want to spread the load on these components - add balancers before them and configure
 `vmalert` with balancer addresses. Please, see more about [VictoriaMetrics cluster architecture](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#architecture-overview).
 
-#### HA vmalert
+### HA vmalert
 
 For High Availability(HA) user can run multiple identically configured `vmalert` instances.
 It means all of them will execute the same rules, write state and results to
@@ -674,12 +659,12 @@ to ensure [high availability](https://github.com/prometheus/alertmanager#high-av
 This example uses single-node VM server for the sake of simplicity.
 Check how to replace it with [cluster VictoriaMetrics](#cluster-victoriametrics) if needed.
 
-#### Downsampling and aggregation via vmalert
+### Downsampling and aggregation via vmalert
 
 _Please note, [stream aggregation](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/) might be more efficient
 for cases when downsampling or aggregation need to be applied **before data gets into the TSDB.**_
 
-`vmalert` can't modify existing data. But it can run arbitrary PromQL/MetricsQL queries
+`vmalert` can't modify existing data. But it can run arbitrary queries
 via [recording rules](#recording-rules) and backfill results to the configured `-remoteWrite.url`.
 This ability allows to aggregate data. For example, the following rule will calculate the average value for
 metric `http_requests` on the `5m` interval:
@@ -732,7 +717,7 @@ Flags `-remoteRead.url` and `-notifier.url` are omitted since we assume only rec
 
 See also [stream aggregation](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/) and [downsampling](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#downsampling).
 
-#### Multiple remote writes
+### Multiple remote writes
 
 For persisting recording or alerting rule results `vmalert` requires `-remoteWrite.url` to be set.
 But this flag supports only one destination. To persist rule results to multiple destinations
@@ -746,7 +731,72 @@ Using `vmagent` as a proxy provides additional benefits such as
 [data persisting when storage is unreachable](https://docs.victoriametrics.com/victoriametrics/vmagent/#replication-and-high-availability),
 or time series modification via [relabeling](https://docs.victoriametrics.com/victoriametrics/relabeling/).
 
-### Web
+## Integrations
+
+vmalert can be integrated with different data sources for alerting and recording rules. But it deliberately allows
+configuring only one `datasource.url`. We recommend running separate instances of vmalert for each datasource type
+with the specified `-rule.defaultRuleType=<datasource_type>` command-line flag.
+
+###### VictoriaMetrics
+
+vmalert natively integrates with [VictoriaMetrics](https://docs.victoriametrics.com/victoriametrics/) for alerting and
+recording rules.
+
+###### VictoriaLogs
+
+vmalert integrates with [VictoriaLogs](https://docs.victoriametrics.com/victorialogs/) and allows configuring alerting and recording rules using [LogsQL](https://docs.victoriametrics.com/victorialogs/logsql/).
+Results of recording rules and alerting state should be persisted to the remote-write compatible storage, such as VictoriaMetrics.
+To enable VictoriaLogs compatibility set the `-rule.defaultRuleType=vlogs` command-line flag.
+
+See [this doc](https://docs.victoriametrics.com/victorialogs/vmalert/) for details.
+
+###### VictoriaTraces
+
+vmalert integrates with [VictoriaTraces](https://docs.victoriametrics.com/victoriatraces/) in exactly the same way as
+with [VictoriaLogs](https://docs.victoriametrics.com/victoriametrics/vmalert/#victorialogs).
+
+###### Graphite
+
+vmalert integrates with [Graphite Render API](https://graphite.readthedocs.io/en/stable/render_api.html) and allows configuring alerting and recording rules.
+During evaluation, vmalert will send requests to `<-datasource.url>/render?format=json`.
+To enable Graphite compatibility set the `-rule.defaultRuleType=graphite` command-line flag.
+
+Since VictoriaMetrics supports both Graphite and Prometheus APIs, it is possible to mix Graphite and VictoriaMetrics rules.
+On the group level, set the `type` field to specify to which datasource type it should belong: `prometheus` (MetricsQL) or `graphite` (GraphiteQL).
+When using vmalert with both `graphite` and `prometheus` rules configured against the cluster version of VictoriaMetrics, don't forget
+to set the `-datasource.appendTypePrefix` flag to `true`, so vmalert can adjust URL prefix automatically based on the query type.
+
+###### Prometheus
+
+vmalert uses [Prometheus HTTP API](https://prometheus.io/docs/prometheus/latest/querying/api/#http-api) for querying
+and [Prometheus Remote Write v1 protocol](https://prometheus.io/docs/specs/prw/remote_write_spec/) for persisting
+recording rules results and alerting state. Hence, it can be integrated with any Prometheus-compatible storage
+that supports these protocols.
+
+###### Grafana
+
+To proxy requests from [Grafana Alerting UI](https://grafana.com/docs/grafana/latest/alerting/) configure `-vmalert.proxyURL`
+on VictoriaMetrics [single-node](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmalert)
+or [vmselect in cluster version](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#vmalert).
+
+###### vmui
+
+To access rules UI through [vmui](https://docs.victoriametrics.com/victoriametrics/#vmui) configure `-vmalert.proxyURL`
+on VictoriaMetrics [single-node](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmalert)
+or [vmselect in cluster version](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#vmalert).
+
+###### vmanomaly
+
+See how to integrate vmalert with [VictoriaMetrics Anomaly Detection](https://docs.victoriametrics.com/anomaly-detection/)
+in the following [guide](https://docs.victoriametrics.com/anomaly-detection/guides/guide-vmanomaly-vmalert/).
+
+###### VictoriaMetrics Cloud
+
+For users of [VictoriaMetrics Cloud](https://console.victoriametrics.cloud/signUp?utm_source=website&utm_campaign=docs_vm_vmalert_config),
+many of the configuration steps (including highly available setup of `vmalert` for cluster deployments) are handled automatically.
+Please refer to the [VictoriaMetrics Cloud documentation](https://docs.victoriametrics.com/victoriametrics-cloud/alertmanager-setup-for-deployment/) for more details.
+
+## Web
 
 `vmalert` runs a web-server (`-httpListenAddr`) for serving metrics and alerts endpoints:
 
@@ -771,28 +821,11 @@ This may be used for better integration with Grafana unified alerting system. Se
 * [How to query vmalert from single-node VictoriaMetrics](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmalert)
 * [How to query vmalert from VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#vmalert)
 
-## Graphite
-
-vmalert sends requests to `<-datasource.url>/render?format=json` during evaluation of alerting and recording rules
-if the corresponding group or rule contains `type: "graphite"` config option. It is expected that the `<-datasource.url>/render`
-implements [Graphite Render API](https://graphite.readthedocs.io/en/stable/render_api.html) for `format=json`.
-When using vmalert with both `graphite` and `prometheus` rules configured against cluster version of VM do not forget
-to set `-datasource.appendTypePrefix` flag to `true`, so vmalert can adjust URL prefix automatically based on the query type.
-
-## VictoriaLogs
-
-vmalert supports [VictoriaLogs](https://docs.victoriametrics.com/victorialogs/) as a datasource for writing alerting and recording rules using [LogsQL](https://docs.victoriametrics.com/victorialogs/logsql/). See [this doc](https://docs.victoriametrics.com/victorialogs/vmalert/) for details.
-
-## VictoriaTraces
-
-vmalert supports [VictoriaTraces](https://docs.victoriametrics.com/victoriatraces/) as a (`vlogs`) datasource for writing alerting and recording rules using [LogsQL](https://docs.victoriametrics.com/victorialogs/logsql/). See [this doc](https://docs.victoriametrics.com/victoriatraces/vmalert/) for details.
-
 ## Rules backfilling
 
 vmalert supports alerting and recording rules backfilling (aka `replay`). In replay mode vmalert
 can read the same rules configuration as normal, evaluate them on the given time range and backfill
-results via remote write to the configured storage. vmalert supports any PromQL/MetricsQL compatible
-data source for backfilling.
+results via remote write to the configured storage. vmalert supports only the `prometheus` datasource type for backfilling.
 
 Please note, that response caching may lead to unexpected results during and after backfilling process.
 In order to avoid this you need to reset cache contents or disable caching when using backfilling
@@ -852,13 +885,9 @@ vmalert respects `evaluationInterval` value set by flag or per-group during the
 vmalert automatically disables caching on VictoriaMetrics side by sending `nocache=1` param. It allows
 to prevent cache pollution and unwanted time range boundaries adjustment during backfilling.
 
-#### Recording rules
+Results of recording rules `replay` should match the results of normal rules evaluation.
 
-The result of recording rules `replay` should match with results of normal rules evaluation.
-
-#### Alerting rules
-
-The result of alerting rules `replay` is time series reflecting [alert's state](#alerts-state-on-restarts).
+Results of alerting rules `replay` are the time series reflecting the [state of the alert](#alerts-state-on-restarts).
 To see if `replayed` alert has fired in the past use the following PromQL/MetricsQL expression:
 
 ```

docs/victoriametrics/vmalert_common_flags.md

@@ -166,7 +166,7 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/vmalert/ .
   -loggerWarnsPerSecondLimit int
      Per-second limit on the number of WARN messages. If more than the given number of warns are emitted per second, then the remaining warns are suppressed. Zero values disable the rate limit
   -memory.allowedBytes size
-     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage
+     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage. The process may behave unexpectedly if this flag is set too small (e.g., 1 byte).
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 0)
   -memory.allowedPercent float
      Allowed percent of system memory VictoriaMetrics caches may occupy. See also -memory.allowedBytes. Too low a value may increase cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache which will result in higher disk IO usage (default 60)

docs/victoriametrics/vmauth_common_flags.md

@@ -135,7 +135,7 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/vmauth/ .
      The maximum request body size to buffer in memory for potential retries at other backends. Request bodies larger than this size cannot be retried if the backend fails. Zero or negative value disables retries. See also -requestBufferSize
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 16384)
   -memory.allowedBytes size
-     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage
+     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage. The process may behave unexpectedly if this flag is set too small (e.g., 1 byte).
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 0)
   -memory.allowedPercent float
      Allowed percent of system memory VictoriaMetrics caches may occupy. See also -memory.allowedBytes. Too low a value may increase cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache which will result in higher disk IO usage (default 60)

docs/victoriametrics/vmctl/vmctl.md

@@ -34,9 +34,9 @@ vmctl command-line tool is available as:
 
 Download and unpack vmctl:
 ```sh
-wget https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/v1.143.0/vmutils-darwin-arm64-v1.143.0.tar.gz
+wget https://github.com/VictoriaMetrics/VictoriaMetrics/releases/download/v1.144.0/vmutils-darwin-arm64-v1.144.0.tar.gz
 
-tar xzf vmutils-darwin-arm64-v1.143.0.tar.gz
+tar xzf vmutils-darwin-arm64-v1.144.0.tar.gz
 ```
 
 Once binary is unpacked, see the full list of supported modes by running the following command:

docs/victoriametrics/vminsert_common_flags.md

@@ -169,7 +169,7 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/cluster-victori
   -maxLabelsPerTimeseries int
      The maximum number of labels per time series to be accepted. Series with superfluous labels are ignored. In this case the vm_rows_ignored_total{reason="too_many_labels"} metric at /metrics page is incremented (default 40)
   -memory.allowedBytes size
-     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage
+     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage. The process may behave unexpectedly if this flag is set too small (e.g., 1 byte).
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 0)
   -memory.allowedPercent float
      Allowed percent of system memory VictoriaMetrics caches may occupy. See also -memory.allowedBytes. Too low a value may increase cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache which will result in higher disk IO usage (default 60)
@@ -184,9 +184,19 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/cluster-victori
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 67108864)
   -opentelemetry.convertMetricNamesToPrometheus
      Whether to convert only metric names into Prometheus-compatible format for the metrics ingested via OpenTelemetry protocol; see https://docs.victoriametrics.com/victoriametrics/integrations/opentelemetry/
+  -opentelemetry.ignoreResourceAttributes array
+     Control which resource attributes to ignore, can only be set when 'opentelemetry.promoteAllResourceAttributes' is true.
+  -opentelemetry.labelNameUnderscoreSanitization
+     Whether to enable prepending of 'key' to labels starting with '_' when -opentelemetry.usePrometheusNaming is enabled. Reserved labels starting with '__' are not modified. See https://docs.victoriametrics.com/victoriametrics/integrations/opentelemetry/ (default true)
   -opentelemetry.maxRequestSize size
      The maximum size in bytes of a single OpenTelemetry request
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 67108864)
+  -opentelemetry.promoteAllResourceAttributes
+     Whether to promote all resource attributes to labels, except for the ones configured with 'opentelemetry.ignoreResourceAttributes'.
+  -opentelemetry.promoteResourceAttributes array
+     Promote specific list of resource attributes to labels.
+  -opentelemetry.promoteScopeMetadata
+     Whether to promote OTel scope metadata (i.e. name, version, schema URL, and attributes) to metric labels.
   -opentelemetry.usePrometheusNaming
      Whether to convert metric names and labels into Prometheus-compatible format for the metrics ingested via OpenTelemetry protocol; see https://docs.victoriametrics.com/victoriametrics/integrations/opentelemetry/
   -opentsdbHTTPListenAddr string

docs/victoriametrics/vminsert_enterprise_flags.md

@@ -18,6 +18,20 @@ sitemap:
      Whether to skip verification of TLS certificates provided by -storageNode nodes if -cluster.tls flag is set. Note that disabled TLS certificate verification breaks security. This flag is available only in VictoriaMetrics enterprise. See https://docs.victoriametrics.com/victoriametrics/enterprise/
   -cluster.tlsKeyFile string
      Path to client-side TLS key file to use when connecting to -storageNode if -cluster.tls flag is set. See https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#mtls-protection . This flag is available only in VictoriaMetrics enterprise. See https://docs.victoriametrics.com/victoriametrics/enterprise/
+  -clusternative.tls
+     Whether to use TLS when accepting connections at -clusternativeListenAddr. See https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#mtls-protection . This flag is available only in VictoriaMetrics enterprise. See https://docs.victoriametrics.com/victoriametrics/enterprise/
+  -clusternative.tlsCAFile string
+     Path to TLS CA file to use for verifying certificates provided by vminsert, which connects at -clusternativeListenAddr if -clusternative.tls flag is set. By default system CA is used. See https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#mtls-protection . This flag is available only in VictoriaMetrics enterprise. See https://docs.victoriametrics.com/victoriametrics/enterprise/
+  -clusternative.tlsCertFile string
+     Path to server-side TLS certificate file to use when accepting connections at -clusternativeListenAddr if -clusternative.tls flag is set. See https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#mtls-protection . This flag is available only in VictoriaMetrics enterprise. See https://docs.victoriametrics.com/victoriametrics/enterprise/
+  -clusternative.tlsCipherSuites array
+     Optional list of TLS cipher suites used for connections at -clusternativeListenAddr if -clusternative.tls flag is set. See the list of supported cipher suites at https://pkg.go.dev/crypto/tls#pkg-constants . This flag is available only in VictoriaMetrics enterprise. See https://docs.victoriametrics.com/victoriametrics/enterprise/
+     Supports an array of values separated by comma or specified via multiple flags.
+     Each array item can contain comma inside single-quoted or double-quoted string, {}, [] and () braces.
+  -clusternative.tlsInsecureSkipVerify
+     Whether to skip verification of TLS certificates provided by vminsert, which connects to -clusternativeListenAddr if -clusternative.tls flag is set. Note that disabled TLS certificate verification breaks security. This flag is available only in VictoriaMetrics enterprise. See https://docs.victoriametrics.com/victoriametrics/enterprise/
+  -clusternative.tlsKeyFile string
+     Path to server-side TLS key file to use when accepting connections at -clusternativeListenAddr if -clusternative.tls flag is set. See https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/#mtls-protection . This flag is available only in VictoriaMetrics enterprise. See https://docs.victoriametrics.com/victoriametrics/enterprise/
   -eula
      Deprecated, please use -license or -licenseFile flags instead. By specifying this flag, you confirm that you have an enterprise license and accept the ESA https://victoriametrics.com/legal/esa/ . This flag is available only in Enterprise binaries. See https://docs.victoriametrics.com/victoriametrics/enterprise/
   -license string

docs/victoriametrics/vmselect_common_flags.md

@@ -126,7 +126,7 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/cluster-victori
   -loggerWarnsPerSecondLimit int
      Per-second limit on the number of WARN messages. If more than the given number of warns are emitted per second, then the remaining warns are suppressed. Zero values disable the rate limit
   -memory.allowedBytes size
-     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage
+     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage. The process may behave unexpectedly if this flag is set too small (e.g., 1 byte).
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 0)
   -memory.allowedPercent float
      Allowed percent of system memory VictoriaMetrics caches may occupy. See also -memory.allowedBytes. Too low a value may increase cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache which will result in higher disk IO usage (default 60)

docs/victoriametrics/vmstorage_common_flags.md

@@ -22,7 +22,7 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/cluster-victori
   -dedup.minScrapeInterval duration
      Leave only the last sample in every time series per each discrete interval equal to -dedup.minScrapeInterval > 0. See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#deduplication for details
   -denyQueriesOutsideRetention
-     Whether to deny queries outside of the configured -retentionPeriod. When set, then /api/v1/query_range would return '503 Service Unavailable' error for queries with 'from' value outside -retentionPeriod. This may be useful when multiple data sources with distinct retentions are hidden behind query-tee
+     Whether to deny queries outside the configured -retentionPeriod and -futureRetention. When set, then /api/v1/query_range will return an error for queries with 'from' value outside -retentionPeriod or 'to' value beyond -futureRetention. This may be useful when multiple data sources with distinct retentions are hidden behind query-tee
   -denyQueryTracing
      Whether to disable the ability to trace queries. See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#query-tracing
   -disablePerDayIndex
@@ -131,7 +131,7 @@ See the docs at https://docs.victoriametrics.com/victoriametrics/cluster-victori
   -maxConcurrentInserts int
      The maximum number of concurrent insert requests. Set higher value when clients send data over slow networks. Default value depends on the number of available CPU cores. It should work fine in most cases since it minimizes resource usage. See also -insert.maxQueueDuration (default 2*cgroup.AvailableCPUs())
   -memory.allowedBytes size
-     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage
+     Allowed size of system memory VictoriaMetrics caches may occupy. This option overrides -memory.allowedPercent if set to a non-zero value. Too low a value may increase the cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache resulting in higher disk IO usage. The process may behave unexpectedly if this flag is set too small (e.g., 1 byte).
      Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 0)
   -memory.allowedPercent float
      Allowed percent of system memory VictoriaMetrics caches may occupy. See also -memory.allowedBytes. Too low a value may increase cache miss rate usually resulting in higher CPU and disk IO usage. Too high a value may evict too much data from the OS page cache which will result in higher disk IO usage (default 60)

go.mod

@@ -33,9 +33,9 @@ require (
 	github.com/valyala/gozstd v1.24.0
 	github.com/valyala/histogram v1.2.0
 	github.com/valyala/quicktemplate v1.8.0
-	golang.org/x/net v0.53.0
+	golang.org/x/net v0.55.0
 	golang.org/x/oauth2 v0.36.0
-	golang.org/x/sys v0.43.0
+	golang.org/x/sys v0.45.0
 	google.golang.org/api v0.276.0
 	gopkg.in/yaml.v2 v2.4.0
 )
@@ -155,11 +155,11 @@ require (
 	go.uber.org/zap v1.27.1 // indirect
 	go.yaml.in/yaml/v2 v2.4.4 // indirect
 	go.yaml.in/yaml/v3 v3.0.4 // indirect
-	golang.org/x/crypto v0.50.0 // indirect
+	golang.org/x/crypto v0.51.0 // indirect
 	golang.org/x/exp v0.0.0-20260410095643-746e56fc9e2f // indirect
 	golang.org/x/sync v0.20.0 // indirect
-	golang.org/x/term v0.42.0 // indirect
-	golang.org/x/text v0.36.0 // indirect
+	golang.org/x/term v0.43.0 // indirect
+	golang.org/x/text v0.37.0 // indirect
 	golang.org/x/time v0.15.0 // indirect
 	google.golang.org/genproto v0.0.0-20260414002931-afd174a4e478 // indirect
 	google.golang.org/genproto/googleapis/api v0.0.0-20260414002931-afd174a4e478 // indirect

go.sum

@@ -376,8 +376,6 @@ github.com/prometheus/otlptranslator v1.0.0 h1:s0LJW/iN9dkIH+EnhiD3BlkkP5QVIUVEo
 github.com/prometheus/otlptranslator v1.0.0/go.mod h1:vRYWnXvI6aWGpsdY/mOT/cbeVRBlPWtBNDb7kGR3uKM=
 github.com/prometheus/procfs v0.20.1 h1:XwbrGOIplXW/AU3YhIhLODXMJYyC1isLFfYCsTEycfc=
 github.com/prometheus/procfs v0.20.1/go.mod h1:o9EMBZGRyvDrSPH1RqdxhojkuXstoe4UlK79eF5TGGo=
-github.com/prometheus/prometheus v0.311.2 h1:6fBxp93y08GAZGNT1o3bIhgV/AMYvBFfU+ltDNEsHg8=
-github.com/prometheus/prometheus v0.311.2/go.mod h1:gjsCxTKtHO1Q8T9333u1s+lUR1OjPyM7ruuGH8RvVyo=
 github.com/prometheus/prometheus v0.311.3 h1:3IrVxQv6v5i/ZCGi6OrYeBhtCwaPTn6Z3DYruXoYm3M=
 github.com/prometheus/prometheus v0.311.3/go.mod h1:gjsCxTKtHO1Q8T9333u1s+lUR1OjPyM7ruuGH8RvVyo=
 github.com/prometheus/sigv4 v0.4.1 h1:EIc3j+8NBea9u1iV6O5ZAN8uvPq2xOIUPcqCTivHuXs=
@@ -511,8 +509,8 @@ go.yaml.in/yaml/v3 v3.0.4/go.mod h1:DhzuOOF2ATzADvBadXxruRBLzYTpT36CKvDb3+aBEFg=
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
 golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
-golang.org/x/crypto v0.50.0 h1:zO47/JPrL6vsNkINmLoo/PH1gcxpls50DNogFvB5ZGI=
-golang.org/x/crypto v0.50.0/go.mod h1:3muZ7vA7PBCE6xgPX7nkzzjiUq87kRItoJQM1Yo8S+Q=
+golang.org/x/crypto v0.51.0 h1:IBPXwPfKxY7cWQZ38ZCIRPI50YLeevDLlLnyC5wRGTI=
+golang.org/x/crypto v0.51.0/go.mod h1:8AdwkbraGNABw2kOX6YFPs3WM22XqI4EXEd8g+x7Oc8=
 golang.org/x/exp v0.0.0-20260410095643-746e56fc9e2f h1:W3F4c+6OLc6H2lb//N1q4WpJkhzJCK5J6kUi1NTVXfM=
 golang.org/x/exp v0.0.0-20260410095643-746e56fc9e2f/go.mod h1:J1xhfL/vlindoeF/aINzNzt2Bket5bjo9sdOYzOsU80=
 golang.org/x/mod v0.2.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
@@ -523,8 +521,8 @@ golang.org/x/net v0.0.0-20190404232315-eb5bcb51f2a3/go.mod h1:t9HGtf8HONx5eT2rtn
 golang.org/x/net v0.0.0-20190620200207-3b0461eec859/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/net v0.0.0-20200226121028-0de0cce0169b/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/net v0.0.0-20201021035429-f5854403a974/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
-golang.org/x/net v0.53.0 h1:d+qAbo5L0orcWAr0a9JweQpjXF19LMXJE8Ey7hwOdUA=
-golang.org/x/net v0.53.0/go.mod h1:JvMuJH7rrdiCfbeHoo3fCQU24Lf5JJwT9W3sJFulfgs=
+golang.org/x/net v0.55.0 h1:bcvxaJn3e1U6InsFWt1JUq1aSjnRxLzT2rtD2KfkDF8=
+golang.org/x/net v0.55.0/go.mod h1:L5U2KuzuOe1lY7Z+aWVIKK6qEeJXnXV9yzGA+WCHJww=
 golang.org/x/oauth2 v0.36.0 h1:peZ/1z27fi9hUOFCAZaHyrpWG5lwe0RJEEEeH0ThlIs=
 golang.org/x/oauth2 v0.36.0/go.mod h1:YDBUJMTkDnJS+A4BP4eZBjCqtokkg1hODuPjwiGPO7Q=
 golang.org/x/sync v0.0.0-20190423024810-112230192c58/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
@@ -536,14 +534,14 @@ golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5h
 golang.org/x/sys v0.0.0-20190412213103-97732733099d/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.1.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.43.0 h1:Rlag2XtaFTxp19wS8MXlJwTvoh8ArU6ezoyFsMyCTNI=
-golang.org/x/sys v0.43.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
-golang.org/x/term v0.42.0 h1:UiKe+zDFmJobeJ5ggPwOshJIVt6/Ft0rcfrXZDLWAWY=
-golang.org/x/term v0.42.0/go.mod h1:Dq/D+snpsbazcBG5+F9Q1n2rXV8Ma+71xEjTRufARgY=
+golang.org/x/sys v0.45.0 h1:dO4czNzziLiiXplLQgBCEpCvXQ3dnkn0SdaZSYdQ+FY=
+golang.org/x/sys v0.45.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
+golang.org/x/term v0.43.0 h1:S4RLU2sB31O/NCl+zFN9Aru9A/Cq2aqKpTZJ6B+DwT4=
+golang.org/x/term v0.43.0/go.mod h1:lrhlHNdQJHO+1qVYiHfFKVuVioJIheAc3fBSMFYEIsk=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
-golang.org/x/text v0.36.0 h1:JfKh3XmcRPqZPKevfXVpI1wXPTqbkE5f7JA92a55Yxg=
-golang.org/x/text v0.36.0/go.mod h1:NIdBknypM8iqVmPiuco0Dh6P5Jcdk8lJL0CUebqK164=
+golang.org/x/text v0.37.0 h1:Cqjiwd9eSg8e0QAkyCaQTNHFIIzWtidPahFWR83rTrc=
+golang.org/x/text v0.37.0/go.mod h1:a5sjxXGs9hsn/AJVwuElvCAo9v8QYLzvavO5z2PiM38=
 golang.org/x/time v0.15.0 h1:bbrp8t3bGUeFOx08pvsMYRTCVSMk89u4tKbNOZbp88U=
 golang.org/x/time v0.15.0/go.mod h1:Y4YMaQmXwGQZoFaVFk4YpCt4FLQMYKZe9oeV/f4MSno=
 golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=

lib/protoparser/opentelemetry/pb/pb.go

@@ -14,6 +14,14 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 )
 
+// DecodeMetricsOptions defines options for DecodeMetricsData
+type DecodeMetricsOptions struct {
+	DisableScopeMetadata      bool
+	DisableResourceAttributes bool
+	// ResourceAttributesList stored a list of resource attributes to ignore or promote based on the value of DisableResourceAttributes.
+	ResourceAttributesList map[string]struct{}
+}
+
 // MetricPusher must push the parsed samples and metric metadata to the underlying storage.
 type MetricPusher interface {
 	// PushSample must store a sample with the given args.
@@ -73,8 +81,8 @@ func (r *MetricsData) marshalProtobuf(mm *easyproto.MessageMarshaler) {
 	}
 }
 
-// DecodeMetricsData decodes metricsData from src and sends the decoded data to mp.
-func DecodeMetricsData(src []byte, mp MetricPusher) (err error) {
+// DecodeMetricsData decodes metricsData with given options from src and sends the decoded data to mp.
+func DecodeMetricsData(src []byte, mp MetricPusher, options DecodeMetricsOptions) (err error) {
 	// See https://github.com/open-telemetry/opentelemetry-proto/blob/049d4332834935792fd4dbd392ecd31904f99ba2/opentelemetry/proto/metrics/v1/metrics.proto#L56
 	//
 	// message MetricsData {
@@ -96,7 +104,7 @@ func DecodeMetricsData(src []byte, mp MetricPusher) (err error) {
 			if !ok {
 				return fmt.Errorf("cannot read ResourceMetrics data")
 			}
-			if err := dctx.decodeResourceMetrics(data); err != nil {
+			if err := dctx.decodeResourceMetrics(data, options); err != nil {
 				return fmt.Errorf("cannot unmarshal ResourceMetrics: %w", err)
 			}
 		}
@@ -121,7 +129,7 @@ func (rm *ResourceMetrics) marshalProtobuf(mm *easyproto.MessageMarshaler) {
 	}
 }
 
-func (dctx *decoderContext) decodeResourceMetrics(src []byte) error {
+func (dctx *decoderContext) decodeResourceMetrics(src []byte, options DecodeMetricsOptions) error {
 	// See https://github.com/open-telemetry/opentelemetry-proto/blob/049d4332834935792fd4dbd392ecd31904f99ba2/opentelemetry/proto/metrics/v1/metrics.proto#L66
 	//
 	// message ResourceMetrics {
@@ -137,7 +145,7 @@ func (dctx *decoderContext) decodeResourceMetrics(src []byte) error {
 		return fmt.Errorf("cannot read Resource data: %w", err)
 	}
 	if ok {
-		if err := dctx.decodeResource(resourceData); err != nil {
+		if err := dctx.decodeResource(resourceData, options.DisableResourceAttributes, options.ResourceAttributesList); err != nil {
 			return fmt.Errorf("cannot decode Resource: %w", err)
 		}
 	}
@@ -157,7 +165,7 @@ func (dctx *decoderContext) decodeResourceMetrics(src []byte) error {
 				return fmt.Errorf("cannot read ScopeMetrics data")
 			}
 
-			if err := dctx.decodeScopeMetrics(data); err != nil {
+			if err := dctx.decodeScopeMetrics(data, options.DisableScopeMetadata); err != nil {
 				return fmt.Errorf("cannot unmarshal ScopeMetrics: %w", err)
 			}
 
@@ -180,7 +188,7 @@ func (r *Resource) marshalProtobuf(mm *easyproto.MessageMarshaler) {
 	}
 }
 
-func (dctx *decoderContext) decodeResource(src []byte) (err error) {
+func (dctx *decoderContext) decodeResource(src []byte, disableResourceAttributes bool, attributeKeys map[string]struct{}) (err error) {
 	// See https://github.com/open-telemetry/opentelemetry-proto/blob/049d4332834935792fd4dbd392ecd31904f99ba2/opentelemetry/proto/resource/v1/resource.proto#L28
 	//
 	// message Resource {
@@ -199,7 +207,34 @@ func (dctx *decoderContext) decodeResource(src []byte) (err error) {
 			if !ok {
 				return fmt.Errorf("cannot read Attributes")
 			}
-			if err := decodeKeyValue(data, &dctx.ls, &dctx.fb, ""); err != nil {
+			keySuffix, ok, err := easyproto.GetString(data, 1)
+			if err != nil {
+				return fmt.Errorf("cannot find Key in KeyValue: %w", err)
+			}
+			if !ok {
+				// Key is missing, skip it.
+				// See https://github.com/VictoriaMetrics/VictoriaLogs/issues/869#issuecomment-3631307996
+				continue
+			}
+			if _, ok := attributeKeys[keySuffix]; ok != disableResourceAttributes {
+				// Skip the attribute if:
+				// 1. it is in the list of ignore attributes when disableResourceAttributes is false,
+				// 2. it isn't in the list of promote attributes when disableResourceAttributes is true.
+				continue
+			}
+			key := dctx.fb.formatSubFieldName("", keySuffix)
+
+			// Decode value
+			value, ok, err := easyproto.GetMessageData(data, 2)
+			if err != nil {
+				return fmt.Errorf("cannot find Value in KeyValue: %w", err)
+			}
+			if !ok {
+				// Value is null, skip it.
+				continue
+			}
+
+			if err := decodeAnyValue(value, &dctx.ls, &dctx.fb, key); err != nil {
 				return fmt.Errorf("cannot unmarshal Attributes: %w", err)
 			}
 		}
@@ -257,7 +292,6 @@ func decodeKeyValue(src []byte, ls *promutil.Labels, fb *fmtBuffer, keyPrefix st
 	if err := decodeAnyValue(valueData, ls, fb, key); err != nil {
 		return fmt.Errorf("cannot decode AnyValue: %w", err)
 	}
-
 	return nil
 }
 
@@ -452,7 +486,7 @@ func (sm *ScopeMetrics) marshalProtobuf(mm *easyproto.MessageMarshaler) {
 	}
 }
 
-func (dctx *decoderContext) decodeScopeMetrics(src []byte) error {
+func (dctx *decoderContext) decodeScopeMetrics(src []byte, disableScopeMetadata bool) error {
 	// See https://github.com/open-telemetry/opentelemetry-proto/blob/049d4332834935792fd4dbd392ecd31904f99ba2/opentelemetry/proto/metrics/v1/metrics.proto#L86
 	//
 	// message ScopeMetrics {
@@ -460,19 +494,22 @@ func (dctx *decoderContext) decodeScopeMetrics(src []byte) error {
 	//   repeated Metric metrics = 2;
 	// }
 
-	scopeData, ok, err := easyproto.GetMessageData(src, 1)
-	if err != nil {
-		return fmt.Errorf("cannot read InstrumentationScope: %w", err)
-	}
-	if ok {
-		if err := dctx.decodeInstrumentationScope(scopeData); err != nil {
-			return fmt.Errorf("cannot decode InstrumentationScope: %w", err)
+	if !disableScopeMetadata {
+		scopeData, ok, err := easyproto.GetMessageData(src, 1)
+		if err != nil {
+			return fmt.Errorf("cannot read InstrumentationScope: %w", err)
+		}
+		if ok {
+			if err := dctx.decodeInstrumentationScope(scopeData); err != nil {
+				return fmt.Errorf("cannot decode InstrumentationScope: %w", err)
+			}
 		}
 	}
 
 	dctxSnapshot := dctx.getSnapshot()
 
 	var fc easyproto.FieldContext
+	var err error
 	for len(src) > 0 {
 		src, err = fc.NextField(src)
 		if err != nil {

lib/protoparser/opentelemetry/pb/pb_test.go

@@ -0,0 +1,166 @@
+package pb
+
+import (
+	"reflect"
+	"testing"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
+)
+
+type testMetricPusher struct {
+	samples  []testSample
+	metadata []MetricMetadata
+}
+
+type testSample struct {
+	mm     MetricMetadata
+	suffix string
+	labels []prompb.Label
+	ts     uint64
+	value  float64
+}
+
+func (p *testMetricPusher) PushSample(mm *MetricMetadata, suffix string, ls *promutil.Labels, timestampNsecs uint64, value float64, _ uint32) {
+	labels := make([]prompb.Label, len(ls.Labels))
+	copy(labels, ls.Labels)
+	p.samples = append(p.samples, testSample{
+		mm:     *mm,
+		suffix: suffix,
+		labels: labels,
+		ts:     timestampNsecs,
+		value:  value,
+	})
+}
+
+func (p *testMetricPusher) PushMetricMetadata(mm *MetricMetadata) {
+	p.metadata = append(p.metadata, *mm)
+}
+
+func TestDecodeScopeMetrics(t *testing.T) {
+	scopeName := "my-scope"
+	scopeVersion := "v1.0"
+	envVal := "prod"
+	intVal := int64(1)
+	md := &MetricsData{
+		ResourceMetrics: []*ResourceMetrics{
+			{
+				Resource: &Resource{
+					Attributes: []*KeyValue{
+						{Key: "job", Value: &AnyValue{StringValue: new("vm")}},
+						{Key: "region", Value: &AnyValue{StringValue: new("us-east-1")}},
+					},
+				},
+				ScopeMetrics: []*ScopeMetrics{
+					{
+						Scope: &InstrumentationScope{
+							Name:    &scopeName,
+							Version: &scopeVersion,
+							Attributes: []*KeyValue{
+								{Key: "env", Value: &AnyValue{StringValue: &envVal}},
+							},
+						},
+						Metrics: []*Metric{
+							{
+								Name:        "my-gauge",
+								Description: "a test gauge",
+								Gauge: &Gauge{
+									DataPoints: []*NumberDataPoint{
+										{
+											Attributes:   []*KeyValue{{Key: "label1", Value: &AnyValue{StringValue: new("value1")}}},
+											IntValue:     &intVal,
+											TimeUnixNano: 1000,
+										},
+									},
+								},
+							},
+						},
+					},
+				},
+			},
+		},
+	}
+	data := md.MarshalProtobuf(nil)
+
+	f := func(options DecodeMetricsOptions, wantLabels map[string]string) {
+		t.Helper()
+		mp := &testMetricPusher{}
+		if err := DecodeMetricsData(data, mp, options); err != nil {
+			t.Fatalf("DecodeMetricsData error: %v", err)
+		}
+		if len(mp.samples) != 1 {
+			t.Fatalf("expected 1 sample, got %d", len(mp.samples))
+		}
+		gotMap := make(map[string]string, len(mp.samples[0].labels))
+		for _, l := range mp.samples[0].labels {
+			gotMap[l.Name] = l.Value
+		}
+		if !reflect.DeepEqual(gotMap, wantLabels) {
+			t.Errorf("unexpected labels:\n got:  %v\n want: %v", gotMap, wantLabels)
+		}
+	}
+
+	// PromoteScopeMetadata=true + PromoteAllResourceAttributes=true:
+	// got all scope labels and resource attrs
+	f(DecodeMetricsOptions{
+		DisableScopeMetadata:      false,
+		DisableResourceAttributes: false,
+	}, map[string]string{
+		"job":                  "vm",
+		"region":               "us-east-1",
+		"scope.name":           "my-scope",
+		"scope.version":        "v1.0",
+		"scope.attributes.env": "prod",
+		"label1":               "value1",
+	})
+
+	// PromoteScopeMetadata=false + PromoteAllResourceAttributes=true:
+	// got all resource attrs, no scope labels
+	f(DecodeMetricsOptions{
+		DisableScopeMetadata:      true,
+		DisableResourceAttributes: false,
+	}, map[string]string{
+		"job":    "vm",
+		"region": "us-east-1",
+		"label1": "value1",
+	})
+
+	// PromoteScopeMetadata=true + PromoteAllResourceAttributes=false + ResourceAttributesList=[region]:
+	// got only the `region` attr from resource
+	f(DecodeMetricsOptions{
+		DisableScopeMetadata:      false,
+		DisableResourceAttributes: true,
+		ResourceAttributesList:    map[string]struct{}{"region": {}},
+	}, map[string]string{
+		"region":               "us-east-1",
+		"scope.name":           "my-scope",
+		"scope.version":        "v1.0",
+		"scope.attributes.env": "prod",
+		"label1":               "value1",
+	})
+
+	// PromoteScopeMetadata=true + PromoteAllResourceAttributes=true + ResourceAttributesList=[region]:
+	// got all resource attrs except `region` (ignore list)
+	f(DecodeMetricsOptions{
+		DisableScopeMetadata:      false,
+		DisableResourceAttributes: false,
+		ResourceAttributesList:    map[string]struct{}{"region": {}},
+	}, map[string]string{
+		"job":                  "vm",
+		"scope.name":           "my-scope",
+		"scope.version":        "v1.0",
+		"scope.attributes.env": "prod",
+		"label1":               "value1",
+	})
+
+	// PromoteScopeMetadata=false + PromoteAllResourceAttributes=false + ResourceAttributesList=[job]:
+	// got only `job` attr
+	f(DecodeMetricsOptions{
+		DisableScopeMetadata:      true,
+		DisableResourceAttributes: true,
+		ResourceAttributesList:    map[string]struct{}{"job": {}},
+	}, map[string]string{
+		"job":    "vm",
+		"label1": "value1",
+	})
+}

lib/protoparser/opentelemetry/stream/streamparser.go

@@ -1,6 +1,7 @@
 package stream
 
 import (
+	"flag"
 	"fmt"
 	"io"
 	"sync"
@@ -10,13 +11,43 @@ import (
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/decimal"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/opentelemetry/pb"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/protoparser/protoparserutil"
 )
 
-var maxRequestSize = flagutil.NewBytes("opentelemetry.maxRequestSize", 64*1024*1024, "The maximum size in bytes of a single OpenTelemetry request")
+var (
+	maxRequestSize = flagutil.NewBytes("opentelemetry.maxRequestSize", 64*1024*1024, "The maximum size in bytes of a single OpenTelemetry request")
+
+	promoteScopeMetadata         = flag.Bool("opentelemetry.promoteScopeMetadata", true, "Whether to promote OTel scope metadata (i.e. name, version, schema URL, and attributes) to metric labels.")
+	promoteAllResourceAttributes = flag.Bool("opentelemetry.promoteAllResourceAttributes", true, "Whether to promote all resource attributes to labels, except for the ones configured with 'opentelemetry.ignoreResourceAttributes'.")
+	promoteResourceAttributes    = flagutil.NewArrayString("opentelemetry.promoteResourceAttributes", "Promote specific list of resource attributes to labels.")
+	ignoreResourceAttributes     = flagutil.NewArrayString("opentelemetry.ignoreResourceAttributes", "Control which resource attributes to ignore, can only be set when 'opentelemetry.promoteAllResourceAttributes' is true.")
+)
+
+// InitDecodeOptions configures decoding settings for the parser
+func InitDecodeOptions() {
+	if *promoteAllResourceAttributes && len(*promoteResourceAttributes) > 0 {
+		logger.Fatalf("cannot set both '-opentelemetry.promoteAllResourceAttributes' and '-opentelemetry.promoteResourceAttributes'")
+	}
+	if !*promoteAllResourceAttributes && len(*ignoreResourceAttributes) > 0 {
+		logger.Fatalf("'-opentelemetry.ignoreResourceAttributes' can only be set when '-opentelemetry.promoteAllResourceAttributes' is true.")
+	}
+	defaultDecodeMetricsOptions.DisableScopeMetadata = !*promoteScopeMetadata
+	defaultDecodeMetricsOptions.DisableResourceAttributes = !*promoteAllResourceAttributes
+	attributes := *ignoreResourceAttributes
+	if !*promoteAllResourceAttributes {
+		attributes = *promoteResourceAttributes
+	}
+	defaultDecodeMetricsOptions.ResourceAttributesList = make(map[string]struct{}, len(attributes))
+	for _, a := range attributes {
+		defaultDecodeMetricsOptions.ResourceAttributesList[a] = struct{}{}
+	}
+}
+
+var defaultDecodeMetricsOptions = pb.DecodeMetricsOptions{}
 
 // ParseStream parses OpenTelemetry protobuf or json data from r and calls callback for the parsed rows.
 //
@@ -47,7 +78,7 @@ func parseData(data []byte, callback func(tss []prompb.TimeSeries, mms []prompb.
 	// the flushFunc will be called multiple time if the request is big, to avoid over allocating memory for such request.
 	wctx.flushFunc = callback
 
-	if err := pb.DecodeMetricsData(data, wctx); err != nil {
+	if err := pb.DecodeMetricsData(data, wctx, defaultDecodeMetricsOptions); err != nil {
 		return fmt.Errorf("cannot unmarshal request from %d bytes: %w", len(data), err)
 	}
 

lib/storage/search.go

@@ -90,6 +90,77 @@ type MetricBlockRef struct {
 	BlockRef *BlockRef
 }
 
+// MetricBlock is a time series block for a single metric.
+type MetricBlock struct {
+	// MetricName is metric name for the given Block.
+	MetricName []byte
+
+	// Block is a block for the given MetricName
+	Block Block
+}
+
+// Marshal marshals MetricBlock to dst
+func (mb *MetricBlock) Marshal(dst []byte) []byte {
+	dst = encoding.MarshalBytes(dst, mb.MetricName)
+	return MarshalBlock(dst, &mb.Block)
+}
+
+// CopyFrom copies src to mb.
+func (mb *MetricBlock) CopyFrom(src *MetricBlock) {
+	mb.MetricName = append(mb.MetricName[:0], src.MetricName...)
+	mb.Block.CopyFrom(&src.Block)
+}
+
+// MarshalBlock marshals b to dst.
+//
+// b.MarshalData must be called on b before calling MarshalBlock.
+func MarshalBlock(dst []byte, b *Block) []byte {
+	dst = b.bh.Marshal(dst)
+	dst = encoding.MarshalBytes(dst, b.timestampsData)
+	dst = encoding.MarshalBytes(dst, b.valuesData)
+	return dst
+}
+
+// Unmarshal unmarshals MetricBlock from src
+func (mb *MetricBlock) Unmarshal(src []byte) ([]byte, error) {
+	mb.Block.Reset()
+	mn, nSize := encoding.UnmarshalBytes(src)
+	if nSize <= 0 {
+		return src, fmt.Errorf("cannot unmarshal MetricName")
+	}
+	src = src[nSize:]
+	mb.MetricName = append(mb.MetricName[:0], mn...)
+
+	return UnmarshalBlock(&mb.Block, src)
+}
+
+// UnmarshalBlock unmarshal Block from src to dst.
+//
+// dst.UnmarshalData isn't called on the block.
+func UnmarshalBlock(dst *Block, src []byte) ([]byte, error) {
+	tail, err := dst.bh.Unmarshal(src)
+	if err != nil {
+		return tail, fmt.Errorf("cannot unmarshal blockHeader: %w", err)
+	}
+	src = tail
+
+	tds, nSize := encoding.UnmarshalBytes(src)
+	if nSize <= 0 {
+		return tail, fmt.Errorf("cannot unmarshal timestampsData")
+	}
+	src = src[nSize:]
+	dst.timestampsData = append(dst.timestampsData[:0], tds...)
+
+	vd, nSize := encoding.UnmarshalBytes(src)
+	if nSize <= 0 {
+		return tail, fmt.Errorf("cannot unmarshal valuesData")
+	}
+	src = src[nSize:]
+	dst.valuesData = append(dst.valuesData[:0], vd...)
+
+	return src, nil
+}
+
 // Search is a search for time series.
 type Search struct {
 	// MetricBlockRef is updated with each Search.NextMetricBlock call.
@@ -290,6 +361,24 @@ func NewSearchQuery(start, end int64, tagFilterss [][]TagFilter, maxMetrics int)
 	}
 }
 
+// TenantToken represents a tenant (accountID, projectID) pair.
+type TenantToken struct {
+	AccountID uint32
+	ProjectID uint32
+}
+
+// String returns string representation of t.
+func (t *TenantToken) String() string {
+	return fmt.Sprintf("{accountID=%d, projectID=%d}", t.AccountID, t.ProjectID)
+}
+
+// Marshal appends marshaled t to dst and returns the result.
+func (t *TenantToken) Marshal(dst []byte) []byte {
+	dst = encoding.MarshalUint32(dst, t.AccountID)
+	dst = encoding.MarshalUint32(dst, t.ProjectID)
+	return dst
+}
+
 // TagFilter represents a single tag filter from SearchQuery.
 type TagFilter struct {
 	Key        []byte

lib/vminsertapi/api.go

@@ -0,0 +1,30 @@
+package vminsertapi
+
+import (
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricsmetadata"
+)
+
+// RPCCall defines rpc call from vminsert to vmstorage
+type RPCCall struct {
+	Name          string
+	VersionedName string
+}
+
+var (
+	MetricRowsRpcCall = RPCCall{
+		Name:          "metric_rows",
+		VersionedName: "writeRows_v1",
+	}
+	MetricMetadataRpcCall = RPCCall{
+		Name:          "metricmetadata_rows",
+		VersionedName: "writeMetadata_v1",
+	}
+)
+
+// API must implement vminsert API.
+type API interface {
+	WriteRows(rows []storage.MetricRow) error
+	WriteMetadata(mrs []metricsmetadata.Row) error
+	IsReadOnly() bool
+}

lib/vmselectapi/api.go

@@ -0,0 +1,68 @@
+package vmselectapi
+
+import (
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/querytracer"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricnamestats"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage/metricsmetadata"
+)
+
+// API must implement vmselect API.
+type API interface {
+	// InitSearch initialize series search for the given sq.
+	//
+	// The returned BlockIterator must be closed with MustClose to free up resources when it is no longer needed.
+	InitSearch(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (BlockIterator, error)
+
+	// SearchMetricNames returns metric names matching the given sq.
+	SearchMetricNames(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) ([]string, error)
+
+	// LabelValues returns values for labelName label acorss series matching the given sq.
+	LabelValues(qt *querytracer.Tracer, sq *storage.SearchQuery, labelName string, maxLabelValues int, deadline uint64) ([]string, error)
+
+	// TagValueSuffixes returns tag value suffixes for the given args.
+	TagValueSuffixes(qt *querytracer.Tracer, accountID, projectID uint32, tr storage.TimeRange, tagKey, tagValuePrefix string, delimiter byte, maxSuffixes int, deadline uint64) ([]string, error)
+
+	// LabelNames returns lable names for series matching the given sq.
+	LabelNames(qt *querytracer.Tracer, sq *storage.SearchQuery, maxLableNames int, deadline uint64) ([]string, error)
+
+	// SeriesCount returns the number of series for the given (accountID, projectID).
+	SeriesCount(qt *querytracer.Tracer, accountID, projectID uint32, deadline uint64) (uint64, error)
+
+	// TSDBStatus returns tsdb status for the given sq.
+	TSDBStatus(qt *querytracer.Tracer, sq *storage.SearchQuery, focusLabel string, topN int, deadline uint64) (*storage.TSDBStatus, error)
+
+	// DeleteSeries deletes series matching the given sq.
+	DeleteSeries(qt *querytracer.Tracer, sq *storage.SearchQuery, deadline uint64) (int, error)
+
+	// RegisterMetricNames registers the given mrs in the storage.
+	RegisterMetricNames(qt *querytracer.Tracer, mrs []storage.MetricRow, deadline uint64) error
+
+	// Tenants returns list of tenants in the storage on the given tr.
+	Tenants(qt *querytracer.Tracer, tr storage.TimeRange, deadline uint64) ([]string, error)
+
+	// GetMetricNamesUsageStats returns statistics for metric names
+	GetMetricNamesUsageStats(qt *querytracer.Tracer, tt *storage.TenantToken, limit, le int, matchPattern string, deadline uint64) (metricnamestats.StatsResult, error)
+
+	// ResetMetricNamesUsageStats resets internal state of metric names tracker
+	ResetMetricNamesUsageStats(qt *querytracer.Tracer, deadline uint64) error
+
+	// GetMetadataRecords returns metrics metadata.
+	GetMetadataRecords(qt *querytracer.Tracer, tt *storage.TenantToken, limit int, metricName string, deadline uint64) ([]*metricsmetadata.Row, error)
+}
+
+// BlockIterator must iterate through series blocks found by VMSelect.InitSearch.
+//
+// MustClose must be called in order to free up allocated resources when BlockIterator is no longer needed.
+type BlockIterator interface {
+	// NextBlock marshals next storage.MetricBlock into dst.
+	//
+	// It returns true on success, false on error or if no blocks to read.
+	NextBlock(dst []byte) ([]byte, bool)
+
+	// MustClose frees up resources allocated by BlockIterator.
+	MustClose()
+
+	// Error returns the last error occurred in NextBlock(), which returns false.
+	Error() error
+}

vendor/golang.org/x/crypto/hkdf/hkdf.go

@@ -11,6 +11,7 @@
 package hkdf
 
 import (
+	"crypto/hkdf"
 	"crypto/hmac"
 	"errors"
 	"hash"
@@ -24,15 +25,19 @@ import (
 // Expand invocations and different context values. Most common scenarios,
 // including the generation of multiple keys, should use New instead.
 func Extract(hash func() hash.Hash, secret, salt []byte) []byte {
-	if salt == nil {
-		salt = make([]byte, hash().Size())
+	// Use the stdlib Extract, which disables FIPS 140 enforcement of the HMAC
+	// key (which in HKDF is the salt). The only possible error is FIPS 140
+	// enforcement of the hash, which had to panic under this API anyway. We
+	// don't use the stdlib Expand, because it switched to returning a []byte
+	// instead of an io.Reader, and Expand uses the HMAC key as a key.
+	out, err := hkdf.Extract(hash, secret, salt)
+	if err != nil {
+		panic(err)
 	}
-	extractor := hmac.New(hash, salt)
-	extractor.Write(secret)
-	return extractor.Sum(nil)
+	return out
 }
 
-type hkdf struct {
+type hkdfReader struct {
 	expander hash.Hash
 	size     int
 
@@ -43,7 +48,7 @@ type hkdf struct {
 	buf  []byte
 }
 
-func (f *hkdf) Read(p []byte) (int, error) {
+func (f *hkdfReader) Read(p []byte) (int, error) {
 	// Check whether enough data can be generated
 	need := len(p)
 	remains := len(f.buf) + int(255-f.counter+1)*f.size
@@ -84,7 +89,7 @@ func (f *hkdf) Read(p []byte) (int, error) {
 // 3.3. Most common scenarios will want to use New instead.
 func Expand(hash func() hash.Hash, pseudorandomKey, info []byte) io.Reader {
 	expander := hmac.New(hash, pseudorandomKey)
-	return &hkdf{expander, expander.Size(), info, 1, nil, nil}
+	return &hkdfReader{expander, expander.Size(), info, 1, nil, nil}
 }
 
 // New returns a Reader, from which keys can be read, using the given hash,

vendor/golang.org/x/net/http2/README.md

@@ -0,0 +1,19 @@
+This package (golang.org/x/net/http2) is the original source of truth
+of the Go HTTP/2 implementation.
+
+As of Go 1.27, the source of truth has moved to the standard library
+package net/http/internal/http2.
+All new feature development should happen in that package.
+Only critical bug fixes and security fixes will be backported to x/net.
+
+The x/net package contains two implementations of the HTTP/2 transport and server:
+
+The original implementation (no longer the source of truth).
+
+A reimplementation of the x/net/http2 APIs in terms of net/http.
+This is called "the wrapping implementation", since it wraps net/http.
+
+The original implementation is used when the Go version is less than 1.27.
+
+The wrapping implementation is used when the Go version is at least 1.27.
+The build tag "http2legacy" may be set to use the original implementation.

vendor/golang.org/x/net/http2/client_conn_pool.go

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build !(go1.27 && !http2legacy)
+
 // Transport code's client connection pooling.
 
 package http2
@@ -14,18 +16,6 @@ import (
 	"sync"
 )
 
-// ClientConnPool manages a pool of HTTP/2 client connections.
-type ClientConnPool interface {
-	// GetClientConn returns a specific HTTP/2 connection (usually
-	// a TLS-TCP connection) to an HTTP/2 server. On success, the
-	// returned ClientConn accounts for the upcoming RoundTrip
-	// call, so the caller should not omit it. If the caller needs
-	// to, ClientConn.RoundTrip can be called with a bogus
-	// new(http.Request) to release the stream reservation.
-	GetClientConn(req *http.Request, addr string) (*ClientConn, error)
-	MarkDead(*ClientConn)
-}
-
 // clientConnPoolIdleCloser is the interface implemented by ClientConnPool
 // implementations which can close their idle connections.
 type clientConnPoolIdleCloser interface {

vendor/golang.org/x/net/http2/clientconn.go

@@ -0,0 +1,57 @@
+// Copyright 2026 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package http2
+
+import (
+	"context"
+	"net/http"
+)
+
+func (cc *ClientConn) RoundTrip(req *http.Request) (*http.Response, error) {
+	return cc.roundTrip(req)
+}
+
+// SetDoNotReuse marks cc as not reusable for future HTTP requests.
+func (cc *ClientConn) SetDoNotReuse() {
+	cc.setDoNotReuse()
+}
+
+// CanTakeNewRequest reports whether the connection can take a new request,
+// meaning it has not been closed or received or sent a GOAWAY.
+//
+// If the caller is going to immediately make a new request on this
+// connection, use ReserveNewRequest instead.
+func (cc *ClientConn) CanTakeNewRequest() bool {
+	return cc.canTakeNewRequest()
+}
+
+// ReserveNewRequest is like CanTakeNewRequest but also reserves a
+// concurrent stream in cc. The reservation is decremented on the
+// next call to RoundTrip.
+func (cc *ClientConn) ReserveNewRequest() bool {
+	return cc.reserveNewRequest()
+}
+
+// State returns a snapshot of cc's state.
+func (cc *ClientConn) State() ClientConnState {
+	return cc.state()
+}
+
+// Shutdown gracefully closes the client connection, waiting for running streams to complete.
+func (cc *ClientConn) Shutdown(ctx context.Context) error {
+	return cc.shutdown(ctx)
+}
+
+// Close closes the client connection immediately.
+//
+// In-flight requests are interrupted. For a graceful shutdown, use Shutdown instead.
+func (cc *ClientConn) Close() error {
+	return cc.close()
+}
+
+// Ping sends a PING frame to the server and waits for the ack.
+func (cc *ClientConn) Ping(ctx context.Context) error {
+	return cc.ping(ctx)
+}

vendor/golang.org/x/net/http2/config.go

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build !(go1.27 && !http2legacy)
+
 package http2
 
 import (

vendor/golang.org/x/net/http2/hpack/tables.go

@@ -6,7 +6,6 @@ package hpack
 
 import (
 	"fmt"
-	"strings"
 )
 
 // headerFieldTable implements a list of HeaderFields.
@@ -55,16 +54,10 @@ func (t *headerFieldTable) len() int {
 
 // addEntry adds a new entry.
 func (t *headerFieldTable) addEntry(f HeaderField) {
-	// Prevent f from escaping to the heap.
-	f2 := HeaderField{
-		Name:      strings.Clone(f.Name),
-		Value:     strings.Clone(f.Value),
-		Sensitive: f.Sensitive,
-	}
 	id := uint64(t.len()) + t.evictCount + 1
-	t.byName[f2.Name] = id
-	t.byNameValue[pairNameValue{f2.Name, f2.Value}] = id
-	t.ents = append(t.ents, f2)
+	t.byName[f.Name] = id
+	t.byNameValue[pairNameValue{f.Name, f.Value}] = id
+	t.ents = append(t.ents, f)
 }
 
 // evictOldest evicts the n oldest entries in the table.

vendor/golang.org/x/net/http2/http2.go

@@ -195,7 +195,7 @@ func (s SettingID) String() string {
 }
 
 // validWireHeaderFieldName reports whether v is a valid header field
-// name (key). See httpguts.ValidHeaderName for the base rules.
+// name (key). See httpguts.ValidHeaderFieldName for the base rules.
 //
 // Further, http2 says:
 //

vendor/golang.org/x/net/http2/server.go

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build !(go1.27 && !http2legacy)
+
 // TODO: turn off the serve goroutine when idle, so
 // an idle conn only has the readFrames goroutine active. (which could
 // also be optimized probably to pin less memory in crypto/tls). This
@@ -88,98 +90,6 @@ var (
 	testHookOnPanic       func(sc *serverConn, panicVal interface{}) (rePanic bool)
 )
 
-// Server is an HTTP/2 server.
-type Server struct {
-	// MaxHandlers limits the number of http.Handler ServeHTTP goroutines
-	// which may run at a time over all connections.
-	// Negative or zero no limit.
-	// TODO: implement
-	MaxHandlers int
-
-	// MaxConcurrentStreams optionally specifies the number of
-	// concurrent streams that each client may have open at a
-	// time. This is unrelated to the number of http.Handler goroutines
-	// which may be active globally, which is MaxHandlers.
-	// If zero, MaxConcurrentStreams defaults to at least 100, per
-	// the HTTP/2 spec's recommendations.
-	MaxConcurrentStreams uint32
-
-	// MaxDecoderHeaderTableSize optionally specifies the http2
-	// SETTINGS_HEADER_TABLE_SIZE to send in the initial settings frame. It
-	// informs the remote endpoint of the maximum size of the header compression
-	// table used to decode header blocks, in octets. If zero, the default value
-	// of 4096 is used.
-	MaxDecoderHeaderTableSize uint32
-
-	// MaxEncoderHeaderTableSize optionally specifies an upper limit for the
-	// header compression table used for encoding request headers. Received
-	// SETTINGS_HEADER_TABLE_SIZE settings are capped at this limit. If zero,
-	// the default value of 4096 is used.
-	MaxEncoderHeaderTableSize uint32
-
-	// MaxReadFrameSize optionally specifies the largest frame
-	// this server is willing to read. A valid value is between
-	// 16k and 16M, inclusive. If zero or otherwise invalid, a
-	// default value is used.
-	MaxReadFrameSize uint32
-
-	// PermitProhibitedCipherSuites, if true, permits the use of
-	// cipher suites prohibited by the HTTP/2 spec.
-	PermitProhibitedCipherSuites bool
-
-	// IdleTimeout specifies how long until idle clients should be
-	// closed with a GOAWAY frame. PING frames are not considered
-	// activity for the purposes of IdleTimeout.
-	// If zero or negative, there is no timeout.
-	IdleTimeout time.Duration
-
-	// ReadIdleTimeout is the timeout after which a health check using a ping
-	// frame will be carried out if no frame is received on the connection.
-	// If zero, no health check is performed.
-	ReadIdleTimeout time.Duration
-
-	// PingTimeout is the timeout after which the connection will be closed
-	// if a response to a ping is not received.
-	// If zero, a default of 15 seconds is used.
-	PingTimeout time.Duration
-
-	// WriteByteTimeout is the timeout after which a connection will be
-	// closed if no data can be written to it. The timeout begins when data is
-	// available to write, and is extended whenever any bytes are written.
-	// If zero or negative, there is no timeout.
-	WriteByteTimeout time.Duration
-
-	// MaxUploadBufferPerConnection is the size of the initial flow
-	// control window for each connections. The HTTP/2 spec does not
-	// allow this to be smaller than 65535 or larger than 2^32-1.
-	// If the value is outside this range, a default value will be
-	// used instead.
-	MaxUploadBufferPerConnection int32
-
-	// MaxUploadBufferPerStream is the size of the initial flow control
-	// window for each stream. The HTTP/2 spec does not allow this to
-	// be larger than 2^32-1. If the value is zero or larger than the
-	// maximum, a default value will be used instead.
-	MaxUploadBufferPerStream int32
-
-	// NewWriteScheduler constructs a write scheduler for a connection.
-	// If nil, a default scheduler is chosen.
-	//
-	// Deprecated: User-provided write schedulers are deprecated.
-	NewWriteScheduler func() WriteScheduler
-
-	// CountError, if non-nil, is called on HTTP/2 server errors.
-	// It's intended to increment a metric for monitoring, such
-	// as an expvar or Prometheus metric.
-	// The errType consists of only ASCII word characters.
-	CountError func(errType string)
-
-	// Internal state. This is a pointer (rather than embedded directly)
-	// so that we don't embed a Mutex in this struct, which will make the
-	// struct non-copyable, which might break some callers.
-	state *serverInternalState
-}
-
 type serverInternalState struct {
 	mu          sync.Mutex
 	activeConns map[*serverConn]struct{}
@@ -187,6 +97,9 @@ type serverInternalState struct {
 	// Pool of error channels. This is per-Server rather than global
 	// because channels can't be reused across synctest bubbles.
 	errChanPool sync.Pool
+
+	// Used in tests.
+	testNewConn func(*serverConn)
 }
 
 func (s *serverInternalState) registerConn(sc *serverConn) {
@@ -239,12 +152,7 @@ func (s *serverInternalState) putErrChan(ch chan error) {
 	s.errChanPool.Put(ch)
 }
 
-// ConfigureServer adds HTTP/2 support to a net/http Server.
-//
-// The configuration conf may be nil.
-//
-// ConfigureServer must be called before s begins serving.
-func ConfigureServer(s *http.Server, conf *Server) error {
+func configureServer(s *http.Server, conf *Server) error {
 	if s == nil {
 		panic("nil *http.Server")
 	}
@@ -349,83 +257,6 @@ func ConfigureServer(s *http.Server, conf *Server) error {
 	return nil
 }
 
-// ServeConnOpts are options for the Server.ServeConn method.
-type ServeConnOpts struct {
-	// Context is the base context to use.
-	// If nil, context.Background is used.
-	Context context.Context
-
-	// BaseConfig optionally sets the base configuration
-	// for values. If nil, defaults are used.
-	BaseConfig *http.Server
-
-	// Handler specifies which handler to use for processing
-	// requests. If nil, BaseConfig.Handler is used. If BaseConfig
-	// or BaseConfig.Handler is nil, http.DefaultServeMux is used.
-	Handler http.Handler
-
-	// UpgradeRequest is an initial request received on a connection
-	// undergoing an h2c upgrade. The request body must have been
-	// completely read from the connection before calling ServeConn,
-	// and the 101 Switching Protocols response written.
-	UpgradeRequest *http.Request
-
-	// Settings is the decoded contents of the HTTP2-Settings header
-	// in an h2c upgrade request.
-	Settings []byte
-
-	// SawClientPreface is set if the HTTP/2 connection preface
-	// has already been read from the connection.
-	SawClientPreface bool
-}
-
-func (o *ServeConnOpts) context() context.Context {
-	if o != nil && o.Context != nil {
-		return o.Context
-	}
-	return context.Background()
-}
-
-func (o *ServeConnOpts) baseConfig() *http.Server {
-	if o != nil && o.BaseConfig != nil {
-		return o.BaseConfig
-	}
-	return new(http.Server)
-}
-
-func (o *ServeConnOpts) handler() http.Handler {
-	if o != nil {
-		if o.Handler != nil {
-			return o.Handler
-		}
-		if o.BaseConfig != nil && o.BaseConfig.Handler != nil {
-			return o.BaseConfig.Handler
-		}
-	}
-	return http.DefaultServeMux
-}
-
-// ServeConn serves HTTP/2 requests on the provided connection and
-// blocks until the connection is no longer readable.
-//
-// ServeConn starts speaking HTTP/2 assuming that c has not had any
-// reads or writes. It writes its initial settings frame and expects
-// to be able to read the preface and settings frame from the
-// client. If c has a ConnectionState method like a *tls.Conn, the
-// ConnectionState is used to verify the TLS ciphersuite and to set
-// the Request.TLS field in Handlers.
-//
-// ServeConn does not support h2c by itself. Any h2c support must be
-// implemented in terms of providing a suitably-behaving net.Conn.
-//
-// The opts parameter is optional. If nil, default values are used.
-func (s *Server) ServeConn(c net.Conn, opts *ServeConnOpts) {
-	if opts == nil {
-		opts = &ServeConnOpts{}
-	}
-	s.serveConn(c, opts, nil)
-}
-
 func (s *Server) serveConn(c net.Conn, opts *ServeConnOpts, newf func(*serverConn)) {
 	baseCtx, cancel := serverConnBaseContext(c, opts)
 	defer cancel()
@@ -461,6 +292,9 @@ func (s *Server) serveConn(c net.Conn, opts *ServeConnOpts, newf func(*serverCon
 	if newf != nil {
 		newf(sc)
 	}
+	if s.state != nil && s.state.testNewConn != nil {
+		s.state.testNewConn(sc)
+	}
 
 	s.state.registerConn(sc)
 	defer s.state.unregisterConn(sc)
@@ -570,15 +404,6 @@ func (s *Server) serveConn(c net.Conn, opts *ServeConnOpts, newf func(*serverCon
 	sc.serve(conf)
 }
 
-func serverConnBaseContext(c net.Conn, opts *ServeConnOpts) (ctx context.Context, cancel func()) {
-	ctx, cancel = context.WithCancel(opts.context())
-	ctx = context.WithValue(ctx, http.LocalAddrContextKey, c.LocalAddr())
-	if hs := opts.baseConfig(); hs != nil {
-		ctx = context.WithValue(ctx, http.ServerContextKey, hs)
-	}
-	return
-}
-
 func (sc *serverConn) rejectConn(err ErrCode, debug string) {
 	sc.vlogf("http2: server rejecting conn: %v, %s", err, debug)
 	// ignoring errors. hanging up anyway.
@@ -2832,21 +2657,6 @@ func (rws *responseWriterState) writeChunk(p []byte) (n int, err error) {
 	return len(p), nil
 }
 
-// TrailerPrefix is a magic prefix for ResponseWriter.Header map keys
-// that, if present, signals that the map entry is actually for
-// the response trailers, and not the response headers. The prefix
-// is stripped after the ServeHTTP call finishes and the values are
-// sent in the trailers.
-//
-// This mechanism is intended only for trailers that are not known
-// prior to the headers being written. If the set of trailers is fixed
-// or known before the header is written, the normal Go trailers mechanism
-// is preferred:
-//
-//	https://golang.org/pkg/net/http/#ResponseWriter
-//	https://golang.org/pkg/net/http/#example_ResponseWriter_trailers
-const TrailerPrefix = "Trailer:"
-
 // promoteUndeclaredTrailers permits http.Handlers to set trailers
 // after the header has already been flushed. Because the Go
 // ResponseWriter interface has no way to set Trailers (only the
@@ -3123,12 +2933,6 @@ func (w *responseWriter) handlerDone() {
 	responseWriterStatePool.Put(rws)
 }
 
-// Push errors.
-var (
-	ErrRecursivePush    = errors.New("http2: recursive push not allowed")
-	ErrPushLimitReached = errors.New("http2: push would exceed peer's SETTINGS_MAX_CONCURRENT_STREAMS")
-)
-
 var _ http.Pusher = (*responseWriter)(nil)
 
 func (w *responseWriter) Push(target string, opts *http.PushOptions) error {

vendor/golang.org/x/net/http2/server_common.go

@@ -0,0 +1,221 @@
+// Copyright 2026 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package http2
+
+import (
+	"context"
+	"errors"
+	"net"
+	"net/http"
+	"time"
+)
+
+// TrailerPrefix is a magic prefix for ResponseWriter.Header map keys
+// that, if present, signals that the map entry is actually for
+// the response trailers, and not the response headers. The prefix
+// is stripped after the ServeHTTP call finishes and the values are
+// sent in the trailers.
+//
+// This mechanism is intended only for trailers that are not known
+// prior to the headers being written. If the set of trailers is fixed
+// or known before the header is written, the normal Go trailers mechanism
+// is preferred:
+//
+//	https://golang.org/pkg/net/http/#ResponseWriter
+//	https://golang.org/pkg/net/http/#example_ResponseWriter_trailers
+const TrailerPrefix = "Trailer:"
+
+// Push errors.
+var (
+	ErrRecursivePush    = errors.New("http2: recursive push not allowed")
+	ErrPushLimitReached = errors.New("http2: push would exceed peer's SETTINGS_MAX_CONCURRENT_STREAMS")
+)
+
+// ConfigureServer adds HTTP/2 support to a net/http Server.
+//
+// The configuration conf may be nil.
+//
+// ConfigureServer must be called before s begins serving.
+func ConfigureServer(s *http.Server, conf *Server) error {
+	return configureServer(s, conf)
+}
+
+// Server is an HTTP/2 server.
+type Server struct {
+	// MaxHandlers limits the number of http.Handler ServeHTTP goroutines
+	// which may run at a time over all connections.
+	// Negative or zero no limit.
+	// TODO: implement
+	MaxHandlers int
+
+	// MaxConcurrentStreams optionally specifies the number of
+	// concurrent streams that each client may have open at a
+	// time. This is unrelated to the number of http.Handler goroutines
+	// which may be active globally, which is MaxHandlers.
+	// If zero, MaxConcurrentStreams defaults to at least 100, per
+	// the HTTP/2 spec's recommendations.
+	MaxConcurrentStreams uint32
+
+	// MaxDecoderHeaderTableSize optionally specifies the http2
+	// SETTINGS_HEADER_TABLE_SIZE to send in the initial settings frame. It
+	// informs the remote endpoint of the maximum size of the header compression
+	// table used to decode header blocks, in octets. If zero, the default value
+	// of 4096 is used.
+	MaxDecoderHeaderTableSize uint32
+
+	// MaxEncoderHeaderTableSize optionally specifies an upper limit for the
+	// header compression table used for encoding request headers. Received
+	// SETTINGS_HEADER_TABLE_SIZE settings are capped at this limit. If zero,
+	// the default value of 4096 is used.
+	MaxEncoderHeaderTableSize uint32
+
+	// MaxReadFrameSize optionally specifies the largest frame
+	// this server is willing to read. A valid value is between
+	// 16k and 16M, inclusive. If zero or otherwise invalid, a
+	// default value is used.
+	MaxReadFrameSize uint32
+
+	// PermitProhibitedCipherSuites, if true, permits the use of
+	// cipher suites prohibited by the HTTP/2 spec.
+	PermitProhibitedCipherSuites bool
+
+	// IdleTimeout specifies how long until idle clients should be
+	// closed with a GOAWAY frame. PING frames are not considered
+	// activity for the purposes of IdleTimeout.
+	// If zero or negative, there is no timeout.
+	IdleTimeout time.Duration
+
+	// ReadIdleTimeout is the timeout after which a health check using a ping
+	// frame will be carried out if no frame is received on the connection.
+	// If zero, no health check is performed.
+	ReadIdleTimeout time.Duration
+
+	// PingTimeout is the timeout after which the connection will be closed
+	// if a response to a ping is not received.
+	// If zero, a default of 15 seconds is used.
+	PingTimeout time.Duration
+
+	// WriteByteTimeout is the timeout after which a connection will be
+	// closed if no data can be written to it. The timeout begins when data is
+	// available to write, and is extended whenever any bytes are written.
+	// If zero or negative, there is no timeout.
+	WriteByteTimeout time.Duration
+
+	// MaxUploadBufferPerConnection is the size of the initial flow
+	// control window for each connections. The HTTP/2 spec does not
+	// allow this to be smaller than 65535 or larger than 2^32-1.
+	// If the value is outside this range, a default value will be
+	// used instead.
+	MaxUploadBufferPerConnection int32
+
+	// MaxUploadBufferPerStream is the size of the initial flow control
+	// window for each stream. The HTTP/2 spec does not allow this to
+	// be larger than 2^32-1. If the value is zero or larger than the
+	// maximum, a default value will be used instead.
+	MaxUploadBufferPerStream int32
+
+	// NewWriteScheduler constructs a write scheduler for a connection.
+	// If nil, a default scheduler is chosen.
+	//
+	// Deprecated: User-provided write schedulers are deprecated.
+	NewWriteScheduler func() WriteScheduler
+
+	// CountError, if non-nil, is called on HTTP/2 server errors.
+	// It's intended to increment a metric for monitoring, such
+	// as an expvar or Prometheus metric.
+	// The errType consists of only ASCII word characters.
+	CountError func(errType string)
+
+	// Internal state. This is a pointer (rather than embedded directly)
+	// so that we don't embed a Mutex in this struct, which will make the
+	// struct non-copyable, which might break some callers.
+	state *serverInternalState
+}
+
+// ServeConnOpts are options for the Server.ServeConn method.
+type ServeConnOpts struct {
+	// Context is the base context to use.
+	// If nil, context.Background is used.
+	Context context.Context
+
+	// BaseConfig optionally sets the base configuration
+	// for values. If nil, defaults are used.
+	BaseConfig *http.Server
+
+	// Handler specifies which handler to use for processing
+	// requests. If nil, BaseConfig.Handler is used. If BaseConfig
+	// or BaseConfig.Handler is nil, http.DefaultServeMux is used.
+	Handler http.Handler
+
+	// UpgradeRequest is an initial request received on a connection
+	// undergoing an h2c upgrade. The request body must have been
+	// completely read from the connection before calling ServeConn,
+	// and the 101 Switching Protocols response written.
+	UpgradeRequest *http.Request
+
+	// Settings is the decoded contents of the HTTP2-Settings header
+	// in an h2c upgrade request.
+	Settings []byte
+
+	// SawClientPreface is set if the HTTP/2 connection preface
+	// has already been read from the connection.
+	SawClientPreface bool
+}
+
+// ServeConn serves HTTP/2 requests on the provided connection and
+// blocks until the connection is no longer readable.
+//
+// ServeConn starts speaking HTTP/2 assuming that c has not had any
+// reads or writes. It writes its initial settings frame and expects
+// to be able to read the preface and settings frame from the
+// client. If c has a ConnectionState method like a *tls.Conn, the
+// ConnectionState is used to verify the TLS ciphersuite and to set
+// the Request.TLS field in Handlers.
+//
+// ServeConn does not support h2c by itself. Any h2c support must be
+// implemented in terms of providing a suitably-behaving net.Conn.
+//
+// The opts parameter is optional. If nil, default values are used.
+func (s *Server) ServeConn(c net.Conn, opts *ServeConnOpts) {
+	if opts == nil {
+		opts = &ServeConnOpts{}
+	}
+	s.serveConn(c, opts, nil)
+}
+
+func (o *ServeConnOpts) context() context.Context {
+	if o != nil && o.Context != nil {
+		return o.Context
+	}
+	return context.Background()
+}
+
+func (o *ServeConnOpts) baseConfig() *http.Server {
+	if o != nil && o.BaseConfig != nil {
+		return o.BaseConfig
+	}
+	return new(http.Server)
+}
+
+func (o *ServeConnOpts) handler() http.Handler {
+	if o != nil {
+		if o.Handler != nil {
+			return o.Handler
+		}
+		if o.BaseConfig != nil && o.BaseConfig.Handler != nil {
+			return o.BaseConfig.Handler
+		}
+	}
+	return http.DefaultServeMux
+}
+
+func serverConnBaseContext(c net.Conn, opts *ServeConnOpts) (ctx context.Context, cancel func()) {
+	ctx, cancel = context.WithCancel(opts.context())
+	ctx = context.WithValue(ctx, http.LocalAddrContextKey, c.LocalAddr())
+	if hs := opts.baseConfig(); hs != nil {
+		ctx = context.WithValue(ctx, http.ServerContextKey, hs)
+	}
+	return
+}

vendor/golang.org/x/net/http2/server_wrap.go

@@ -0,0 +1,201 @@
+// Copyright 2026 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build go1.27 && !http2legacy
+
+// Server wrapping a net/http.Server.
+
+package http2
+
+import (
+	"context"
+	"errors"
+	"net"
+	"net/http"
+	"sync"
+	"time"
+)
+
+type serverInternalState struct {
+	s1            *http.Server
+	initOnce      sync.Once
+	serveConnFunc func(context.Context, net.Conn, http.Handler, bool, *http.Request, []byte)
+}
+
+func configureServer(s *http.Server, conf *Server) error {
+	if s == nil {
+		panic("nil *http.Server")
+	}
+	if conf == nil {
+		conf = new(Server)
+	}
+	if conf.state != nil {
+		// This isn't a panic in the pre-wrapping implementation,
+		// but calling ConfigureServer twice with the same http2.Server
+		// overwrites internal state on the server.
+		// Make the error explicit and early here.
+		panic("ConfigureServer may be called only once per Server")
+	}
+	if h1, h2 := s, conf; h2.IdleTimeout == 0 {
+		if h1.IdleTimeout != 0 {
+			h2.IdleTimeout = h1.IdleTimeout
+		} else {
+			h2.IdleTimeout = h1.ReadTimeout
+		}
+	}
+	conf.state = &serverInternalState{
+		s1: s,
+	}
+	sconfig := &serverConfig{s: conf}
+	if err := s.Serve(sconfig); err != nil || sconfig.serveConnFunc == nil {
+		panic("http2: net/http does not support this version of x/net/http2")
+	}
+	conf.state.serveConnFunc = sconfig.serveConnFunc
+	return nil
+}
+
+type serverConfig struct {
+	s             *Server
+	serveConnFunc func(context.Context, net.Conn, http.Handler, bool, *http.Request, []byte)
+}
+
+func (*serverConfig) Accept() (net.Conn, error) {
+	return nil, errors.New("unexpected call to Accept")
+}
+func (*serverConfig) Close() error {
+	return nil
+}
+func (*serverConfig) Addr() net.Addr {
+	return nil
+}
+
+func (s *serverConfig) ServeConnFunc(f func(context.Context, net.Conn, http.Handler, bool, *http.Request, []byte)) {
+	s.serveConnFunc = f
+}
+
+func (s *serverConfig) HTTP2Config() http.HTTP2Config {
+	return http.HTTP2Config{
+		MaxConcurrentStreams:          int(s.s.MaxConcurrentStreams),
+		MaxDecoderHeaderTableSize:     int(s.s.MaxDecoderHeaderTableSize),
+		MaxEncoderHeaderTableSize:     int(s.s.MaxEncoderHeaderTableSize),
+		MaxReadFrameSize:              int(s.s.MaxReadFrameSize),
+		PermitProhibitedCipherSuites:  s.s.PermitProhibitedCipherSuites,
+		MaxReceiveBufferPerConnection: int(s.s.MaxUploadBufferPerConnection),
+		MaxReceiveBufferPerStream:     int(s.s.MaxUploadBufferPerStream),
+		SendPingTimeout:               s.s.ReadIdleTimeout,
+		PingTimeout:                   s.s.PingTimeout,
+		WriteByteTimeout:              s.s.WriteByteTimeout,
+		CountError:                    s.s.CountError,
+	}
+}
+
+func (s *serverConfig) IdleTimeout() time.Duration {
+	return s.s.IdleTimeout
+}
+
+type serverConn struct{}
+
+func (s *Server) serveConn(c net.Conn, opts *ServeConnOpts, _ func(*serverConn)) {
+	var serveConnFunc func(context.Context, net.Conn, http.Handler, bool, *http.Request, []byte)
+	switch {
+	case opts.BaseConfig != nil:
+		// The user has provided us with an http.Server to take configuration from.
+		//
+		// We can't send our request to opts.BaseConfig, because an http.Server can
+		// only be associated with a single http2.Server and the user might
+		// use this one with several http.Servers.
+		//
+		// We can't send our request to s.state.s1, because it doesn't contain
+		// the right configuration.
+		//
+		// So create a one-off copy of opts.BaseConfig and use it.
+		h1 := &http.Server{
+			TLSConfig:         opts.BaseConfig.TLSConfig,
+			ReadTimeout:       opts.BaseConfig.ReadTimeout,
+			ReadHeaderTimeout: opts.BaseConfig.ReadHeaderTimeout,
+			WriteTimeout:      opts.BaseConfig.WriteTimeout,
+			IdleTimeout:       opts.BaseConfig.IdleTimeout,
+			MaxHeaderBytes:    opts.BaseConfig.MaxHeaderBytes,
+			ConnState:         opts.BaseConfig.ConnState,
+			ErrorLog:          opts.BaseConfig.ErrorLog,
+			BaseContext:       opts.BaseConfig.BaseContext,
+			ConnContext:       opts.BaseConfig.ConnContext,
+			HTTP2:             opts.BaseConfig.HTTP2,
+		}
+		sconfig := &serverConfig{s: s}
+		if err := h1.Serve(sconfig); err != nil || sconfig.serveConnFunc == nil {
+			panic("http2: net/http does not support this version of x/net/http2")
+		}
+		serveConnFunc = sconfig.serveConnFunc
+	case s.state != nil:
+		serveConnFunc = s.state.serveConnFunc
+	default:
+		// Strange-but-true: Server has no concurrency-safe way to initialize
+		// its internal state, so historically ServeConn just doesn't use any
+		// persistent state if you don't call ConfigureServer first.
+		//
+		// If ConfigureServer hasn't been called, create a one-off http.Server
+		// for the connection, since we don't have any way to keep one around for reuse.
+		h1 := &http.Server{}
+		sconfig := &serverConfig{s: s}
+		if err := h1.Serve(sconfig); err != nil || sconfig.serveConnFunc == nil {
+			panic("http2: net/http does not support this version of x/net/http2")
+		}
+		serveConnFunc = sconfig.serveConnFunc
+	}
+
+	ctx, cancel := serverConnBaseContext(c, opts)
+	defer cancel()
+	serveConnFunc(ctx, c, opts.handler(), opts.SawClientPreface, opts.UpgradeRequest, opts.Settings)
+
+}
+
+// FrameWriteRequest is a request to write a frame.
+//
+// Deprecated: User-provided write schedulers are deprecated.
+type FrameWriteRequest struct {
+	// Ideally we'd define this in writesched_common.go,
+	// to avoid duplicating an exported symbol across two files,
+	// but the changes required to make this work are fairly large.
+}
+
+func (wr FrameWriteRequest) StreamID() uint32 {
+	return 0
+}
+
+func (wr FrameWriteRequest) DataSize() int {
+	return 0
+}
+
+func (wr FrameWriteRequest) Consume(n int32) (FrameWriteRequest, FrameWriteRequest, int) {
+	return FrameWriteRequest{}, FrameWriteRequest{}, 0
+}
+
+func (wr FrameWriteRequest) String() string {
+	return ""
+}
+
+// NewPriorityWriteScheduler is deprecated.
+//
+// Deprecated: User-provided write schedulers are deprecated.
+func NewPriorityWriteScheduler(cfg *PriorityWriteSchedulerConfig) WriteScheduler {
+	return unsupportedWriteScheduler{}
+}
+
+// NewRandomWriteScheduler is deprecated.
+//
+// Deprecated: User-provided write schedulers are deprecated.
+func NewRandomWriteScheduler() WriteScheduler {
+	return unsupportedWriteScheduler{}
+}
+
+type unsupportedWriteScheduler struct{}
+
+func (unsupportedWriteScheduler) OpenStream(streamID uint32, options OpenStreamOptions) {}
+func (unsupportedWriteScheduler) CloseStream(streamID uint32)                           {}
+func (unsupportedWriteScheduler) AdjustStream(streamID uint32, priority PriorityParam)  {}
+func (unsupportedWriteScheduler) Push(wr FrameWriteRequest)                             {}
+func (unsupportedWriteScheduler) Pop() (wr FrameWriteRequest, ok bool) {
+	return FrameWriteRequest{}, false
+}

vendor/golang.org/x/net/http2/transport.go

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build !(go1.27 && !http2legacy)
+
 // Transport code.
 
 package http2
@@ -21,20 +23,17 @@ import (
 	"log"
 	"math"
 	"math/bits"
-	mathrand "math/rand"
 	"net"
 	"net/http"
 	"net/http/httptrace"
 	"net/textproto"
 	"strconv"
-	"strings"
 	"sync"
 	"sync/atomic"
 	"time"
 
 	"golang.org/x/net/http/httpguts"
 	"golang.org/x/net/http2/hpack"
-	"golang.org/x/net/idna"
 	"golang.org/x/net/internal/httpcommon"
 )
 
@@ -60,123 +59,7 @@ const (
 	defaultMaxConcurrentStreams = 1000
 )
 
-// Transport is an HTTP/2 Transport.
-//
-// A Transport internally caches connections to servers. It is safe
-// for concurrent use by multiple goroutines.
-type Transport struct {
-	// DialTLSContext specifies an optional dial function with context for
-	// creating TLS connections for requests.
-	//
-	// If DialTLSContext and DialTLS is nil, tls.Dial is used.
-	//
-	// If the returned net.Conn has a ConnectionState method like tls.Conn,
-	// it will be used to set http.Response.TLS.
-	DialTLSContext func(ctx context.Context, network, addr string, cfg *tls.Config) (net.Conn, error)
-
-	// DialTLS specifies an optional dial function for creating
-	// TLS connections for requests.
-	//
-	// If DialTLSContext and DialTLS is nil, tls.Dial is used.
-	//
-	// Deprecated: Use DialTLSContext instead, which allows the transport
-	// to cancel dials as soon as they are no longer needed.
-	// If both are set, DialTLSContext takes priority.
-	DialTLS func(network, addr string, cfg *tls.Config) (net.Conn, error)
-
-	// TLSClientConfig specifies the TLS configuration to use with
-	// tls.Client. If nil, the default configuration is used.
-	TLSClientConfig *tls.Config
-
-	// ConnPool optionally specifies an alternate connection pool to use.
-	// If nil, the default is used.
-	ConnPool ClientConnPool
-
-	// DisableCompression, if true, prevents the Transport from
-	// requesting compression with an "Accept-Encoding: gzip"
-	// request header when the Request contains no existing
-	// Accept-Encoding value. If the Transport requests gzip on
-	// its own and gets a gzipped response, it's transparently
-	// decoded in the Response.Body. However, if the user
-	// explicitly requested gzip it is not automatically
-	// uncompressed.
-	DisableCompression bool
-
-	// AllowHTTP, if true, permits HTTP/2 requests using the insecure,
-	// plain-text "http" scheme. Note that this does not enable h2c support.
-	AllowHTTP bool
-
-	// MaxHeaderListSize is the http2 SETTINGS_MAX_HEADER_LIST_SIZE to
-	// send in the initial settings frame. It is how many bytes
-	// of response headers are allowed. Unlike the http2 spec, zero here
-	// means to use a default limit (currently 10MB). If you actually
-	// want to advertise an unlimited value to the peer, Transport
-	// interprets the highest possible value here (0xffffffff or 1<<32-1)
-	// to mean no limit.
-	MaxHeaderListSize uint32
-
-	// MaxReadFrameSize is the http2 SETTINGS_MAX_FRAME_SIZE to send in the
-	// initial settings frame. It is the size in bytes of the largest frame
-	// payload that the sender is willing to receive. If 0, no setting is
-	// sent, and the value is provided by the peer, which should be 16384
-	// according to the spec:
-	// https://datatracker.ietf.org/doc/html/rfc7540#section-6.5.2.
-	// Values are bounded in the range 16k to 16M.
-	MaxReadFrameSize uint32
-
-	// MaxDecoderHeaderTableSize optionally specifies the http2
-	// SETTINGS_HEADER_TABLE_SIZE to send in the initial settings frame. It
-	// informs the remote endpoint of the maximum size of the header compression
-	// table used to decode header blocks, in octets. If zero, the default value
-	// of 4096 is used.
-	MaxDecoderHeaderTableSize uint32
-
-	// MaxEncoderHeaderTableSize optionally specifies an upper limit for the
-	// header compression table used for encoding request headers. Received
-	// SETTINGS_HEADER_TABLE_SIZE settings are capped at this limit. If zero,
-	// the default value of 4096 is used.
-	MaxEncoderHeaderTableSize uint32
-
-	// StrictMaxConcurrentStreams controls whether the server's
-	// SETTINGS_MAX_CONCURRENT_STREAMS should be respected
-	// globally. If false, new TCP connections are created to the
-	// server as needed to keep each under the per-connection
-	// SETTINGS_MAX_CONCURRENT_STREAMS limit. If true, the
-	// server's SETTINGS_MAX_CONCURRENT_STREAMS is interpreted as
-	// a global limit and callers of RoundTrip block when needed,
-	// waiting for their turn.
-	StrictMaxConcurrentStreams bool
-
-	// IdleConnTimeout is the maximum amount of time an idle
-	// (keep-alive) connection will remain idle before closing
-	// itself.
-	// Zero means no limit.
-	IdleConnTimeout time.Duration
-
-	// ReadIdleTimeout is the timeout after which a health check using ping
-	// frame will be carried out if no frame is received on the connection.
-	// Note that a ping response will is considered a received frame, so if
-	// there is no other traffic on the connection, the health check will
-	// be performed every ReadIdleTimeout interval.
-	// If zero, no health check is performed.
-	ReadIdleTimeout time.Duration
-
-	// PingTimeout is the timeout after which the connection will be closed
-	// if a response to Ping is not received.
-	// Defaults to 15s.
-	PingTimeout time.Duration
-
-	// WriteByteTimeout is the timeout after which the connection will be
-	// closed no data can be written to it. The timeout begins when data is
-	// available to write, and is extended whenever any bytes are written.
-	WriteByteTimeout time.Duration
-
-	// CountError, if non-nil, is called on HTTP/2 transport errors.
-	// It's intended to increment a metric for monitoring, such
-	// as an expvar or Prometheus metric.
-	// The errType consists of only ASCII word characters.
-	CountError func(errType string)
-
+type transportInternal struct {
 	// t1, if non-nil, is the standard library Transport using
 	// this transport. Its settings are used (but not its
 	// RoundTrip method, etc).
@@ -217,27 +100,18 @@ func (t *Transport) disableCompression() bool {
 	return t.DisableCompression || (t.t1 != nil && t.t1.DisableCompression)
 }
 
-// ConfigureTransport configures a net/http HTTP/1 Transport to use HTTP/2.
-// It returns an error if t1 has already been HTTP/2-enabled.
-//
-// Use ConfigureTransports instead to configure the HTTP/2 Transport.
-func ConfigureTransport(t1 *http.Transport) error {
-	_, err := ConfigureTransports(t1)
+func configureTransport(t1 *http.Transport) error {
+	_, err := configureTransports(t1)
 	return err
 }
 
-// ConfigureTransports configures a net/http HTTP/1 Transport to use HTTP/2.
-// It returns a new HTTP/2 Transport for further configuration.
-// It returns an error if t1 has already been HTTP/2-enabled.
-func ConfigureTransports(t1 *http.Transport) (*Transport, error) {
-	return configureTransports(t1)
-}
-
 func configureTransports(t1 *http.Transport) (*Transport, error) {
 	connPool := new(clientConnPool)
 	t2 := &Transport{
 		ConnPool: noDialClientConnPool{connPool},
-		t1:       t1,
+		transportInternal: transportInternal{
+			t1: t1,
+		},
 	}
 	connPool.t = t2
 	if err := registerHTTPSProtocol(t1, noDialH2RoundTripper{t2}); err != nil {
@@ -525,68 +399,7 @@ func (sew stickyErrWriter) Write(p []byte) (n int, err error) {
 	return n, err
 }
 
-// noCachedConnError is the concrete type of ErrNoCachedConn, which
-// needs to be detected by net/http regardless of whether it's its
-// bundled version (in h2_bundle.go with a rewritten type name) or
-// from a user's x/net/http2. As such, as it has a unique method name
-// (IsHTTP2NoCachedConnError) that net/http sniffs for via func
-// isNoCachedConnError.
-type noCachedConnError struct{}
-
-func (noCachedConnError) IsHTTP2NoCachedConnError() {}
-func (noCachedConnError) Error() string             { return "http2: no cached connection was available" }
-
-// isNoCachedConnError reports whether err is of type noCachedConnError
-// or its equivalent renamed type in net/http2's h2_bundle.go. Both types
-// may coexist in the same running program.
-func isNoCachedConnError(err error) bool {
-	_, ok := err.(interface{ IsHTTP2NoCachedConnError() })
-	return ok
-}
-
-var ErrNoCachedConn error = noCachedConnError{}
-
-// RoundTripOpt are options for the Transport.RoundTripOpt method.
-type RoundTripOpt struct {
-	// OnlyCachedConn controls whether RoundTripOpt may
-	// create a new TCP connection. If set true and
-	// no cached connection is available, RoundTripOpt
-	// will return ErrNoCachedConn.
-	OnlyCachedConn bool
-
-	allowHTTP bool // allow http:// URLs
-}
-
-func (t *Transport) RoundTrip(req *http.Request) (*http.Response, error) {
-	return t.RoundTripOpt(req, RoundTripOpt{})
-}
-
-// authorityAddr returns a given authority (a host/IP, or host:port / ip:port)
-// and returns a host:port. The port 443 is added if needed.
-func authorityAddr(scheme string, authority string) (addr string) {
-	host, port, err := net.SplitHostPort(authority)
-	if err != nil { // authority didn't have a port
-		host = authority
-		port = ""
-	}
-	if port == "" { // authority's port was empty
-		port = "443"
-		if scheme == "http" {
-			port = "80"
-		}
-	}
-	if a, err := idna.ToASCII(host); err == nil {
-		host = a
-	}
-	// IPv6 address literal, without a port:
-	if strings.HasPrefix(host, "[") && strings.HasSuffix(host, "]") {
-		return host + ":" + port
-	}
-	return net.JoinHostPort(host, port)
-}
-
-// RoundTripOpt is like RoundTrip, but takes options.
-func (t *Transport) RoundTripOpt(req *http.Request, opt RoundTripOpt) (*http.Response, error) {
+func (t *Transport) roundTripOpt(req *http.Request, opt RoundTripOpt) (*http.Response, error) {
 	switch req.URL.Scheme {
 	case "https":
 		// Always okay.
@@ -597,126 +410,15 @@ func (t *Transport) RoundTripOpt(req *http.Request, opt RoundTripOpt) (*http.Res
 	default:
 		return nil, errors.New("http2: unsupported scheme")
 	}
-
-	addr := authorityAddr(req.URL.Scheme, req.URL.Host)
-	for retry := 0; ; retry++ {
-		cc, err := t.connPool().GetClientConn(req, addr)
-		if err != nil {
-			t.vlogf("http2: Transport failed to get client conn for %s: %v", addr, err)
-			return nil, err
-		}
-		reused := !atomic.CompareAndSwapUint32(&cc.atomicReused, 0, 1)
-		traceGotConn(req, cc, reused)
-		res, err := cc.RoundTrip(req)
-		if err != nil && retry <= 6 {
-			roundTripErr := err
-			if req, err = shouldRetryRequest(req, err); err == nil {
-				// After the first retry, do exponential backoff with 10% jitter.
-				if retry == 0 {
-					t.vlogf("RoundTrip retrying after failure: %v", roundTripErr)
-					continue
-				}
-				backoff := float64(uint(1) << (uint(retry) - 1))
-				backoff += backoff * (0.1 * mathrand.Float64())
-				d := time.Second * time.Duration(backoff)
-				tm := time.NewTimer(d)
-				select {
-				case <-tm.C:
-					t.vlogf("RoundTrip retrying after failure: %v", roundTripErr)
-					continue
-				case <-req.Context().Done():
-					tm.Stop()
-					err = req.Context().Err()
-				}
-			}
-		}
-		if err == errClientConnNotEstablished {
-			// This ClientConn was created recently,
-			// this is the first request to use it,
-			// and the connection is closed and not usable.
-			//
-			// In this state, cc.idleTimer will remove the conn from the pool
-			// when it fires. Stop the timer and remove it here so future requests
-			// won't try to use this connection.
-			//
-			// If the timer has already fired and we're racing it, the redundant
-			// call to MarkDead is harmless.
-			if cc.idleTimer != nil {
-				cc.idleTimer.Stop()
-			}
-			t.connPool().MarkDead(cc)
-		}
-		if err != nil {
-			t.vlogf("RoundTrip failure: %v", err)
-			return nil, err
-		}
-		return res, nil
-	}
+	return t.roundTripViaPool(req, opt, t.connPool())
 }
 
-// CloseIdleConnections closes any connections which were previously
-// connected from previous requests but are now sitting idle.
-// It does not interrupt any connections currently in use.
-func (t *Transport) CloseIdleConnections() {
+func (t *Transport) closeIdleConnections() {
 	if cp, ok := t.connPool().(clientConnPoolIdleCloser); ok {
 		cp.closeIdleConnections()
 	}
 }
 
-var (
-	errClientConnClosed         = errors.New("http2: client conn is closed")
-	errClientConnUnusable       = errors.New("http2: client conn not usable")
-	errClientConnNotEstablished = errors.New("http2: client conn could not be established")
-	errClientConnGotGoAway      = errors.New("http2: Transport received Server's graceful shutdown GOAWAY")
-	errClientConnForceClosed    = errors.New("http2: client connection force closed via ClientConn.Close")
-)
-
-// shouldRetryRequest is called by RoundTrip when a request fails to get
-// response headers. It is always called with a non-nil error.
-// It returns either a request to retry (either the same request, or a
-// modified clone), or an error if the request can't be replayed.
-func shouldRetryRequest(req *http.Request, err error) (*http.Request, error) {
-	if !canRetryError(err) {
-		return nil, err
-	}
-	// If the Body is nil (or http.NoBody), it's safe to reuse
-	// this request and its Body.
-	if req.Body == nil || req.Body == http.NoBody {
-		return req, nil
-	}
-
-	// If the request body can be reset back to its original
-	// state via the optional req.GetBody, do that.
-	if req.GetBody != nil {
-		body, err := req.GetBody()
-		if err != nil {
-			return nil, err
-		}
-		newReq := *req
-		newReq.Body = body
-		return &newReq, nil
-	}
-
-	// The Request.Body can't reset back to the beginning, but we
-	// don't seem to have started to read from it yet, so reuse
-	// the request directly.
-	if err == errClientConnUnusable {
-		return req, nil
-	}
-
-	return nil, fmt.Errorf("http2: Transport: cannot retry err [%v] after Request.Body was written; define Request.GetBody to avoid this error", err)
-}
-
-func canRetryError(err error) bool {
-	if err == errClientConnUnusable || err == errClientConnGotGoAway {
-		return true
-	}
-	if se, ok := err.(StreamError); ok {
-		return se.Code == ErrCodeRefusedStream
-	}
-	return false
-}
-
 func (t *Transport) dialClientConn(ctx context.Context, addr string, singleUse bool) (*ClientConn, error) {
 	host, _, err := net.SplitHostPort(addr)
 	if err != nil {
@@ -743,27 +445,6 @@ func (t *Transport) newTLSConfig(host string) *tls.Config {
 	return cfg
 }
 
-func (t *Transport) dialTLS(ctx context.Context, network, addr string, tlsCfg *tls.Config) (net.Conn, error) {
-	if t.DialTLSContext != nil {
-		return t.DialTLSContext(ctx, network, addr, tlsCfg)
-	} else if t.DialTLS != nil {
-		return t.DialTLS(network, addr, tlsCfg)
-	}
-
-	tlsCn, err := t.dialTLSWithContext(ctx, network, addr, tlsCfg)
-	if err != nil {
-		return nil, err
-	}
-	state := tlsCn.ConnectionState()
-	if p := state.NegotiatedProtocol; p != NextProtoTLS {
-		return nil, fmt.Errorf("http2: unexpected ALPN protocol %q; want %q", p, NextProtoTLS)
-	}
-	if !state.NegotiatedProtocolIsMutual {
-		return nil, errors.New("http2: could not negotiate protocol mutually")
-	}
-	return tlsCn, nil
-}
-
 // disableKeepAlives reports whether connections should be closed as
 // soon as possible after handling the first request.
 func (t *Transport) disableKeepAlives() bool {
@@ -777,7 +458,7 @@ func (t *Transport) expectContinueTimeout() time.Duration {
 	return t.t1.ExpectContinueTimeout
 }
 
-func (t *Transport) NewClientConn(c net.Conn) (*ClientConn, error) {
+func (t *Transport) newUserClientConn(c net.Conn) (*ClientConn, error) {
 	return t.newClientConn(c, t.disableKeepAlives(), nil)
 }
 
@@ -890,8 +571,7 @@ func (cc *ClientConn) healthCheck() {
 	}
 }
 
-// SetDoNotReuse marks cc as not reusable for future HTTP requests.
-func (cc *ClientConn) SetDoNotReuse() {
+func (cc *ClientConn) setDoNotReuse() {
 	cc.mu.Lock()
 	defer cc.mu.Unlock()
 	cc.doNotReuse = true
@@ -932,21 +612,13 @@ func (cc *ClientConn) setGoAway(f *GoAwayFrame) {
 	}
 }
 
-// CanTakeNewRequest reports whether the connection can take a new request,
-// meaning it has not been closed or received or sent a GOAWAY.
-//
-// If the caller is going to immediately make a new request on this
-// connection, use ReserveNewRequest instead.
-func (cc *ClientConn) CanTakeNewRequest() bool {
+func (cc *ClientConn) canTakeNewRequest() bool {
 	cc.mu.Lock()
 	defer cc.mu.Unlock()
 	return cc.canTakeNewRequestLocked()
 }
 
-// ReserveNewRequest is like CanTakeNewRequest but also reserves a
-// concurrent stream in cc. The reservation is decremented on the
-// next call to RoundTrip.
-func (cc *ClientConn) ReserveNewRequest() bool {
+func (cc *ClientConn) reserveNewRequest() bool {
 	cc.mu.Lock()
 	defer cc.mu.Unlock()
 	if st := cc.idleStateLocked(); !st.canTakeNewRequest {
@@ -956,41 +628,7 @@ func (cc *ClientConn) ReserveNewRequest() bool {
 	return true
 }
 
-// ClientConnState describes the state of a ClientConn.
-type ClientConnState struct {
-	// Closed is whether the connection is closed.
-	Closed bool
-
-	// Closing is whether the connection is in the process of
-	// closing. It may be closing due to shutdown, being a
-	// single-use connection, being marked as DoNotReuse, or
-	// having received a GOAWAY frame.
-	Closing bool
-
-	// StreamsActive is how many streams are active.
-	StreamsActive int
-
-	// StreamsReserved is how many streams have been reserved via
-	// ClientConn.ReserveNewRequest.
-	StreamsReserved int
-
-	// StreamsPending is how many requests have been sent in excess
-	// of the peer's advertised MaxConcurrentStreams setting and
-	// are waiting for other streams to complete.
-	StreamsPending int
-
-	// MaxConcurrentStreams is how many concurrent streams the
-	// peer advertised as acceptable. Zero means no SETTINGS
-	// frame has been received yet.
-	MaxConcurrentStreams uint32
-
-	// LastIdle, if non-zero, is when the connection last
-	// transitioned to idle state.
-	LastIdle time.Time
-}
-
-// State returns a snapshot of cc's state.
-func (cc *ClientConn) State() ClientConnState {
+func (cc *ClientConn) state() ClientConnState {
 	cc.wmu.Lock()
 	maxConcurrent := cc.maxConcurrentStreams
 	if !cc.seenSettings {
@@ -1161,6 +799,12 @@ func (cc *ClientConn) closeIfIdle() {
 	cc.closeConn()
 }
 
+func (cc *ClientConn) stopIdleTimer() {
+	if cc.idleTimer != nil {
+		cc.idleTimer.Stop()
+	}
+}
+
 func (cc *ClientConn) isDoNotReuseAndIdle() bool {
 	cc.mu.Lock()
 	defer cc.mu.Unlock()
@@ -1169,8 +813,7 @@ func (cc *ClientConn) isDoNotReuseAndIdle() bool {
 
 var shutdownEnterWaitStateHook = func() {}
 
-// Shutdown gracefully closes the client connection, waiting for running streams to complete.
-func (cc *ClientConn) Shutdown(ctx context.Context) error {
+func (cc *ClientConn) shutdown(ctx context.Context) error {
 	if err := cc.sendGoAway(); err != nil {
 		return err
 	}
@@ -1244,10 +887,7 @@ func (cc *ClientConn) closeForError(err error) {
 	cc.closeConn()
 }
 
-// Close closes the client connection immediately.
-//
-// In-flight requests are interrupted. For a graceful shutdown, use Shutdown instead.
-func (cc *ClientConn) Close() error {
+func (cc *ClientConn) close() error {
 	cc.closeForError(errClientConnForceClosed)
 	return nil
 }
@@ -1301,11 +941,11 @@ func (cc *ClientConn) decrStreamReservationsLocked() {
 	}
 }
 
-func (cc *ClientConn) RoundTrip(req *http.Request) (*http.Response, error) {
-	return cc.roundTrip(req, nil)
+func (cc *ClientConn) roundTrip(req *http.Request) (*http.Response, error) {
+	return cc.internalRoundTrip(req, nil)
 }
 
-func (cc *ClientConn) roundTrip(req *http.Request, streamf func(*clientStream)) (*http.Response, error) {
+func (cc *ClientConn) internalRoundTrip(req *http.Request, streamf func(*clientStream)) (*http.Response, error) {
 	ctx := req.Context()
 	cs := &clientStream{
 		cc:                   cc,
@@ -2125,19 +1765,6 @@ func (cc *ClientConn) readLoop() {
 	}
 }
 
-// GoAwayError is returned by the Transport when the server closes the
-// TCP connection after sending a GOAWAY frame.
-type GoAwayError struct {
-	LastStreamID uint32
-	ErrCode      ErrCode
-	DebugData    string
-}
-
-func (e GoAwayError) Error() string {
-	return fmt.Sprintf("http2: server sent GOAWAY and closed the connection; LastStreamID=%v, ErrCode=%v, debug=%q",
-		e.LastStreamID, e.ErrCode, e.DebugData)
-}
-
 func isEOFOrNetReadError(err error) bool {
 	if err == io.EOF {
 		return true
@@ -2978,7 +2605,7 @@ func (rl *clientConnReadLoop) processResetStream(f *RSTStreamFrame) error {
 }
 
 // Ping sends a PING frame to the server and waits for the ack.
-func (cc *ClientConn) Ping(ctx context.Context) error {
+func (cc *ClientConn) ping(ctx context.Context) error {
 	c := make(chan struct{})
 	// Generate a random payload
 	var p [8]byte
@@ -3092,16 +2719,6 @@ func (cc *ClientConn) vlogf(format string, args ...interface{}) {
 	cc.t.vlogf(format, args...)
 }
 
-func (t *Transport) vlogf(format string, args ...interface{}) {
-	if VerboseLogs {
-		t.logf(format, args...)
-	}
-}
-
-func (t *Transport) logf(format string, args ...interface{}) {
-	log.Printf(format, args...)
-}
-
 var noBody io.ReadCloser = noBodyReader{}
 
 type noBodyReader struct{}
@@ -3417,17 +3034,3 @@ func traceGot1xxResponseFunc(trace *httptrace.ClientTrace) func(int, textproto.M
 	}
 	return nil
 }
-
-// dialTLSWithContext uses tls.Dialer, added in Go 1.15, to open a TLS
-// connection.
-func (t *Transport) dialTLSWithContext(ctx context.Context, network, addr string, cfg *tls.Config) (*tls.Conn, error) {
-	dialer := &tls.Dialer{
-		Config: cfg,
-	}
-	cn, err := dialer.DialContext(ctx, network, addr)
-	if err != nil {
-		return nil, err
-	}
-	tlsCn := cn.(*tls.Conn) // DialContext comment promises this will always succeed
-	return tlsCn, nil
-}

vendor/golang.org/x/net/http2/transport_common.go

@@ -0,0 +1,447 @@
+// Copyright 2026 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package http2
+
+import (
+	"context"
+	"crypto/tls"
+	"errors"
+	"fmt"
+	"log"
+	mathrand "math/rand"
+	"net"
+	"net/http"
+	"strings"
+	"sync/atomic"
+	"time"
+
+	"golang.org/x/net/idna"
+)
+
+// ConfigureTransport configures a net/http HTTP/1 Transport to use HTTP/2.
+// It returns an error if t1 has already been HTTP/2-enabled.
+//
+// Use ConfigureTransports instead to configure the HTTP/2 Transport.
+func ConfigureTransport(t1 *http.Transport) error {
+	return configureTransport(t1)
+}
+
+// ConfigureTransports configures a net/http HTTP/1 Transport to use HTTP/2.
+// It returns a new HTTP/2 Transport for further configuration.
+// It returns an error if t1 has already been HTTP/2-enabled.
+func ConfigureTransports(t1 *http.Transport) (*Transport, error) {
+	return configureTransports(t1)
+}
+
+// Transport is an HTTP/2 Transport.
+//
+// A Transport internally caches connections to servers. It is safe
+// for concurrent use by multiple goroutines.
+type Transport struct {
+	// DialTLSContext specifies an optional dial function with context for
+	// creating TLS connections for requests.
+	//
+	// If DialTLSContext and DialTLS is nil, tls.Dial is used.
+	//
+	// If the returned net.Conn has a ConnectionState method like tls.Conn,
+	// it will be used to set http.Response.TLS.
+	DialTLSContext func(ctx context.Context, network, addr string, cfg *tls.Config) (net.Conn, error)
+
+	// DialTLS specifies an optional dial function for creating
+	// TLS connections for requests.
+	//
+	// If DialTLSContext and DialTLS is nil, tls.Dial is used.
+	//
+	// Deprecated: Use DialTLSContext instead, which allows the transport
+	// to cancel dials as soon as they are no longer needed.
+	// If both are set, DialTLSContext takes priority.
+	DialTLS func(network, addr string, cfg *tls.Config) (net.Conn, error)
+
+	// TLSClientConfig specifies the TLS configuration to use with
+	// tls.Client. If nil, the default configuration is used.
+	TLSClientConfig *tls.Config
+
+	// ConnPool optionally specifies an alternate connection pool to use.
+	// If nil, the default is used.
+	ConnPool ClientConnPool
+
+	// DisableCompression, if true, prevents the Transport from
+	// requesting compression with an "Accept-Encoding: gzip"
+	// request header when the Request contains no existing
+	// Accept-Encoding value. If the Transport requests gzip on
+	// its own and gets a gzipped response, it's transparently
+	// decoded in the Response.Body. However, if the user
+	// explicitly requested gzip it is not automatically
+	// uncompressed.
+	DisableCompression bool
+
+	// AllowHTTP, if true, permits HTTP/2 requests using the insecure,
+	// plain-text "http" scheme. Note that this does not enable h2c support.
+	AllowHTTP bool
+
+	// MaxHeaderListSize is the http2 SETTINGS_MAX_HEADER_LIST_SIZE to
+	// send in the initial settings frame. It is how many bytes
+	// of response headers are allowed. Unlike the http2 spec, zero here
+	// means to use a default limit (currently 10MB). If you actually
+	// want to advertise an unlimited value to the peer, Transport
+	// interprets the highest possible value here (0xffffffff or 1<<32-1)
+	// to mean no limit.
+	MaxHeaderListSize uint32
+
+	// MaxReadFrameSize is the http2 SETTINGS_MAX_FRAME_SIZE to send in the
+	// initial settings frame. It is the size in bytes of the largest frame
+	// payload that the sender is willing to receive. If 0, no setting is
+	// sent, and the value is provided by the peer, which should be 16384
+	// according to the spec:
+	// https://datatracker.ietf.org/doc/html/rfc7540#section-6.5.2.
+	// Values are bounded in the range 16k to 16M.
+	MaxReadFrameSize uint32
+
+	// MaxDecoderHeaderTableSize optionally specifies the http2
+	// SETTINGS_HEADER_TABLE_SIZE to send in the initial settings frame. It
+	// informs the remote endpoint of the maximum size of the header compression
+	// table used to decode header blocks, in octets. If zero, the default value
+	// of 4096 is used.
+	MaxDecoderHeaderTableSize uint32
+
+	// MaxEncoderHeaderTableSize optionally specifies an upper limit for the
+	// header compression table used for encoding request headers. Received
+	// SETTINGS_HEADER_TABLE_SIZE settings are capped at this limit. If zero,
+	// the default value of 4096 is used.
+	MaxEncoderHeaderTableSize uint32
+
+	// StrictMaxConcurrentStreams controls whether the server's
+	// SETTINGS_MAX_CONCURRENT_STREAMS should be respected
+	// globally. If false, new TCP connections are created to the
+	// server as needed to keep each under the per-connection
+	// SETTINGS_MAX_CONCURRENT_STREAMS limit. If true, the
+	// server's SETTINGS_MAX_CONCURRENT_STREAMS is interpreted as
+	// a global limit and callers of RoundTrip block when needed,
+	// waiting for their turn.
+	StrictMaxConcurrentStreams bool
+
+	// IdleConnTimeout is the maximum amount of time an idle
+	// (keep-alive) connection will remain idle before closing
+	// itself.
+	// Zero means no limit.
+	IdleConnTimeout time.Duration
+
+	// ReadIdleTimeout is the timeout after which a health check using ping
+	// frame will be carried out if no frame is received on the connection.
+	// Note that a ping response will is considered a received frame, so if
+	// there is no other traffic on the connection, the health check will
+	// be performed every ReadIdleTimeout interval.
+	// If zero, no health check is performed.
+	ReadIdleTimeout time.Duration
+
+	// PingTimeout is the timeout after which the connection will be closed
+	// if a response to Ping is not received.
+	// Defaults to 15s.
+	PingTimeout time.Duration
+
+	// WriteByteTimeout is the timeout after which the connection will be
+	// closed no data can be written to it. The timeout begins when data is
+	// available to write, and is extended whenever any bytes are written.
+	WriteByteTimeout time.Duration
+
+	// CountError, if non-nil, is called on HTTP/2 transport errors.
+	// It's intended to increment a metric for monitoring, such
+	// as an expvar or Prometheus metric.
+	// The errType consists of only ASCII word characters.
+	CountError func(errType string)
+
+	// Internal state, differs between wrapped and non-wrapped implementations.
+	transportInternal
+}
+
+var (
+	errClientConnClosed         = errors.New("http2: client conn is closed")
+	errClientConnNotEstablished = errors.New("http2: client conn could not be established")
+	errClientConnGotGoAway      = errors.New("http2: Transport received Server's graceful shutdown GOAWAY")
+	errClientConnForceClosed    = errors.New("http2: client connection force closed via ClientConn.Close")
+	errClientConnUnusable       = errors.New("http2: client conn not usable")
+)
+
+// ClientConnPool manages a pool of HTTP/2 client connections.
+type ClientConnPool interface {
+	// GetClientConn returns a specific HTTP/2 connection (usually
+	// a TLS-TCP connection) to an HTTP/2 server. On success, the
+	// returned ClientConn accounts for the upcoming RoundTrip
+	// call, so the caller should not omit it. If the caller needs
+	// to, ClientConn.RoundTrip can be called with a bogus
+	// new(http.Request) to release the stream reservation.
+	GetClientConn(req *http.Request, addr string) (*ClientConn, error)
+	MarkDead(*ClientConn)
+}
+
+// ClientConnState describes the state of a ClientConn.
+type ClientConnState struct {
+	// Closed is whether the connection is closed.
+	Closed bool
+
+	// Closing is whether the connection is in the process of
+	// closing. It may be closing due to shutdown, being a
+	// single-use connection, being marked as DoNotReuse, or
+	// having received a GOAWAY frame.
+	Closing bool
+
+	// StreamsActive is how many streams are active.
+	StreamsActive int
+
+	// StreamsReserved is how many streams have been reserved via
+	// ClientConn.ReserveNewRequest.
+	StreamsReserved int
+
+	// StreamsPending is how many requests have been sent in excess
+	// of the peer's advertised MaxConcurrentStreams setting and
+	// are waiting for other streams to complete.
+	StreamsPending int
+
+	// MaxConcurrentStreams is how many concurrent streams the
+	// peer advertised as acceptable. Zero means no SETTINGS
+	// frame has been received yet.
+	MaxConcurrentStreams uint32
+
+	// LastIdle, if non-zero, is when the connection last
+	// transitioned to idle state.
+	LastIdle time.Time
+}
+
+// RoundTripOpt are options for the Transport.RoundTripOpt method.
+type RoundTripOpt struct {
+	// OnlyCachedConn controls whether RoundTripOpt may
+	// create a new TCP connection. If set true and
+	// no cached connection is available, RoundTripOpt
+	// will return ErrNoCachedConn.
+
+	// OnlyCachedConn was broken in https://go.dev/cl/16699.
+	OnlyCachedConn bool
+
+	allowHTTP bool // allow http:// URLs
+}
+
+func (t *Transport) RoundTrip(req *http.Request) (*http.Response, error) {
+	return t.RoundTripOpt(req, RoundTripOpt{})
+}
+
+// RoundTripOpt is like RoundTrip, but takes options.
+func (t *Transport) RoundTripOpt(req *http.Request, opt RoundTripOpt) (*http.Response, error) {
+	return t.roundTripOpt(req, opt)
+}
+
+// CloseIdleConnections closes any connections which were previously
+// connected from previous requests but are now sitting idle.
+// It does not interrupt any connections currently in use.
+func (t *Transport) CloseIdleConnections() {
+	t.closeIdleConnections()
+}
+
+func (t *Transport) NewClientConn(c net.Conn) (*ClientConn, error) {
+	return t.newUserClientConn(c)
+}
+
+// authorityAddr returns a given authority (a host/IP, or host:port / ip:port)
+// and returns a host:port. The port 443 is added if needed.
+func authorityAddr(scheme string, authority string) (addr string) {
+	host, port, err := net.SplitHostPort(authority)
+	if err != nil { // authority didn't have a port
+		host = authority
+		port = ""
+	}
+	if port == "" { // authority's port was empty
+		port = "443"
+		if scheme == "http" {
+			port = "80"
+		}
+	}
+	if a, err := idna.ToASCII(host); err == nil {
+		host = a
+	}
+	// IPv6 address literal, without a port:
+	if strings.HasPrefix(host, "[") && strings.HasSuffix(host, "]") {
+		return host + ":" + port
+	}
+	return net.JoinHostPort(host, port)
+}
+
+func (t *Transport) roundTripViaPool(req *http.Request, opt RoundTripOpt, pool ClientConnPool) (*http.Response, error) {
+	addr := authorityAddr(req.URL.Scheme, req.URL.Host)
+	for retry := 0; ; retry++ {
+		cc, err := pool.GetClientConn(req, addr)
+		if err != nil {
+			t.vlogf("http2: Transport failed to get client conn for %s: %v", addr, err)
+			return nil, err
+		}
+		reused := !atomic.CompareAndSwapUint32(&cc.atomicReused, 0, 1)
+		traceGotConn(req, cc, reused)
+		res, err := cc.RoundTrip(req)
+		if err != nil && retry <= 6 {
+			roundTripErr := err
+			if req, err = shouldRetryRequest(req, err); err == nil {
+				// After the first retry, do exponential backoff with 10% jitter.
+				if retry == 0 {
+					t.vlogf("RoundTrip retrying after failure: %v", roundTripErr)
+					continue
+				}
+				backoff := float64(uint(1) << (uint(retry) - 1))
+				backoff += backoff * (0.1 * mathrand.Float64())
+				d := time.Second * time.Duration(backoff)
+				tm := time.NewTimer(d)
+				select {
+				case <-tm.C:
+					t.vlogf("RoundTrip retrying after failure: %v", roundTripErr)
+					continue
+				case <-req.Context().Done():
+					tm.Stop()
+					err = req.Context().Err()
+				}
+			}
+		}
+		if err == errClientConnNotEstablished {
+			// This ClientConn was created recently,
+			// this is the first request to use it,
+			// and the connection is closed and not usable.
+			//
+			// In this state, cc.idleTimer will remove the conn from the pool
+			// when it fires. Stop the timer and remove it here so future requests
+			// won't try to use this connection.
+			//
+			// If the timer has already fired and we're racing it, the redundant
+			// call to MarkDead is harmless.
+			cc.stopIdleTimer()
+			pool.MarkDead(cc)
+		}
+		if err != nil {
+			t.vlogf("RoundTrip failure: %v", err)
+			return nil, err
+		}
+		return res, nil
+	}
+}
+
+// shouldRetryRequest is called by RoundTrip when a request fails to get
+// response headers. It is always called with a non-nil error.
+// It returns either a request to retry (either the same request, or a
+// modified clone), or an error if the request can't be replayed.
+func shouldRetryRequest(req *http.Request, err error) (*http.Request, error) {
+	if !canRetryError(err) {
+		return nil, err
+	}
+	// If the Body is nil (or http.NoBody), it's safe to reuse
+	// this request and its Body.
+	if req.Body == nil || req.Body == http.NoBody {
+		return req, nil
+	}
+
+	// If the request body can be reset back to its original
+	// state via the optional req.GetBody, do that.
+	if req.GetBody != nil {
+		body, err := req.GetBody()
+		if err != nil {
+			return nil, err
+		}
+		newReq := *req
+		newReq.Body = body
+		return &newReq, nil
+	}
+
+	// The Request.Body can't reset back to the beginning, but we
+	// don't seem to have started to read from it yet, so reuse
+	// the request directly.
+	if err == errClientConnUnusable {
+		return req, nil
+	}
+
+	return nil, fmt.Errorf("http2: Transport: cannot retry err [%v] after Request.Body was written; define Request.GetBody to avoid this error", err)
+}
+
+func canRetryError(err error) bool {
+	if err == errClientConnUnusable || err == errClientConnGotGoAway {
+		return true
+	}
+	if se, ok := err.(StreamError); ok {
+		return se.Code == ErrCodeRefusedStream
+	}
+	return false
+}
+
+func (t *Transport) vlogf(format string, args ...interface{}) {
+	if VerboseLogs {
+		t.logf(format, args...)
+	}
+}
+
+func (t *Transport) logf(format string, args ...interface{}) {
+	log.Printf(format, args...)
+}
+
+func (t *Transport) dialTLS(ctx context.Context, network, addr string, tlsCfg *tls.Config) (net.Conn, error) {
+	if t.DialTLSContext != nil {
+		return t.DialTLSContext(ctx, network, addr, tlsCfg)
+	} else if t.DialTLS != nil {
+		return t.DialTLS(network, addr, tlsCfg)
+	}
+
+	tlsCn, err := t.dialTLSWithContext(ctx, network, addr, tlsCfg)
+	if err != nil {
+		return nil, err
+	}
+	state := tlsCn.ConnectionState()
+	if p := state.NegotiatedProtocol; p != NextProtoTLS {
+		return nil, fmt.Errorf("http2: unexpected ALPN protocol %q; want %q", p, NextProtoTLS)
+	}
+	if !state.NegotiatedProtocolIsMutual {
+		return nil, errors.New("http2: could not negotiate protocol mutually")
+	}
+	return tlsCn, nil
+}
+
+// dialTLSWithContext uses tls.Dialer, added in Go 1.15, to open a TLS
+// connection.
+func (t *Transport) dialTLSWithContext(ctx context.Context, network, addr string, cfg *tls.Config) (*tls.Conn, error) {
+	dialer := &tls.Dialer{
+		Config: cfg,
+	}
+	cn, err := dialer.DialContext(ctx, network, addr)
+	if err != nil {
+		return nil, err
+	}
+	tlsCn := cn.(*tls.Conn) // DialContext comment promises this will always succeed
+	return tlsCn, nil
+}
+
+// GoAwayError is returned by the Transport when the server closes the
+// TCP connection after sending a GOAWAY frame.
+type GoAwayError struct {
+	LastStreamID uint32
+	ErrCode      ErrCode
+	DebugData    string
+}
+
+func (e GoAwayError) Error() string {
+	return fmt.Sprintf("http2: server sent GOAWAY and closed the connection; LastStreamID=%v, ErrCode=%v, debug=%q",
+		e.LastStreamID, e.ErrCode, e.DebugData)
+}
+
+// noCachedConnError is the concrete type of ErrNoCachedConn, which
+// needs to be detected by net/http regardless of whether it's its
+// bundled version (in h2_bundle.go with a rewritten type name) or
+// from a user's x/net/http2. As such, as it has a unique method name
+// (IsHTTP2NoCachedConnError) that net/http sniffs for via func
+// isNoCachedConnError.
+type noCachedConnError struct{}
+
+func (noCachedConnError) IsHTTP2NoCachedConnError() {}
+func (noCachedConnError) Error() string             { return "http2: no cached connection was available" }
+
+// isNoCachedConnError reports whether err is of type noCachedConnError
+// or its equivalent renamed type in net/http2's h2_bundle.go. Both types
+// may coexist in the same running program.
+func isNoCachedConnError(err error) bool {
+	_, ok := err.(interface{ IsHTTP2NoCachedConnError() })
+	return ok
+}
+
+var ErrNoCachedConn error = noCachedConnError{}

vendor/golang.org/x/net/http2/transport_wrap.go

@@ -0,0 +1,381 @@
+// Copyright 2026 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build go1.27 && !http2legacy
+
+// Transport wrapping a net/http.Transport.
+
+package http2
+
+import (
+	"context"
+	"crypto/tls"
+	"errors"
+	"math"
+	"net"
+	"net/http"
+	"net/http/httptrace"
+	"slices"
+	"sync"
+	"time"
+)
+
+func configureTransport(t1 *http.Transport) error {
+	// ConfigureTransport is a no-op: The http.Transport already supports HTTP/2.
+	return nil
+}
+
+func configureTransports(t1 *http.Transport) (*Transport, error) {
+	// ConfigureTransport returns an http2.Transport with a configuration
+	// linked to the http.Transport's.
+	tr2 := &Transport{}
+	tr2.configure(t1)
+	return tr2, nil
+}
+
+// transportConfig is passed to net/http.Transport.RegisterProtocol("http/2", config).
+// It provides the net/http.Transport with access to the configuration in the
+// x/net/http2.Transport.
+type transportConfig struct {
+	t *Transport
+}
+
+// Registered is called by net/http.Transport.RegisterProtocol,
+// to let us know that it understands the registration mechanism we're using.
+func (t transportConfig) Registered(t1 *http.Transport) {
+	t.t.t1 = t1
+}
+
+func (t transportConfig) DisableCompression() bool {
+	return t.t.DisableCompression
+}
+
+func (t transportConfig) MaxHeaderListSize() int64 {
+	return int64(t.t.MaxHeaderListSize)
+}
+
+func (t transportConfig) IdleConnTimeout() time.Duration {
+	return t.t.IdleConnTimeout
+}
+
+func (t transportConfig) HTTP2Config() http.HTTP2Config {
+	return http.HTTP2Config{
+		StrictMaxConcurrentRequests: t.t.StrictMaxConcurrentStreams,
+		MaxDecoderHeaderTableSize:   int(t.t.MaxDecoderHeaderTableSize),
+		MaxEncoderHeaderTableSize:   int(t.t.MaxEncoderHeaderTableSize),
+		MaxReadFrameSize:            int(t.t.MaxReadFrameSize),
+		SendPingTimeout:             t.t.ReadIdleTimeout,
+		PingTimeout:                 t.t.PingTimeout,
+		WriteByteTimeout:            t.t.WriteByteTimeout,
+		CountError:                  t.t.CountError,
+	}
+}
+
+// ExternalRoundTrip reports whether the Transport wants to take control of the RoundTrip call.
+// If the user hasn't configured a custom connection pool, we leave the RoundTrip up to net/http.
+func (t transportConfig) ExternalRoundTrip() bool {
+	return t.t.ConnPool != nil
+}
+
+// RoundTrip is used when the http.Transport is passing control of the full
+// RoundTrip to us--connection pooling, retries, etc.
+//
+// This is only used when the http2.Transport has a user-provided ConnPool.
+// Any other time, net/http handles everything.
+func (t transportConfig) RoundTrip(req *http.Request) (*http.Response, error) {
+	if t.t.ConnPool == nil {
+		return nil, http.ErrSkipAltProtocol
+	}
+	return t.t.RoundTrip(req)
+}
+
+// netConnContextKey passes a net.Conn to http.Transport.NewClientConn.
+// See http2.Transport.NewClientConn.
+type netConnContextKey struct{}
+
+// ConnFromContext lets the http.Transport fetch a net.Conn out of a context
+// passed to NewClientConn. See http2.Transport.NewClientConn.
+func (t transportConfig) ConnFromContext(ctx context.Context) net.Conn {
+	nc, _ := ctx.Value(netConnContextKey{}).(net.Conn)
+	return nc
+}
+
+// http2TransportContextKey marks a RoundTrip as needing its dial handled by the http2.Transport.
+// We set this for http2.RoundTrip calls, where the historical behavior is to use the
+// http2.Transport's dialer.
+type http2TransportContextKey struct{}
+
+// DialFromContext dials a new connection using the http2.Transport's DialTLS/DialTLSContext.
+func (t transportConfig) DialFromContext(ctx context.Context, network, address string) (net.Conn, error) {
+	if ctx.Value(http2TransportContextKey{}) == nil {
+		// We're being called from a RoundTrip that did not start with an http2.Transport.
+		// Use the http.Transport's dialer.
+		return nil, errors.ErrUnsupported
+	}
+
+	tlsConf := t.t.TLSClientConfig
+	if tlsConf == nil {
+		tlsConf = &tls.Config{}
+	} else {
+		tlsConf = tlsConf.Clone()
+	}
+	if !slices.Contains(tlsConf.NextProtos, "h2") {
+		tlsConf.NextProtos = append([]string{"h2"}, tlsConf.NextProtos...)
+	}
+	if tlsConf.ServerName == "" {
+		host, _, err := net.SplitHostPort(address)
+		if err == nil {
+			tlsConf.ServerName = host
+		}
+	}
+	return t.t.dialTLS(ctx, network, address, tlsConf)
+}
+
+type transportInternal struct {
+	initOnce sync.Once
+	t1       *http.Transport
+}
+
+func (t *Transport) init() {
+	t.initOnce.Do(func() {
+		if t.t1 != nil {
+			return
+		}
+		t1 := &http.Transport{}
+		t.configure(t1)
+	})
+}
+
+func (t *Transport) configure(t1 *http.Transport) {
+	t1.RegisterProtocol("http/2", transportConfig{t})
+	// tr2.t1 is set by transportConfig.Registered.
+	if t.t1 != t1 {
+		panic("http2: net/http does not support this version of x/net/http2")
+	}
+}
+
+func (t *Transport) roundTripOpt(req *http.Request, opt RoundTripOpt) (*http.Response, error) {
+	t.init()
+
+	if req.URL.Scheme == "http" && !t.AllowHTTP {
+		return nil, errors.New("http2: unencrypted HTTP/2 not enabled")
+	}
+
+	// When the Transport has a user-provided connection pool (unusual, deprecated),
+	// we need to handle picking a connection, retrys, etc.
+	if t.ConnPool != nil {
+		return t.roundTripViaPool(req, opt, t.ConnPool)
+	}
+
+	// Setting this context key lets net/http know that if it is necessary to dial
+	// a new connection, we should handle the net.Dial.
+	//
+	// Both http.Transport and http2.Transport allow the user to provide a custom
+	// dial function, and historically you only get the dial function from the
+	// Transport you're calling RoundTrip on.
+	ctx := context.WithValue(req.Context(), http2TransportContextKey{}, t)
+	req = req.WithContext(ctx)
+
+	return t.t1.RoundTrip(req)
+}
+
+func (t *Transport) closeIdleConnections() {
+	t.init()
+	t.t1.CloseIdleConnections()
+}
+
+func (t *Transport) newUserClientConn(c net.Conn) (*ClientConn, error) {
+	// http.Transport's NewClientConn doesn't provide a supported way to create
+	// a connection from a net.Conn. (This might be useful to add in the future?)
+	// We're going to craftily sneak one in via the context key, with the
+	// scheme of "http/2" telling NewClientConn to look for it.
+	ctx := context.WithValue(context.Background(), netConnContextKey{}, c)
+
+	nhcc, err := t.t1.NewClientConn(ctx, "http/2", "")
+	if err != nil {
+		return nil, err
+	}
+	cc := &ClientConn{cc: nhcc, tr: t, tconn: c}
+	nhcc.SetStateHook(cc.stateHook)
+	return cc, nil
+}
+
+// ClientConn is the state of a single HTTP/2 client connection to an
+// HTTP/2 server.
+type ClientConn struct {
+	cc         *http.ClientConn
+	tconn      net.Conn
+	tr         *Transport
+	doNotReuse bool
+
+	mu            sync.Mutex
+	closing       bool
+	closed        bool
+	roundTrips    int
+	reserved      int
+	starting      int
+	pending       int
+	maxConcurrent int
+	lastIdle      time.Time
+	shutdownc     chan struct{}
+
+	atomicReused uint32 // whether conn is being reused; atomic
+}
+
+func (cc *ClientConn) roundTrip(req *http.Request) (*http.Response, error) {
+	err := func() error {
+		cc.mu.Lock()
+		defer cc.mu.Unlock()
+		if cc.doNotReuse {
+			return errClientConnUnusable
+		}
+		cc.roundTrips++
+		if cc.reserved > 0 {
+			// We've already reserved a concurrency slot for this request.
+			cc.reserved--
+		} else if cc.cc.Reserve() != nil {
+			// We don't seem to have an available concurrency slot,
+			// so bump the pending count (requests waiting for a slot).
+			cc.pending++
+		}
+		// ClientConn.Shutdown will not shut down the conn while
+		// cc.starting > 0 or cc.cc.InFlight() > 0.
+		//
+		// The starting state covers the gap between us deciding to
+		// start sending the request, and actually sending it.
+		cc.starting++
+		return nil
+	}()
+	if err != nil {
+		return nil, err
+	}
+	resp, err := cc.cc.RoundTrip(req)
+	cc.mu.Lock()
+	cc.starting--
+	if cc.pending > 0 {
+		// A request completing frees up a concurrency slot for
+		// a pending request to start.
+		cc.pending--
+	}
+	cc.updateStateLocked()
+	cc.mu.Unlock()
+	return resp, err
+}
+
+func (cc *ClientConn) canTakeNewRequest() bool {
+	return cc.cc.Available() > 0 && !cc.doNotReuse
+}
+
+func (cc *ClientConn) close() error {
+	return cc.cc.Close()
+}
+
+func (cc *ClientConn) ping(ctx context.Context) error {
+	// Ask net/http to ping its connection by sending a request with a method of ":ping".
+	_, err := cc.cc.RoundTrip((&http.Request{
+		Method: ":ping",
+	}).WithContext(ctx))
+	return err
+}
+
+func (cc *ClientConn) reserveNewRequest() bool {
+	cc.mu.Lock()
+	defer cc.mu.Unlock()
+	if cc.doNotReuse {
+		return false
+	}
+	if err := cc.cc.Reserve(); err != nil {
+		return false
+	}
+	cc.reserved++
+	return true
+}
+
+func (cc *ClientConn) setDoNotReuse() {
+	cc.mu.Lock()
+	defer cc.mu.Unlock()
+	cc.doNotReuse = true
+	cc.closing = true
+}
+
+func (cc *ClientConn) shutdown(ctx context.Context) error {
+	cc.mu.Lock()
+	inFlight := cc.cc.InFlight() + cc.starting
+	if inFlight > 0 && cc.shutdownc == nil {
+		cc.shutdownc = make(chan struct{})
+	}
+	shutdownc := cc.shutdownc
+	cc.mu.Unlock()
+	if shutdownc != nil {
+		// Wait for in-flight requests to finish.
+		select {
+		case <-shutdownc:
+		case <-ctx.Done():
+			return ctx.Err()
+		}
+	}
+	cc.cc.Close()
+	return nil
+}
+
+func (cc *ClientConn) state() ClientConnState {
+	cc.mu.Lock()
+	defer cc.mu.Unlock()
+	cc.updateStateLocked()
+	return ClientConnState{
+		Closed:               cc.closed,
+		Closing:              cc.closing,
+		StreamsActive:        cc.cc.InFlight() - cc.reserved,
+		StreamsReserved:      cc.reserved,
+		StreamsPending:       cc.pending,
+		MaxConcurrentStreams: uint32(min(int64(cc.maxConcurrent), math.MaxUint32)),
+		LastIdle:             cc.lastIdle,
+	}
+}
+
+// stateHook is the http.ClientConn's state hook.
+func (cc *ClientConn) stateHook(*http.ClientConn) {
+	cc.mu.Lock()
+	defer cc.mu.Unlock()
+	cc.updateStateLocked()
+}
+
+func (cc *ClientConn) updateStateLocked() {
+	if cc.cc.Err() != nil && !cc.closed {
+		cc.closing = true
+		cc.closed = true
+		if cc.tr.ConnPool != nil {
+			// Do the ConnPool update in another goroutine,
+			// to avoid holding the conn mutex while it runs.
+			go cc.tr.ConnPool.MarkDead(cc)
+		}
+	}
+	if cc.cc.InFlight() == 0 && cc.roundTrips > 0 && cc.starting == 0 {
+		cc.lastIdle = time.Now()
+	}
+	if !cc.closed {
+		// This is slightly racy (a request could start or finish in between
+		// the Available and InFlight calls), but the best we can do given that
+		// the net/http ClientConn API doesn't expose the conn's max concurrency.
+		cc.maxConcurrent = cc.cc.Available() + cc.cc.InFlight()
+	}
+	if cc.shutdownc != nil && cc.cc.InFlight()+cc.starting == 0 {
+		close(cc.shutdownc)
+		cc.shutdownc = nil
+	}
+}
+
+func (cc *ClientConn) stopIdleTimer() {}
+
+// traceGotConn is (when http2legacy is not enabled) only used for tracing
+// connections acquired while using a user-provided ClientConnPool.
+func traceGotConn(req *http.Request, cc *ClientConn, reused bool) {
+	trace := httptrace.ContextClientTrace(req.Context())
+	if trace == nil || trace.GotConn == nil {
+		return
+	}
+	ci := httptrace.GotConnInfo{Conn: cc.tconn}
+	ci.Reused = reused
+	trace.GotConn(ci)
+}

vendor/golang.org/x/net/http2/writesched.go

@@ -2,54 +2,12 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build !(go1.27 && !http2legacy)
+
 package http2
 
 import "fmt"
 
-// WriteScheduler is the interface implemented by HTTP/2 write schedulers.
-// Methods are never called concurrently.
-//
-// Deprecated: User-provided write schedulers are deprecated.
-type WriteScheduler interface {
-	// OpenStream opens a new stream in the write scheduler.
-	// It is illegal to call this with streamID=0 or with a streamID that is
-	// already open -- the call may panic.
-	OpenStream(streamID uint32, options OpenStreamOptions)
-
-	// CloseStream closes a stream in the write scheduler. Any frames queued on
-	// this stream should be discarded. It is illegal to call this on a stream
-	// that is not open -- the call may panic.
-	CloseStream(streamID uint32)
-
-	// AdjustStream adjusts the priority of the given stream. This may be called
-	// on a stream that has not yet been opened or has been closed. Note that
-	// RFC 7540 allows PRIORITY frames to be sent on streams in any state. See:
-	// https://tools.ietf.org/html/rfc7540#section-5.1
-	AdjustStream(streamID uint32, priority PriorityParam)
-
-	// Push queues a frame in the scheduler. In most cases, this will not be
-	// called with wr.StreamID()!=0 unless that stream is currently open. The one
-	// exception is RST_STREAM frames, which may be sent on idle or closed streams.
-	Push(wr FrameWriteRequest)
-
-	// Pop dequeues the next frame to write. Returns false if no frames can
-	// be written. Frames with a given wr.StreamID() are Pop'd in the same
-	// order they are Push'd, except RST_STREAM frames. No frames should be
-	// discarded except by CloseStream.
-	Pop() (wr FrameWriteRequest, ok bool)
-}
-
-// OpenStreamOptions specifies extra options for WriteScheduler.OpenStream.
-//
-// Deprecated: User-provided write schedulers are deprecated.
-type OpenStreamOptions struct {
-	// PusherID is zero if the stream was initiated by the client. Otherwise,
-	// PusherID names the stream that pushed the newly opened stream.
-	PusherID uint32
-	// priority is used to set the priority of the newly opened stream.
-	priority PriorityParam
-}
-
 // FrameWriteRequest is a request to write a frame.
 //
 // Deprecated: User-provided write schedulers are deprecated.

vendor/golang.org/x/net/http2/writesched_common.go

@@ -0,0 +1,90 @@
+// Copyright 2026 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package http2
+
+// WriteScheduler is the interface implemented by HTTP/2 write schedulers.
+// Methods are never called concurrently.
+//
+// Deprecated: User-provided write schedulers are deprecated.
+type WriteScheduler interface {
+	// OpenStream opens a new stream in the write scheduler.
+	// It is illegal to call this with streamID=0 or with a streamID that is
+	// already open -- the call may panic.
+	OpenStream(streamID uint32, options OpenStreamOptions)
+
+	// CloseStream closes a stream in the write scheduler. Any frames queued on
+	// this stream should be discarded. It is illegal to call this on a stream
+	// that is not open -- the call may panic.
+	CloseStream(streamID uint32)
+
+	// AdjustStream adjusts the priority of the given stream. This may be called
+	// on a stream that has not yet been opened or has been closed. Note that
+	// RFC 7540 allows PRIORITY frames to be sent on streams in any state. See:
+	// https://tools.ietf.org/html/rfc7540#section-5.1
+	AdjustStream(streamID uint32, priority PriorityParam)
+
+	// Push queues a frame in the scheduler. In most cases, this will not be
+	// called with wr.StreamID()!=0 unless that stream is currently open. The one
+	// exception is RST_STREAM frames, which may be sent on idle or closed streams.
+	Push(wr FrameWriteRequest)
+
+	// Pop dequeues the next frame to write. Returns false if no frames can
+	// be written. Frames with a given wr.StreamID() are Pop'd in the same
+	// order they are Push'd, except RST_STREAM frames. No frames should be
+	// discarded except by CloseStream.
+	Pop() (wr FrameWriteRequest, ok bool)
+}
+
+// OpenStreamOptions specifies extra options for WriteScheduler.OpenStream.
+//
+// Deprecated: User-provided write schedulers are deprecated.
+type OpenStreamOptions struct {
+	// PusherID is zero if the stream was initiated by the client. Otherwise,
+	// PusherID names the stream that pushed the newly opened stream.
+	PusherID uint32
+	// priority is used to set the priority of the newly opened stream.
+	priority PriorityParam
+}
+
+// PriorityWriteSchedulerConfig configures a priorityWriteScheduler.
+//
+// Deprecated: User-provided write schedulers are deprecated.
+type PriorityWriteSchedulerConfig struct {
+	// MaxClosedNodesInTree controls the maximum number of closed streams to
+	// retain in the priority tree. Setting this to zero saves a small amount
+	// of memory at the cost of performance.
+	//
+	// See RFC 7540, Section 5.3.4:
+	//   "It is possible for a stream to become closed while prioritization
+	//   information ... is in transit. ... This potentially creates suboptimal
+	//   prioritization, since the stream could be given a priority that is
+	//   different from what is intended. To avoid these problems, an endpoint
+	//   SHOULD retain stream prioritization state for a period after streams
+	//   become closed. The longer state is retained, the lower the chance that
+	//   streams are assigned incorrect or default priority values."
+	MaxClosedNodesInTree int
+
+	// MaxIdleNodesInTree controls the maximum number of idle streams to
+	// retain in the priority tree. Setting this to zero saves a small amount
+	// of memory at the cost of performance.
+	//
+	// See RFC 7540, Section 5.3.4:
+	//   Similarly, streams that are in the "idle" state can be assigned
+	//   priority or become a parent of other streams. This allows for the
+	//   creation of a grouping node in the dependency tree, which enables
+	//   more flexible expressions of priority. Idle streams begin with a
+	//   default priority (Section 5.3.5).
+	MaxIdleNodesInTree int
+
+	// ThrottleOutOfOrderWrites enables write throttling to help ensure that
+	// data is delivered in priority order. This works around a race where
+	// stream B depends on stream A and both streams are about to call Write
+	// to queue DATA frames. If B wins the race, a naive scheduler would eagerly
+	// write as much data from B as possible, but this is suboptimal because A
+	// is a higher-priority stream. With throttling enabled, we write a small
+	// amount of data from B to minimize the amount of bandwidth that B can
+	// steal from A.
+	ThrottleOutOfOrderWrites bool
+}

vendor/golang.org/x/net/http2/writesched_priority_rfc7540.go

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build !(go1.27 && !http2legacy)
+
 package http2
 
 import (
@@ -13,47 +15,6 @@ import (
 // RFC 7540, Section 5.3.5: the default weight is 16.
 const priorityDefaultWeightRFC7540 = 15 // 16 = 15 + 1
 
-// PriorityWriteSchedulerConfig configures a priorityWriteScheduler.
-//
-// Deprecated: User-provided write schedulers are deprecated.
-type PriorityWriteSchedulerConfig struct {
-	// MaxClosedNodesInTree controls the maximum number of closed streams to
-	// retain in the priority tree. Setting this to zero saves a small amount
-	// of memory at the cost of performance.
-	//
-	// See RFC 7540, Section 5.3.4:
-	//   "It is possible for a stream to become closed while prioritization
-	//   information ... is in transit. ... This potentially creates suboptimal
-	//   prioritization, since the stream could be given a priority that is
-	//   different from what is intended. To avoid these problems, an endpoint
-	//   SHOULD retain stream prioritization state for a period after streams
-	//   become closed. The longer state is retained, the lower the chance that
-	//   streams are assigned incorrect or default priority values."
-	MaxClosedNodesInTree int
-
-	// MaxIdleNodesInTree controls the maximum number of idle streams to
-	// retain in the priority tree. Setting this to zero saves a small amount
-	// of memory at the cost of performance.
-	//
-	// See RFC 7540, Section 5.3.4:
-	//   Similarly, streams that are in the "idle" state can be assigned
-	//   priority or become a parent of other streams. This allows for the
-	//   creation of a grouping node in the dependency tree, which enables
-	//   more flexible expressions of priority. Idle streams begin with a
-	//   default priority (Section 5.3.5).
-	MaxIdleNodesInTree int
-
-	// ThrottleOutOfOrderWrites enables write throttling to help ensure that
-	// data is delivered in priority order. This works around a race where
-	// stream B depends on stream A and both streams are about to call Write
-	// to queue DATA frames. If B wins the race, a naive scheduler would eagerly
-	// write as much data from B as possible, but this is suboptimal because A
-	// is a higher-priority stream. With throttling enabled, we write a small
-	// amount of data from B to minimize the amount of bandwidth that B can
-	// steal from A.
-	ThrottleOutOfOrderWrites bool
-}
-
 // NewPriorityWriteScheduler constructs a WriteScheduler that schedules
 // frames by following HTTP/2 priorities as described in RFC 7540 Section 5.3.
 // If cfg is nil, default options are used.

vendor/golang.org/x/net/http2/writesched_priority_rfc9218.go

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build !(go1.27 && !http2legacy)
+
 package http2
 
 import (

vendor/golang.org/x/net/http2/writesched_random.go

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build !(go1.27 && !http2legacy)
+
 package http2
 
 import "math"

vendor/golang.org/x/net/http2/writesched_roundrobin.go

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+//go:build !(go1.27 && !http2legacy)
+
 package http2
 
 import (

vendor/golang.org/x/net/idna/go118.go

@@ -1,13 +0,0 @@
-// Code generated by running "go generate" in golang.org/x/text. DO NOT EDIT.
-
-// Copyright 2021 The Go Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-//go:build go1.18
-
-package idna
-
-// Transitional processing is disabled by default in Go 1.18.
-// https://golang.org/issue/47510
-const transitionalLookup = false

vendor/golang.org/x/net/idna/idna.go

@@ -4,8 +4,6 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-//go:build go1.10
-
 // Package idna implements IDNA2008 using the compatibility processing
 // defined by UTS (Unicode Technical Standard) #46, which defines a standard to
 // deal with the transition from IDNA2003.
@@ -20,6 +18,7 @@ package idna // import "golang.org/x/net/idna"
 import (
 	"fmt"
 	"strings"
+	"unicode"
 	"unicode/utf8"
 
 	"golang.org/x/text/secure/bidirule"
@@ -27,6 +26,8 @@ import (
 	"golang.org/x/text/unicode/norm"
 )
 
+const unicode16 = unicode.Version >= "16.0.0"
+
 // NOTE: Unlike common practice in Go APIs, the functions will return a
 // sanitized domain name in case of errors. Browsers sometimes use a partially
 // evaluated string as lookup.
@@ -101,6 +102,11 @@ func ValidateLabels(enable bool) Option {
 	}
 }
 
+// validateLabels reports whether the ValidateLabels option is enabled.
+func (p *Profile) validateLabels() bool {
+	return p.fromPuny != nil
+}
+
 // CheckHyphens sets whether to check for correct use of hyphens ('-') in
 // labels. Most web browsers do not have this option set, since labels such as
 // "r3---sn-apo3qvuoxuxbt-j5pe" are in common use.
@@ -263,6 +269,10 @@ func (p *Profile) String() string {
 	return s
 }
 
+// Transitional processing is disabled by default as of Go 1.18.
+// https://golang.org/issue/47510
+const transitionalLookup = false
+
 var (
 	// Punycode is a Profile that does raw punycode processing with a minimum
 	// of validation.
@@ -324,15 +334,30 @@ func (e labelError) Error() string {
 	return fmt.Sprintf("idna: invalid label %q", e.label)
 }
 
-type runeError rune
-
-func (e runeError) code() string { return "P1" }
-func (e runeError) Error() string {
-	return fmt.Sprintf("idna: disallowed rune %U", e)
+type runeError struct {
+	r     rune
+	code_ string
 }
 
-// process implements the algorithm described in section 4 of UTS #46,
-// see https://www.unicode.org/reports/tr46.
+func (e runeError) code() string { return e.code_ }
+func (e runeError) Error() string {
+	return fmt.Sprintf("idna: disallowed rune %U", e.r)
+}
+
+// code16 returns old for Unicode < 16, new for Unicode >= 16.
+func code16(old, new string) string {
+	if unicode16 {
+		return new
+	}
+	return old
+}
+
+// process10 implements the algorithm described in section 4 of UTS #46.
+// It implements both the Unicode 10 algorithm
+// (https://www.unicode.org/reports/tr46/tr46-19.html)
+// and the Unicode 16 algorithm
+// (https://www.unicode.org/reports/tr46/tr46-35.html)
+// depending on unicode16, which in turn depends on unicode.Version.
 func (p *Profile) process(s string, toASCII bool) (string, error) {
 	var err error
 	var isBidi bool
@@ -347,8 +372,12 @@ func (p *Profile) process(s string, toASCII bool) (string, error) {
 	// TODO: allow for a quick check of the tables data.
 	// It seems like we should only create this error on ToASCII, but the
 	// UTS 46 conformance tests suggests we should always check this.
+	labelCode := "X4_2"
+	if !unicode16 || toASCII {
+		labelCode = "A4"
+	}
 	if err == nil && p.verifyDNSLength && s == "" {
-		err = &labelError{s, "A4"}
+		err = labelError{s, labelCode}
 	}
 	labels := labelIter{orig: s}
 	for ; !labels.done(); labels.next() {
@@ -357,12 +386,13 @@ func (p *Profile) process(s string, toASCII bool) (string, error) {
 			// Empty labels are not okay. The label iterator skips the last
 			// label if it is empty.
 			if err == nil && p.verifyDNSLength {
-				err = &labelError{s, "A4"}
+				err = labelError{s, labelCode}
 			}
 			continue
 		}
 		if strings.HasPrefix(label, acePrefix) {
-			u, err2 := decode(label[len(acePrefix):])
+			enc := label[len(acePrefix):]
+			u, err2 := decode(enc)
 			if err2 != nil {
 				if err == nil {
 					err = err2
@@ -370,6 +400,9 @@ func (p *Profile) process(s string, toASCII bool) (string, error) {
 				// Spec says keep the old label.
 				continue
 			}
+			if unicode16 && err == nil && len(u) > 0 && isASCII(u) {
+				err = punyError(enc)
+			}
 			isBidi = isBidi || bidirule.DirectionString(u) != bidi.LeftToRight
 			labels.set(u)
 			if err == nil && p.fromPuny != nil {
@@ -379,16 +412,16 @@ func (p *Profile) process(s string, toASCII bool) (string, error) {
 				// This should be called on NonTransitional, according to the
 				// spec, but that currently does not have any effect. Use the
 				// original profile to preserve options.
-				err = p.validateLabel(u)
+				err = p.validateLabel(u, labelCode)
 			}
 		} else if err == nil {
-			err = p.validateLabel(label)
+			err = p.validateLabel(label, labelCode)
 		}
 	}
 	if isBidi && p.bidirule != nil && err == nil {
 		for labels.reset(); !labels.done(); labels.next() {
 			if !p.bidirule(labels.label()) {
-				err = &labelError{s, "B"}
+				err = labelError{s, "B"}
 				break
 			}
 		}
@@ -406,24 +439,36 @@ func (p *Profile) process(s string, toASCII bool) (string, error) {
 			}
 			n := len(label)
 			if p.verifyDNSLength && err == nil && (n == 0 || n > 63) {
-				err = &labelError{label, "A4"}
+				err = labelError{label, labelCode}
 			}
 		}
 	}
 	s = labels.result()
 	if toASCII && p.verifyDNSLength && err == nil {
+		if unicode16 && strings.HasSuffix(s, ".") {
+			err = labelError{s, labelCode}
+		}
 		// Compute the length of the domain name minus the root label and its dot.
 		n := len(s)
 		if n > 0 && s[n-1] == '.' {
 			n--
 		}
 		if len(s) < 1 || n > 253 {
-			err = &labelError{s, "A4"}
+			err = labelError{s, labelCode}
 		}
 	}
 	return s, err
 }
 
+func isASCII(s string) bool {
+	for _, c := range []byte(s) {
+		if c >= 0x80 {
+			return false
+		}
+	}
+	return true
+}
+
 func normalize(p *Profile, s string) (mapped string, isBidi bool, err error) {
 	// TODO: consider first doing a quick check to see if any of these checks
 	// need to be done. This will make it slower in the general case, but
@@ -436,12 +481,12 @@ func normalize(p *Profile, s string) (mapped string, isBidi bool, err error) {
 func validateRegistration(p *Profile, s string) (idem string, bidi bool, err error) {
 	// TODO: filter need for normalization in loop below.
 	if !norm.NFC.IsNormalString(s) {
-		return s, false, &labelError{s, "V1"}
+		return s, false, labelError{s, "V1"}
 	}
 	for i := 0; i < len(s); {
 		v, sz := trie.lookupString(s[i:])
 		if sz == 0 {
-			return s, bidi, runeError(utf8.RuneError)
+			return s, bidi, runeError{utf8.RuneError, "P1"}
 		}
 		bidi = bidi || info(v).isBidi(s[i:])
 		// Copy bytes not copied so far.
@@ -449,9 +494,12 @@ func validateRegistration(p *Profile, s string) (idem string, bidi bool, err err
 		// TODO: handle the NV8 defined in the Unicode idna data set to allow
 		// for strict conformance to IDNA2008.
 		case valid, deviation:
+			if sz == 1 && p.useSTD3Rules && !allowedSTD3(rune(s[i])) {
+				return s, bidi, runeError{rune(s[i]), "P1"}
+			}
 		case disallowed, mapped, unknown, ignored:
 			r, _ := utf8.DecodeRuneInString(s[i:])
-			return s, bidi, runeError(r)
+			return s, bidi, runeError{r, "P1"}
 		}
 		i += sz
 	}
@@ -489,7 +537,7 @@ func validateAndMap(p *Profile, s string) (vm string, bidi bool, err error) {
 			b = append(b, "\ufffd"...)
 			k = len(s)
 			if err == nil {
-				err = runeError(utf8.RuneError)
+				err = runeError{utf8.RuneError, "P1"}
 			}
 			break
 		}
@@ -502,14 +550,26 @@ func validateAndMap(p *Profile, s string) (vm string, bidi bool, err error) {
 		case valid:
 			continue
 		case disallowed:
-			if err == nil {
+			// Unicode 16 delays the error until validateLabels.
+			// Unicode 10 gave an error now.
+			if !unicode16 && err == nil {
 				r, _ := utf8.DecodeRuneInString(s[start:])
-				err = runeError(r)
+				err = runeError{r, "P1"}
 			}
 			continue
-		case mapped, deviation:
+		case deviation:
+			if unicode16 && !p.transitional {
+				break
+			}
+			fallthrough
+		case mapped:
 			b = append(b, s[k:start]...)
-			b = info(v).appendMapping(b, s[start:i])
+			// Unicode 16 requires a special case to handle ẞ -> ss in transitional mode.
+			if unicode16 && p.transitional && s[start:start+sz] == "ẞ" {
+				b = append(b, "ss"...)
+			} else {
+				b = info(v).appendMapping(b, s[start:i])
+			}
 		case ignored:
 			b = append(b, s[k:start]...)
 			// drop the rune
@@ -600,13 +660,13 @@ const acePrefix = "xn--"
 
 func (p *Profile) simplify(cat category) category {
 	switch cat {
-	case disallowedSTD3Mapped:
+	case disallowedSTD3Mapped: // only happens for pre-Unicode 16
 		if p.useSTD3Rules {
 			cat = disallowed
 		} else {
 			cat = mapped
 		}
-	case disallowedSTD3Valid:
+	case disallowedSTD3Valid: // only happens for pre-Unicode 16
 		if p.useSTD3Rules {
 			cat = disallowed
 		} else {
@@ -625,17 +685,18 @@ func (p *Profile) simplify(cat category) category {
 
 func validateFromPunycode(p *Profile, s string) error {
 	if !norm.NFC.IsNormalString(s) {
-		return &labelError{s, "V1"}
+		return labelError{s, "V1"}
 	}
 	// TODO: detect whether string may have to be normalized in the following
 	// loop.
 	for i := 0; i < len(s); {
 		v, sz := trie.lookupString(s[i:])
 		if sz == 0 {
-			return runeError(utf8.RuneError)
+			return runeError{utf8.RuneError, "P1"}
 		}
-		if c := p.simplify(info(v).category()); c != valid && c != deviation {
-			return &labelError{s, "V6"}
+		cat := info(v).category()
+		if c := p.simplify(cat); c != valid && c != deviation {
+			return labelError{s, code16("V6", "V7")}
 		}
 		i += sz
 	}
@@ -704,23 +765,51 @@ var joinStates = [][numJoinTypes]joinState{
 	},
 }
 
+// allowedSTD3 reports whether r is a rune that can appear in a domain name
+// according to STD3. We allow all non-ASCII runes and then letters, digits, hyphens.
+// We also add dot so that this can be run against the whole name and not just
+// a single name element (label). The surrounding code checks dots well enough.
+func allowedSTD3(r rune) bool {
+	return r >= 0x80 || 'a' <= r && r <= 'z' || '0' <= r && r <= '9' || r == '-' || r == '.'
+}
+
 // validateLabel validates the criteria from Section 4.1. Item 1, 4, and 6 are
 // already implicitly satisfied by the overall implementation.
-func (p *Profile) validateLabel(s string) (err error) {
+func (p *Profile) validateLabel(s string, labelCode string) (err error) {
 	if s == "" {
 		if p.verifyDNSLength {
-			return &labelError{s, "A4"}
+			return labelError{s, labelCode}
 		}
 		return nil
 	}
 	if p.checkHyphens {
 		if len(s) > 4 && s[2] == '-' && s[3] == '-' {
-			return &labelError{s, "V2"}
+			return labelError{s, "V2"}
 		}
 		if s[0] == '-' || s[len(s)-1] == '-' {
-			return &labelError{s, "V3"}
+			return labelError{s, "V3"}
 		}
 	}
+
+	// Unicode 16's TR 46 delays the rune validity checks until after the label is decoded.
+	// (validateAndMap did not reject them earlier.)
+	if unicode16 && p.validateLabels() {
+		for i := 0; i < len(s); {
+			v, sz := trie.lookupString(s[i:])
+			if sz == 0 {
+				return runeError{utf8.RuneError, "P1"}
+			}
+			cat := info(v).category()
+			if c := p.simplify(cat); c != valid && (!p.transitional || c != deviation) {
+				return labelError{s, "V7"}
+			}
+			if sz == 1 && p.useSTD3Rules && !allowedSTD3(rune(s[i])) {
+				return runeError{rune(s[i]), "U1"}
+			}
+			i += sz
+		}
+	}
+
 	if !p.checkJoiners {
 		return nil
 	}
@@ -729,7 +818,7 @@ func (p *Profile) validateLabel(s string) (err error) {
 	v, sz := trie.lookupString(s)
 	x := info(v)
 	if x.isModifier() {
-		return &labelError{s, "V5"}
+		return labelError{s, code16("V5", "V6")}
 	}
 	// Quickly return in the absence of zero-width (non) joiners.
 	if strings.Index(s, zwj) == -1 && strings.Index(s, zwnj) == -1 {
@@ -754,8 +843,9 @@ func (p *Profile) validateLabel(s string) (err error) {
 		x = info(v)
 	}
 	if st == stateFAIL || st == stateAfter {
-		return &labelError{s, "C"}
+		return labelError{s, "C"}
 	}
+
 	return nil
 }
 
@@ -767,3 +857,24 @@ func ascii(s string) bool {
 	}
 	return true
 }
+
+// appendMapping appends the mapping for the respective rune. isMapped must be
+// true. A mapping is a categorization of a rune as defined in UTS #46.
+func (c info) appendMapping(b []byte, s string) []byte {
+	index := int(c >> indexShift)
+	if c&xorBit == 0 {
+		p := index
+		return append(b, mappings[mappingIndex[p]:mappingIndex[p+1]]...)
+	}
+	b = append(b, s...)
+	if c&inlineXOR == inlineXOR {
+		// TODO: support and handle two-byte inline masks
+		b[len(b)-1] ^= byte(index)
+	} else {
+		for p := len(b) - int(xorData[index]); p < len(b); p++ {
+			index++
+			b[p] ^= xorData[index]
+		}
+	}
+	return b
+}

vendor/golang.org/x/net/idna/idna9.0.0.go

@@ -1,717 +0,0 @@
-// Code generated by running "go generate" in golang.org/x/text. DO NOT EDIT.
-
-// Copyright 2016 The Go Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-//go:build !go1.10
-
-// Package idna implements IDNA2008 using the compatibility processing
-// defined by UTS (Unicode Technical Standard) #46, which defines a standard to
-// deal with the transition from IDNA2003.
-//
-// IDNA2008 (Internationalized Domain Names for Applications), is defined in RFC
-// 5890, RFC 5891, RFC 5892, RFC 5893 and RFC 5894.
-// UTS #46 is defined in https://www.unicode.org/reports/tr46.
-// See https://unicode.org/cldr/utility/idna.jsp for a visualization of the
-// differences between these two standards.
-package idna // import "golang.org/x/net/idna"
-
-import (
-	"fmt"
-	"strings"
-	"unicode/utf8"
-
-	"golang.org/x/text/secure/bidirule"
-	"golang.org/x/text/unicode/norm"
-)
-
-// NOTE: Unlike common practice in Go APIs, the functions will return a
-// sanitized domain name in case of errors. Browsers sometimes use a partially
-// evaluated string as lookup.
-// TODO: the current error handling is, in my opinion, the least opinionated.
-// Other strategies are also viable, though:
-// Option 1) Return an empty string in case of error, but allow the user to
-//    specify explicitly which errors to ignore.
-// Option 2) Return the partially evaluated string if it is itself a valid
-//    string, otherwise return the empty string in case of error.
-// Option 3) Option 1 and 2.
-// Option 4) Always return an empty string for now and implement Option 1 as
-//    needed, and document that the return string may not be empty in case of
-//    error in the future.
-// I think Option 1 is best, but it is quite opinionated.
-
-// ToASCII is a wrapper for Punycode.ToASCII.
-func ToASCII(s string) (string, error) {
-	return Punycode.process(s, true)
-}
-
-// ToUnicode is a wrapper for Punycode.ToUnicode.
-func ToUnicode(s string) (string, error) {
-	return Punycode.process(s, false)
-}
-
-// An Option configures a Profile at creation time.
-type Option func(*options)
-
-// Transitional sets a Profile to use the Transitional mapping as defined in UTS
-// #46. This will cause, for example, "ß" to be mapped to "ss". Using the
-// transitional mapping provides a compromise between IDNA2003 and IDNA2008
-// compatibility. It is used by some browsers when resolving domain names. This
-// option is only meaningful if combined with MapForLookup.
-func Transitional(transitional bool) Option {
-	return func(o *options) { o.transitional = transitional }
-}
-
-// VerifyDNSLength sets whether a Profile should fail if any of the IDN parts
-// are longer than allowed by the RFC.
-//
-// This option corresponds to the VerifyDnsLength flag in UTS #46.
-func VerifyDNSLength(verify bool) Option {
-	return func(o *options) { o.verifyDNSLength = verify }
-}
-
-// RemoveLeadingDots removes leading label separators. Leading runes that map to
-// dots, such as U+3002 IDEOGRAPHIC FULL STOP, are removed as well.
-func RemoveLeadingDots(remove bool) Option {
-	return func(o *options) { o.removeLeadingDots = remove }
-}
-
-// ValidateLabels sets whether to check the mandatory label validation criteria
-// as defined in Section 5.4 of RFC 5891. This includes testing for correct use
-// of hyphens ('-'), normalization, validity of runes, and the context rules.
-// In particular, ValidateLabels also sets the CheckHyphens and CheckJoiners flags
-// in UTS #46.
-func ValidateLabels(enable bool) Option {
-	return func(o *options) {
-		// Don't override existing mappings, but set one that at least checks
-		// normalization if it is not set.
-		if o.mapping == nil && enable {
-			o.mapping = normalize
-		}
-		o.trie = trie
-		o.checkJoiners = enable
-		o.checkHyphens = enable
-		if enable {
-			o.fromPuny = validateFromPunycode
-		} else {
-			o.fromPuny = nil
-		}
-	}
-}
-
-// CheckHyphens sets whether to check for correct use of hyphens ('-') in
-// labels. Most web browsers do not have this option set, since labels such as
-// "r3---sn-apo3qvuoxuxbt-j5pe" are in common use.
-//
-// This option corresponds to the CheckHyphens flag in UTS #46.
-func CheckHyphens(enable bool) Option {
-	return func(o *options) { o.checkHyphens = enable }
-}
-
-// CheckJoiners sets whether to check the ContextJ rules as defined in Appendix
-// A of RFC 5892, concerning the use of joiner runes.
-//
-// This option corresponds to the CheckJoiners flag in UTS #46.
-func CheckJoiners(enable bool) Option {
-	return func(o *options) {
-		o.trie = trie
-		o.checkJoiners = enable
-	}
-}
-
-// StrictDomainName limits the set of permissible ASCII characters to those
-// allowed in domain names as defined in RFC 1034 (A-Z, a-z, 0-9 and the
-// hyphen). This is set by default for MapForLookup and ValidateForRegistration,
-// but is only useful if ValidateLabels is set.
-//
-// This option is useful, for instance, for browsers that allow characters
-// outside this range, for example a '_' (U+005F LOW LINE). See
-// http://www.rfc-editor.org/std/std3.txt for more details.
-//
-// This option corresponds to the UseSTD3ASCIIRules flag in UTS #46.
-func StrictDomainName(use bool) Option {
-	return func(o *options) { o.useSTD3Rules = use }
-}
-
-// NOTE: the following options pull in tables. The tables should not be linked
-// in as long as the options are not used.
-
-// BidiRule enables the Bidi rule as defined in RFC 5893. Any application
-// that relies on proper validation of labels should include this rule.
-//
-// This option corresponds to the CheckBidi flag in UTS #46.
-func BidiRule() Option {
-	return func(o *options) { o.bidirule = bidirule.ValidString }
-}
-
-// ValidateForRegistration sets validation options to verify that a given IDN is
-// properly formatted for registration as defined by Section 4 of RFC 5891.
-func ValidateForRegistration() Option {
-	return func(o *options) {
-		o.mapping = validateRegistration
-		StrictDomainName(true)(o)
-		ValidateLabels(true)(o)
-		VerifyDNSLength(true)(o)
-		BidiRule()(o)
-	}
-}
-
-// MapForLookup sets validation and mapping options such that a given IDN is
-// transformed for domain name lookup according to the requirements set out in
-// Section 5 of RFC 5891. The mappings follow the recommendations of RFC 5894,
-// RFC 5895 and UTS 46. It does not add the Bidi Rule. Use the BidiRule option
-// to add this check.
-//
-// The mappings include normalization and mapping case, width and other
-// compatibility mappings.
-func MapForLookup() Option {
-	return func(o *options) {
-		o.mapping = validateAndMap
-		StrictDomainName(true)(o)
-		ValidateLabels(true)(o)
-		RemoveLeadingDots(true)(o)
-	}
-}
-
-type options struct {
-	transitional      bool
-	useSTD3Rules      bool
-	checkHyphens      bool
-	checkJoiners      bool
-	verifyDNSLength   bool
-	removeLeadingDots bool
-
-	trie *idnaTrie
-
-	// fromPuny calls validation rules when converting A-labels to U-labels.
-	fromPuny func(p *Profile, s string) error
-
-	// mapping implements a validation and mapping step as defined in RFC 5895
-	// or UTS 46, tailored to, for example, domain registration or lookup.
-	mapping func(p *Profile, s string) (string, error)
-
-	// bidirule, if specified, checks whether s conforms to the Bidi Rule
-	// defined in RFC 5893.
-	bidirule func(s string) bool
-}
-
-// A Profile defines the configuration of a IDNA mapper.
-type Profile struct {
-	options
-}
-
-func apply(o *options, opts []Option) {
-	for _, f := range opts {
-		f(o)
-	}
-}
-
-// New creates a new Profile.
-//
-// With no options, the returned Profile is the most permissive and equals the
-// Punycode Profile. Options can be passed to further restrict the Profile. The
-// MapForLookup and ValidateForRegistration options set a collection of options,
-// for lookup and registration purposes respectively, which can be tailored by
-// adding more fine-grained options, where later options override earlier
-// options.
-func New(o ...Option) *Profile {
-	p := &Profile{}
-	apply(&p.options, o)
-	return p
-}
-
-// ToASCII converts a domain or domain label to its ASCII form. For example,
-// ToASCII("bücher.example.com") is "xn--bcher-kva.example.com", and
-// ToASCII("golang") is "golang". If an error is encountered it will return
-// an error and a (partially) processed result.
-func (p *Profile) ToASCII(s string) (string, error) {
-	return p.process(s, true)
-}
-
-// ToUnicode converts a domain or domain label to its Unicode form. For example,
-// ToUnicode("xn--bcher-kva.example.com") is "bücher.example.com", and
-// ToUnicode("golang") is "golang". If an error is encountered it will return
-// an error and a (partially) processed result.
-func (p *Profile) ToUnicode(s string) (string, error) {
-	pp := *p
-	pp.transitional = false
-	return pp.process(s, false)
-}
-
-// String reports a string with a description of the profile for debugging
-// purposes. The string format may change with different versions.
-func (p *Profile) String() string {
-	s := ""
-	if p.transitional {
-		s = "Transitional"
-	} else {
-		s = "NonTransitional"
-	}
-	if p.useSTD3Rules {
-		s += ":UseSTD3Rules"
-	}
-	if p.checkHyphens {
-		s += ":CheckHyphens"
-	}
-	if p.checkJoiners {
-		s += ":CheckJoiners"
-	}
-	if p.verifyDNSLength {
-		s += ":VerifyDNSLength"
-	}
-	return s
-}
-
-var (
-	// Punycode is a Profile that does raw punycode processing with a minimum
-	// of validation.
-	Punycode *Profile = punycode
-
-	// Lookup is the recommended profile for looking up domain names, according
-	// to Section 5 of RFC 5891. The exact configuration of this profile may
-	// change over time.
-	Lookup *Profile = lookup
-
-	// Display is the recommended profile for displaying domain names.
-	// The configuration of this profile may change over time.
-	Display *Profile = display
-
-	// Registration is the recommended profile for checking whether a given
-	// IDN is valid for registration, according to Section 4 of RFC 5891.
-	Registration *Profile = registration
-
-	punycode = &Profile{}
-	lookup   = &Profile{options{
-		transitional:      true,
-		removeLeadingDots: true,
-		useSTD3Rules:      true,
-		checkHyphens:      true,
-		checkJoiners:      true,
-		trie:              trie,
-		fromPuny:          validateFromPunycode,
-		mapping:           validateAndMap,
-		bidirule:          bidirule.ValidString,
-	}}
-	display = &Profile{options{
-		useSTD3Rules:      true,
-		removeLeadingDots: true,
-		checkHyphens:      true,
-		checkJoiners:      true,
-		trie:              trie,
-		fromPuny:          validateFromPunycode,
-		mapping:           validateAndMap,
-		bidirule:          bidirule.ValidString,
-	}}
-	registration = &Profile{options{
-		useSTD3Rules:    true,
-		verifyDNSLength: true,
-		checkHyphens:    true,
-		checkJoiners:    true,
-		trie:            trie,
-		fromPuny:        validateFromPunycode,
-		mapping:         validateRegistration,
-		bidirule:        bidirule.ValidString,
-	}}
-
-	// TODO: profiles
-	// Register: recommended for approving domain names: don't do any mappings
-	// but rather reject on invalid input. Bundle or block deviation characters.
-)
-
-type labelError struct{ label, code_ string }
-
-func (e labelError) code() string { return e.code_ }
-func (e labelError) Error() string {
-	return fmt.Sprintf("idna: invalid label %q", e.label)
-}
-
-type runeError rune
-
-func (e runeError) code() string { return "P1" }
-func (e runeError) Error() string {
-	return fmt.Sprintf("idna: disallowed rune %U", e)
-}
-
-// process implements the algorithm described in section 4 of UTS #46,
-// see https://www.unicode.org/reports/tr46.
-func (p *Profile) process(s string, toASCII bool) (string, error) {
-	var err error
-	if p.mapping != nil {
-		s, err = p.mapping(p, s)
-	}
-	// Remove leading empty labels.
-	if p.removeLeadingDots {
-		for ; len(s) > 0 && s[0] == '.'; s = s[1:] {
-		}
-	}
-	// It seems like we should only create this error on ToASCII, but the
-	// UTS 46 conformance tests suggests we should always check this.
-	if err == nil && p.verifyDNSLength && s == "" {
-		err = &labelError{s, "A4"}
-	}
-	labels := labelIter{orig: s}
-	for ; !labels.done(); labels.next() {
-		label := labels.label()
-		if label == "" {
-			// Empty labels are not okay. The label iterator skips the last
-			// label if it is empty.
-			if err == nil && p.verifyDNSLength {
-				err = &labelError{s, "A4"}
-			}
-			continue
-		}
-		if strings.HasPrefix(label, acePrefix) {
-			u, err2 := decode(label[len(acePrefix):])
-			if err2 != nil {
-				if err == nil {
-					err = err2
-				}
-				// Spec says keep the old label.
-				continue
-			}
-			labels.set(u)
-			if err == nil && p.fromPuny != nil {
-				err = p.fromPuny(p, u)
-			}
-			if err == nil {
-				// This should be called on NonTransitional, according to the
-				// spec, but that currently does not have any effect. Use the
-				// original profile to preserve options.
-				err = p.validateLabel(u)
-			}
-		} else if err == nil {
-			err = p.validateLabel(label)
-		}
-	}
-	if toASCII {
-		for labels.reset(); !labels.done(); labels.next() {
-			label := labels.label()
-			if !ascii(label) {
-				a, err2 := encode(acePrefix, label)
-				if err == nil {
-					err = err2
-				}
-				label = a
-				labels.set(a)
-			}
-			n := len(label)
-			if p.verifyDNSLength && err == nil && (n == 0 || n > 63) {
-				err = &labelError{label, "A4"}
-			}
-		}
-	}
-	s = labels.result()
-	if toASCII && p.verifyDNSLength && err == nil {
-		// Compute the length of the domain name minus the root label and its dot.
-		n := len(s)
-		if n > 0 && s[n-1] == '.' {
-			n--
-		}
-		if len(s) < 1 || n > 253 {
-			err = &labelError{s, "A4"}
-		}
-	}
-	return s, err
-}
-
-func normalize(p *Profile, s string) (string, error) {
-	return norm.NFC.String(s), nil
-}
-
-func validateRegistration(p *Profile, s string) (string, error) {
-	if !norm.NFC.IsNormalString(s) {
-		return s, &labelError{s, "V1"}
-	}
-	for i := 0; i < len(s); {
-		v, sz := trie.lookupString(s[i:])
-		// Copy bytes not copied so far.
-		switch p.simplify(info(v).category()) {
-		// TODO: handle the NV8 defined in the Unicode idna data set to allow
-		// for strict conformance to IDNA2008.
-		case valid, deviation:
-		case disallowed, mapped, unknown, ignored:
-			r, _ := utf8.DecodeRuneInString(s[i:])
-			return s, runeError(r)
-		}
-		i += sz
-	}
-	return s, nil
-}
-
-func validateAndMap(p *Profile, s string) (string, error) {
-	var (
-		err error
-		b   []byte
-		k   int
-	)
-	for i := 0; i < len(s); {
-		v, sz := trie.lookupString(s[i:])
-		start := i
-		i += sz
-		// Copy bytes not copied so far.
-		switch p.simplify(info(v).category()) {
-		case valid:
-			continue
-		case disallowed:
-			if err == nil {
-				r, _ := utf8.DecodeRuneInString(s[start:])
-				err = runeError(r)
-			}
-			continue
-		case mapped, deviation:
-			b = append(b, s[k:start]...)
-			b = info(v).appendMapping(b, s[start:i])
-		case ignored:
-			b = append(b, s[k:start]...)
-			// drop the rune
-		case unknown:
-			b = append(b, s[k:start]...)
-			b = append(b, "\ufffd"...)
-		}
-		k = i
-	}
-	if k == 0 {
-		// No changes so far.
-		s = norm.NFC.String(s)
-	} else {
-		b = append(b, s[k:]...)
-		if norm.NFC.QuickSpan(b) != len(b) {
-			b = norm.NFC.Bytes(b)
-		}
-		// TODO: the punycode converters require strings as input.
-		s = string(b)
-	}
-	return s, err
-}
-
-// A labelIter allows iterating over domain name labels.
-type labelIter struct {
-	orig     string
-	slice    []string
-	curStart int
-	curEnd   int
-	i        int
-}
-
-func (l *labelIter) reset() {
-	l.curStart = 0
-	l.curEnd = 0
-	l.i = 0
-}
-
-func (l *labelIter) done() bool {
-	return l.curStart >= len(l.orig)
-}
-
-func (l *labelIter) result() string {
-	if l.slice != nil {
-		return strings.Join(l.slice, ".")
-	}
-	return l.orig
-}
-
-func (l *labelIter) label() string {
-	if l.slice != nil {
-		return l.slice[l.i]
-	}
-	p := strings.IndexByte(l.orig[l.curStart:], '.')
-	l.curEnd = l.curStart + p
-	if p == -1 {
-		l.curEnd = len(l.orig)
-	}
-	return l.orig[l.curStart:l.curEnd]
-}
-
-// next sets the value to the next label. It skips the last label if it is empty.
-func (l *labelIter) next() {
-	l.i++
-	if l.slice != nil {
-		if l.i >= len(l.slice) || l.i == len(l.slice)-1 && l.slice[l.i] == "" {
-			l.curStart = len(l.orig)
-		}
-	} else {
-		l.curStart = l.curEnd + 1
-		if l.curStart == len(l.orig)-1 && l.orig[l.curStart] == '.' {
-			l.curStart = len(l.orig)
-		}
-	}
-}
-
-func (l *labelIter) set(s string) {
-	if l.slice == nil {
-		l.slice = strings.Split(l.orig, ".")
-	}
-	l.slice[l.i] = s
-}
-
-// acePrefix is the ASCII Compatible Encoding prefix.
-const acePrefix = "xn--"
-
-func (p *Profile) simplify(cat category) category {
-	switch cat {
-	case disallowedSTD3Mapped:
-		if p.useSTD3Rules {
-			cat = disallowed
-		} else {
-			cat = mapped
-		}
-	case disallowedSTD3Valid:
-		if p.useSTD3Rules {
-			cat = disallowed
-		} else {
-			cat = valid
-		}
-	case deviation:
-		if !p.transitional {
-			cat = valid
-		}
-	case validNV8, validXV8:
-		// TODO: handle V2008
-		cat = valid
-	}
-	return cat
-}
-
-func validateFromPunycode(p *Profile, s string) error {
-	if !norm.NFC.IsNormalString(s) {
-		return &labelError{s, "V1"}
-	}
-	for i := 0; i < len(s); {
-		v, sz := trie.lookupString(s[i:])
-		if c := p.simplify(info(v).category()); c != valid && c != deviation {
-			return &labelError{s, "V6"}
-		}
-		i += sz
-	}
-	return nil
-}
-
-const (
-	zwnj = "\u200c"
-	zwj  = "\u200d"
-)
-
-type joinState int8
-
-const (
-	stateStart joinState = iota
-	stateVirama
-	stateBefore
-	stateBeforeVirama
-	stateAfter
-	stateFAIL
-)
-
-var joinStates = [][numJoinTypes]joinState{
-	stateStart: {
-		joiningL:   stateBefore,
-		joiningD:   stateBefore,
-		joinZWNJ:   stateFAIL,
-		joinZWJ:    stateFAIL,
-		joinVirama: stateVirama,
-	},
-	stateVirama: {
-		joiningL: stateBefore,
-		joiningD: stateBefore,
-	},
-	stateBefore: {
-		joiningL:   stateBefore,
-		joiningD:   stateBefore,
-		joiningT:   stateBefore,
-		joinZWNJ:   stateAfter,
-		joinZWJ:    stateFAIL,
-		joinVirama: stateBeforeVirama,
-	},
-	stateBeforeVirama: {
-		joiningL: stateBefore,
-		joiningD: stateBefore,
-		joiningT: stateBefore,
-	},
-	stateAfter: {
-		joiningL:   stateFAIL,
-		joiningD:   stateBefore,
-		joiningT:   stateAfter,
-		joiningR:   stateStart,
-		joinZWNJ:   stateFAIL,
-		joinZWJ:    stateFAIL,
-		joinVirama: stateAfter, // no-op as we can't accept joiners here
-	},
-	stateFAIL: {
-		0:          stateFAIL,
-		joiningL:   stateFAIL,
-		joiningD:   stateFAIL,
-		joiningT:   stateFAIL,
-		joiningR:   stateFAIL,
-		joinZWNJ:   stateFAIL,
-		joinZWJ:    stateFAIL,
-		joinVirama: stateFAIL,
-	},
-}
-
-// validateLabel validates the criteria from Section 4.1. Item 1, 4, and 6 are
-// already implicitly satisfied by the overall implementation.
-func (p *Profile) validateLabel(s string) error {
-	if s == "" {
-		if p.verifyDNSLength {
-			return &labelError{s, "A4"}
-		}
-		return nil
-	}
-	if p.bidirule != nil && !p.bidirule(s) {
-		return &labelError{s, "B"}
-	}
-	if p.checkHyphens {
-		if len(s) > 4 && s[2] == '-' && s[3] == '-' {
-			return &labelError{s, "V2"}
-		}
-		if s[0] == '-' || s[len(s)-1] == '-' {
-			return &labelError{s, "V3"}
-		}
-	}
-	if !p.checkJoiners {
-		return nil
-	}
-	trie := p.trie // p.checkJoiners is only set if trie is set.
-	// TODO: merge the use of this in the trie.
-	v, sz := trie.lookupString(s)
-	x := info(v)
-	if x.isModifier() {
-		return &labelError{s, "V5"}
-	}
-	// Quickly return in the absence of zero-width (non) joiners.
-	if strings.Index(s, zwj) == -1 && strings.Index(s, zwnj) == -1 {
-		return nil
-	}
-	st := stateStart
-	for i := 0; ; {
-		jt := x.joinType()
-		if s[i:i+sz] == zwj {
-			jt = joinZWJ
-		} else if s[i:i+sz] == zwnj {
-			jt = joinZWNJ
-		}
-		st = joinStates[st][jt]
-		if x.isViramaModifier() {
-			st = joinStates[st][joinVirama]
-		}
-		if i += sz; i == len(s) {
-			break
-		}
-		v, sz = trie.lookupString(s[i:])
-		x = info(v)
-	}
-	if st == stateFAIL || st == stateAfter {
-		return &labelError{s, "C"}
-	}
-	return nil
-}
-
-func ascii(s string) bool {
-	for i := 0; i < len(s); i++ {
-		if s[i] >= utf8.RuneSelf {
-			return false
-		}
-	}
-	return true
-}

vendor/golang.org/x/net/idna/pre_go118.go

@@ -1,11 +0,0 @@
-// Code generated by running "go generate" in golang.org/x/text. DO NOT EDIT.
-
-// Copyright 2021 The Go Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-//go:build !go1.18
-
-package idna
-
-const transitionalLookup = true

vendor/golang.org/x/net/idna/punycode.go

@@ -28,7 +28,7 @@ const (
 	tmin        int32 = 1
 )
 
-func punyError(s string) error { return &labelError{s, "A3"} }
+func punyError(s string) error { return &labelError{s, code16("A3", "P4")} }
 
 // decode decodes a string as specified in section 6.2.
 func decode(encoded string) (string, error) {
@@ -108,6 +108,9 @@ func encode(prefix, s string) (string, error) {
 	delta, n, bias := int32(0), initialN, initialBias
 	b, remaining := int32(0), int32(0)
 	for _, r := range s {
+		if unicode16 && r == 0xfffd {
+			return s, &labelError{s, "A3"}
+		}
 		if r < 0x80 {
 			b++
 			output = append(output, byte(r))