Merge branch 'main' of github.com:thanos-io/thanos into limits-per-te…

…nant-config-file

.busybox-versions

@@ -1,6 +1,6 @@
 # Auto generated by busybox-updater.sh. DO NOT EDIT
-amd64=60ded79a99eb70aa36d57c598707c961ffa4f9f7b63237823a780eaf6d437a78
-arm64=d2e2aa8b7fd3d11412c4918ed35bcabc13027ed4812932a94b29940805a87176
-arm=53e81f95dc293546ed76257636bd4ec60cba74297ea637aefd37fe9446b66a59
-ppc64le=d3f30bf50b2d6e7097ad02a1c3d64c5ffe535b969d80d24bb9543391122d4b8c
-s390x=c12940cb7882554a5cb5c582facc2a01a9f30447c1aa067c9c1cca7c9f16c7ee
+amd64=d8d3654786836cad8c09543704807c7a6d75de53b9e9cd21a1bbd8cb1a607004
+arm64=a3435ee186dbf88238388c112761488ecd2c264dbff8957ab73f804be62a9080
+arm=b063a2176f23a13007de5c447ab3552f8e355162ac54fc2a545b00b612d4c81e
+ppc64le=203c3f97bc34c4d5df50bd61beaa397f2a4c7cbd470c84fe7ec3db12409435d3
+s390x=1a6eb305bd08bd1d38cb85a097ad776a78dd72b7c1a35094bb080788a39b174c

CHANGELOG.md

@@ -9,12 +9,16 @@ NOTE: As semantic versioning states all 0.y.z releases can contain breaking chan
 We use *breaking :warning:* to mark changes that are not backward compatible (relates only to v0.y.z releases.)
 
 ## Unreleased
+- [#5607](https://github.com/thanos-io/thanos/pull/5607) Query: Support custom lookback delta from request in query api.
 - [#5453](https://github.com/thanos-io/thanos/pull/5453) Compact: Skip erroneous empty non `*AggrChunk` chunks during 1h downsampling of 5m resolution blocks.
 
 ### Fixed
 - [#5502](https://github.com/thanos-io/thanos/pull/5502) Receive: Handle exemplar storage errors as conflict error.
 - [#5534](https://github.com/thanos-io/thanos/pull/5534) Query: Set struct return by query api alerts same as prometheus api.
 - [#5554](https://github.com/thanos-io/thanos/pull/5554) Query/Receiver: Fix querying exemplars from multi-tenant receivers.
+- [#5583](https://github.com/thanos-io/thanos/pull/5583) Query: fix data race between Respond() and query/queryRange functions. Fixes [#5410](https://github.com/thanos-io/thanos/pull/5410).
+- [#5642](https://github.com/thanos-io/thanos/pull/5642) Receive: Log labels correctly in writer debug messages.
+- [#5655](https://github.com/thanos-io/thanos/pull/5655) Receive: Fix recreating already pruned tenants.
 
 ### Added
 
@@ -25,12 +29,16 @@ We use *breaking :warning:* to mark changes that are not backward compatible (re
 - [#5475](https://github.com/thanos-io/thanos/pull/5475) Compact/Store: Added `--block-files-concurrency` allowing to configure number of go routines for download/upload block files during compaction.
 - [#5470](https://github.com/thanos-io/thanos/pull/5470) Receive: Implement exposing TSDB stats for all tenants
 - [#5493](https://github.com/thanos-io/thanos/pull/5493) Compact: Added `--compact.blocks-fetch-concurrency` allowing to configure number of go routines for download blocks during compactions.
+- [#5480](https://github.com/thanos-io/thanos/pull/5480) Query: Expose endpoint info timeout as a hidden flag.
 - [#5527](https://github.com/thanos-io/thanos/pull/5527) Receive: Add per request limits for remote write.
 - [#5520](https://github.com/thanos-io/thanos/pull/5520) Receive: Meta-monitoring based active series limiting.
 - [#5555](https://github.com/thanos-io/thanos/pull/5555) Query: Added `--query.active-query-path` flag, allowing the user to configure the directory to create an active query tracking file, `queries.active`, for different resolution.
 - [#5566](https://github.com/thanos-io/thanos/pull/5566) Receive: Added experimental support to enable chunk write queue via `--tsdb.write-queue-size` flag.
 - [#5575](https://github.com/thanos-io/thanos/pull/5575) Receive: Add support for gRPC compression with snappy.
+- [#5508](https://github.com/thanos-io/thanos/pull/5508) Receive: Validate labels in write requests.
+- [#5439](https://github.com/thanos-io/thanos/pull/5439) Mixin: Add Alert ThanosQueryOverload to Mixin.
 - [#5439](https://github.com/thanos-io/thanos/pull/5439) Add Alert ThanosQueryOverload to Mixin.
+- [#5561](https://github.com/thanos-io/thanos/pull/5561) Query Frontend: Support instant query sharding.
 - [#5565](https://github.com/thanos-io/thanos/pull/5565) Receive: Allow remote write request limits to be defined per file and tenant.
 
 ### Changed
@@ -39,6 +47,7 @@ We use *breaking :warning:* to mark changes that are not backward compatible (re
 - [#5451](https://github.com/thanos-io/thanos/pull/5451) Azure: Reduce memory usage by not buffering file downloads entirely in memory.
 - [#5484](https://github.com/thanos-io/thanos/pull/5484) Update Prometheus deps to v2.36.2.
 - [#5511](https://github.com/thanos-io/thanos/pull/5511) Update Prometheus deps to v2.37.0.
+- [#5588](https://github.com/thanos-io/thanos/pull/5588) Store: improve index header reading performance by sorting values first.
 
 ### Removed
 

cmd/thanos/query.go

@@ -8,8 +8,6 @@ import (
 	"fmt"
 	"math"
 	"net/http"
-	"path/filepath"
-	"strconv"
 	"strings"
 	"time"
 
@@ -147,6 +145,8 @@ func registerQuery(app *extkingpin.App) {
 
 	unhealthyStoreTimeout := extkingpin.ModelDuration(cmd.Flag("store.unhealthy-timeout", "Timeout before an unhealthy store is cleaned from the store UI page.").Default("5m"))
 
+	endpointInfoTimeout := extkingpin.ModelDuration(cmd.Flag("endpoint.info-timeout", "Timeout of gRPC Info requests.").Default("5s").Hidden())
+
 	enableAutodownsampling := cmd.Flag("query.auto-downsampling", "Enable automatic adjustment (step / 5) to what source of data should be used in store gateways if no max_source_resolution param is specified.").
 		Default("false").Bool()
 
@@ -279,6 +279,7 @@ func registerQuery(app *extkingpin.App) {
 			time.Duration(*dnsSDInterval),
 			*dnsSDResolver,
 			time.Duration(*unhealthyStoreTimeout),
+			time.Duration(*endpointInfoTimeout),
 			time.Duration(*instantDefaultMaxSourceResolution),
 			*defaultMetadataTimeRange,
 			*strictStores,
@@ -347,6 +348,7 @@ func runQuery(
 	dnsSDInterval time.Duration,
 	dnsSDResolver string,
 	unhealthyStoreTimeout time.Duration,
+	endpointInfoTimeout time.Duration,
 	instantDefaultMaxSourceResolution time.Duration,
 	defaultMetadataTimeRange time.Duration,
 	strictStores []string,
@@ -459,6 +461,7 @@ func runQuery(
 			},
 			dialOpts,
 			unhealthyStoreTimeout,
+			endpointInfoTimeout,
 		)
 		proxy            = store.NewProxyStore(logger, reg, endpoints.GetStoreClients, component.Query, selectorLset, storeResponseTimeout)
 		rulesProxy       = rules.NewProxy(logger, endpoints.GetRulesClients)
@@ -581,8 +584,14 @@ func runQuery(
 		grpcProbe,
 		prober.NewInstrumentation(comp, logger, extprom.WrapRegistererWithPrefix("thanos_", reg)),
 	)
-	engineCreator := engineFactory(promql.NewEngine, engineOpts, dynamicLookbackDelta, activeQueryDir,
-		maxConcurrentQueries, logger)
+
+	// An active query tracker will be added only if the user specifies a non-default path.
+	// Otherwise, the nil active query tracker from existing engine options will be used.
+	if activeQueryDir != "" {
+		engineOpts.ActiveQueryTracker = promql.NewActiveQueryTracker(activeQueryDir, maxConcurrentQueries, logger)
+	}
+	engine := promql.NewEngine(engineOpts)
+	lookbackDeltaCreator := LookbackDeltaFactory(engineOpts, dynamicLookbackDelta)
 
 	// Start query API + UI HTTP server.
 	{
@@ -612,7 +621,8 @@ func runQuery(
 		api := apiv1.NewQueryAPI(
 			logger,
 			endpoints.GetEndpointStatus,
-			engineCreator,
+			engine,
+			lookbackDeltaCreator,
 			queryableCreator,
 			// NOTE: Will share the same replica label as the query for now.
 			rules.NewGRPCClientWithDedup(rulesProxy, queryReplicaLabels),
@@ -687,7 +697,7 @@ func runQuery(
 			info.WithQueryAPIInfoFunc(),
 		)
 
-		grpcAPI := apiv1.NewGRPCAPI(time.Now, queryReplicaLabels, queryableCreator, engineCreator, instantDefaultMaxSourceResolution)
+		grpcAPI := apiv1.NewGRPCAPI(time.Now, queryReplicaLabels, queryableCreator, engine, lookbackDeltaCreator, instantDefaultMaxSourceResolution)
 		s := grpcserver.New(logger, reg, tracer, grpcLogOpts, tagOpts, comp, grpcProbe,
 			grpcserver.WithServer(apiv1.RegisterQueryServer(grpcAPI)),
 			grpcserver.WithServer(store.RegisterStoreServer(proxy)),
@@ -732,82 +742,40 @@ func removeDuplicateEndpointSpecs(logger log.Logger, duplicatedStores prometheus
 	return deduplicated
 }
 
-// engineFactory creates from 1 to 3 promql.Engines depending on
+// LookbackDeltaFactory creates from 1 to 3 lookback deltas depending on
 // dynamicLookbackDelta and eo.LookbackDelta and returns a function
-// that returns appropriate engine for given maxSourceResolutionMillis.
-//
-// TODO: it seems like a good idea to tweak Prometheus itself
-// instead of creating several Engines here.
-func engineFactory(
-	newEngine func(promql.EngineOpts) *promql.Engine,
+// that returns appropriate lookback delta for given maxSourceResolutionMillis.
+func LookbackDeltaFactory(
 	eo promql.EngineOpts,
 	dynamicLookbackDelta bool,
-	activeQueryDir string,
-	maxConcurrentQueries int,
-	logger log.Logger,
-) func(int64) *promql.Engine {
+) func(int64) time.Duration {
 	resolutions := []int64{downsample.ResLevel0}
 	if dynamicLookbackDelta {
 		resolutions = []int64{downsample.ResLevel0, downsample.ResLevel1, downsample.ResLevel2}
 	}
 	var (
-		engines = make([]*promql.Engine, len(resolutions))
-		ld      = eo.LookbackDelta.Milliseconds()
+		lds = make([]time.Duration, len(resolutions))
+		ld  = eo.LookbackDelta.Milliseconds()
 	)
-	wrapReg := func(engineNum int) prometheus.Registerer {
-		return extprom.WrapRegistererWith(map[string]string{"engine": strconv.Itoa(engineNum)}, eo.Reg)
-	}
 
 	lookbackDelta := eo.LookbackDelta
 	for i, r := range resolutions {
 		if ld < r {
 			lookbackDelta = time.Duration(r) * time.Millisecond
 		}
 
-		newEngineOpts := promql.EngineOpts{
-			Logger:                   eo.Logger,
-			Reg:                      wrapReg(i),
-			MaxSamples:               eo.MaxSamples,
-			Timeout:                  eo.Timeout,
-			LookbackDelta:            lookbackDelta,
-			NoStepSubqueryIntervalFn: eo.NoStepSubqueryIntervalFn,
-			EnableAtModifier:         eo.EnableAtModifier,
-			EnableNegativeOffset:     eo.EnableNegativeOffset,
-		}
-		// An active query tracker will be added only if the user specifies a non-default path.
-		// Otherwise, the nil active query tracker from existing engine options will be used.
-		if activeQueryDir != "" {
-			resActiveQueryDir := filepath.Join(activeQueryDir, getActiveQueryDirBasedOnResolution(r))
-			newEngineOpts.ActiveQueryTracker = promql.NewActiveQueryTracker(resActiveQueryDir, maxConcurrentQueries, logger)
-		} else {
-			newEngineOpts.ActiveQueryTracker = eo.ActiveQueryTracker
-		}
-
-		engines[i] = newEngine(newEngineOpts)
+		lds[i] = lookbackDelta
 	}
-	return func(maxSourceResolutionMillis int64) *promql.Engine {
+	return func(maxSourceResolutionMillis int64) time.Duration {
 		for i := len(resolutions) - 1; i >= 1; i-- {
 			left := resolutions[i-1]
 			if resolutions[i-1] < ld {
 				left = ld
 			}
 			if left < maxSourceResolutionMillis {
-				return engines[i]
+				return lds[i]
 			}
 		}
-		return engines[0]
-	}
-}
-
-func getActiveQueryDirBasedOnResolution(resolution int64) string {
-	if resolution == downsample.ResLevel0 {
-		return "raw"
-	}
-	if resolution == downsample.ResLevel1 {
-		return "5m"
-	}
-	if resolution == downsample.ResLevel2 {
-		return "1h"
+		return lds[0]
 	}
-	return ""
 }

cmd/thanos/query_test.go

@@ -7,31 +7,15 @@ import (
 	"testing"
 	"time"
 
-	"github.com/go-kit/log"
 	"github.com/prometheus/prometheus/promql"
 
 	"github.com/thanos-io/thanos/pkg/testutil"
 )
 
-func TestEngineFactory(t *testing.T) {
-	var (
-		engineRaw = promql.NewEngine(promql.EngineOpts{})
-		engine5m  = promql.NewEngine(promql.EngineOpts{LookbackDelta: 5 * time.Minute})
-		engine1h  = promql.NewEngine(promql.EngineOpts{LookbackDelta: 1 * time.Hour})
-	)
-	mockNewEngine := func(opts promql.EngineOpts) *promql.Engine {
-		switch opts.LookbackDelta {
-		case 1 * time.Hour:
-			return engine1h
-		case 5 * time.Minute:
-			return engine5m
-		default:
-			return engineRaw
-		}
-	}
+func TestLookbackDeltaFactory(t *testing.T) {
 	type testCase struct {
 		stepMillis int64
-		expect     *promql.Engine
+		expect     time.Duration
 	}
 	var (
 		minute = time.Minute.Milliseconds()
@@ -46,67 +30,66 @@ func TestEngineFactory(t *testing.T) {
 				lookbackDelta:        0,
 				dynamicLookbackDelta: false,
 				tcs: []testCase{
-					{0, engineRaw},
-					{5 * minute, engineRaw},
-					{1 * hour, engineRaw},
+					{0, time.Duration(0)},
+					{5 * minute, time.Duration(0)},
+					{1 * hour, time.Duration(0)},
 				},
 			},
 			{
 
 				lookbackDelta:        3 * time.Minute,
 				dynamicLookbackDelta: true,
 				tcs: []testCase{
-					{2 * minute, engineRaw},
-					{3 * minute, engineRaw},
-					{4 * minute, engine5m},
-					{5 * minute, engine5m},
-					{6 * minute, engine1h},
-					{1 * hour, engine1h},
-					{2 * hour, engine1h},
+					{2 * minute, time.Duration(3) * time.Minute},
+					{3 * minute, time.Duration(3) * time.Minute},
+					{4 * minute, time.Duration(5) * time.Minute},
+					{5 * minute, time.Duration(5) * time.Minute},
+					{6 * minute, time.Duration(1) * time.Hour},
+					{1 * hour, time.Duration(1) * time.Hour},
+					{2 * hour, time.Duration(1) * time.Hour},
 				},
 			},
 			{
 				lookbackDelta:        5 * time.Minute,
 				dynamicLookbackDelta: true,
 				tcs: []testCase{
-					{0, engine5m},
-					{5 * minute, engine5m},
-					{6 * minute, engine1h},
-					{59 * minute, engine1h},
-					{1 * hour, engine1h},
-					{2 * hour, engine1h},
+					{0, time.Duration(5) * time.Minute},
+					{5 * minute, time.Duration(5) * time.Minute},
+					{6 * minute, time.Duration(1) * time.Hour},
+					{59 * minute, time.Duration(1) * time.Hour},
+					{1 * hour, time.Duration(1) * time.Hour},
+					{2 * hour, time.Duration(1) * time.Hour},
 				},
 			},
 			{
 				lookbackDelta:        30 * time.Minute,
 				dynamicLookbackDelta: true,
 				tcs: []testCase{
-					{0, engineRaw},
-					{5 * minute, engineRaw},
-					{30 * minute, engineRaw},
-					{31 * minute, engine1h},
-					{59 * minute, engine1h},
-					{1 * hour, engine1h},
-					{2 * hour, engine1h},
+					{0, time.Duration(30) * time.Minute},
+					{5 * minute, time.Duration(30) * time.Minute},
+					{30 * minute, time.Duration(30) * time.Minute},
+					{31 * minute, time.Duration(1) * time.Hour},
+					{59 * minute, time.Duration(1) * time.Hour},
+					{1 * hour, time.Duration(1) * time.Hour},
+					{2 * hour, time.Duration(1) * time.Hour},
 				},
 			},
 			{
 				lookbackDelta:        1 * time.Hour,
 				dynamicLookbackDelta: true,
 				tcs: []testCase{
-					{0, engine1h},
-					{5 * minute, engine1h},
-					{1 * hour, engine1h},
-					{2 * hour, engine1h},
+					{0, time.Duration(1) * time.Hour},
+					{5 * minute, time.Duration(1) * time.Hour},
+					{1 * hour, time.Duration(1) * time.Hour},
+					{2 * hour, time.Duration(1) * time.Hour},
 				},
 			},
 		}
 	)
 	for _, td := range tData {
-		e := engineFactory(mockNewEngine, promql.EngineOpts{LookbackDelta: td.lookbackDelta}, td.dynamicLookbackDelta,
-			"", 1, log.NewNopLogger())
+		lookbackCreate := LookbackDeltaFactory(promql.EngineOpts{LookbackDelta: td.lookbackDelta}, td.dynamicLookbackDelta)
 		for _, tc := range td.tcs {
-			got := e(tc.stepMillis)
+			got := lookbackCreate(tc.stepMillis)
 			testutil.Equals(t, tc.expect, got)
 		}
 	}