vmselect: use NewRequest to create cache reset propagate request

Co-authored-by: cubic-dev-ai[bot] <191113872+cubic-dev-ai[bot]@users.noreply.github.com>
Signed-off-by: Zhu Jiekun <jiekun@victoriametrics.com>

app/vmagent/remotewrite/client.go

@@ -187,7 +187,7 @@ func newHTTPClient(argIdx int, remoteWriteURL, sanitizedURL string, fq *persiste
 	return c
 }
 
-func (c *client) init(argIdx int, sanitizedURL string) {
+func (c *client) init(argIdx, concurrency int, sanitizedURL string) {
 	limitReached := metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_rate_limit_reached_total{url=%q}`, c.sanitizedURL))
 	if bytesPerSec := rateLimit.GetOptionalArg(argIdx); bytesPerSec > 0 {
 		logger.Infof("applying %d bytes per second rate limit for -remoteWrite.url=%q", bytesPerSec, sanitizedURL)
@@ -204,20 +204,11 @@ func (c *client) init(argIdx int, sanitizedURL string) {
 	c.packetsDropped = metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_packets_dropped_total{url=%q}`, c.sanitizedURL))
 	c.retriesCount = metrics.GetOrCreateCounter(fmt.Sprintf(`vmagent_remotewrite_retries_count_total{url=%q}`, c.sanitizedURL))
 	c.sendDuration = metrics.GetOrCreateFloatCounter(fmt.Sprintf(`vmagent_remotewrite_send_duration_seconds_total{url=%q}`, c.sanitizedURL))
-	workers := queues.GetOptionalArg(argIdx)
-	if workers <= 0 {
-		workers = 1
-	}
-	inmemoryWorkers := inmemoryQueues.GetOptionalArg(argIdx)
-	for range inmemoryWorkers {
-		c.wg.Go(func() {
-			c.runWorker(c.fq.MustReadInMemoryBlockBlocking)
-		})
-	}
-	for range workers {
-		c.wg.Go(func() {
-			c.runWorker(c.fq.MustReadBlock)
-		})
+	metrics.GetOrCreateGauge(fmt.Sprintf(`vmagent_remotewrite_queues{url=%q}`, c.sanitizedURL), func() float64 {
+		return float64(concurrency)
+	})
+	for range concurrency {
+		c.wg.Go(c.runWorker)
 	}
 	logger.Infof("initialized client for -remoteWrite.url=%q", c.sanitizedURL)
 }
@@ -311,12 +302,12 @@ func getAWSAPIConfig(argIdx int) (*awsapi.Config, error) {
 	return cfg, nil
 }
 
-func (c *client) runWorker(readBlock func(dst []byte) ([]byte, bool)) {
+func (c *client) runWorker() {
 	var ok bool
 	var block []byte
 	ch := make(chan bool, 1)
 	for {
-		block, ok = readBlock(block[:0])
+		block, ok = c.fq.MustReadBlock(block[:0])
 		if !ok {
 			return
 		}

app/vmagent/remotewrite/pendingseries.go

@@ -209,12 +209,13 @@ func (wr *writeRequest) tryPushMetadata(mms []prompb.MetricMetadata) bool {
 func (wr *writeRequest) copyMetadata(dst, src *prompb.MetricMetadata) {
 	// Direct copy for non-string fields, which are safe by value.
 	dst.Type = src.Type
+	dst.Unit = src.Unit
 
 	dst.AccountID = src.AccountID
 	dst.ProjectID = src.ProjectID
 
 	// Pre-allocate memory for all string fields.
-	neededBufLen := len(src.MetricFamilyName) + len(src.Help) + len(src.Unit)
+	neededBufLen := len(src.MetricFamilyName) + len(src.Help)
 	bufLen := len(wr.metadatabuf)
 	wr.metadatabuf = slicesutil.SetLength(wr.metadatabuf, bufLen+neededBufLen)
 	buf := wr.metadatabuf[:bufLen]
@@ -229,11 +230,6 @@ func (wr *writeRequest) copyMetadata(dst, src *prompb.MetricMetadata) {
 	buf = append(buf, src.Help...)
 	dst.Help = bytesutil.ToUnsafeString(buf[bufLen:])
 
-	// Copy Unit
-	bufLen = len(buf)
-	buf = append(buf, src.Unit...)
-	dst.Unit = bytesutil.ToUnsafeString(buf[bufLen:])
-
 	wr.metadatabuf = buf
 }
 

app/vmagent/remotewrite/remotewrite.go

@@ -66,9 +66,6 @@ var (
 	queues = flagutil.NewArrayInt("remoteWrite.queues", cgroup.AvailableCPUs()*2, "The number of concurrent queues to each -remoteWrite.url. Set more queues if default number of queues "+
 		"isn't enough for sending high volume of collected data to remote storage. "+
 		"Default value depends on the number of available CPU cores. It should work fine in most cases since it minimizes resource usage")
-	inmemoryQueues = flagutil.NewArrayInt("remoteWrite.inmemoryQueues", 0, "The number of additional workers per each -remoteWrite.url, which send only recently ingested data from the in-memory queue, "+
-		"while the file-based queue at -remoteWrite.tmpDataPath is drained by workers configured via -remoteWrite.queues. "+
-		"This reduces delivery lag for fresh samples when the file-based queue contains a backlog accumulated during remote storage outages.")
 	showRemoteWriteURL = flag.Bool("remoteWrite.showURL", false, "Whether to show -remoteWrite.url in the exported metrics. "+
 		"It is hidden by default, since it can contain sensitive info such as auth key")
 	maxPendingBytesPerURL = flagutil.NewArrayBytes("remoteWrite.maxDiskUsagePerURL", 0, "The maximum file-based buffer size in bytes at -remoteWrite.tmpDataPath "+
@@ -909,8 +906,7 @@ func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, sanitizedURL string)
 	}
 
 	isPQDisabled := disableOnDiskQueue.GetOptionalArg(argIdx)
-	inmemoryQueueSize := inmemoryQueues.GetOptionalArg(argIdx)
-	queuesSize := queues.GetOptionalArg(argIdx) + inmemoryQueueSize
+	queuesSize := queues.GetOptionalArg(argIdx)
 	if queuesSize > maxQueues {
 		queuesSize = maxQueues
 	} else if queuesSize <= 0 {
@@ -927,13 +923,7 @@ func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, sanitizedURL string)
 	if maxInmemoryBlocks < 2 {
 		maxInmemoryBlocks = 2
 	}
-	fqOpts := persistentqueue.OpenFastQueueOpts{
-		MaxInmemoryBlocks:      maxInmemoryBlocks,
-		MaxPendingBytes:        maxPendingBytes,
-		IsPQDisabled:           isPQDisabled,
-		PrioritizeInmemoryData: inmemoryQueueSize > 0,
-	}
-	fq := persistentqueue.MustOpenFastQueueWithOpts(queuePath, sanitizedURL, fqOpts)
+	fq := persistentqueue.MustOpenFastQueue(queuePath, sanitizedURL, maxInmemoryBlocks, maxPendingBytes, isPQDisabled)
 	_ = metrics.GetOrCreateGauge(fmt.Sprintf(`vmagent_remotewrite_pending_data_bytes{path=%q, url=%q}`, queuePath, sanitizedURL), func() float64 {
 		return float64(fq.GetPendingBytes())
 	})
@@ -946,9 +936,6 @@ func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, sanitizedURL string)
 		}
 		return 0
 	})
-	metrics.GetOrCreateGauge(fmt.Sprintf(`vmagent_remotewrite_queues{url=%q}`, sanitizedURL), func() float64 {
-		return float64(queuesSize)
-	})
 
 	var c *client
 	switch remoteWriteURL.Scheme {
@@ -957,7 +944,7 @@ func newRemoteWriteCtx(argIdx int, remoteWriteURL *url.URL, sanitizedURL string)
 	default:
 		logger.Fatalf("unsupported scheme: %s for remoteWriteURL: %s, want `http`, `https`", remoteWriteURL.Scheme, sanitizedURL)
 	}
-	c.init(argIdx, sanitizedURL)
+	c.init(argIdx, queuesSize, sanitizedURL)
 
 	// Initialize pss
 	sf := significantFigures.GetOptionalArg(argIdx)

app/vmauth/auth_config.go

@@ -911,8 +911,7 @@ func reloadAuthConfigData(data []byte) (bool, error) {
 		return false, fmt.Errorf("failed to parse auth config: %w", err)
 	}
 
-	oidcDP := &oidcDiscovererPool{}
-	jui, err := parseJWTUsers(ac, oidcDP)
+	jui, oidcDP, err := parseJWTUsers(ac)
 	if err != nil {
 		return false, fmt.Errorf("failed to parse JWT users from auth config: %w", err)
 	}

app/vmauth/jwt.go

@@ -72,8 +72,9 @@ type JWTConfig struct {
 	verifierPool atomic.Pointer[jwt.VerifierPool]
 }
 
-func parseJWTUsers(ac *AuthConfig, oidcDP *oidcDiscovererPool) ([]*UserInfo, error) {
+func parseJWTUsers(ac *AuthConfig) ([]*UserInfo, *oidcDiscovererPool, error) {
 	jui := make([]*UserInfo, 0, len(ac.Users))
+	oidcDP := &oidcDiscovererPool{}
 
 	uniqClaims := make(map[string]*UserInfo)
 	var sortedClaims []string
@@ -84,10 +85,10 @@ func parseJWTUsers(ac *AuthConfig, oidcDP *oidcDiscovererPool) ([]*UserInfo, err
 		}
 
 		if ui.AuthToken != "" || ui.BearerToken != "" || ui.Username != "" || ui.Password != "" {
-			return nil, fmt.Errorf("auth_token, bearer_token, username and password cannot be specified if jwt is set")
+			return nil, nil, fmt.Errorf("auth_token, bearer_token, username and password cannot be specified if jwt is set")
 		}
 		if len(jwtToken.PublicKeys) == 0 && len(jwtToken.PublicKeyFiles) == 0 && !jwtToken.SkipVerify && jwtToken.OIDC == nil {
-			return nil, fmt.Errorf("jwt must contain at least a single public key, public_key_files, oidc or have skip_verify=true")
+			return nil, nil, fmt.Errorf("jwt must contain at least a single public key, public_key_files, oidc or have skip_verify=true")
 		}
 		var claimsString string
 		sortedClaims = sortedClaims[:0]
@@ -96,7 +97,7 @@ func parseJWTUsers(ac *AuthConfig, oidcDP *oidcDiscovererPool) ([]*UserInfo, err
 			sortedClaims = append(sortedClaims, fmt.Sprintf("%s=%s", ck, cv))
 			pc, err := jwt.NewClaim(ck, cv)
 			if err != nil {
-				return nil, fmt.Errorf("incorrect match claim, key=%q, value regex=%q: %w", ck, cv, err)
+				return nil, nil, fmt.Errorf("incorrect match claim, key=%q, value regex=%q: %w", ck, cv, err)
 			}
 			parsedClaims = append(parsedClaims, pc)
 		}
@@ -105,7 +106,7 @@ func parseJWTUsers(ac *AuthConfig, oidcDP *oidcDiscovererPool) ([]*UserInfo, err
 		claimsString = strings.Join(sortedClaims, ",")
 
 		if oldUI, ok := uniqClaims[claimsString]; ok {
-			return nil, fmt.Errorf("duplicate match claims=%q found for name=%q at idx=%d; the previous one is set for name=%q", claimsString, ui.Name, idx, oldUI.Name)
+			return nil, nil, fmt.Errorf("duplicate match claims=%q found for name=%q at idx=%d; the previous one is set for name=%q", claimsString, ui.Name, idx, oldUI.Name)
 		}
 		uniqClaims[claimsString] = &ui
 		if len(jwtToken.PublicKeys) > 0 || len(jwtToken.PublicKeyFiles) > 0 {
@@ -114,7 +115,7 @@ func parseJWTUsers(ac *AuthConfig, oidcDP *oidcDiscovererPool) ([]*UserInfo, err
 			for i := range jwtToken.PublicKeys {
 				k, err := jwt.ParseKey([]byte(jwtToken.PublicKeys[i]))
 				if err != nil {
-					return nil, err
+					return nil, nil, err
 				}
 				keys = append(keys, k)
 			}
@@ -122,52 +123,52 @@ func parseJWTUsers(ac *AuthConfig, oidcDP *oidcDiscovererPool) ([]*UserInfo, err
 			for _, filePath := range jwtToken.PublicKeyFiles {
 				keyData, err := os.ReadFile(filePath)
 				if err != nil {
-					return nil, fmt.Errorf("cannot read public key from file %q: %w", filePath, err)
+					return nil, nil, fmt.Errorf("cannot read public key from file %q: %w", filePath, err)
 				}
 				k, err := jwt.ParseKey(keyData)
 				if err != nil {
-					return nil, fmt.Errorf("cannot parse public key from file %q: %w", filePath, err)
+					return nil, nil, fmt.Errorf("cannot parse public key from file %q: %w", filePath, err)
 				}
 				keys = append(keys, k)
 			}
 
 			vp, err := jwt.NewVerifierPool(keys)
 			if err != nil {
-				return nil, err
+				return nil, nil, err
 			}
 
 			jwtToken.verifierPool.Store(vp)
 		}
 		if jwtToken.OIDC != nil {
 			if len(jwtToken.PublicKeys) > 0 || len(jwtToken.PublicKeyFiles) > 0 || jwtToken.SkipVerify {
-				return nil, fmt.Errorf("jwt with oidc cannot contain public keys or have skip_verify=true")
+				return nil, nil, fmt.Errorf("jwt with oidc cannot contain public keys or have skip_verify=true")
 			}
 
 			if jwtToken.OIDC.Issuer == "" {
-				return nil, fmt.Errorf("oidc issuer cannot be empty")
+				return nil, nil, fmt.Errorf("oidc issuer cannot be empty")
 			}
 			isserURL, err := url.Parse(jwtToken.OIDC.Issuer)
 			if err != nil {
-				return nil, fmt.Errorf("oidc issuer %q must be a valid URL", jwtToken.OIDC.Issuer)
+				return nil, nil, fmt.Errorf("oidc issuer %q must be a valid URL", jwtToken.OIDC.Issuer)
 			}
 			if isserURL.Scheme != "https" && isserURL.Scheme != "http" {
-				return nil, fmt.Errorf("oidc issuer %q must have http or https scheme", jwtToken.OIDC.Issuer)
+				return nil, nil, fmt.Errorf("oidc issuer %q must have http or https scheme", jwtToken.OIDC.Issuer)
 			}
 
 			oidcDP.createOrAdd(ui.JWT.OIDC.Issuer, &ui.JWT.verifierPool)
 		}
 
 		if err := parseJWTPlaceholdersForUserInfo(&ui, true); err != nil {
-			return nil, err
+			return nil, nil, err
 		}
 
 		if err := ui.initURLs(); err != nil {
-			return nil, err
+			return nil, nil, err
 		}
 
 		metricLabels, err := ui.getMetricLabels()
 		if err != nil {
-			return nil, fmt.Errorf("cannot parse metric_labels: %w", err)
+			return nil, nil, fmt.Errorf("cannot parse metric_labels: %w", err)
 		}
 		ui.requests = ac.ms.GetOrCreateCounter(`vmauth_user_requests_total` + metricLabels)
 		ui.requestErrors = ac.ms.GetOrCreateCounter(`vmauth_user_request_errors_total` + metricLabels)
@@ -186,7 +187,7 @@ func parseJWTUsers(ac *AuthConfig, oidcDP *oidcDiscovererPool) ([]*UserInfo, err
 
 		rt, err := newRoundTripper(ui.TLSCAFile, ui.TLSCertFile, ui.TLSKeyFile, ui.TLSServerName, ui.TLSInsecureSkipVerify)
 		if err != nil {
-			return nil, fmt.Errorf("cannot initialize HTTP RoundTripper: %w", err)
+			return nil, nil, fmt.Errorf("cannot initialize HTTP RoundTripper: %w", err)
 		}
 		ui.rt = rt
 
@@ -199,7 +200,7 @@ func parseJWTUsers(ac *AuthConfig, oidcDP *oidcDiscovererPool) ([]*UserInfo, err
 		return len(jui[i].JWT.MatchClaims) > len(jui[j].JWT.MatchClaims)
 	})
 
-	return jui, nil
+	return jui, oidcDP, nil
 }
 
 var tokenPool sync.Pool

app/vmauth/jwt_test.go

@@ -39,14 +39,16 @@ XOtclIk1uhc03oL9nOQ=
 			}
 			return
 		}
-		oidcDP := &oidcDiscovererPool{}
-		users, err := parseJWTUsers(ac, oidcDP)
+		users, oidcDP, err := parseJWTUsers(ac)
 		if err == nil {
 			t.Fatalf("expecting non-nil error; got %v", users)
 		}
 		if expErr != err.Error() {
 			t.Fatalf("unexpected error; got\n%q\nwant \n%q", err.Error(), expErr)
 		}
+		if oidcDP != nil {
+			t.Fatalf("expecting nil oidcDP; got %v", oidcDP)
+		}
 	}
 
 	// unauthorized_user cannot be used with jwt
@@ -324,8 +326,7 @@ XOtclIk1uhc03oL9nOQ=
 			t.Fatalf("unexpected error: %s", err)
 		}
 
-		oidcDP := &oidcDiscovererPool{}
-		jui, err := parseJWTUsers(ac, oidcDP)
+		jui, oidcDP, err := parseJWTUsers(ac)
 		if err != nil {
 			t.Fatalf("unexpected error: %s", err)
 		}

app/vmctl/influx/influx.go

@@ -259,7 +259,7 @@ func (cr *ChunkedResponse) Next() ([]int64, []float64, error) {
 
 	fieldValues, ok := r.values[cr.field]
 	if !ok {
-		return nil, nil, fmt.Errorf("response doesn't contain field %q", cr.field)
+		return nil, nil, fmt.Errorf("response doesn't contain filed %q", cr.field)
 	}
 	values := make([]float64, len(fieldValues))
 	for i, fv := range fieldValues {

app/vmctl/main.go

@@ -563,11 +563,11 @@ func main() {
 	}()
 
 	err = app.Run(os.Args)
-	pushmetrics.StopAndPush()
 	if err != nil {
 		log.Fatalln(err)
 	}
 	log.Printf("Total time: %v", time.Since(start))
+	pushmetrics.StopAndPush()
 }
 
 func initConfigVM(c *cli.Context) (vm.Config, error) {

app/vmctl/vm_native.go

@@ -405,16 +405,7 @@ func buildMatchWithFilter(filter string, metricName string) (string, error) {
 			if len(tf.Key) == 0 {
 				continue
 			}
-			switch {
-			case tf.IsNegative && tf.IsRegexp:
-				a = append(a, fmt.Sprintf("%s!~%q", tf.Key, tf.Value))
-			case tf.IsNegative:
-				a = append(a, fmt.Sprintf("%s!=%q", tf.Key, tf.Value))
-			case tf.IsRegexp:
-				a = append(a, fmt.Sprintf("%s=~%q", tf.Key, tf.Value))
-			default:
-				a = append(a, fmt.Sprintf("%s=%q", tf.Key, tf.Value))
-			}
+			a = append(a, tf.String())
 		}
 		a = append(a, nameFilter)
 		filters = append(filters, strings.Join(a, ","))

app/vmselect/main.go

@@ -56,7 +56,6 @@ var (
 	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 https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#deduplication for details")
 	deleteAuthKey                = flagutil.NewPassword("deleteAuthKey", "authKey for metrics' deletion via /prometheus/api/v1/admin/tsdb/delete_series and /graphite/tags/delSeries. It could be passed via authKey query arg. It overrides -httpAuth.*")
-	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.*")
 	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")
 
@@ -260,11 +259,10 @@ func requestHandler(w http.ResponseWriter, r *http.Request) bool {
 	}
 
 	if path == "/internal/resetRollupResultCache" {
-		if !httpserver.CheckAuthFlag(w, r, resetCacheAuthKey) {
+		if !httpserver.CheckAuthFlag(w, r, prometheus.GetResetCacheAuthKey()) {
 			return true
 		}
-		promql.ResetRollupResultCache()
-		return true
+		return prometheus.ResetRollupResultCacheHandler(w, r)
 	}
 	if path == "/admin/tenants" {
 		tenantsRequests.Inc()

app/vmselect/netstorage/tenant_cache.go

@@ -47,10 +47,6 @@ func TenantsCached(qt *querytracer.Tracer, tr storage.TimeRange, deadline search
 	}
 
 	qtL.Printf("fetched %d tenants from storage", len(tenants))
-	if len(tenants) == 0 {
-		qtL.Printf("no tenants found - don't cache this fetch")
-		return nil, nil
-	}
 
 	tt := make([]storage.TenantToken, len(tenants))
 	for i, t := range tenants {

app/vmselect/netstorage/tenant_filters.go

@@ -2,8 +2,7 @@ package netstorage
 
 import (
 	"fmt"
-
-	"github.com/VictoriaMetrics/metricsql"
+	"strings"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vmselect/searchutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
@@ -71,6 +70,14 @@ func ApplyTenantFiltersToTagFilters(tts []storage.TenantToken, tfs [][]storage.T
 	return tts, otherFilters
 }
 
+func tagFiltersToString(tfs []storage.TagFilter) string {
+	a := make([]string, len(tfs))
+	for i, tf := range tfs {
+		a[i] = tf.String()
+	}
+	return "{" + strings.Join(a, ",") + "}"
+}
+
 // applyFiltersToTenants applies the given filters to the given tenants.
 // It returns the filtered tenants.
 func applyFiltersToTenants(tenants []storage.TenantToken, filters [][]storage.TagFilter) ([]storage.TenantToken, error) {
@@ -95,16 +102,21 @@ func applyFiltersToTenants(tenants []storage.TenantToken, filters [][]storage.Ta
 		lbs = append(lbs, lbsAux[lbsAuxLen:])
 	}
 
-	me := &metricsql.MetricExpr{
-		LabelFilterss: toLabelFilterss(filters),
-	}
-	var promIf promrelabel.IfExpression
-	if err := promIf.ParseFromMetricExpr(me); err != nil {
-		return nil, fmt.Errorf("cannot parse if expression from filters %v: %w", filters, err)
+	promIfs := make([]promrelabel.IfExpression, len(filters))
+	for i, tags := range filters {
+		filter := tagFiltersToString(tags)
+		err := promIfs[i].Parse(filter)
+		if err != nil {
+			return nil, fmt.Errorf("cannot parse if expression from filters %v: %w", filter, err)
+		}
 	}
+
 	for i, lb := range lbs {
-		if promIf.Match(lb) {
-			resultingTokens = append(resultingTokens, tenants[i])
+		for _, promIf := range promIfs {
+			if promIf.Match(lb) {
+				resultingTokens = append(resultingTokens, tenants[i])
+				break
+			}
 		}
 	}
 
@@ -115,26 +127,3 @@ func applyFiltersToTenants(tenants []storage.TenantToken, filters [][]storage.Ta
 func isTenancyLabel(name string) bool {
 	return name == "vm_account_id" || name == "vm_project_id"
 }
-
-func toLabelFilterss(tfss [][]storage.TagFilter) [][]metricsql.LabelFilter {
-	lfss := make([][]metricsql.LabelFilter, len(tfss))
-	for i, tfs := range tfss {
-		lfs := make([]metricsql.LabelFilter, len(tfs))
-		for j := range tfs {
-			toLabelFilter(&lfs[j], &tfs[j])
-		}
-		lfss[i] = lfs
-	}
-	return lfss
-}
-
-func toLabelFilter(dst *metricsql.LabelFilter, src *storage.TagFilter) {
-	if src.Key == nil {
-		dst.Label = "__name__"
-	} else {
-		dst.Label = string(src.Key)
-	}
-	dst.Value = string(src.Value)
-	dst.IsRegexp = src.IsRegexp
-	dst.IsNegative = src.IsNegative
-}

app/vmselect/netstorage/tenant_filters_test.go

@@ -27,12 +27,6 @@ func TestApplyFiltersToTenants(t *testing.T) {
 	f([][]storage.TagFilter{{{Key: []byte("vm_account_id"), Value: []byte("1[0-9]+"), IsNegative: false, IsRegexp: true}}}, []storage.TenantToken{{AccountID: 1, ProjectID: 1}, {AccountID: 12323, ProjectID: 0}, {AccountID: 12323, ProjectID: 3}, {AccountID: 345, ProjectID: 0}}, []storage.TenantToken{{AccountID: 12323, ProjectID: 0}, {AccountID: 12323, ProjectID: 3}})
 
 	f([][]storage.TagFilter{{{Key: []byte("vm_account_id"), Value: []byte("1"), IsNegative: false, IsRegexp: false}, {Key: []byte("vm_project_id"), Value: []byte("0"), IsNegative: true, IsRegexp: false}}}, []storage.TenantToken{{AccountID: 1, ProjectID: 1}, {AccountID: 1, ProjectID: 0}}, []storage.TenantToken{{AccountID: 1, ProjectID: 1}})
-
-	f(
-		[][]storage.TagFilter{{{Key: []byte("vm_account_id"), Value: []byte("100|101|102|103|104|105|106|107|108|109|110|111|112|113|114|115|116"), IsNegative: false, IsRegexp: true}}},
-		[]storage.TenantToken{{AccountID: 100, ProjectID: 0}, {AccountID: 108, ProjectID: 0}, {AccountID: 116, ProjectID: 0}, {AccountID: 1239, ProjectID: 0}, {AccountID: 200, ProjectID: 0}},
-		[]storage.TenantToken{{AccountID: 100, ProjectID: 0}, {AccountID: 108, ProjectID: 0}, {AccountID: 116, ProjectID: 0}},
-	)
 }
 
 func TestIsTenancyLabel(t *testing.T) {

app/vmselect/prometheus/metadata_response.qtpl

@@ -16,7 +16,7 @@ See https://prometheus.io/docs/prometheus/latest/querying/api/#querying-metric-m
 			currentItem := 0
 		 %}
 		{% for _, row := range result %}
-			{%q= string(row.MetricFamilyName) %}: [
+			"{%s string(row.MetricFamilyName) %}": [
 					{
 						"type": {%q= row.Type.String() %},
             {% if len(row.Unit) > 0 -%}

app/vmselect/prometheus/metadata_response.qtpl.go

@@ -47,10 +47,12 @@ func StreamMetadataResponse(qw422016 *qt422016.Writer, isPartial bool, result []
 
 //line app/vmselect/prometheus/metadata_response.qtpl:18
 	for _, row := range result {
+//line app/vmselect/prometheus/metadata_response.qtpl:18
+		qw422016.N().S(`"`)
 //line app/vmselect/prometheus/metadata_response.qtpl:19
-		qw422016.N().Q(string(row.MetricFamilyName))
+		qw422016.E().S(string(row.MetricFamilyName))
 //line app/vmselect/prometheus/metadata_response.qtpl:19
-		qw422016.N().S(`: [{"type":`)
+		qw422016.N().S(`": [{"type":`)
 //line app/vmselect/prometheus/metadata_response.qtpl:21
 		qw422016.N().Q(row.Type.String())
 //line app/vmselect/prometheus/metadata_response.qtpl:21

app/vmselect/prometheus/prometheus.go

@@ -51,9 +51,10 @@ 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")
-	selectNodes = flagutil.NewArrayString("selectNode", "A list of vmselect node addresses to propagate the '/internal/resetRollupResultCache' call. "+
-		"If this flag isn't set, then cache need to be purged from each vmselect individually. "+
+	selectNodes = flagutil.NewArrayString("selectNode", "A list of vmselect node addresses to propagate the '/internal/resetRollupResultCache' call with 'propagate=1' argument. "+
+		"If this flag or the 'propagate' argument isn't set, then cache need to be purged from each vmselect individually. "+
 		"Comma-separated addresses of vmselect nodes; usage: -selectNode=vmselect-host1,...,vmselect-hostN")
+	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.*. It'll be used when reset request is propagate to other -selectNode.")
 
 	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. "+
 		"The limit can't exceed the explicitly set corresponding value `-search.maxUniqueTimeseries` on vmstorage side.")
@@ -562,7 +563,7 @@ func DeleteHandler(startTime time.Time, at *auth.Token, r *http.Request) error {
 		// Reset rollup result cache on all the vmselect nodes,
 		// since the cache may contain deleted data.
 		// TODO: reset only cache for (account, project)
-		resetRollupResultCaches()
+		resetRollupResultCachesAndPropagate()
 	}
 	logger.Infof("/api/v1/admin/tsdb/delete_series has been called for %q. Deleted %d series.", sq.FiltersString(), deletedCount)
 	return nil
@@ -570,33 +571,69 @@ func DeleteHandler(startTime time.Time, at *auth.Token, r *http.Request) error {
 
 var deleteDuration = metrics.NewSummary(`vm_request_duration_seconds{path="/api/v1/admin/tsdb/delete_series"}`)
 
+// ResetRollupResultCacheHandler handle request for `/internal/resetRollupResultCache` API.
+// It propagates the request if `propagate` argument is set.
+func ResetRollupResultCacheHandler(w http.ResponseWriter, r *http.Request) bool {
+	// check if this is a propagated request from another vmselect, by propagate argument.
+	// - if yes: simply execute and return.
+	propagate := httputil.GetBool(r, "propagate")
+	if !propagate {
+		resetRollupResultCaches()
+		return true
+	}
+	// - if no: it's manual request and need to propagate to other vmselect(s).
+	resetRollupResultCachesAndPropagate()
+	return true
+}
+
+// GetResetCacheAuthKey returns resetCacheAuthKey value in *Password.
+func GetResetCacheAuthKey() *flagutil.Password {
+	return resetCacheAuthKey
+}
+
 func resetRollupResultCaches() {
 	resetRollupResultCacheCalls.Inc()
 	// Reset local cache before checking whether selectNodes list is empty.
 	// This guarantees that at least local cache is reset if selectNodes list is empty.
 	promql.ResetRollupResultCache()
+}
+
+func resetRollupResultCachesAndPropagate() {
+	resetRollupResultCaches()
 	if len(*selectNodes) == 0 {
 		logger.Warnf("missing -selectNode flag, cache reset request wont be propagated to the other vmselect nodes." +
 			"This can be fixed by enumerating all the vmselect node addresses in `-selectNode` command line flag. " +
 			" For example: -selectNode=select-addr-1:8481,select-addr-2:8481")
 		return
 	}
+	rcAuthKey := GetResetCacheAuthKey().Get()
 	for _, selectNode := range *selectNodes {
 		normalizedAddr, err := netutil.NormalizeAddr(selectNode, 8481)
 		if err != nil {
 			logger.Fatalf("cannot normalize -selectNode=%q: %s", selectNode, err)
 		}
 		selectNode = normalizedAddr
-		callURL := fmt.Sprintf("http://%s/internal/resetRollupResultCache", selectNode)
-		resp, err := httpClient.Get(callURL)
+		req, err := http.NewRequest(http.MethodGet, fmt.Sprintf("http://%s/internal/resetRollupResultCache", selectNode), nil)
 		if err != nil {
-			logger.Errorf("error when accessing %q: %s", callURL, err)
+			logger.Errorf("cannot create cache reset request for %q: %s", selectNode, err)
+			resetRollupResultCacheErrors.Inc()
+			continue
+		}
+		// usually `-search.resetCacheAuthKey` is set to the same on each vmselect. it's good to propagate with this argument.
+		if rcAuthKey != "" {
+			q := req.URL.Query()
+			q.Add("authKey", rcAuthKey)
+			req.URL.RawQuery = q.Encode()
+		}
+		resp, err := httpClient.Do(req)
+		if err != nil {
+			logger.Errorf("error when accessing %q: %s", req.URL.String(), err)
 			resetRollupResultCacheErrors.Inc()
 			continue
 		}
 		if resp.StatusCode != http.StatusOK {
 			_ = resp.Body.Close()
-			logger.Errorf("unexpected status code at %q; got %d; want %d", callURL, resp.StatusCode, http.StatusOK)
+			logger.Errorf("unexpected status code at %q; got %d; want %d", req.URL.String(), resp.StatusCode, http.StatusOK)
 			resetRollupResultCacheErrors.Inc()
 			continue
 		}

app/vmselect/vmui/assets/rolldown-runtime-COnpUsM8.js

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

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

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

app/vmselect/vmui/index.html

@@ -37,9 +37,9 @@
   <meta property="og:title" content="UI for VictoriaMetrics">
   <meta property="og:url" content="https://victoriametrics.com/">
   <meta property="og:description" content="Explore and troubleshoot your VictoriaMetrics data">
-  <script type="module" crossorigin src="./assets/index-CusQvJzs.js"></script>
-  <link rel="modulepreload" crossorigin href="./assets/rolldown-runtime-Cyuzqnbw.js">
-  <link rel="modulepreload" crossorigin href="./assets/vendor-B83wxFqK.js">
+  <script type="module" crossorigin src="./assets/index-CoGukb-x.js"></script>
+  <link rel="modulepreload" crossorigin href="./assets/rolldown-runtime-COnpUsM8.js">
+  <link rel="modulepreload" crossorigin href="./assets/vendor-C8Kwp93_.js">
   <link rel="stylesheet" crossorigin href="./assets/vendor-CnsZ1jie.css">
   <link rel="stylesheet" crossorigin href="./assets/index-BBUnmLOr.css">
 </head>

apptest/tests/metricsmetadata_test.go

@@ -45,13 +45,11 @@ func TestSingleMetricsMetadata(t *testing.T) {
 			{Labels: []prompb.Label{{Name: "__name__", Value: "metric_name_4"}}, Samples: []prompb.Sample{{Value: 40, Timestamp: ingestTimestamp}}},
 			{Labels: []prompb.Label{{Name: "__name__", Value: "metric_name_5"}}, Samples: []prompb.Sample{{Value: 40, Timestamp: ingestTimestamp}}},
 			{Labels: []prompb.Label{{Name: "__name__", Value: "metric_name_6"}}, Samples: []prompb.Sample{{Value: 40, Timestamp: ingestTimestamp}}},
-			{Labels: []prompb.Label{{Name: "__name__", Value: `metric_name_7_!@"_suffix`}}, Samples: []prompb.Sample{{Value: 40, Timestamp: ingestTimestamp}}},
 		},
 		Metadata: []prompb.MetricMetadata{
 			{MetricFamilyName: "metric_name_4", Help: "some help message", Type: prompb.MetricTypeSummary},
 			{MetricFamilyName: "metric_name_5", Help: "some help message", Type: prompb.MetricTypeSummary},
 			{MetricFamilyName: "metric_name_6", Help: "some help message", Type: prompb.MetricTypeStateset},
-			{MetricFamilyName: `metric_name_7_!@"_suffix`, Help: "some help message", Type: prompb.MetricTypeStateset},
 		},
 	}
 
@@ -61,13 +59,12 @@ func TestSingleMetricsMetadata(t *testing.T) {
 	expected := &apptest.PrometheusAPIV1Metadata{
 		Status: "success",
 		Data: map[string][]apptest.MetadataEntry{
-			"metric_name_1":            {{Help: "some help message", Type: "gauge"}},
-			"metric_name_2":            {{Help: "some help message", Type: "counter"}},
-			"metric_name_3":            {{Help: "some help message", Type: "gauge"}},
-			"metric_name_4":            {{Help: "some help message", Type: "summary"}},
-			"metric_name_5":            {{Help: "some help message", Type: "summary"}},
-			"metric_name_6":            {{Help: "some help message", Type: "stateset"}},
-			`metric_name_7_!@"_suffix`: {{Help: "some help message", Type: "stateset"}},
+			"metric_name_1": {{Help: "some help message", Type: "gauge"}},
+			"metric_name_2": {{Help: "some help message", Type: "counter"}},
+			"metric_name_3": {{Help: "some help message", Type: "gauge"}},
+			"metric_name_4": {{Help: "some help message", Type: "summary"}},
+			"metric_name_5": {{Help: "some help message", Type: "summary"}},
+			"metric_name_6": {{Help: "some help message", Type: "stateset"}},
 		},
 	}
 	gotStats := sut.PrometheusAPIV1Metadata(t, "", 0, apptest.QueryOpts{})
@@ -157,13 +154,11 @@ func TestClusterMetricsMetadata(t *testing.T) {
 			{Labels: []prompb.Label{{Name: "__name__", Value: "metric_name_4"}}, Samples: []prompb.Sample{{Value: 40, Timestamp: ingestTimestamp}}},
 			{Labels: []prompb.Label{{Name: "__name__", Value: "metric_name_5"}}, Samples: []prompb.Sample{{Value: 40, Timestamp: ingestTimestamp}}},
 			{Labels: []prompb.Label{{Name: "__name__", Value: "metric_name_6"}}, Samples: []prompb.Sample{{Value: 40, Timestamp: ingestTimestamp}}},
-			{Labels: []prompb.Label{{Name: "__name__", Value: `metric_name_7_!@"_suffix`}}, Samples: []prompb.Sample{{Value: 40, Timestamp: ingestTimestamp}}},
 		},
 		Metadata: []prompb.MetricMetadata{
 			{MetricFamilyName: "metric_name_4", Help: "some help message", Type: prompb.MetricTypeSummary},
 			{MetricFamilyName: "metric_name_5", Help: "some help message", Type: prompb.MetricTypeSummary},
 			{MetricFamilyName: "metric_name_6", Help: "some help message", Type: prompb.MetricTypeStateset},
-			{MetricFamilyName: `metric_name_7_!@"_suffix`, Help: "some help message", Type: prompb.MetricTypeStateset},
 		},
 	}
 
@@ -176,13 +171,12 @@ func TestClusterMetricsMetadata(t *testing.T) {
 		expected := &apptest.PrometheusAPIV1Metadata{
 			Status: "success",
 			Data: map[string][]apptest.MetadataEntry{
-				"metric_name_1":            {{Help: "some help message", Type: "gauge"}},
-				"metric_name_2":            {{Help: "some help message", Type: "counter"}},
-				"metric_name_3":            {{Help: "some help message", Type: "gauge"}},
-				"metric_name_4":            {{Help: "some help message", Type: "summary"}},
-				"metric_name_5":            {{Help: "some help message", Type: "summary"}},
-				"metric_name_6":            {{Help: "some help message", Type: "stateset"}},
-				`metric_name_7_!@"_suffix`: {{Help: "some help message", Type: "stateset"}},
+				"metric_name_1": {{Help: "some help message", Type: "gauge"}},
+				"metric_name_2": {{Help: "some help message", Type: "counter"}},
+				"metric_name_3": {{Help: "some help message", Type: "gauge"}},
+				"metric_name_4": {{Help: "some help message", Type: "summary"}},
+				"metric_name_5": {{Help: "some help message", Type: "summary"}},
+				"metric_name_6": {{Help: "some help message", Type: "stateset"}},
 			},
 		}
 		gotStats := vmselect.PrometheusAPIV1Metadata(t, "", 0, apptest.QueryOpts{Tenant: tenantID})

apptest/tests/vmagent_remotewrite_test.go

@@ -8,7 +8,6 @@ import (
 	"net/http/httptest"
 	"strings"
 	"sync"
-	"sync/atomic"
 	"testing"
 	"time"
 
@@ -333,11 +332,13 @@ func TestSingleVMAgentDropOnOverload(t *testing.T) {
 	vmagent.APIV1ImportPrometheusNoWaitFlush(t, []string{
 		"foo_bar 1 1652169600000", // 2022-05-10T08:00:00Z
 	}, apptest.QueryOpts{})
+
 	waitFor(
 		func() bool {
 			return vmagent.RemoteWriteRequests(t, url1) == 1 && vmagent.RemoteWriteRequests(t, url2) == 1
 		},
 	)
+
 	// Send 2 more requests, the first RW endpoint should receive everything, the second should add them to the queue
 	// since worker is busy with the first request.
 	for i := range 2 {
@@ -640,116 +641,3 @@ func TestSingleVMAgentMultitenancy(t *testing.T) {
 		t.Fatalf("expected vmagent_tenant_inserted_rows_total to have value 1 for accountID=5, projectID=0")
 	}
 }
-
-func TestSingleVMAgentPriorizeRecentData(t *testing.T) {
-	tc := apptest.NewTestCase(t)
-	defer tc.Stop()
-
-	remoteWriteSrv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		w.WriteHeader(http.StatusNoContent)
-	}))
-	defer remoteWriteSrv.Close()
-
-	var mustRW2ReturnError atomic.Bool
-	mustRW2ReturnError.Store(true)
-
-	remoteWriteSrv2 := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if mustRW2ReturnError.Load() {
-			w.WriteHeader(http.StatusServiceUnavailable)
-			return
-		}
-		w.WriteHeader(http.StatusNoContent)
-	}))
-	defer remoteWriteSrv2.Close()
-
-	vmagent := tc.MustStartDefaultRWVmagent("vmagent", []string{
-		fmt.Sprintf(`-remoteWrite.url=%s/api/v1/write`, remoteWriteSrv.URL),
-		fmt.Sprintf(`-remoteWrite.url=%s/api/v1/write`, remoteWriteSrv2.URL),
-		"-remoteWrite.disableOnDiskQueue=true",
-		// use only 1 worker to get a full queue faster
-		"-remoteWrite.queues=1",
-		"-remoteWrite.flushInterval=1ms",
-		"-remoteWrite.inmemoryQueues=1",
-		// fastqueue size is roughly memory.Allowed() / len(urls) / *maxRowsPerBlock / 100
-		// Use very large maxRowsPerBlock to get fastqueue of minimal length(2).
-		// See initRemoteWriteCtxs function in remotewrite.go for details.
-		"-remoteWrite.maxRowsPerBlock=1000000000",
-		"-remoteWrite.tmpDataPath=" + tc.Dir() + "/vmagent",
-
-		// Delay retry logic to avoid race conditions with waitFor assertions.
-		// It improves the test stability on resource-constrained runners.
-		"-remoteWrite.retryMinInterval=3s",
-		"-remoteWrite.retryMaxTime=3s",
-	})
-
-	const (
-		retries = 20
-		period  = 200 * time.Millisecond
-	)
-
-	waitFor := func(f func() bool) {
-		t.Helper()
-		for range retries {
-			if f() {
-				return
-			}
-			time.Sleep(period)
-		}
-		t.Fatalf("timed out waiting for retry #%d", retries)
-	}
-
-	// Real remote write URLs are hidden in metrics
-	url1 := "1:secret-url"
-	url2 := "2:secret-url"
-
-	// Wait until first request got flushed to remote write server
-	vmagent.APIV1ImportPrometheusNoWaitFlush(t, []string{
-		"foo_bar 1 1652169600000", // 2022-05-10T08:00:00Z
-	}, apptest.QueryOpts{})
-	waitFor(
-		func() bool {
-			return vmagent.RemoteWriteRequests(t, url1) == 1 && vmagent.RemoteWriteRequests(t, url2) == 1
-		},
-	)
-	// Wait until second request got flushed to remote write server
-	// since there are 2 independent queues (general and in-memory) with minimal capacity of 1
-	vmagent.APIV1ImportPrometheusNoWaitFlush(t, []string{
-		"foo_bar 1 1652169600000", // 2022-05-10T08:00:00Z
-	}, apptest.QueryOpts{})
-	waitFor(
-		func() bool {
-			return vmagent.RemoteWriteRequests(t, url1) == 2 && vmagent.RemoteWriteRequests(t, url2) == 2
-		},
-	)
-	// Send 2 more requests, the first RW endpoint should receive everything, the second should add them to the queue
-	// since worker is busy with the first request.
-	for i := range 2 {
-		vmagent.APIV1ImportPrometheusNoWaitFlush(t, []string{
-			"foo_bar 1 1652169600000", // 2022-05-10T08:00:00Z
-		}, apptest.QueryOpts{})
-
-		waitFor(
-			func() bool {
-				return vmagent.RemoteWriteRequests(t, url1) == 3+i && vmagent.RemoteWritePendingInmemoryBlocks(t, url2) == 1+i
-			},
-		)
-	}
-
-	// Send one more request.
-	vmagent.APIV1ImportPrometheusNoWaitFlush(t, []string{
-		"foo_bar 1 1652169600000", // 2022-05-10T08:00:00Z
-	}, apptest.QueryOpts{})
-
-	waitFor(
-		func() bool {
-			return vmagent.RemoteWriteRequests(t, url1) == 5 && vmagent.RemoteWriteSamplesDropped(t, url2) > 0
-		},
-	)
-	mustRW2ReturnError.Store(false)
-	// ensure that inmemory data correctly flushed to the remote write
-	waitFor(
-		func() bool {
-			return vmagent.RemoteWritePendingInmemoryBlocks(t, url2) == 0
-		},
-	)
-}

dashboards/vm/vmagent.json

@@ -6201,7 +6201,7 @@
             "type": "victoriametrics-metrics-datasource",
             "uid": "$ds"
           },
-          "description": "The rate of dropped samples during aggregation. \nStream aggregation will drop samples with NaN values, too old timestamps or samples identified as duplicates during deduplication. See https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#ignoring-old-samples ",
+          "description": "The rate of ignored samples during aggregation. \nStream aggregation will drop samples with NaN values, or samples with too old timestamps. See https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#ignoring-old-samples ",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -6282,14 +6282,14 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(rate({__name__=~\"vm_streamaggr_ignored_samples_total|vm_streamaggr_dedup_dropped_samples_total\", job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval]) > 0) without (instance, pod)",
+              "expr": "sum(rate(vm_streamaggr_ignored_samples_total{job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval]) > 0) without (instance, pod)",
               "instant": false,
               "legendFormat": "__auto",
               "range": true,
               "refId": "A"
             }
           ],
-          "title": "Dropped samples ($instance)",
+          "title": "Ignored samples ($instance)",
           "type": "timeseries"
         },
         {

dashboards/vmagent.json

@@ -6200,7 +6200,7 @@
             "type": "prometheus",
             "uid": "$ds"
           },
-          "description": "The rate of dropped samples during aggregation. \nStream aggregation will drop samples with NaN values, too old timestamps or samples identified as duplicates during deduplication. See https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#ignoring-old-samples ",
+          "description": "The rate of ignored samples during aggregation. \nStream aggregation will drop samples with NaN values, or samples with too old timestamps. See https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#ignoring-old-samples ",
           "fieldConfig": {
             "defaults": {
               "color": {
@@ -6281,14 +6281,14 @@
                 "uid": "$ds"
               },
               "editorMode": "code",
-              "expr": "sum(rate({__name__=~\"vm_streamaggr_ignored_samples_total|vm_streamaggr_dedup_dropped_samples_total\", job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval]) > 0) without (instance, pod)",
+              "expr": "sum(rate(vm_streamaggr_ignored_samples_total{job=~\"$job\",instance=~\"$instance\", url=~\"$url\"}[$__rate_interval]) > 0) without (instance, pod)",
               "instant": false,
               "legendFormat": "__auto",
               "range": true,
               "refId": "A"
             }
           ],
-          "title": "Dropped samples ($instance)",
+          "title": "Ignored samples ($instance)",
           "type": "timeseries"
         },
         {

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

@@ -59,7 +59,7 @@ services:
       - '--external.alert.source=explore?orgId=1&left=["now-1h","now","VictoriaMetrics",{"expr": },{"mode":"Metrics"},{"ui":[true,true,true,"none"]}]'
     restart: always
   vmanomaly:
-    image: victoriametrics/vmanomaly:v1.29.6
+    image: victoriametrics/vmanomaly:v1.29.5
     depends_on:
       - "victoriametrics"
     ports:

docs/anomaly-detection/CHANGELOG.md

@@ -14,13 +14,6 @@ aliases:
 ---
 Please find the changelog for VictoriaMetrics Anomaly Detection below.
 
-## v1.29.6
-Released: 2026-06-17
-
-- BUGFIX: Fixed `VLogsReader` startup and query execution when `tenant_id` is omitted or provided in short account-only form such as `"0"`. Omitted or empty tenant IDs are treated as single-node/no-tenant mode, and account-only tenant IDs are expanded to `accountID:0` before adding VictoriaLogs `AccountID`/`ProjectID` params or VM tenant labels.
-
-- BUGFIX: Hardened [`OnlineMADModel`](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-mad) anomaly scoring for perfectly constant time series (all values identical). The model now keeps a small deterministic prediction interval when the learned MAD is zero, so values deviating from an unknown constant baseline can produce `anomaly_score > 1` (previously, all anomaly scores were `0`).
-
 ## v1.29.5
 Released: 2026-06-11
 

docs/anomaly-detection/FAQ.md

@@ -423,7 +423,7 @@ services:
   # ...
   vmanomaly:
     container_name: vmanomaly
-    image: victoriametrics/vmanomaly:v1.29.6
+    image: victoriametrics/vmanomaly:v1.29.5
     # ...
     restart: always
     volumes:
@@ -641,7 +641,7 @@ options:
 Here’s an example of using the config splitter to divide configurations based on the `extra_filters` argument from the reader section:
 
 ```sh
-docker pull victoriametrics/vmanomaly:v1.29.6 && docker image tag victoriametrics/vmanomaly:v1.29.6 vmanomaly
+docker pull victoriametrics/vmanomaly:v1.29.5 && docker image tag victoriametrics/vmanomaly:v1.29.5 vmanomaly
 ```
 
 ```sh

docs/anomaly-detection/Migration.md

@@ -45,7 +45,7 @@ There are 2 types of compatibility to consider when migrating in stateful mode:
 
 | Group start | Group end | Compatibility | Notes |
 |---------|--------- |------------|-------|
-| [v1.29.1](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1291) | [v1.29.6](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1296) | Fully Compatible | - |
+| [v1.29.1](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1291) | [v1.29.5](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1295) | Fully Compatible | - |
 | [v1.28.7](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1287) | [v1.29.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1290) | Partially compatible* | Dumped models of class [prophet](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet) and [seasonal quantile](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-seasonal-quantile) have problems with loading to [v1.29.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1290) due to dropped `pytz` library. **Upgrading directly from v1.28.7 to [v1.29.1](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1291) with a fix is suggested** |
 | [v1.26.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1262) | [v1.28.7](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1287) | Fully Compatible | [v1.28.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1280) introduced [rolling](https://docs.victoriametrics.com/anomaly-detection/components/models/#rolling-models) model class drop in favor of [online](https://docs.victoriametrics.com/anomaly-detection/components/models/#online-models) models (`rolling_quantile` and `std` models), however, it does not impact compatibility, as artifacts were not produced by default for rolling models. Also, offline `mad` and `zscore` models are redirecting to their respective online counterparts since [v1.28.4](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1284). |
 | [v1.25.3](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1253) | [v1.26.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1270) | Partially Compatible* | [v1.25.3](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1253) introduced `forecast_at` argument for base [univariate](https://docs.victoriametrics.com/anomaly-detection/components/models/#univariate-models) and `Prophet` [models](https://docs.victoriametrics.com/anomaly-detection/components/models/#prophet), however, itself remains backward-reversible from newer states like [v1.26.2](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1262), [v1.27.0](https://docs.victoriametrics.com/anomaly-detection/changelog/#v1270). (All models except `isolation_forest_multivariate` class will be dropped) |

docs/anomaly-detection/QuickStart.md

@@ -132,7 +132,7 @@ Below are the steps to get `vmanomaly` up and running inside a Docker container:
 1. Pull Docker image:
 
 ```sh
-docker pull victoriametrics/vmanomaly:v1.29.6
+docker pull victoriametrics/vmanomaly:v1.29.5
 ```
 
 2. Create the license file with your license key.
@@ -152,7 +152,7 @@ docker run -it \
     -v ./license:/license \
     -v ./config.yaml:/config.yaml \
     -p 8490:8490 \
-    victoriametrics/vmanomaly:v1.29.6 \
+    victoriametrics/vmanomaly:v1.29.5 \
     /config.yaml \
     --licenseFile=/license \
     --loggerLevel=INFO \
@@ -169,7 +169,7 @@ docker run -it \
     -e VMANOMALY_DATA_DUMPS_DIR=/tmp/vmanomaly/data \
     -e VMANOMALY_MODEL_DUMPS_DIR=/tmp/vmanomaly/models \
     -p 8490:8490 \
-    victoriametrics/vmanomaly:v1.29.6 \
+    victoriametrics/vmanomaly:v1.29.5 \
     /config.yaml \
     --licenseFile=/license \
     --loggerLevel=INFO \
@@ -182,7 +182,7 @@ services:
   # ...
   vmanomaly:
     container_name: vmanomaly
-    image: victoriametrics/vmanomaly:v1.29.6
+    image: victoriametrics/vmanomaly:v1.29.5
     # ...
     restart: always
     volumes:

docs/anomaly-detection/UI.md

@@ -315,7 +315,7 @@ docker run -it --rm \
   -e VMANOMALY_MCP_SERVER_URL=http://mcp-vmanomaly:8081/mcp \
   -p 8080:8080 \
   -p 8490:8490 \
-  victoriametrics/vmanomaly:v1.29.6 \
+  victoriametrics/vmanomaly:v1.29.5 \
   vmanomaly_config.yaml
 ```
 

docs/anomaly-detection/components/models.md

@@ -1265,7 +1265,7 @@ monitoring:
 Let's pull the docker image for `vmanomaly`:
 
 ```sh
-docker pull victoriametrics/vmanomaly:v1.29.6
+docker pull victoriametrics/vmanomaly:v1.29.5
 ```
 
 Now we can run the docker container putting as volumes both config and model file:
@@ -1279,7 +1279,7 @@ docker run -it \
 -v $(PWD)/license:/license \
 -v $(PWD)/custom_model.py:/vmanomaly/model/custom.py \
 -v $(PWD)/custom.yaml:/config.yaml \
-victoriametrics/vmanomaly:v1.29.6 /config.yaml \
+victoriametrics/vmanomaly:v1.29.5 /config.yaml \
 --licenseFile=/license
 --watch
 ```

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

@@ -395,7 +395,7 @@ services:
     restart: always
   vmanomaly:
     container_name: vmanomaly
-    image: victoriametrics/vmanomaly:v1.29.6
+    image: victoriametrics/vmanomaly:v1.29.5
     depends_on:
       - "victoriametrics"
     ports:

docs/victoriametrics/Articles.md

@@ -19,7 +19,6 @@ See also [case studies](https://docs.victoriametrics.com/victoriametrics/casestu
 * [Datanami: Why Roblox Picked VictoriaMetrics for Observability Data Overhaul](https://www.hpcwire.com/bigdatawire/2023/05/30/why-roblox-picked-victoriametrics-for-observability-data-overhaul/)
 * [Cloudflare: Introducing notifications for HTTP Traffic Anomalies](https://blog.cloudflare.com/introducing-http-traffic-anomalies-notifications/)
 * [Grammarly: Better, Faster, Cheaper: How Grammarly Improved Monitoring by Over 10x with VictoriaMetrics](https://www.grammarly.com/blog/engineering/monitoring-with-victoriametrics/)
-* [Xata: How we rebuilt PostgreSQL branch metrics on VictoriaMetrics, per cell](https://xata.io/blog/how-we-rebuilt-postgresql-branch-metrics-on-victoriametrics-per-cell)
 * [CERN: CMS monitoring R&D: Real-time monitoring and alerts](https://indico.cern.ch/event/877333/contributions/3696707/attachments/1972189/3281133/CMS_mon_RD_for_opInt.pdf)
 * [CERN: The CMS monitoring infrastructure and applications](https://arxiv.org/pdf/2007.03630.pdf)
 * [Forbes: The (Almost) Infinitely Scalable Open Source Monitoring Dream](https://www.forbes.com/sites/adrianbridgwater/2022/08/16/the-almost-infinitely-scalable-open-source-monitoring-dream/)

docs/victoriametrics/changelog/CHANGELOG.md

@@ -30,26 +30,12 @@ See also [LTS releases](https://docs.victoriametrics.com/victoriametrics/lts-rel
 * FEATURE: [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): log calls to [/api/v1/admin/tsdb/delete_series](https://docs.victoriametrics.com/victoriametrics/url-examples/#apiv1admintsdbdelete_series) API handler. This should help to identify events of metrics deletion from the database. See [#11104](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11104).
 * FEATURE: [vmctl](https://docs.victoriametrics.com/victoriametrics/vmctl/): add `-vm-headers` and `-vm-bearer-token` flags for authenticating requests to the VictoriaMetrics import destination. The flags are available in `opentsdb`, `influx`, `remote-read`, `prometheus`, `mimir`, and `thanos` vmctl sub-commands. See [#8897](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8897).
 * FEATURE: [vmui](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmui): add the `last` value to graph legend statistics. See [#10759](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10759).
-* FEATURE: [stream aggregation](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/): expose `vm_streamaggr_dedup_dropped_samples_total` to allow tracking dropped old samples during [deduplication](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#deduplication).
-* FEATURE: [stream aggregation](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/): use the aggregation rule interval as the default [staleness_interval](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#staleness) instead of `2*interval`, to reduce spikes when there are gaps between received samples. See [#11102](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11102).
-* FEATURE: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): add a new flag `-remoteWrite.inmemoryQueues` to prioritize recently ingested data over historical data stored at file-based [persistent queue](https://docs.victoriametrics.com/victoriametrics/vmagent/#on-disk-persistence-and-data-processing-order). See [#8833](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8833)
-*  FEATURE: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): add `-promscrape.cluster.shardByLabels` command-line flag for selecting target labels used for sharding scrape targets among `vmagent` instances in cluster mode. See [#11044](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11044).
-* FEATURE: [vmctl](https://docs.victoriametrics.com/victoriametrics/vmctl/): add `-vm-headers` and `-vm-bearer-token` flags for authenticating requests to the VictoriaMetrics import destination. The flags are available in `opentsdb`, `influx`, `remote-read`, `prometheus`, `mimir`, and `thanos` vmctl sub-commands. See [#8897](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8897).
-* FEATURE: [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): log calls to [/api/v1/admin/tsdb/delete_series](https://docs.victoriametrics.com/victoriametrics/url-examples/#apiv1admintsdbdelete_series) API handler. This should help to identify events of metrics deletion from the database. See [#11104](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11104).
-* FEATURE: [vmui](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmui): add the `last` value to graph legend statistics. See [#10759](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10759).
 
-* BUGFIX: [enterprise](https://docs.victoriametrics.com/enterprise/) [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vmstorage` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): properly expose metric `vm_retention_filters_partitions_scheduled_rows`. See [#11138](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11138)
 * BUGFIX: [stream aggregation](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/): fix issue with producing aggregated samples with identical timestamps between flushes. See [#10808](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10808).
-* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): fix potential corruption of remote-write metadata `Unit` values. See [#11120](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11120). Thanks for @fxrlv for the contribution.
 * BUGFIX: [vmalert](https://docs.victoriametrics.com/victoriametrics/vmalert/),[vmauth](https://docs.victoriametrics.com/victoriametrics/vmauth/),[vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/) and [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/): fix rare unbounded shutdown delay when config reload takes longer than `-configCheckInterval`. See [#11107](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11107). Thanks to @PleasingFungus for contribution.
-* BUGFIX: [vmbackup](https://docs.victoriametrics.com/vmbackup/), [vmbackupmanager](https://docs.victoriametrics.com/victoriametrics/vmbackupmanager/): do not fail backup list if directory is absent while using `fs://` destination to align with other protocols. See [6c3c548d](https://github.com/VictoriaMetrics/VictoriaMetrics/commit/6c3c548ddb0385b749e731f52276f130e2a4e4a8).
-* BUGFIX: [vmctl](https://docs.victoriametrics.com/victoriametrics/vmctl/): push metrics to configured `-pushmetrics.url` on shutdown when migration fails. Previously, metrics were not pushed if vmctl exited with an error. See [#11081](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11081). Thanks to @zasdaym for contribution.
-* BUGFIX: [vmrestore](https://docs.victoriametrics.com/victoriametrics/vmrestore/): disallow restoring parts outside the configured `-storageDataPath` directory. See [710c920d](https://github.com/VictoriaMetrics/VictoriaMetrics/commit/710c920d6083327042a309e449fae4383617d817).
-* BUGFIX: `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): correctly apply long tenant filters. Previously, such filters could be truncated, causing tenants to be matched incorrectly. See [#11096](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11096). Thanks for @fxrlv for the contribution.
 * BUGFIX: `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): fix corrupted metrics metadata when a response contains multiple rows. See [#11115](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11115). Thanks for @fxrlv for the contribution.
-* BUGFIX: `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): don't cache empty responses for tenant IDs discovery during [multitenant queries](https://docs.victoriametrics.com/Cluster-VictoriaMetrics.html#multitenant-reads). This problem was visible during integration tests when multitenant queries were executed before the first ingestion happened. See [#10982](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/10982)
-* BUGFIX: [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): properly escape `metricFamilyName` at metrics metadata response. See [#11129](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11129). Thanks for @fxrlv for the contribution.
-* BUGFIX: [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vmstorage` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): prevent more cases of panic during directory deletion on `NFS`-based mounts. See [#11060](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11060).
+* BUGFIX: [vmbackup](https://docs.victoriametrics.com/vmbackup/), [vmbackupmanager](https://docs.victoriametrics.com/victoriametrics/vmbackupmanager/): do not fail backup list if directory is absent while using `fs://` destination to align with other protocols. See [6c3c548](https://github.com/VictoriaMetrics/VictoriaMetrics/commit/6c3c548ddb0385b749e731f52276f130e2a4e4a8)
+* BUGFIX: `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): propagate cache reset operation to `selectNode` when `/internal/resetRollupResultCache` is called. Previously, the propagation only happened when the `delete_series` API was called. See [#11112](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11112).
 
 ## [v1.145.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.145.0)
 

docs/victoriametrics/stream-aggregation/README.md

@@ -76,7 +76,7 @@ It is better to substitute the slow recording rule with the following [stream ag
   outputs: [rate_sum]
 ```
 
-> It is recommended to set the `interval` field to a value at least 2 times the matched metrics collection interval.
+> Field `interval` should be set to a value at least several times higher than the matched metrics collection interval. 
 
 This stream aggregation generates `http_request_duration_seconds_bucket:1m_without_instance_rate_sum` output series according to [output metric naming](#output-metric-names).
 Then these series can be used in [alerting rules](https://docs.victoriametrics.com/victoriametrics/vmalert/#alerting-rules):
@@ -396,7 +396,7 @@ before sending them to the configured `-remoteWrite.url`. The deduplication can
 
 Labels can be dropped before deduplication is applied. See [these docs](#dropping-unneeded-labels).
 
-Stream aggregation deduplication is applied before aggregation rules, so duplicate samples are dropped before aggregation. The dropped old samples can be tracked with the `vm_streamaggr_dedup_dropped_samples_total` metric.
+Stream aggregation deduplication is applied before aggregation rules, so duplicate samples are dropped before aggregation.
 
 # Relabeling
 
@@ -444,9 +444,7 @@ outside the current [aggregation interval](https://docs.victoriametrics.com/vict
 
 - To enable [aggregation windows](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#aggregation-windows).
 
-- To enable [deduplication](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#deduplication).
-
-The dropped old samples can be tracked with the `vm_streamaggr_ignored_samples_total{reason="too_old"}` and `vm_streamaggr_dedup_dropped_samples_total` metrics.
+The dropped old samples can be tracked with the `vm_streamaggr_ignored_samples_total{reason="too_old"}` metric.
 
 ## Ignore aggregation intervals on start
 
@@ -644,9 +642,9 @@ See also [why you shouldn't put an aggregator behind a load balancer](https://do
 
 # Troubleshooting
 
-- [Unexpected spikes for `total` or `increase` outputs](#data-delay-and-staleness).
+- [Unexpected spikes for `total` or `increase` outputs](#staleness).
 - [Excessively large values for `total*`, `increase*`, and `rate*` outputs](#counter-resets).
-- [Lower than expected values for `total_prometheus` and `increase_prometheus` outputs](#data-delay-and-staleness).
+- [Lower than expected values for `total_prometheus` and `increase_prometheus` outputs](#staleness).
 - [High memory usage and CPU usage](#high-resource-usage).
 - [Unexpected results in vmagent cluster mode](#cluster-mode).
 - [Inaccurate aggregation results for histograms](#aggregation-windows)
@@ -679,19 +677,11 @@ the following settings:
 
 If counter-specific outputs, such as `total*`, `rate*`, and `increase*`, produce values that are significantly higher than anticipated, then check the `vm_streamaggr_counter_resets_total` metric. This metric increments each time when [counter reset event](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#counter) happens and could be caused by duplication or collision of raw samples. If you observe duplication or collision, try solving this problem by either fixing the source of these metrics or by [deduplicating](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#deduplication) these samples before aggregation.
 
-## Data delay and staleness {#staleness}
+## Staleness
 
-Stream aggregation processes input samples in a streaming manner and flushes results once per specified `interval`. Because of this, aggregation results can be heavily affected by data delays (see `vm_streamaggr_samples_lag_seconds_bucket` metric).
-
-In particular:
-1. Stream aggregation won't produce results if input samples are delayed for multiple aggregation intervals, causing gaps in the output.
-2. Delayed and out-of-order samples can inflate or skew correctness of aggregation results.
-
-Dropping delayed samples can result in missed observations in the results, while keeping delayed samples may inflate the results. It is up to the user to decide what they prefer in the produced results:
-1. If you prefer consistency in aggregation results and do not want delayed data to affect the next aggregation window, drop all potentially delayed samples via [ignore_old_samples](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#ignoring-old-samples).
-2. If you prefer to have the accumulated changes from delayed data reflected in aggregation windows after the delay, increase `staleness_interval` in the [stream aggregation config](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#stream-aggregation-config).
-This is especially important for outputs that track the last seen per-series values in order to properly calculate output values:
+The following outputs track the last seen per-series values in order to properly calculate output values:
 
+- [histogram_bucket](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#histogram_bucket)
 - [increase](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#increase)
 - [increase_prometheus](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#increase_prometheus)
 - [rate_avg](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#rate_avg)
@@ -699,19 +689,21 @@ This is especially important for outputs that track the last seen per-series val
 - [total](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#total)
 - [total_prometheus](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#total_prometheus)
 
-For these outputs, the last seen per-series value is dropped if no new samples are received for the given time series during consecutive aggregation intervals specified in the [stream aggregation config](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#stream-aggregation-config) via `interval` option.
+The last seen per-series value is dropped if no new samples are received for the given time series during two consecutive aggregations
+intervals specified in [stream aggregation config](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#stream-aggregation-config) via `interval` option.
 If a new sample for the existing time series is received after that, then it is treated as the first sample for a new time series.
-This may lead to the following issues when data is delayed:
+This may lead to the following issues:
 
-- [total](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#total) and [increase](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#increase) may produce unexpected spikes, since they assume that a new time series starts from `0`.
-- [total_prometheus](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#total_prometheus) and [increase_prometheus](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#increase_prometheus) may produce lower than expected results, if you expect to see the accumulated changes reflected after the delay, since they ignore the first sample in a new time series.
+- Lower than expected results for [total_prometheus](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#total_prometheus) and [increase_prometheus](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#increase_prometheus) outputs,
+  since they ignore the first sample in a new time series.
+- Unexpected spikes for [total](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#total) and [increase](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#increase) outputs, since they assume that new time series start from 0.
 
-These issues can be improved in the following ways:
+These issues can be fixed in the following ways:
 
 - By increasing the `interval` option at [stream aggregation config](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#stream-aggregation-config), so it covers the expected
-  delays in data ingestion pipelines. It is recommended to set `interval` to at least 2× the scrape or push interval of the input. Set it to a higher value if the input pipeline is prone to large delays.
-- By increasing the `staleness_interval` option in the [stream aggregation config](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#stream-aggregation-config), so it covers the expected
-  delays in data ingestion pipelines. By default, the `staleness_interval` is equal to `interval`.
+  delays in data ingestion pipelines.
+- By specifying the `staleness_interval` option at [stream aggregation config](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/configuration/#stream-aggregation-config), so it covers the expected
+  delays in data ingestion pipelines. By default, the `staleness_interval` is equal to `2 x interval`.
 
 ## High resource usage
 

docs/victoriametrics/stream-aggregation/configuration.md

@@ -66,8 +66,6 @@ specified individually per each `-remoteWrite.url`:
 
   # interval is the interval for the aggregation.
   # The aggregated stats is sent to remote storage once per interval.
-  # It is recommended to set `interval` to at least 2× the scrape or push interval of the input.
-  # Set it to a higher value if the input pipeline is prone to large delays.
   #
   interval: 1m
 
@@ -96,7 +94,7 @@ specified individually per each `-remoteWrite.url`:
   # - total_prometheus
   # See https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#staleness for more details.
   #
-  # staleness_interval: 1m
+  # staleness_interval: 2m
 
   # ignore_first_sample_interval specifies the interval after which the agent begins sending samples.
   # By default, it is set to the staleness interval, and it helps reduce the initial sample load after an agent restart.
@@ -293,6 +291,9 @@ The results of `histogram_bucket` is equal to the following [MetricsQL](https://
 sum(histogram_over_time(some_histogram_bucket[interval])) by (vmrange)
 ```
 
+Aggregating irregular and sporadic metrics (received from [Lambdas](https://aws.amazon.com/lambda/)
+or [Cloud Functions](https://cloud.google.com/functions)) can be controlled via [staleness_interval](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#staleness) option.
+
 See also:
 - [quantiles](#quantiles)
 - [avg](#avg)
@@ -506,19 +507,6 @@ See also:
 - [count_samples](#count_samples)
 - [count_series](#count_series)
 
-### `sum_samples_total`
-
-`sum_samples_total` sums input delta values into a cumulative [counter](https://docs.victoriametrics.com/victoriametrics/keyconcepts/index.html#counter) and outputs the result at the given `interval`.
-`sum_samples_total` makes sense only for aggregating delta values from clients such as [StatsD counter](https://github.com/statsd/statsd/blob/master/docs/metric_types.md#counting).
-
-The results of `sum_samples_total` is roughly equal to the following [MetricsQL](https://docs.victoriametrics.com/victoriametrics/metricsql/) query:
-
-```metricsql
-sum(running_sum(some_delta_values))
-```
-
->Note: The aggregator will forget the cumulative counter if it has not seen input samples for `staleness_interval`(set to `interval` by default) per output result, so the output counter will start from `0` the next time it sees the input again. Increase the `staleness_interval` option if you want to extend the window to tolerate bigger gaps.
-
 ### total
 
 `total` generates output [counter](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#counter) by summing the input counters over the given `interval`.

docs/victoriametrics/url-examples.md

@@ -622,11 +622,13 @@ curl -Is http://localhost:8428/internal/resetRollupResultCache
 Cluster version of VictoriaMetrics:
 
 ```sh
-curl -Is http://<vmselect>:8481/internal/resetRollupResultCache
+curl -Is http://<vmselect>:8481/internal/resetRollupResultCache?propagate=1
 ```
 
-vmselect will propagate this call to the rest of the vmselects listed in its `-selectNode` cmd-line flag. If this
-flag isn't set, then cache need to be purged from each vmselect individually.
+vmselect will propagate this call to the rest of the vmselects listed in its `-selectNode` cmd-line flag when `propagate=1` argument is set.
+If this flag or the `propagate` argument isn't set, then cache need to be purged from each vmselect individually.
+
+If `-search.resetCacheAuthKey` is set, it will be attached to the propagation request as query argument.
 
 ### TCP and UDP
 

docs/victoriametrics/vmagent.md

@@ -797,12 +797,6 @@ For example, the following commands spread scrape targets among a cluster of two
 The `-promscrape.cluster.memberNum` can be set to a StatefulSet pod name when `vmagent` runs in Kubernetes.
 The pod name must end with a number in the range `0 ... promscrape.cluster.membersCount-1`. For example, `-promscrape.cluster.memberNum=vmagent-0`.
 
-By default, targets are sharded among `vmagent` instances by all target labels after relabeling.
-Use `-promscrape.cluster.shardByLabels` {{% available_from "v1.146.0" %}} to shard targets by specified labels instead.
-For example, with `-promscrape.cluster.shardByLabels=service`, the targets with the same `service` label value will be scraped by the same `vmagent` instance, 
-which is useful when perform stream aggregation that requires all metrics with the same `service` label value to be processed on the same `vmagent` instance. 
-If none of the specified labels are present in the target labels, then all target labels will be used for sharding.
-
 By default, each scrape target is scraped only by a single `vmagent` instance in the cluster. If there is a need for replicating scrape targets among multiple `vmagent` instances,
 then `-promscrape.cluster.replicationFactor` command-line flag must be set to the desired number of replicas. For example, the following commands
 start a cluster of three `vmagent` instances, where two `vmagent` instances scrape each target:
@@ -934,29 +928,6 @@ vmagent will generate the following persistent queue folders:
 2_0AAFDF53E314A72A
 ```
 
-### On-disk persistence and data processing order
-
-By default, vmagent processes data in FIFO order. If data has been written to the on-disk queue,
-it must be flushed to the remote storage before newly ingested data can be forwarded there.
-During long outages, vmagent may accumulate large amounts of data in the file-based queue,
-which can introduce a significant lag between the moment data is collected by vmagent and the
-moment it becomes visible at the remote storage.
-
-This behavior can be changed with the `-remoteWrite.inmemoryQueues` {{% available_from "v1.146.0" %}} command-line flag.
-When set to a non-zero value, vmagent starts the given number of additional workers,
-which send only recently ingested data from the in-memory queue, while the workers configured via `-remoteWrite.queues` drain the file-based backlog concurrently.
-This reduces the delivery lag for fresh samples after remote storage outages or slowdowns. The flag can be set individually per each `-remoteWrite.url`.
-
-Note that these workers are started in addition to the workers configured via `-remoteWrite.queues`, so the total number of concurrent connections to
-the remote storage becomes the sum of both flags. Take this into account if the remote storage limits the number of concurrent requests.
-
-This flag has the following possible limitations:
-
-* Samples may arrive at the remote storage out of order, since recent data can be delivered before the older backlogged data.
-  Do not use this option if the remote storage doesn't accept out-of-order samples.
-* Recent data isn't guaranteed to take the fast path: if the in-memory queue  is full,
-  newly ingested data is still written to the file-based queue and is delivered in FIFO order by the generic workers.
-
 ### Disabling On-disk persistence
 
 There are cases when it is better to disable on-disk persistence for pending data on the `vmagent` side:

go.mod

@@ -10,7 +10,7 @@ require (
 	github.com/VictoriaMetrics/VictoriaLogs v1.121.1-0.20260616132739-c901a1e31cb3
 	github.com/VictoriaMetrics/easyproto v1.2.0
 	github.com/VictoriaMetrics/fastcache v1.13.3
-	github.com/VictoriaMetrics/metrics v1.44.0
+	github.com/VictoriaMetrics/metrics v1.43.2
 	github.com/VictoriaMetrics/metricsql v0.87.1
 	github.com/aws/aws-sdk-go-v2 v1.42.0
 	github.com/aws/aws-sdk-go-v2/config v1.32.25

go.sum

@@ -62,8 +62,6 @@ github.com/VictoriaMetrics/fastcache v1.13.3 h1:rBabE0iIxcqKEMCwUmwHZ9dgEqXerg8F
 github.com/VictoriaMetrics/fastcache v1.13.3/go.mod h1:hHXhl4DA2fTL2HTZDJFXWgW0LNjo6B+4aj2Wmng3TjU=
 github.com/VictoriaMetrics/metrics v1.43.2 h1:+8pIQEGwchKS5CYFyvv3LKvNXGi7baZ9hmIV4RHqibY=
 github.com/VictoriaMetrics/metrics v1.43.2/go.mod h1:xDM82ULLYCYdFRgQ2JBxi8Uf1+8En1So9YUwlGTOqTc=
-github.com/VictoriaMetrics/metrics v1.44.0 h1:Fr8yqQSV+ZfYaDD/anqk1E8e9YPgfleSleJmAI0M0Tw=
-github.com/VictoriaMetrics/metrics v1.44.0/go.mod h1:xDM82ULLYCYdFRgQ2JBxi8Uf1+8En1So9YUwlGTOqTc=
 github.com/VictoriaMetrics/metricsql v0.87.1 h1:GdIblCDgXsrBJcBSDtFT8SLK7P+QHijdQmcr4L/f0Go=
 github.com/VictoriaMetrics/metricsql v0.87.1/go.mod h1:d4EisFO6ONP/HIGDYTAtwrejJBBeKGQYiRl095bS4QQ=
 github.com/VividCortex/ewma v1.2.0 h1:f58SaIzcDXrSy3kWaHNvuJgJ3Nmz59Zji6XoJR/q1ow=

lib/backup/actions/restore.go

@@ -91,11 +91,6 @@ func (r *Restore) Run(ctx context.Context) error {
 	if err != nil {
 		return fmt.Errorf("cannot list src parts: %w", err)
 	}
-	for _, srcPart := range srcParts {
-		if !srcPart.IsLocalPathInsideDir(r.Dst.Dir) {
-			return fmt.Errorf("part file %s would be written outside storage directory %s", srcPart.Path, r.Dst.Dir)
-		}
-	}
 	logger.Infof("obtaining list of parts at %s", dst)
 	dstParts, err := dst.ListParts()
 	if err != nil {

lib/backup/common/part.go

@@ -120,17 +120,6 @@ func (p *Part) ParseFromRemotePath(remotePath string) bool {
 	return true
 }
 
-// IsLocalPathInsideDir returns true if the part's local path resolves inside dir.
-// It resolves ../../ sequences and prevents path traversal outside dir.
-func (p *Part) IsLocalPathInsideDir(dir string) bool {
-	dir = filepath.Clean(dir)
-	if dir == `/` {
-		return true
-	}
-
-	return strings.HasPrefix(p.LocalPath(dir), dir+string(filepath.Separator))
-}
-
 // MaxPartSize is the maximum size for each part.
 //
 // The MaxPartSize reduces bandwidth usage during retires on network errors

lib/backup/common/part_test.go

@@ -1,54 +0,0 @@
-package common
-
-import (
-	"testing"
-)
-
-func TestIsLocalPathInsideDir(t *testing.T) {
-	f := func(dir, path string, expected bool) {
-		t.Helper()
-		p := Part{Path: path}
-		if got := p.IsLocalPathInsideDir(dir); got != expected {
-			t.Fatalf("IsLocalPathInsideDir(%q, %q): got %v, want %v", dir, path, got, expected)
-		}
-	}
-
-	// normal path inside dir
-	f("/data/storage", "parts/segment1/data.bin", true)
-
-	// dir with trailing slash is normalized
-	f("/data/storage/", "parts/segment1/data.bin", true)
-
-	// deeply nested path
-	f("/data/storage", "a/b/c/d/e/file.dat", true)
-
-	// traversal that stays inside dir
-	f("/data/storage", "foo/../bar/file.dat", true)
-
-	// root dir allows any path
-	f("/", "any/path/here", true)
-
-	// root dir allows traversal attempts since nothing is outside /
-	f("/", "../outside/marker.txt", true)
-
-	// path with leading slash is treated as relative by filepath.Join and stays inside dir
-	f("/data/storage", "/outside/marker.txt", true)
-
-	// dir with .. components is normalized; path inside resolved dir
-	f("/data/storage/../foo", "parts/file.dat", true)
-
-	// dir with .. components is normalized; traversal outside resolved dir
-	f("/data/storage/../foo", "../storage/evil.txt", false)
-
-	// simple traversal
-	f("/data/storage", "../outside/marker.txt", false)
-
-	// traversal with trailing slash in dir
-	f("/data/storage/", "../outside/marker.txt", false)
-
-	// deep traversal
-	f("/data/storage", "a/../../outside/marker.txt", false)
-
-	// sibling directory whose name shares a prefix with dir
-	f("/data/storage", "../storagefoo/evil.txt", false)
-}

lib/backup/fslocal/fslocal.go

@@ -129,10 +129,6 @@ func (fs *FS) NewReadCloser(p common.Part) (io.ReadCloser, error) {
 // On platforms with preallocation, writes go to a .tmp file that must be
 // finalized with FinalizeFile.
 func (fs *FS) NewDirectWriteCloser(p common.Part) (io.WriteCloser, error) {
-	if !p.IsLocalPathInsideDir(fs.Dir) {
-		logger.Fatalf("BUG: part file %s would be written outside storage directory %s", p.Path, fs.Dir)
-	}
-
 	path := fs.writePath(p)
 	if err := fs.mkdirAll(path); err != nil {
 		return nil, err

lib/fs/dir_remover.go

@@ -179,9 +179,6 @@ func tryRemoveDir(dirPath string) bool {
 			// times simultaneously and properly close it, fs caching may still
 			// confuse NFS client.
 			if err := os.RemoveAll(dirEntryPath); err != nil {
-				if os.IsNotExist(err) {
-					return
-				}
 				if !isTemporaryNFSError(err) {
 					logger.Fatalf("FATAL: cannot remove %q: %s", dirEntryPath, err)
 				}
@@ -206,9 +203,8 @@ func tryRemoveDir(dirPath string) bool {
 
 	deleteFilePath := filepath.Join(dirPath, deleteDirFilename)
 	// Remove the deleteDirFilename file, since there are no other entries left in the directory.
-	if !tryRemovePath(deleteFilePath) {
-		return false
-	}
+	MustRemovePath(deleteFilePath)
+
 	// Sync the directory after the removing deletDirFilename file in order to make sure
 	// all the metadata files are removed at some exotic filesystems such as OSSFS2.
 	// See https://github.com/VictoriaMetrics/VictoriaLogs/issues/649
@@ -216,9 +212,7 @@ func tryRemoveDir(dirPath string) bool {
 	MustSyncPath(dirPath)
 
 	// Remove the dirPath itself
-	if !tryRemovePath(dirPath) {
-		return false
-	}
+	MustRemovePath(dirPath)
 
 	// Do not sync the parent directory for the dirPath - the caller can do this if needed.
 	// It is OK if the dirPath will remain undeleted after unclean shutdown - it will be deleted
@@ -227,23 +221,6 @@ func tryRemoveDir(dirPath string) bool {
 	return true
 }
 
-// tryRemovePath removes given path and returns true on success
-// or false if error is temporary NFS error
-func tryRemovePath(path string) bool {
-	if err := os.Remove(path); err != nil {
-		if os.IsNotExist(err) {
-			return true
-		}
-		if !isTemporaryNFSError(err) {
-			logger.Fatalf("FATAL: cannot remove %q: %s", path, err)
-		}
-		nfsDirRemoveFailedAttempts.Inc()
-		return false
-	}
-
-	return true
-}
-
 var (
 	dirRemoverWG               sync.WaitGroup
 	nfsDirRemoveFailedAttempts = metrics.NewCounter(`vm_nfs_dir_remove_failed_attempts_total`)

lib/httpserver/httpserver.go

@@ -811,20 +811,10 @@ func LogError(req *http.Request, errStr string) {
 	logger.Errorf("uri: %s, remote address: %q: %s", uri, remoteAddr, errStr)
 }
 
-// tlsErrorSkipLogger must be passed as the out argument to log.New only.
-// It suppresses noisy TCP probe errors on TLS connections to avoid log pollution.
-//
-// This cannot be implemented in net.Listener because a TLS handshake may take seconds,
-// during which no other connections can be accepted. Therefor, the implementation inside net.Listener can lead to DoS.
-// Once a connection is passed to the conn serve goroutine, there is no direct access to the handshake logic, so this indirect
-// approach is used instead.
 type tlsErrorSkipLogger struct{}
 
-// Write filters out TLS handshake errors from health-check probes.
-// log.Logger guarantees that each complete message is delivered in a single Write call
-// and that calls are serialized, so we can safely inspect p for a "TLS handshake error".
-// See https://github.com/golang/go/blob/38e988efb4b8f5e73e887027f386a342c138b649/src/log/log.go#L53-L57
 func (*tlsErrorSkipLogger) Write(p []byte) (int, error) {
+	// skip common health check errors produced by Kubernetes and other tools
 	if bytes.Contains(p, []byte("TLS handshake error")) &&
 		(bytes.Contains(p, []byte("EOF")) || bytes.Contains(p, []byte("connection reset by peer"))) {
 		return len(p), nil

lib/persistentqueue/fastqueue.go

@@ -26,8 +26,6 @@ type FastQueue struct {
 	// isPQDisabled is set to true when pq is disabled.
 	isPQDisabled bool
 
-	prioritizeInMemoryData bool
-
 	// pq is file-based queue
 	pq *queue
 
@@ -41,31 +39,6 @@ type FastQueue struct {
 	stopDeadline uint64
 }
 
-// OpenFastQueueOpts defines options for FastQueue
-type OpenFastQueueOpts struct {
-	// MaxInmemoryBlocks defines amount of blocks to hold in memory before falling back to file-based persistence.
-	MaxInmemoryBlocks int
-	// MaxPendingBytes limits file-based size of the queue.
-	// If MaxPendingBytes is 0, then the queue size is unlimited.
-	// The oldest data is dropped when the queue
-	// reaches MaxPendingSize.
-	MaxPendingBytes int64
-	// IsPQDisabled defines whether file-based queue could be used.
-	// If it is set to true, then write requests that exceed in-memory buffer capacity are rejected.
-	// in-memory queue part can be stored on disk during graceful shutdown.
-	IsPQDisabled bool
-	// PrioritizeInMemoryData instructs FastQueue to write data into the in-memory queue
-	// even if the file-based queue is not empty.
-	// This is useful when data order doesn't matter and getting the most recent data
-	// as fast as possible is more important.
-	PrioritizeInmemoryData bool
-}
-
-// MustOpenFastQueueWithOpts opens persistent queue at the given path with given opts
-func MustOpenFastQueueWithOpts(path, name string, opts OpenFastQueueOpts) *FastQueue {
-	return mustOpenFastQueue(path, name, opts)
-}
-
 // MustOpenFastQueue opens persistent queue at the given path.
 //
 // It holds up to maxInmemoryBlocks in memory before falling back to file-based persistence.
@@ -76,22 +49,11 @@ func MustOpenFastQueueWithOpts(path, name string, opts OpenFastQueueOpts) *FastQ
 // if isPQDisabled is set to true, then write requests that exceed in-memory buffer capacity are rejected.
 // in-memory queue part can be stored on disk during graceful shutdown.
 func MustOpenFastQueue(path, name string, maxInmemoryBlocks int, maxPendingBytes int64, isPQDisabled bool) *FastQueue {
-	opts := OpenFastQueueOpts{
-		MaxInmemoryBlocks: maxInmemoryBlocks,
-		MaxPendingBytes:   maxPendingBytes,
-		IsPQDisabled:      isPQDisabled,
-	}
-	return mustOpenFastQueue(path, name, opts)
-}
-func mustOpenFastQueue(path, name string, opts OpenFastQueueOpts) *FastQueue {
-	maxPendingBytes := opts.MaxPendingBytes
-	isPQDisabled := opts.IsPQDisabled
 	pq := mustOpen(path, name, maxPendingBytes)
 	fq := &FastQueue{
-		pq:                     pq,
-		isPQDisabled:           isPQDisabled,
-		prioritizeInMemoryData: opts.PrioritizeInmemoryData,
-		ch:                     make(chan *bytesutil.ByteBuffer, opts.MaxInmemoryBlocks),
+		pq:           pq,
+		isPQDisabled: isPQDisabled,
+		ch:           make(chan *bytesutil.ByteBuffer, maxInmemoryBlocks),
 	}
 	fq.cond.L = &fq.mu
 	fq.lastInmemoryBlockReadTime = fasttime.UnixTimestamp()
@@ -119,7 +81,7 @@ func mustOpenFastQueue(path, name string, opts OpenFastQueueOpts) *FastQueue {
 	if isPQDisabled {
 		persistenceStatus = "disabled"
 	}
-	logger.Infof("opened fast queue at %q with maxInmemoryBlocks=%d, it contains %d pending bytes, persistence is %s", path, opts.MaxInmemoryBlocks, pendingBytes, persistenceStatus)
+	logger.Infof("opened fast queue at %q with maxInmemoryBlocks=%d, it contains %d pending bytes, persistence is %s", path, maxInmemoryBlocks, pendingBytes, persistenceStatus)
 	return fq
 }
 
@@ -135,7 +97,7 @@ func (fq *FastQueue) IsWriteBlocked() bool {
 	}
 	fq.mu.Lock()
 	defer fq.mu.Unlock()
-	return len(fq.ch) == cap(fq.ch) || (fq.pq.GetPendingBytes() > 0 && !fq.prioritizeInMemoryData)
+	return len(fq.ch) == cap(fq.ch) || fq.pq.GetPendingBytes() > 0
 }
 
 // UnblockAllReaders unblocks all the readers.
@@ -231,24 +193,19 @@ func (fq *FastQueue) tryWriteBlock(block []byte, ignoreDisabledPQ bool) bool {
 	defer fq.mu.Unlock()
 
 	isPQWriteAllowed := !fq.isPQDisabled || ignoreDisabledPQ
-	if !isPQWriteAllowed && fq.pq.GetPendingBytes() > 0 {
-		// fast path: there is pending data at file-based queue,
-		// it must be drained before in-memory queue could be used.
-		// File-based queue could be non-empty after vmagent restart
-		// and vmagent couldn't flush in-memory queue during shutdown.
-		return false
-	}
-	if !fq.prioritizeInMemoryData {
-		fq.flushInmemoryBlocksToFileIfNeededLocked()
-		if n := fq.pq.GetPendingBytes(); n > 0 {
-			// The file-based queue isn't drained yet. This means that in-memory queue cannot be used yet.
-			// So put the block to file-based queue.
-			if len(fq.ch) > 0 {
-				logger.Panicf("BUG: the in-memory queue must be empty when the file-based queue is non-empty; it contains %d pending bytes", n)
-			}
-			fq.pq.MustWriteBlock(block)
-			return true
+
+	fq.flushInmemoryBlocksToFileIfNeededLocked()
+	if n := fq.pq.GetPendingBytes(); n > 0 {
+		// The file-based queue isn't drained yet. This means that in-memory queue cannot be used yet.
+		// So put the block to file-based queue.
+		if len(fq.ch) > 0 {
+			logger.Panicf("BUG: the in-memory queue must be empty when the file-based queue is non-empty; it contains %d pending bytes", n)
 		}
+		if !isPQWriteAllowed {
+			return false
+		}
+		fq.pq.MustWriteBlock(block)
+		return true
 	}
 	if len(fq.ch) == cap(fq.ch) {
 		// There is no space left in the in-memory queue. Put the data to file-based queue.
@@ -259,7 +216,7 @@ func (fq *FastQueue) tryWriteBlock(block []byte, ignoreDisabledPQ bool) bool {
 		fq.pq.MustWriteBlock(block)
 		return true
 	}
-
+	// Fast path - put the block to in-memory queue.
 	bb := blockBufPool.Get()
 	bb.B = append(bb.B[:0], block...)
 	fq.ch <- bb
@@ -272,7 +229,7 @@ func (fq *FastQueue) tryWriteBlock(block []byte, ignoreDisabledPQ bool) bool {
 }
 
 // MustReadBlock reads the next block from fq into dst and returns it.
-// It first reads from the file-based queue, then checks in-memory queue.
+// It first reads from the in-memory queue, then checks file-based queue.
 // It blocks until a block is available or the stop deadline is exceeded, in which case it returns (dst, false).
 func (fq *FastQueue) MustReadBlock(dst []byte) ([]byte, bool) {
 	fq.mu.Lock()
@@ -282,15 +239,16 @@ func (fq *FastQueue) MustReadBlock(dst []byte) ([]byte, bool) {
 		if fq.stopDeadline > 0 && fasttime.UnixTimestamp() > fq.stopDeadline {
 			return dst, false
 		}
+		if len(fq.ch) > 0 {
+			return fq.mustReadInMemoryBlockLocked(dst), true
+		}
 		if n := fq.pq.GetPendingBytes(); n > 0 {
 			data, ok := fq.pq.MustReadBlockNonblocking(dst)
 			if ok {
 				return data, true
 			}
 			dst = data
-		}
-		if len(fq.ch) > 0 {
-			return fq.mustReadInMemoryBlockLocked(dst), true
+			continue
 		}
 		if fq.stopDeadline > 0 {
 			return dst, false
@@ -301,27 +259,6 @@ func (fq *FastQueue) MustReadBlock(dst []byte) ([]byte, bool) {
 	}
 }
 
-// MustReadInMemoryBlockBlocking reads the next block from the in-memory queue into dst and returns it.
-// It blocks until a block is available or the stop deadline is exceeded, in which case it returns (dst, false).
-func (fq *FastQueue) MustReadInMemoryBlockBlocking(dst []byte) ([]byte, bool) {
-	fq.mu.Lock()
-	defer fq.mu.Unlock()
-
-	for {
-		if fq.stopDeadline > 0 && fasttime.UnixTimestamp() > fq.stopDeadline {
-			return dst, false
-		}
-		if len(fq.ch) > 0 {
-			return fq.mustReadInMemoryBlockLocked(dst), true
-		}
-		if fq.stopDeadline > 0 {
-			return dst, false
-		}
-		// There are no blocks. Wait for new block.
-		fq.cond.Wait()
-	}
-}
-
 // MustReadInMemoryBlock reads the next block from the in-memory queue into dst and returns it.
 // It returns (dst, true) if a block was available, or (nil, false) if the in-memory queue is empty.
 // It does not block waiting for new blocks.
@@ -340,6 +277,9 @@ func (fq *FastQueue) mustReadInMemoryBlockLocked(dst []byte) []byte {
 	if len(fq.ch) == 0 {
 		logger.Panicf("BUG: the function must not be called when in-memory queue is empty. Caller should verify the queue len upfront")
 	}
+	if n := fq.pq.GetPendingBytes(); n > 0 {
+		logger.Panicf("BUG: the file-based queue must be empty when the in-memory queue is non-empty; it contains %d pending bytes", n)
+	}
 	bb := <-fq.ch
 	fq.pendingInmemoryBytes -= uint64(len(bb.B))
 	fq.lastInmemoryBlockReadTime = fasttime.UnixTimestamp()

lib/persistentqueue/fastqueue_test.go

@@ -364,64 +364,3 @@ func TestFastQueueWriteReadWithIgnoreDisabledPQ(t *testing.T) {
 	fq.MustClose()
 	fs.MustRemoveDir(path)
 }
-
-func TestFastQueueWriteReadWithPrioritizeInmemory(t *testing.T) {
-	path := "fast-queue-write-read-inmemory-disabled-pq-force-write"
-	fs.MustRemoveDir(path)
-
-	capacity := 20
-	opts := OpenFastQueueOpts{
-		MaxInmemoryBlocks:      capacity,
-		PrioritizeInmemoryData: true,
-	}
-	fq := MustOpenFastQueueWithOpts(path, "foobar", opts)
-	if n := fq.GetInmemoryQueueLen(); n != 0 {
-		t.Fatalf("unexpected non-zero inmemory queue size:  %d", n)
-	}
-	var blocks []string
-	for i := range capacity {
-		block := fmt.Sprintf("block %d", i)
-		if !fq.TryWriteBlock([]byte(block)) {
-			t.Fatalf("TryWriteBlock must return true in this context")
-		}
-		blocks = append(blocks, block)
-	}
-	if n := fq.GetInmemoryQueueLen(); n != capacity {
-		t.Fatalf("unexpected non-zero inmemory queue size:  %d: %d", n, capacity)
-	}
-	for i := range capacity {
-		block := fmt.Sprintf("block %d-%d", i, i)
-		if !fq.TryWriteBlock([]byte(block)) {
-			t.Fatalf("TryWriteBlock must return true in this context")
-		}
-		blocks = append(blocks, block)
-	}
-
-	// in case of capacity exceed last element is written into file-based queue
-	if n := fq.GetInmemoryQueueLen(); n != capacity-1 {
-		t.Fatalf("unexpected non-zero inmemory queue size:  %d: %d", n, capacity)
-	}
-
-	// make sure that recently ingested elemements returned first
-	for idx := capacity + 1; idx < capacity*2; idx++ {
-		buf, ok := fq.MustReadInMemoryBlockBlocking(nil)
-		if !ok {
-			t.Fatalf("unexpected ok=false")
-		}
-		if string(buf) != blocks[idx] {
-			t.Fatalf("unexpected block read; got %q; want %q: %d", buf, blocks[idx], idx)
-		}
-	}
-	blocks = blocks[:capacity+1]
-	for _, block := range blocks {
-		buf, ok := fq.MustReadBlock(nil)
-		if !ok {
-			t.Fatalf("unexpected ok=false")
-		}
-		if string(buf) != block {
-			t.Fatalf("unexpected block read; got %q; want %q", buf, block)
-		}
-	}
-	fq.MustClose()
-	fs.MustRemoveDir(path)
-}

lib/promrelabel/if_expression.go

@@ -50,17 +50,6 @@ func (ie *IfExpression) Parse(s string) error {
 	return nil
 }
 
-// ParseFromMetricExpr parses if from given MetricExpr
-func (ie *IfExpression) ParseFromMetricExpr(me *metricsql.MetricExpr) error {
-	var ieLocal ifExpression
-	if err := ieLocal.parseFromMetricExpr(me); err != nil {
-		return err
-	}
-
-	ie.ies = []*ifExpression{&ieLocal}
-	return nil
-}
-
 // UnmarshalJSON unmarshals ie from JSON data.
 func (ie *IfExpression) UnmarshalJSON(data []byte) error {
 	var v any
@@ -193,16 +182,6 @@ func (ie *ifExpression) Parse(s string) error {
 	return nil
 }
 
-func (ie *ifExpression) parseFromMetricExpr(me *metricsql.MetricExpr) error {
-	lfss, err := metricExprToLabelFilterss(me)
-	if err != nil {
-		return fmt.Errorf("cannot parse series selector: %w", err)
-	}
-	ie.s = string(me.AppendString(nil))
-	ie.lfss = lfss
-	return nil
-}
-
 // UnmarshalJSON unmarshals ie from JSON data.
 func (ie *ifExpression) UnmarshalJSON(data []byte) error {
 	var s string

lib/promscrape/config.go

@@ -76,9 +76,6 @@ var (
 		"Every %d occurrence in the template is substituted with -promscrape.cluster.memberNum at urls to vmagent instances responsible for scraping the given target "+
 		"at /service-discovery page. For example -promscrape.cluster.memberURLTemplate='http://vmagent-%d:8429/targets'. "+
 		"See https://docs.victoriametrics.com/victoriametrics/vmagent/#scraping-big-number-of-targets for more details")
-	clusterShardByLabels = flagutil.NewArrayString("promscrape.cluster.shardByLabels", "Optional list of target labels, which will be used for sharding targets among cluster members "+
-		"if -promscrape.cluster.membersCount is greater than 1. If none of the specified labels are found in a target, then all the target labels will be used for sharding. "+
-		"See https://docs.victoriametrics.com/victoriametrics/vmagent/#scraping-big-number-of-targets for more info")
 	clusterReplicationFactor = flag.Int("promscrape.cluster.replicationFactor", 1, "The number of members in the cluster, which scrape the same targets. "+
 		"If the replication factor is greater than 1, then the deduplication must be enabled at remote storage side. "+
 		"See https://docs.victoriametrics.com/victoriametrics/vmagent/#scraping-big-number-of-targets for more info")
@@ -89,10 +86,7 @@ var (
 		"Bigger uncompressed responses are rejected. See also max_scrape_size option at https://docs.victoriametrics.com/victoriametrics/sd_configs/#scrape_configs")
 )
 
-var (
-	clusterMemberID            int
-	clusterShardByLabelsSorted []string
-)
+var clusterMemberID int
 
 func mustInitClusterMemberID() {
 	s := *clusterMemberNum
@@ -116,15 +110,6 @@ func mustInitClusterMemberID() {
 	clusterMemberID = n
 }
 
-func initClusterShardByLabels() {
-	if len(*clusterShardByLabels) == 0 {
-		clusterShardByLabelsSorted = nil
-		return
-	}
-	clusterShardByLabelsSorted = slices.Clone(*clusterShardByLabels)
-	slices.Sort(clusterShardByLabelsSorted)
-}
-
 // Config represents essential parts from Prometheus config defined at https://prometheus.io/docs/prometheus/latest/configuration/configuration/
 type Config struct {
 	Global            GlobalConfig    `yaml:"global,omitempty"`
@@ -1153,28 +1138,12 @@ func (stc *StaticConfig) appendScrapeWork(dst []*ScrapeWork, swc *scrapeWorkConf
 }
 
 func appendScrapeWorkKey(dst []byte, labels *promutil.Labels) []byte {
-	originalDstLen := len(dst)
-	for _, targetLabelName := range clusterShardByLabelsSorted {
-		for _, label := range labels.GetLabels() {
-			if label.Name == targetLabelName {
-				// Do not use strconv.AppendQuote, since it is slow according to CPU profile.
-				dst = append(dst, label.Name...)
-				dst = append(dst, '=')
-				dst = append(dst, label.Value...)
-				dst = append(dst, ',')
-				break
-			}
-		}
-	}
-	// Use all labels to compute the key if `promscrape.cluster.shardByLabels` is not configured
-	if len(dst) == originalDstLen {
-		for _, label := range labels.GetLabels() {
-			dst = append(dst, label.Name...)
-			dst = append(dst, '=')
-			dst = append(dst, label.Value...)
-			dst = append(dst, ',')
-		}
-		return dst
+	for _, label := range labels.GetLabels() {
+		// Do not use strconv.AppendQuote, since it is slow according to CPU profile.
+		dst = append(dst, label.Name...)
+		dst = append(dst, '=')
+		dst = append(dst, label.Value...)
+		dst = append(dst, ',')
 	}
 	return dst
 }

lib/promscrape/config_test.go

@@ -148,77 +148,6 @@ func TestGetClusterMemberNumsForScrapeWork(t *testing.T) {
 	f("foo", 3, 2, []int{2, 0})
 }
 
-func TestAppendScrapeWorkKeyShardByLabels(t *testing.T) {
-	f := func(labels map[string]string, shardByLabels []string, expectedKey string) {
-		t.Helper()
-		originValue := *clusterShardByLabels
-		*clusterShardByLabels = shardByLabels
-		defer func() {
-			*clusterShardByLabels = originValue
-		}()
-		initClusterShardByLabels()
-		outputKey := string(appendScrapeWorkKey(nil, promutil.NewLabelsFromMap(labels)))
-		if expectedKey != outputKey {
-			t.Fatalf("unexpected sharding key:%q for target labels:%v with shardByLabels=%q, expect: %q",
-				outputKey, labels, shardByLabels, expectedKey)
-		}
-	}
-
-	// didn't specify -promscrape.cluster.shardByLabels, so all labels will be used for sharding
-	f(
-		map[string]string{
-			"a": "aa",
-			"b": "bb",
-			"c": "cc",
-			"d": "dd"},
-		[]string{},
-		"a=aa,b=bb,c=cc,d=dd,",
-	)
-
-	// match all labels in -promscrape.cluster.shardByLabels, so label "a" and "c" will be used for sharding
-	f(
-		map[string]string{
-			"a": "aa",
-			"b": "bb",
-			"c": "cc",
-			"d": "dd"},
-		[]string{"a", "c"},
-		"a=aa,c=cc,",
-	)
-
-	// match all labels in -promscrape.cluster.shardByLabels, so label "a" and "c" will be used for sharding even if they're not in order in -promscrape.cluster.shardByLabels.
-	f(
-		map[string]string{
-			"a": "aa",
-			"b": "bb",
-			"c": "cc",
-			"d": "dd"},
-		[]string{"c", "a"},
-		"a=aa,c=cc,",
-	)
-
-	// match part of labels in -promscrape.cluster.shardByLabels, label "a" and "c" will be used for sharding
-	f(
-		map[string]string{
-			"a": "aa",
-			"c": "cc",
-			"d": "dd"},
-
-		[]string{"a", "b", "c"},
-		"a=aa,c=cc,",
-	)
-
-	// none of labels in -promscrape.cluster.shardByLabels is matched, so all labels will be used for sharding
-	f(
-		map[string]string{
-			"d": "dd",
-			"e": "ee"},
-		[]string{"a", "b", "c"},
-		"d=dd,e=ee,",
-	)
-
-}
-
 func TestLoadStaticConfigs(t *testing.T) {
 	scs, err := loadStaticConfigs("testdata/file_sd.json")
 	if err != nil {

lib/promscrape/scraper.go

@@ -66,7 +66,6 @@ func CheckConfig() error {
 // Scraped data is passed to pushData.
 func Init(pushData func(at *auth.Token, wr *prompb.WriteRequest)) {
 	mustInitClusterMemberID()
-	initClusterShardByLabels()
 	globalStopChan = make(chan struct{})
 	scraperWG.Go(func() {
 		runScraper(*promscrapeConfigFile, pushData, globalStopChan)

lib/streamaggr/dedup.go

@@ -1,7 +1,6 @@
 package streamaggr
 
 import (
-	"fmt"
 	"sync"
 	"sync/atomic"
 	"unsafe"
@@ -18,10 +17,9 @@ import (
 const dedupAggrShardsCount = 128
 
 type dedupAggr struct {
-	shards         []dedupAggrShard
-	flushDuration  *metrics.Histogram
-	flushTimeouts  *metrics.Counter
-	droppedSamples *metrics.Counter
+	shards        []dedupAggrShard
+	flushDuration *metrics.Histogram
+	flushTimeouts *metrics.Counter
 }
 
 type dedupAggrShard struct {
@@ -49,20 +47,10 @@ type dedupAggrSample struct {
 	timestamp int64
 }
 
-func newDedupAggr(ms *metrics.Set, metricLabels string) *dedupAggr {
-	var d dedupAggr
-	d.shards = make([]dedupAggrShard, dedupAggrShardsCount)
-	_ = ms.NewGauge(fmt.Sprintf(`vm_streamaggr_dedup_state_size_bytes{%s}`, metricLabels), func() float64 {
-		return float64(d.sizeBytes())
-	})
-	_ = ms.NewGauge(fmt.Sprintf(`vm_streamaggr_dedup_state_items_count{%s}`, metricLabels), func() float64 {
-		return float64(d.itemsCount())
-	})
-
-	d.flushDuration = ms.NewHistogram(fmt.Sprintf(`vm_streamaggr_dedup_flush_duration_seconds{%s}`, metricLabels))
-	d.flushTimeouts = ms.NewCounter(fmt.Sprintf(`vm_streamaggr_dedup_flush_timeouts_total{%s}`, metricLabels))
-	d.droppedSamples = ms.NewCounter(fmt.Sprintf(`vm_streamaggr_dedup_dropped_samples_total{%s}`, metricLabels))
-	return &d
+func newDedupAggr() *dedupAggr {
+	return &dedupAggr{
+		shards: make([]dedupAggrShard, dedupAggrShardsCount),
+	}
 }
 
 func (da *dedupAggr) sizeBytes() uint64 {
@@ -99,8 +87,7 @@ func (da *dedupAggr) pushSamples(samples []pushSample, _ int64, isGreen bool) {
 		if len(shardSamples) == 0 {
 			continue
 		}
-		deduplicatedSamples := da.shards[i].pushSamples(shardSamples, isGreen)
-		da.droppedSamples.Add(deduplicatedSamples)
+		da.shards[i].pushSamples(shardSamples, isGreen)
 	}
 	putPerShardSamples(pss)
 }
@@ -180,9 +167,8 @@ func putPerShardSamples(pss *perShardSamples) {
 
 var perShardSamplesPool sync.Pool
 
-func (das *dedupAggrShard) pushSamples(samples []pushSample, isGreen bool) int {
+func (das *dedupAggrShard) pushSamples(samples []pushSample, isGreen bool) {
 	var state *dedupAggrState
-	var deduplicatedSamples int
 
 	if isGreen {
 		state = &das.green
@@ -212,10 +198,8 @@ func (das *dedupAggrShard) pushSamples(samples []pushSample, isGreen bool) int {
 			continue
 		}
 		s.timestamp, s.value = deduplicateSamples(s.timestamp, sample.timestamp, s.value, sample.value)
-		deduplicatedSamples++
 	}
 	state.samplesBuf = samplesBuf
-	return deduplicatedSamples
 }
 
 // deduplicateSamples returns deduplicated timestamp and value results.

lib/streamaggr/dedup_test.go

@@ -7,13 +7,11 @@ import (
 	"testing"
 	"time"
 
-	"github.com/VictoriaMetrics/metrics"
-
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/decimal"
 )
 
 func TestDedupAggrSerial(t *testing.T) {
-	da := newDedupAggr(metrics.NewSet(), "")
+	da := newDedupAggr()
 
 	const seriesCount = 100_000
 	expectedSamplesMap := make(map[string]pushSample)
@@ -61,7 +59,7 @@ func TestDedupAggrSerial(t *testing.T) {
 func TestDedupAggrConcurrent(_ *testing.T) {
 	const concurrency = 5
 	const seriesCount = 10_000
-	da := newDedupAggr(metrics.NewSet(), "")
+	da := newDedupAggr()
 
 	var wg sync.WaitGroup
 	for range concurrency {

lib/streamaggr/dedup_timing_test.go

@@ -5,8 +5,6 @@ import (
 	"sync/atomic"
 	"testing"
 
-	"github.com/VictoriaMetrics/metrics"
-
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
 )
 
@@ -25,7 +23,7 @@ func benchmarkDedupAggr(b *testing.B, samplesPerPush int) {
 
 	const loops = 2
 	benchSamples := newBenchSamples(samplesPerPush)
-	da := newDedupAggr(metrics.NewSet(), "")
+	da := newDedupAggr()
 
 	b.ResetTimer()
 	b.ReportAllocs()

lib/streamaggr/deduplicator.go

@@ -44,6 +44,7 @@ type Deduplicator struct {
 // MustStop must be called on the returned deduplicator in order to free up occupied resources.
 func NewDeduplicator(pushFunc PushFunc, enableWindows bool, interval time.Duration, dropLabels []string, alias string) *Deduplicator {
 	d := &Deduplicator{
+		da:            newDedupAggr(),
 		dropLabels:    dropLabels,
 		interval:      interval,
 		enableWindows: enableWindows,
@@ -63,7 +64,16 @@ func NewDeduplicator(pushFunc PushFunc, enableWindows bool, interval time.Durati
 	ms := d.ms
 
 	metricLabels := fmt.Sprintf(`name="dedup",url=%q`, alias)
-	d.da = newDedupAggr(ms, metricLabels)
+
+	_ = ms.NewGauge(fmt.Sprintf(`vm_streamaggr_dedup_state_size_bytes{%s}`, metricLabels), func() float64 {
+		return float64(d.da.sizeBytes())
+	})
+	_ = ms.NewGauge(fmt.Sprintf(`vm_streamaggr_dedup_state_items_count{%s}`, metricLabels), func() float64 {
+		return float64(d.da.itemsCount())
+	})
+
+	d.da.flushDuration = ms.NewHistogram(fmt.Sprintf(`vm_streamaggr_dedup_flush_duration_seconds{%s}`, metricLabels))
+	d.da.flushTimeouts = ms.NewCounter(fmt.Sprintf(`vm_streamaggr_dedup_flush_timeouts_total{%s}`, metricLabels))
 
 	metrics.RegisterSet(ms)
 
@@ -110,7 +120,6 @@ func (d *Deduplicator) Push(tss []prompb.TimeSeries) {
 		key := bytesutil.ToUnsafeString(buf[bufLen:])
 		for _, s := range ts.Samples {
 			if d.enableWindows && minDeadline > s.Timestamp {
-				d.da.droppedSamples.Inc()
 				continue
 			} else if d.enableWindows && s.Timestamp <= cs.maxDeadline == cs.isGreen {
 				ctx.green = append(ctx.green, pushSample{

lib/streamaggr/increase.go

@@ -31,16 +31,12 @@ type increaseAggrValue struct {
 }
 
 func (av *increaseAggrValue) pushSample(c aggrConfig, sample *pushSample, key string, deleteDeadline int64) {
-	if av.total == nil {
-		av.total = new(float64)
-	}
 	ac := c.(*increaseAggrConfig)
 	currentTime := fasttime.UnixTimestamp()
 	keepFirstSample := ac.keepFirstSample && currentTime >= ac.ignoreFirstSampleDeadline
 	lv, ok := av.shared[key]
-	// The last value is stale, reset it.
-	if ok && lv.deleteDeadline < int64(currentTime)*1000 {
-		ok = false
+	if av.total == nil {
+		av.total = new(float64)
 	}
 	if ok {
 		if sample.timestamp < lv.timestamp {

lib/streamaggr/rate.go

@@ -7,7 +7,6 @@ import (
 	"github.com/VictoriaMetrics/metrics"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
 )
 
 var rateAggrSharedValuePool sync.Pool
@@ -100,12 +99,6 @@ func (av *rateAggrValue) pushSample(c aggrConfig, sample *pushSample, key string
 	ac := c.(*rateAggrConfig)
 	var state *rateAggrStateValue
 	sv, ok := av.shared[key]
-	// The last value is stale, reset it.
-	if ok && sv.deleteDeadline < int64(fasttime.UnixTimestamp())*1000 {
-		delete(av.shared, key)
-		putRateAggrSharedValue(sv)
-		ok = false
-	}
 	if ok {
 		state = sv.getState(av.isGreen)
 		if sample.timestamp < state.timestamp {

lib/streamaggr/streamaggr.go

@@ -43,7 +43,6 @@ var supportedOutputs = []string{
 	"stddev",
 	"stdvar",
 	"sum_samples",
-	"sum_samples_total",
 	"total",
 	"total_prometheus",
 	"unique_samples",
@@ -173,12 +172,12 @@ type Config struct {
 	DedupInterval string `yaml:"dedup_interval,omitempty"`
 
 	// Staleness interval is interval after which the series state will be reset if no samples have been sent during it.
-	// The parameter is only relevant for outputs: total, total_prometheus, increase, increase_prometheus, rate_avg and rate_sum.
+	// The parameter is only relevant for outputs: total, total_prometheus, increase, increase_prometheus and histogram_bucket.
 	StalenessInterval string `yaml:"staleness_interval,omitempty"`
 
 	// IgnoreFirstSampleInterval specifies the interval after which the agent begins sending samples.
 	// By default, it is set to the staleness interval, and it helps reduce the initial sample load after an agent restart.
-	// This parameter is relevant only for the following outputs: total, total_prometheus, increase and increase_prometheus.
+	// This parameter is relevant only for the following outputs: total, total_prometheus, increase, increase_prometheus, and histogram_bucket.
 	IgnoreFirstSampleInterval string `yaml:"ignore_first_sample_interval,omitempty"`
 
 	// Outputs is a list of output aggregate functions to produce.
@@ -502,9 +501,8 @@ func newAggregator(cfg *Config, path string, pushFunc PushFunc, ms *metrics.Set,
 		return nil, fmt.Errorf("interval=%s must be a multiple of dedup_interval=%s", interval, dedupInterval)
 	}
 
-	// set the default staleness interval as the aggregation interval, to be consistent with query lookbehind window in metricsQL,
-	// see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11102
-	stalenessInterval := interval
+	// check cfg.StalenessInterval
+	stalenessInterval := interval * 2
 	if cfg.StalenessInterval != "" {
 		stalenessInterval, err = time.ParseDuration(cfg.StalenessInterval)
 		if err != nil {
@@ -670,7 +668,18 @@ func newAggregator(cfg *Config, path string, pushFunc PushFunc, ms *metrics.Set,
 	}
 
 	if dedupInterval > 0 {
-		a.da = newDedupAggr(ms, metricLabels)
+		a.da = newDedupAggr()
+		a.da.flushTimeouts = ms.NewCounter(fmt.Sprintf(`vm_streamaggr_dedup_flush_timeouts_total{%s}`, metricLabels))
+		a.da.flushDuration = ms.NewHistogram(fmt.Sprintf(`vm_streamaggr_dedup_flush_duration_seconds{%s}`, metricLabels))
+
+		_ = ms.NewGauge(fmt.Sprintf(`vm_streamaggr_dedup_state_size_bytes{%s}`, metricLabels), func() float64 {
+			n := a.da.sizeBytes()
+			return float64(n)
+		})
+		_ = ms.NewGauge(fmt.Sprintf(`vm_streamaggr_dedup_state_items_count{%s}`, metricLabels), func() float64 {
+			n := a.da.itemsCount()
+			return float64(n)
+		})
 	}
 
 	alignFlushToInterval := !opts.NoAlignFlushToInterval
@@ -771,9 +780,7 @@ func newOutputConfig(ms *metrics.Set, metricLabels, output string, outputsSeen m
 	case "stdvar":
 		return newStdvarAggrConfig(), nil
 	case "sum_samples":
-		return newSumSamplesAggrConfig(true), nil
-	case "sum_samples_total":
-		return newSumSamplesAggrConfig(false), nil
+		return newSumSamplesAggrConfig(), nil
 	case "total":
 		return newTotalAggrConfig(ms, metricLabels, ignoreFirstSampleIntervalSecs, true), nil
 	case "total_prometheus":

lib/streamaggr/streamaggr_synctest_test.go

@@ -1,5 +1,3 @@
-//go:build synctest
-
 package streamaggr
 
 import (
@@ -477,125 +475,26 @@ foo:1m_increase_prometheus{baz="qwe"} 15
   outputs: [increase_prometheus]
 `, "11111111")
 
-	// increase, increase_prometheus, total, total_prometheus outputs with different staleness intervals
-	f([]string{`
-foo 5
-bar 200
-`, `
-foo 10
-bar 201
-`, ``, `
-foo 7
-bar 205
-`}, time.Minute, `bar:1m_increase 200
-bar:1m_increase 1
-bar:1m_increase 205
-bar:1m_increase_prometheus 0
-bar:1m_increase_prometheus 1
-bar:1m_increase_prometheus 0
-bar:1m_total 200
-bar:1m_total 201
-bar:1m_total 205
-bar:1m_total_prometheus 0
-bar:1m_total_prometheus 1
-bar:1m_total_prometheus 0
-bar:1m_without_non_existing_label_increase 0
-bar:1m_without_non_existing_label_increase 1
-bar:1m_without_non_existing_label_increase 4
-bar:1m_without_non_existing_label_increase_prometheus 0
-bar:1m_without_non_existing_label_increase_prometheus 1
-bar:1m_without_non_existing_label_increase_prometheus 4
-bar:1m_without_non_existing_label_total 0
-bar:1m_without_non_existing_label_total 1
-bar:1m_without_non_existing_label_total 1
-bar:1m_without_non_existing_label_total 5
-bar:1m_without_non_existing_label_total_prometheus 0
-bar:1m_without_non_existing_label_total_prometheus 1
-bar:1m_without_non_existing_label_total_prometheus 1
-bar:1m_without_non_existing_label_total_prometheus 5
-foo:1m_increase 5
-foo:1m_increase 5
-foo:1m_increase 7
-foo:1m_increase_prometheus 0
-foo:1m_increase_prometheus 5
-foo:1m_increase_prometheus 0
-foo:1m_total 5
-foo:1m_total 10
-foo:1m_total 7
-foo:1m_total_prometheus 0
-foo:1m_total_prometheus 5
-foo:1m_total_prometheus 0
-foo:1m_without_non_existing_label_increase 0
-foo:1m_without_non_existing_label_increase 5
-foo:1m_without_non_existing_label_increase 7
-foo:1m_without_non_existing_label_increase_prometheus 0
-foo:1m_without_non_existing_label_increase_prometheus 5
-foo:1m_without_non_existing_label_increase_prometheus 7
-foo:1m_without_non_existing_label_total 0
-foo:1m_without_non_existing_label_total 5
-foo:1m_without_non_existing_label_total 5
-foo:1m_without_non_existing_label_total 12
-foo:1m_without_non_existing_label_total_prometheus 0
-foo:1m_without_non_existing_label_total_prometheus 5
-foo:1m_without_non_existing_label_total_prometheus 5
-foo:1m_without_non_existing_label_total_prometheus 12
-`, `
-- interval: 1m
-  ignore_first_sample_interval: 0s
-  outputs: [increase, increase_prometheus, total, total_prometheus]
-- interval: 1m
-  staleness_interval: 2m
-  without: [non_existing_label]
-  outputs: [increase, increase_prometheus, total, total_prometheus]
-`, "111111")
-
-	// sum_sample and sum_samples_total outputs with different staleness intervals
+	// multiple aggregate configs
 	f([]string{`
 foo 1
-foo 2 1
 foo{bar="baz"} 2
-`, `
-foo 4
-`, ``, ``, `
-foo 6
-`, ``, ``}, time.Minute, `foo:1m_sum_samples 3
-foo:1m_sum_samples 4
-foo:1m_sum_samples 6
-foo:1m_sum_samples_total 3
-foo:1m_sum_samples_total 7
-foo:1m_sum_samples_total 6
-foo:1m_sum_samples_total{bar="baz"} 2
+foo 3.3
+`, ``, ``, ``, ``}, time.Minute, `foo:1m_count_series 1
+foo:1m_count_series{bar="baz"} 1
+foo:1m_sum_samples 0
+foo:1m_sum_samples 4.3
+foo:1m_sum_samples{bar="baz"} 0
 foo:1m_sum_samples{bar="baz"} 2
-foo:1m_without_non-existing-label_sum_samples 3
-foo:1m_without_non-existing-label_sum_samples 4
-foo:1m_without_non-existing-label_sum_samples 0
-foo:1m_without_non-existing-label_sum_samples 6
-foo:1m_without_non-existing-label_sum_samples 0
-foo:1m_without_non-existing-label_sum_samples_total 3
-foo:1m_without_non-existing-label_sum_samples_total 7
-foo:1m_without_non-existing-label_sum_samples_total 7
-foo:1m_without_non-existing-label_sum_samples_total 6
-foo:1m_without_non-existing-label_sum_samples_total 6
-foo:1m_without_non-existing-label_sum_samples_total{bar="baz"} 2
-foo:1m_without_non-existing-label_sum_samples_total{bar="baz"} 2
-foo:1m_without_non-existing-label_sum_samples{bar="baz"} 2
-foo:1m_without_non-existing-label_sum_samples{bar="baz"} 0
-foo:5m_by_bar_sum_samples 13
-foo:5m_by_bar_sum_samples_total 13
-foo:5m_by_bar_sum_samples_total{bar="baz"} 2
+foo:5m_by_bar_sum_samples 4.3
 foo:5m_by_bar_sum_samples{bar="baz"} 2
 `, `
 - interval: 1m
-  staleness_interval: 1m
-  outputs: [ sum_samples, sum_samples_total]
-- interval: 1m
-  staleness_interval: 2m
-  without: [non-existing-label]
-  outputs: [ sum_samples, sum_samples_total]
+  outputs: [count_series, sum_samples]
 - interval: 5m
   by: [bar]
-  outputs: [sum_samples, sum_samples_total]
-`, "11111")
+  outputs: [sum_samples]
+`, "111")
 
 	// min and max outputs
 	f([]string{`
@@ -789,39 +688,30 @@ foo:1m_by_cde_rate_sum{cde="1"} 0.125
   outputs: [rate_sum, rate_avg]
 `, "11111")
 
-	// test rate_sum and rate_avg with different staleness intervals
+	// test rate_sum and rate_avg, when two aggregation intervals are empty
 	f([]string{`
 foo{abc="123", cde="1"} 1
 foo{abc="123", cde="1"} 2 1
-foo{abc="456", cde="1"} 3
-foo{abc="456", cde="1"} 4 1
-foo{abc="777", cde="1"} 5
-foo{abc="777", cde="1"} 6 1
-`, ``, `
-foo{abc="123", cde="1"} 121
-foo{abc="123", cde="1"} 122 1
-foo{abc="456", cde="1"} 123
-foo{abc="456", cde="1"} 124 1
-foo{abc="777", cde="1"} 125
-foo{abc="777", cde="1"} 126 1
+foo{abc="456", cde="1"} 7
+foo{abc="456", cde="1"} 8 1
+foo{abc="777", cde="1"} 8
+foo{abc="777", cde="1"} 9 1
+`, ``, ``, `
+foo{abc="123", cde="1"} 19
+foo{abc="123", cde="1"} 20 1
+foo{abc="456", cde="1"} 26
+foo{abc="456", cde="1"} 27 1
+foo{abc="777", cde="1"} 27
+foo{abc="777", cde="1"} 28 1
 `}, time.Minute, `foo:1m_by_cde_rate_avg{cde="1"} 1
 foo:1m_by_cde_rate_avg{cde="1"} 1
 foo:1m_by_cde_rate_sum{cde="1"} 3
 foo:1m_by_cde_rate_sum{cde="1"} 3
-foo:1m_without_abc_rate_avg{cde="1"} 1
-foo:1m_without_abc_rate_avg{cde="1"} 1
-foo:1m_without_abc_rate_sum{cde="1"} 3
-foo:1m_without_abc_rate_sum{cde="1"} 3
 `, `            
 - interval: 1m
   by: [cde]
   outputs: [rate_sum, rate_avg]
   enable_windows: true
-- interval: 1m
-  staleness_interval: 2m
-  without: [abc]
-  outputs: [rate_sum, rate_avg]
-  enable_windows: true
 `, "111111111111")
 
 	// rate_sum and rate_avg with duplicated events

lib/streamaggr/streamaggr_test.go

@@ -252,15 +252,11 @@ func TestAggregatorsEqual(t *testing.T) {
 }
 
 func timeSeriessToString(tss []prompb.TimeSeries) string {
-	sorted := make([]prompb.TimeSeries, len(tss))
-	copy(sorted, tss)
-	sort.SliceStable(sorted, func(i, j int) bool {
-		return promrelabel.LabelsToString(sorted[i].Labels) < promrelabel.LabelsToString(sorted[j].Labels)
-	})
-	a := make([]string, len(sorted))
-	for i, ts := range sorted {
+	a := make([]string, len(tss))
+	for i, ts := range tss {
 		a[i] = timeSeriesToString(ts)
 	}
+	sort.Strings(a)
 	return strings.Join(a, "")
 }
 

lib/streamaggr/streamaggr_timing_test.go

@@ -27,7 +27,6 @@ var benchOutputs = []string{
 	"stddev",
 	"stdvar",
 	"sum_samples",
-	"sum_samples_total",
 	"total",
 	"total_prometheus",
 	"unique_samples",

lib/streamaggr/sum_samples.go

@@ -1,44 +1,27 @@
 package streamaggr
 
-import (
-	"math"
-)
-
 type sumSamplesAggrValue struct {
 	sum float64
 }
 
 func (av *sumSamplesAggrValue) pushSample(_ aggrConfig, sample *pushSample, _ string, _ int64) {
-	if math.Abs(av.sum) >= (1 << 53) {
-		// It is time to reset the entry, since it starts losing float64 precision
-		av.sum = 0
-	}
 	av.sum += sample.value
 }
 
-func (av *sumSamplesAggrValue) flush(c aggrConfig, ctx *flushCtx, key string, _ bool) {
-	ac := c.(*sumSamplesAggrConfig)
-	if ac.resetTotalOnFlush {
-		ctx.appendSeries(key, "sum_samples", av.sum)
-		av.sum = 0
-		return
-	}
-	ctx.appendSeries(key, "sum_samples_total", av.sum)
+func (av *sumSamplesAggrValue) flush(_ aggrConfig, ctx *flushCtx, key string, _ bool) {
+	ctx.appendSeries(key, "sum_samples", av.sum)
+	av.sum = 0
 }
 
 func (*sumSamplesAggrValue) state() any {
 	return nil
 }
 
-func newSumSamplesAggrConfig(resetTotalOnFlush bool) aggrConfig {
-	return &sumSamplesAggrConfig{
-		resetTotalOnFlush: resetTotalOnFlush,
-	}
+func newSumSamplesAggrConfig() aggrConfig {
+	return &sumSamplesAggrConfig{}
 }
 
-type sumSamplesAggrConfig struct {
-	resetTotalOnFlush bool
-}
+type sumSamplesAggrConfig struct{}
 
 func (*sumSamplesAggrConfig) getValue(_ any) aggrValue {
 	return &sumSamplesAggrValue{}

lib/streamaggr/total.go

@@ -31,11 +31,7 @@ func (av *totalAggrValue) pushSample(c aggrConfig, sample *pushSample, key strin
 	currentTime := fasttime.UnixTimestamp()
 	keepFirstSample := ac.keepFirstSample && currentTime >= ac.ignoreFirstSampleDeadline
 	lv, ok := av.shared.lastValues[key]
-	// The last value is stale, reset it.
-	if ok && lv.deleteDeadline < int64(currentTime)*1000 {
-		ok = false
-	}
-	if ok {
+	if ok || keepFirstSample {
 		if sample.timestamp < lv.timestamp {
 			// Skip out of order sample
 			return
@@ -47,8 +43,6 @@ func (av *totalAggrValue) pushSample(c aggrConfig, sample *pushSample, key strin
 			av.total += sample.value
 			ac.counterResetsTotal.Inc()
 		}
-	} else if keepFirstSample {
-		av.total += sample.value
 	}
 	lv.value = sample.value
 	lv.timestamp = sample.timestamp

vendor/github.com/VictoriaMetrics/metrics/process_metrics_linux.go

@@ -7,7 +7,6 @@ import (
 	"io/ioutil"
 	"log"
 	"os"
-	"path"
 	"strconv"
 	"strings"
 	"sync/atomic"
@@ -405,113 +404,16 @@ func readPSITotals(cgroupPath, statsName string) (uint64, uint64, error) {
 }
 
 func getCgroupV2Path() string {
-	cgroupData, err := os.ReadFile("/proc/self/cgroup")
+	data, err := ioutil.ReadFile("/proc/self/cgroup")
 	if err != nil {
 		return ""
 	}
-	// Read /proc/self/mountinfo with a timeout. Generating the mountinfo contents
-	// can block in the kernel when a backing filesystem (e.g. a hung NFS or FUSE
-	// mount) is unresponsive. Since this runs at program init via psiMetricsStart,
-	// a blocking read would hang startup, so fall back to disabling PSI metrics instead.
-	mountinfoData, _ := readFileWithTimeout("/proc/self/mountinfo", time.Second)
-	return getCgroupV2PathInternal(string(cgroupData), mountinfoData)
-}
-
-// readFileWithTimeout reads the file at path, returning ("", false) if the read
-// doesn't complete within timeout.
-//
-// A timed-out read leaks the reading goroutine until the read eventually unblocks
-// (if ever). This is an acceptable safeguard against a read of a pseudo-file such
-// as /proc/self/mountinfo hanging on an unresponsive mount.
-func readFileWithTimeout(path string, timeout time.Duration) (string, bool) {
-	type result struct {
-		data []byte
-		err  error
-	}
-	// The channel is buffered so the goroutine can always send and exit,
-	// even after this function has returned on timeout.
-	ch := make(chan result, 1)
-	go func() {
-		data, err := os.ReadFile(path)
-		ch <- result{data: data, err: err}
-	}()
-	timer := time.NewTimer(timeout)
-	defer timer.Stop()
-	select {
-	case r := <-ch:
-		if r.err != nil {
-			return "", false
-		}
-		return string(r.data), true
-	case <-timer.C:
-		return "", false
-	}
-}
-
-func getCgroupV2PathInternal(cgroupData, mountinfoData string) string {
-	rel := getCgroupV2RelativePath(cgroupData)
-	if rel == "" {
-		// The process doesn't run under cgroup v2.
+	tmp := strings.SplitN(string(data), "::", 2)
+	if len(tmp) != 2 {
 		return ""
 	}
+	path := "/sys/fs/cgroup" + strings.TrimSpace(tmp[1])
 
-	// Determine the actual cgroup v2 mountpoint instead of assuming /sys/fs/cgroup.
-	// On systems with a hybrid cgroup hierarchy the unified cgroup v2 is mounted
-	// at a different location such as /sys/fs/cgroup/unified.
-	// See https://github.com/VictoriaMetrics/metrics/issues/127
-	mountpoint := getCgroupV2Mountpoint(mountinfoData)
-	if mountpoint == "" {
-		// fallback to assumed path
-		mountpoint = "/sys/fs/cgroup"
-	}
-	cgroupPath := path.Join(mountpoint, rel)
-	// Drop trailing slash if it exists. This prevents from '//' in the constructed paths by the caller.
-	return strings.TrimSuffix(cgroupPath, "/")
-}
-
-// getCgroupV2RelativePath returns the cgroup v2 path of the process relative to
-// the cgroup v2 mountpoint, or an empty string if the process doesn't run under cgroup v2.
-//
-// The cgroup v2 entry in /proc/self/cgroup has an empty controllers field, e.g. "0::/the/path".
-// See https://man7.org/linux/man-pages/man7/cgroups.7.html
-func getCgroupV2RelativePath(cgroupData string) string {
-	for _, line := range strings.Split(cgroupData, "\n") {
-		// Each line has the form "hierarchy-ID:controller-list:cgroup-path".
-		// The cgroup v2 line has an empty hierarchy-ID and controller-list, i.e. it starts with "0::".
-		tmp := strings.SplitN(line, "::", 2)
-		if len(tmp) == 2 && strings.HasPrefix(line, "0::") {
-			return strings.TrimSpace(tmp[1])
-		}
-	}
-	return ""
-}
-
-// getCgroupV2Mountpoint returns the mountpoint of the cgroup v2 (unified) hierarchy
-// parsed from the contents of /proc/self/mountinfo, or an empty string if cgroup v2 isn't mounted.
-func getCgroupV2Mountpoint(mountinfoData string) string {
-	for _, line := range strings.Split(mountinfoData, "\n") {
-		if !strings.Contains(line, "cgroup2") {
-			// fast path
-			continue
-		}
-		// mountinfo lines have the form:
-		//   36 35 98:0 / /sys/fs/cgroup/unified rw,... - cgroup2 cgroup2 rw,...
-		// The optional fields preceding the filesystem type are terminated by " - ".
-		// See https://man7.org/linux/man-pages/man5/proc_pid_mountinfo.5.html
-		tmp := strings.SplitN(line, " - ", 2)
-		if len(tmp) != 2 {
-			continue
-		}
-		after := strings.Fields(tmp[1])
-		if len(after) < 1 || after[0] != "cgroup2" {
-			continue
-		}
-		before := strings.Fields(tmp[0])
-		if len(before) < 5 {
-			continue
-		}
-		// before[4] is the mount point.
-		return before[4]
-	}
-	return ""
+	// Drop trailing slash if it exsits. This prevents from '//' in the constructed paths by the caller.
+	return strings.TrimSuffix(path, "/")
 }

vendor/modules.txt

@@ -143,7 +143,7 @@ github.com/VictoriaMetrics/easyproto
 # github.com/VictoriaMetrics/fastcache v1.13.3
 ## explicit; go 1.24.0
 github.com/VictoriaMetrics/fastcache
-# github.com/VictoriaMetrics/metrics v1.44.0
+# github.com/VictoriaMetrics/metrics v1.43.2
 ## explicit; go 1.24.0
 github.com/VictoriaMetrics/metrics
 # github.com/VictoriaMetrics/metricsql v0.87.1