Merge pull request #14676 from prometheus/krajo/update-pr-14546-from-main

update pr 14546 from main

.github/workflows/ci.yml

@@ -186,7 +186,7 @@ jobs:
         with:
           args: --verbose
           # Make sure to sync this with Makefile.common and scripts/golangci-lint.yml.
-          version: v1.59.1
+          version: v1.60.2
   fuzzing:
     uses: ./.github/workflows/fuzzing.yml
     if: github.event_name == 'pull_request'

.golangci.yml

@@ -25,15 +25,34 @@ linters:
     - loggercheck
 
 issues:
+  max-issues-per-linter: 0
   max-same-issues: 0
+  # The default exclusions are too aggressive. For one, they
+  # essentially disable any linting on doc comments. We disable
+  # default exclusions here and add exclusions fitting our codebase
+  # further down.
+  exclude-use-default: false
   exclude-files:
     # Skip autogenerated files.
     - ^.*\.(pb|y)\.go$
   exclude-dirs:
-    # Copied it from a different source
+    # Copied it from a different source.
     - storage/remote/otlptranslator/prometheusremotewrite
     - storage/remote/otlptranslator/prometheus
   exclude-rules:
+    - linters:
+        - errcheck
+      # Taken from the default exclusions (that are otherwise disabled above).
+      text: Error return value of .((os\.)?std(out|err)\..*|.*Close|.*Flush|os\.Remove(All)?|.*print(f|ln)?|os\.(Un)?Setenv). is not checked
+    - linters:
+        - govet
+      # We use many Seek methods that do not follow the usual pattern.
+      text: "stdmethods: method Seek.* should have signature Seek"
+    - linters:
+        - revive
+      # We have stopped at some point to write doc comments on exported symbols.
+      # TODO(beorn7): Maybe we should enforce this again? There are ~500 offenders right now.
+      text: exported (.+) should have comment( \(or a comment on this block\))? or be unexported
     - linters:
         - gocritic
       text: "appendAssign"
@@ -94,15 +113,14 @@ linters-settings:
     errorf: false
   revive:
     # By default, revive will enable only the linting rules that are named in the configuration file.
-    # So, it's needed to explicitly set in configuration all required rules.
+    # So, it's needed to explicitly enable all required rules here.
-    # The following configuration enables all the rules from the defaults.toml
-    # https://github.com/mgechev/revive/blob/master/defaults.toml
     rules:
       # https://github.com/mgechev/revive/blob/master/RULES_DESCRIPTIONS.md
       - name: blank-imports
+      - name: comment-spacings
       - name: context-as-argument
         arguments:
-          # allow functions with test or bench signatures
+          # Allow functions with test or bench signatures.
           - allowTypesBefore: "*testing.T,testing.TB"
       - name: context-keys-type
       - name: dot-imports
@@ -118,6 +136,8 @@ linters-settings:
       - name: increment-decrement
       - name: indent-error-flow
       - name: package-comments
+        # TODO(beorn7): Currently, we have a lot of missing package doc comments. Maybe we should have them.
+        disabled: true
       - name: range
       - name: receiver-naming
       - name: redefines-builtin-id

CHANGELOG.md

@@ -3,6 +3,8 @@
 ## unreleased
 
 * [FEATURE] OTLP receiver: Add new option `otlp.promote_resource_attributes`, for any OTel resource attributes that should be promoted to metric labels. #14200
+* [ENHANCEMENT] OTLP receiver: Warn when encountering exponential histograms with zero count and non-zero sum. #14706
+* [BUGFIX] tsdb/wlog.Watcher.readSegmentForGC: Only count unknown record types against record_decode_failures_total metric. #14042
 
 ## 2.54.0-rc.1 / 2024-08-05
 

Makefile.common

@@ -61,7 +61,7 @@ PROMU_URL     := https://github.com/prometheus/promu/releases/download/v$(PROMU_
 SKIP_GOLANGCI_LINT :=
 GOLANGCI_LINT :=
 GOLANGCI_LINT_OPTS ?=
-GOLANGCI_LINT_VERSION ?= v1.59.1
+GOLANGCI_LINT_VERSION ?= v1.60.2
 # golangci-lint only supports linux, darwin and windows platforms on i386/amd64/arm64.
 # windows isn't included here because of the path separator being different.
 ifeq ($(GOHOSTOS),$(filter $(GOHOSTOS),linux darwin))

cmd/prometheus/main.go

@@ -152,6 +152,7 @@ type flagConfig struct {
 	queryConcurrency    int
 	queryMaxSamples     int
 	RemoteFlushDeadline model.Duration
+	nameEscapingScheme  string
 
 	featureList   []string
 	memlimitRatio float64
@@ -240,6 +241,9 @@ func (c *flagConfig) setFeatureListOptions(logger log.Logger) error {
 			case "delayed-compaction":
 				c.tsdb.EnableDelayedCompaction = true
 				level.Info(logger).Log("msg", "Experimental delayed compaction is enabled.")
+			case "utf8-names":
+				model.NameValidationScheme = model.UTF8Validation
+				level.Info(logger).Log("msg", "Experimental UTF-8 support enabled")
 			case "":
 				continue
 			case "promql-at-modifier", "promql-negative-offset":
@@ -484,7 +488,9 @@ func main() {
 	a.Flag("scrape.discovery-reload-interval", "Interval used by scrape manager to throttle target groups updates.").
 		Hidden().Default("5s").SetValue(&cfg.scrape.DiscoveryReloadInterval)
 
-	a.Flag("enable-feature", "Comma separated feature names to enable. Valid options: agent, auto-gomemlimit, exemplar-storage, expand-external-labels, memory-snapshot-on-shutdown, promql-per-step-stats, promql-experimental-functions, remote-write-receiver (DEPRECATED), extra-scrape-metrics, new-service-discovery-manager, auto-gomaxprocs, no-default-scrape-port, native-histograms, otlp-write-receiver, created-timestamp-zero-ingestion, concurrent-rule-eval, delayed-compaction. See https://prometheus.io/docs/prometheus/latest/feature_flags/ for more details.").
+	a.Flag("scrape.name-escaping-scheme", `Method for escaping legacy invalid names when sending to Prometheus that does not support UTF-8. Can be one of "values", "underscores", or "dots".`).Default(scrape.DefaultNameEscapingScheme.String()).StringVar(&cfg.nameEscapingScheme)
+
+	a.Flag("enable-feature", "Comma separated feature names to enable. Valid options: agent, auto-gomemlimit, exemplar-storage, expand-external-labels, memory-snapshot-on-shutdown, promql-per-step-stats, promql-experimental-functions, remote-write-receiver (DEPRECATED), extra-scrape-metrics, new-service-discovery-manager, auto-gomaxprocs, no-default-scrape-port, native-histograms, otlp-write-receiver, created-timestamp-zero-ingestion, concurrent-rule-eval, delayed-compaction, utf8-names. See https://prometheus.io/docs/prometheus/latest/feature_flags/ for more details.").
 		Default("").StringsVar(&cfg.featureList)
 
 	promlogflag.AddFlags(a, &cfg.promlogConfig)
@@ -512,6 +518,15 @@ func main() {
 		os.Exit(1)
 	}
 
+	if cfg.nameEscapingScheme != "" {
+		scheme, err := model.ToEscapingScheme(cfg.nameEscapingScheme)
+		if err != nil {
+			fmt.Fprintf(os.Stderr, `Invalid name escaping scheme: %q; Needs to be one of "values", "underscores", or "dots"`, cfg.nameEscapingScheme)
+			os.Exit(1)
+		}
+		model.NameEscapingScheme = scheme
+	}
+
 	if agentMode && len(serverOnlyFlags) > 0 {
 		fmt.Fprintf(os.Stderr, "The following flag(s) can not be used in agent mode: %q", serverOnlyFlags)
 		os.Exit(3)

cmd/promtool/main.go

@@ -471,7 +471,7 @@ func (ls lintConfig) lintDuplicateRules() bool {
 	return ls.all || ls.duplicateRules
 }
 
-// Check server status - healthy & ready.
+// CheckServerStatus - healthy & ready.
 func CheckServerStatus(serverURL *url.URL, checkEndpoint string, roundTripper http.RoundTripper) error {
 	if serverURL.Scheme == "" {
 		serverURL.Scheme = "http"

cmd/promtool/metrics.go

@@ -31,7 +31,7 @@ import (
 	"github.com/prometheus/prometheus/util/fmtutil"
 )
 
-// Push metrics to a prometheus remote write (for testing purpose only).
+// PushMetrics to a prometheus remote write (for testing purpose only).
 func PushMetrics(url *url.URL, roundTripper http.RoundTripper, headers map[string]string, timeout time.Duration, labels map[string]string, files ...string) int {
 	addressURL, err := url.Parse(url.String())
 	if err != nil {

cmd/promtool/tsdb.go

@@ -866,16 +866,16 @@ func displayHistogram(dataType string, datas []int, total int) {
 	fmt.Println()
 }
 
-func generateBucket(min, max int) (start, end, step int) {
+func generateBucket(minVal, maxVal int) (start, end, step int) {
-	s := (max - min) / 10
+	s := (maxVal - minVal) / 10
 
 	step = 10
 	for step < s && step <= 10000 {
 		step *= 10
 	}
 
-	start = min - min%step
+	start = minVal - minVal%step
-	end = max - max%step + step
+	end = maxVal - maxVal%step + step
 
 	return
 }

cmd/promtool/tsdb_test.go

@@ -20,6 +20,7 @@ import (
 	"math"
 	"os"
 	"runtime"
+	"slices"
 	"strings"
 	"testing"
 	"time"
@@ -152,12 +153,18 @@ func TestTSDBDump(t *testing.T) {
 			expectedMetrics, err := os.ReadFile(tt.expectedDump)
 			require.NoError(t, err)
 			expectedMetrics = normalizeNewLine(expectedMetrics)
-			// even though in case of one matcher samples are not sorted, the order in the cases above should stay the same.
+			// Sort both, because Prometheus does not guarantee the output order.
-			require.Equal(t, string(expectedMetrics), dumpedMetrics)
+			require.Equal(t, sortLines(string(expectedMetrics)), sortLines(dumpedMetrics))
 		})
 	}
 }
 
+func sortLines(buf string) string {
+	lines := strings.Split(buf, "\n")
+	slices.Sort(lines)
+	return strings.Join(lines, "\n")
+}
+
 func TestTSDBDumpOpenMetrics(t *testing.T) {
 	storage := promqltest.LoadedStorage(t, `
 		load 1m
@@ -169,7 +176,7 @@ func TestTSDBDumpOpenMetrics(t *testing.T) {
 	require.NoError(t, err)
 	expectedMetrics = normalizeNewLine(expectedMetrics)
 	dumpedMetrics := getDumpedSamples(t, storage.Dir(), math.MinInt64, math.MaxInt64, []string{"{__name__=~'(?s:.*)'}"}, formatSeriesSetOpenMetrics)
-	require.Equal(t, string(expectedMetrics), dumpedMetrics)
+	require.Equal(t, sortLines(string(expectedMetrics)), sortLines(dumpedMetrics))
 }
 
 func TestTSDBDumpOpenMetricsRoundTrip(t *testing.T) {

config/config.go

@@ -67,6 +67,11 @@ var (
 	}
 )
 
+const (
+	LegacyValidationConfig = "legacy"
+	UTF8ValidationConfig   = "utf8"
+)
+
 // Load parses the YAML input s into a Config.
 func Load(s string, expandExternalLabels bool, logger log.Logger) (*Config, error) {
 	cfg := &Config{}
@@ -446,6 +451,8 @@ type GlobalConfig struct {
 	// Keep no more than this many dropped targets per job.
 	// 0 means no limit.
 	KeepDroppedTargets uint `yaml:"keep_dropped_targets,omitempty"`
+	// Allow UTF8 Metric and Label Names.
+	MetricNameValidationScheme string `yaml:"metric_name_validation_scheme,omitempty"`
 }
 
 // ScrapeProtocol represents supported protocol for scraping metrics.
@@ -471,6 +478,7 @@ var (
 	PrometheusText0_0_4  ScrapeProtocol = "PrometheusText0.0.4"
 	OpenMetricsText0_0_1 ScrapeProtocol = "OpenMetricsText0.0.1"
 	OpenMetricsText1_0_0 ScrapeProtocol = "OpenMetricsText1.0.0"
+	UTF8NamesHeader      string         = model.EscapingKey + "=" + model.AllowUTF8
 
 	ScrapeProtocolsHeaders = map[ScrapeProtocol]string{
 		PrometheusProto:      "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=delimited",
@@ -656,6 +664,8 @@ type ScrapeConfig struct {
 	// Keep no more than this many dropped targets per job.
 	// 0 means no limit.
 	KeepDroppedTargets uint `yaml:"keep_dropped_targets,omitempty"`
+	// Allow UTF8 Metric and Label Names.
+	MetricNameValidationScheme string `yaml:"metric_name_validation_scheme,omitempty"`
 
 	// We cannot do proper Go type embedding below as the parser will then parse
 	// values arbitrarily into the overflow maps of further-down types.
@@ -762,6 +772,17 @@ func (c *ScrapeConfig) Validate(globalConfig GlobalConfig) error {
 		return fmt.Errorf("%w for scrape config with job name %q", err, c.JobName)
 	}
 
+	switch globalConfig.MetricNameValidationScheme {
+	case "", LegacyValidationConfig:
+	case UTF8ValidationConfig:
+		if model.NameValidationScheme != model.UTF8Validation {
+			return fmt.Errorf("utf8 name validation requested but feature not enabled via --enable-feature=utf8-names")
+		}
+	default:
+		return fmt.Errorf("unknown name validation method specified, must be either 'legacy' or 'utf8', got %s", globalConfig.MetricNameValidationScheme)
+	}
+	c.MetricNameValidationScheme = globalConfig.MetricNameValidationScheme
+
 	return nil
 }
 

discovery/discoverer_metrics_noop.go

@@ -13,7 +13,7 @@
 
 package discovery
 
-// Create a dummy metrics struct, because this SD doesn't have any metrics.
+// NoopDiscovererMetrics creates a dummy metrics struct, because this SD doesn't have any metrics.
 type NoopDiscovererMetrics struct{}
 
 var _ DiscovererMetrics = (*NoopDiscovererMetrics)(nil)

discovery/discovery.go

@@ -39,7 +39,7 @@ type Discoverer interface {
 	Run(ctx context.Context, up chan<- []*targetgroup.Group)
 }
 
-// Internal metrics of service discovery mechanisms.
+// DiscovererMetrics are internal metrics of service discovery mechanisms.
 type DiscovererMetrics interface {
 	Register() error
 	Unregister()
@@ -56,7 +56,7 @@ type DiscovererOptions struct {
 	HTTPClientOptions []config.HTTPClientOption
 }
 
-// Metrics used by the "refresh" package.
+// RefreshMetrics are used by the "refresh" package.
 // We define them here in the "discovery" package in order to avoid a cyclic dependency between
 // "discovery" and "refresh".
 type RefreshMetrics struct {
@@ -64,17 +64,18 @@ type RefreshMetrics struct {
 	Duration prometheus.Observer
 }
 
-// Instantiate the metrics used by the "refresh" package.
+// RefreshMetricsInstantiator instantiates the metrics used by the "refresh" package.
 type RefreshMetricsInstantiator interface {
 	Instantiate(mech string) *RefreshMetrics
 }
 
-// An interface for registering, unregistering, and instantiating metrics for the "refresh" package.
+// RefreshMetricsManager is an interface for registering, unregistering, and
-// Refresh metrics are registered and unregistered outside of the service discovery mechanism.
+// instantiating metrics for the "refresh" package. Refresh metrics are
-// This is so that the same metrics can be reused across different service discovery mechanisms.
+// registered and unregistered outside of the service discovery mechanism. This
-// To manage refresh metrics inside the SD mechanism, we'd need to use const labels which are
+// is so that the same metrics can be reused across different service discovery
-// specific to that SD. However, doing so would also expose too many unused metrics on
+// mechanisms. To manage refresh metrics inside the SD mechanism, we'd need to
-// the Prometheus /metrics endpoint.
+// use const labels which are specific to that SD. However, doing so would also
+// expose too many unused metrics on the Prometheus /metrics endpoint.
 type RefreshMetricsManager interface {
 	DiscovererMetrics
 	RefreshMetricsInstantiator
@@ -145,7 +146,8 @@ func (c StaticConfig) NewDiscoverer(DiscovererOptions) (Discoverer, error) {
 	return staticDiscoverer(c), nil
 }
 
-// No metrics are needed for this service discovery mechanism.
+// NewDiscovererMetrics returns NoopDiscovererMetrics because no metrics are
+// needed for this service discovery mechanism.
 func (c StaticConfig) NewDiscovererMetrics(prometheus.Registerer, RefreshMetricsInstantiator) DiscovererMetrics {
 	return &NoopDiscovererMetrics{}
 }

discovery/eureka/client.go

@@ -97,7 +97,6 @@ func fetchApps(ctx context.Context, server string, client *http.Client) (*Applic
 		resp.Body.Close()
 	}()
 
-	//nolint:usestdlibvars
 	if resp.StatusCode/100 != 2 {
 		return nil, fmt.Errorf("non 2xx status '%d' response during eureka service discovery", resp.StatusCode)
 	}

discovery/hetzner/robot.go

@@ -87,7 +87,6 @@ func (d *robotDiscovery) refresh(context.Context) ([]*targetgroup.Group, error)
 		resp.Body.Close()
 	}()
 
-	//nolint:usestdlibvars
 	if resp.StatusCode/100 != 2 {
 		return nil, fmt.Errorf("non 2xx status '%d' response during hetzner service discovery with role robot", resp.StatusCode)
 	}

discovery/kubernetes/endpoints_test.go

@@ -970,7 +970,7 @@ func TestEndpointsDiscoveryEmptyPodStatus(t *testing.T) {
 	}.Run(t)
 }
 
-// TestEndpointsUpdatePod makes sure that Endpoints discovery detects underlying Pods changes.
+// TestEndpointsDiscoveryUpdatePod makes sure that Endpoints discovery detects underlying Pods changes.
 // See https://github.com/prometheus/prometheus/issues/11305 for more details.
 func TestEndpointsDiscoveryUpdatePod(t *testing.T) {
 	pod := &v1.Pod{

discovery/kubernetes/kubernetes_test.go

@@ -154,7 +154,7 @@ func (d k8sDiscoveryTest) Run(t *testing.T) {
 
 // readResultWithTimeout reads all targetgroups from channel with timeout.
 // It merges targetgroups by source and sends the result to result channel.
-func readResultWithTimeout(t *testing.T, ctx context.Context, ch <-chan []*targetgroup.Group, max int, stopAfter time.Duration, resChan chan<- map[string]*targetgroup.Group) {
+func readResultWithTimeout(t *testing.T, ctx context.Context, ch <-chan []*targetgroup.Group, maxGroups int, stopAfter time.Duration, resChan chan<- map[string]*targetgroup.Group) {
 	res := make(map[string]*targetgroup.Group)
 	timeout := time.After(stopAfter)
 Loop:
@@ -167,7 +167,7 @@ Loop:
 				}
 				res[tg.Source] = tg
 			}
-			if len(res) == max {
+			if len(res) == maxGroups {
 				// Reached max target groups we may get, break fast.
 				break Loop
 			}
@@ -175,10 +175,10 @@ Loop:
 			// Because we use queue, an object that is created then
 			// deleted or updated may be processed only once.
 			// So possibly we may skip events, timed out here.
-			t.Logf("timed out, got %d (max: %d) items, some events are skipped", len(res), max)
+			t.Logf("timed out, got %d (max: %d) items, some events are skipped", len(res), maxGroups)
 			break Loop
 		case <-ctx.Done():
-			t.Logf("stopped, got %d (max: %d) items", len(res), max)
+			t.Logf("stopped, got %d (max: %d) items", len(res), maxGroups)
 			break Loop
 		}
 	}

discovery/manager.go

@@ -64,7 +64,7 @@ func (p *Provider) Config() interface{} {
 	return p.config
 }
 
-// Registers the metrics needed for SD mechanisms.
+// CreateAndRegisterSDMetrics registers the metrics needed for SD mechanisms.
 // Does not register the metrics for the Discovery Manager.
 // TODO(ptodev): Add ability to unregister the metrics?
 func CreateAndRegisterSDMetrics(reg prometheus.Registerer) (map[string]DiscovererMetrics, error) {

discovery/metrics_refresh.go

@@ -17,7 +17,7 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 )
 
-// Metric vectors for the "refresh" package.
+// RefreshMetricsVecs are metric vectors for the "refresh" package.
 // We define them here in the "discovery" package in order to avoid a cyclic dependency between
 // "discovery" and "refresh".
 type RefreshMetricsVecs struct {

discovery/util.go

@@ -19,8 +19,8 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 )
 
-// A utility to be used by implementations of discovery.Discoverer
+// MetricRegisterer is used by implementations of discovery.Discoverer that need
-// which need to manage the lifetime of their metrics.
+// to manage the lifetime of their metrics.
 type MetricRegisterer interface {
 	RegisterMetrics() error
 	UnregisterMetrics()
@@ -34,7 +34,7 @@ type metricRegistererImpl struct {
 
 var _ MetricRegisterer = &metricRegistererImpl{}
 
-// Creates an instance of a MetricRegisterer.
+// NewMetricRegisterer creates an instance of a MetricRegisterer.
 // Typically called inside the implementation of the NewDiscoverer() method.
 func NewMetricRegisterer(reg prometheus.Registerer, metrics []prometheus.Collector) MetricRegisterer {
 	return &metricRegistererImpl{

docs/command-line/prometheus.md

@@ -56,7 +56,8 @@ The Prometheus monitoring server
 | <code class="text-nowrap">--query.timeout</code> | Maximum time a query may take before being aborted. Use with server mode only. | `2m` |
 | <code class="text-nowrap">--query.max-concurrency</code> | Maximum number of queries executed concurrently. Use with server mode only. | `20` |
 | <code class="text-nowrap">--query.max-samples</code> | Maximum number of samples a single query can load into memory. Note that queries will fail if they try to load more samples than this into memory, so this also limits the number of samples a query can return. Use with server mode only. | `50000000` |
-| <code class="text-nowrap">--enable-feature</code> | Comma separated feature names to enable. Valid options: agent, auto-gomemlimit, exemplar-storage, expand-external-labels, memory-snapshot-on-shutdown, promql-per-step-stats, promql-experimental-functions, remote-write-receiver (DEPRECATED), extra-scrape-metrics, new-service-discovery-manager, auto-gomaxprocs, no-default-scrape-port, native-histograms, otlp-write-receiver, created-timestamp-zero-ingestion, concurrent-rule-eval, delayed-compaction. See https://prometheus.io/docs/prometheus/latest/feature_flags/ for more details. |  |
+| <code class="text-nowrap">--scrape.name-escaping-scheme</code> | Method for escaping legacy invalid names when sending to Prometheus that does not support UTF-8. Can be one of "values", "underscores", or "dots". | `values` |
+| <code class="text-nowrap">--enable-feature</code> | Comma separated feature names to enable. Valid options: agent, auto-gomemlimit, exemplar-storage, expand-external-labels, memory-snapshot-on-shutdown, promql-per-step-stats, promql-experimental-functions, remote-write-receiver (DEPRECATED), extra-scrape-metrics, new-service-discovery-manager, auto-gomaxprocs, no-default-scrape-port, native-histograms, otlp-write-receiver, created-timestamp-zero-ingestion, concurrent-rule-eval, delayed-compaction, utf8-names. See https://prometheus.io/docs/prometheus/latest/feature_flags/ for more details. |  |
 | <code class="text-nowrap">--log.level</code> | Only log messages with the given severity or above. One of: [debug, info, warn, error] | `info` |
 | <code class="text-nowrap">--log.format</code> | Output format of log messages. One of: [logfmt, json] | `logfmt` |
 

docs/configuration/configuration.md

@@ -121,6 +121,11 @@ global:
   # that will be kept in memory. 0 means no limit.
   [ keep_dropped_targets: <int> | default = 0 ]
 
+  # Specifies the validation scheme for metric and label names. Either blank or
+  # "legacy" for letters, numbers, colons, and underscores; or "utf8" for full
+  # UTF-8 support.
+  [ metric_name_validation_scheme <string> | default "legacy" ]
+
 runtime:
   # Configure the Go garbage collector GOGC parameter
   # See: https://tip.golang.org/doc/gc-guide#GOGC
@@ -461,6 +466,11 @@ metric_relabel_configs:
 # that will be kept in memory. 0 means no limit.
 [ keep_dropped_targets: <int> | default = 0 ]
 
+# Specifies the validation scheme for metric and label names. Either blank or
+# "legacy" for letters, numbers, colons, and underscores; or "utf8" for full
+# UTF-8 support.
+[ metric_name_validation_scheme <string> | default "legacy" ]
+
 # Limit on total number of positive and negative buckets allowed in a single
 # native histogram. The resolution of a histogram with more buckets will be
 # reduced until the number of buckets is within the limit. If the limit cannot
@@ -3265,12 +3275,16 @@ Initially, aside from the configured per-target labels, a target's `job`
 label is set to the `job_name` value of the respective scrape configuration.
 The `__address__` label is set to the `<host>:<port>` address of the target.
 After relabeling, the `instance` label is set to the value of `__address__` by default if
-it was not set during relabeling. The `__scheme__` and `__metrics_path__` labels
+it was not set during relabeling.
-are set to the scheme and metrics path of the target respectively. The `__param_<name>`
+
-label is set to the value of the first passed URL parameter called `<name>`.
+The `__scheme__` and `__metrics_path__` labels
+are set to the scheme and metrics path of the target respectively, as specified in `scrape_config`.
+
+The `__param_<name>`
+label is set to the value of the first passed URL parameter called `<name>`, as defined in `scrape_config`.
 
 The `__scrape_interval__` and `__scrape_timeout__` labels are set to the target's
-interval and timeout.
+interval and timeout, as specified in `scrape_config`.
 
 Additional labels prefixed with `__meta_` may be available during the
 relabeling phase. They are set by the service discovery mechanism that provided

docs/feature_flags.md

@@ -249,3 +249,11 @@ In the event of multiple consecutive Head compactions being possible, only the f
 Note that during this delay, the Head continues its usual operations, which include serving and appending series.
 
 Despite the delay in compaction, the blocks produced are time-aligned in the same manner as they would be if the delay was not in place.
+
+## UTF-8 Name Support
+
+`--enable-feature=utf8-names`
+
+When enabled, changes the metric and label name validation scheme inside Prometheus to allow the full UTF-8 character set.
+By itself, this flag does not enable the request of UTF-8 names via content negotiation.
+Users will also have to set `metric_name_validation_scheme` in scrape configs to enable the feature either on the global config or on a per-scrape config basis.

docs/querying/functions.md

@@ -617,9 +617,9 @@ Like `sort`, `sort_desc` only affects the results of instant queries, as range q
 
 ## `sort_by_label()`
 
-**This function has to be enabled via the [feature flag](../feature_flags.md) `--enable-feature=promql-experimental-functions`.**
+**This function has to be enabled via the [feature flag](../feature_flags.md#experimental-promql-functions) `--enable-feature=promql-experimental-functions`.**
 
-`sort_by_label(v instant-vector, label string, ...)` returns vector elements sorted by their label values and sample value in case of label values being equal, in ascending order.
+`sort_by_label(v instant-vector, label string, ...)` returns vector elements sorted by the values of the given labels in ascending order. In case these label values are equal, elements are sorted by their full label sets.
 
 Please note that the sort by label functions only affect the results of instant queries, as range query results always have a fixed output ordering.
 
@@ -627,7 +627,7 @@ This function uses [natural sort order](https://en.wikipedia.org/wiki/Natural_so
 
 ## `sort_by_label_desc()`
 
-**This function has to be enabled via the [feature flag](../feature_flags.md) `--enable-feature=promql-experimental-functions`.**
+**This function has to be enabled via the [feature flag](../feature_flags.md#experimental-promql-functions) `--enable-feature=promql-experimental-functions`.**
 
 Same as `sort_by_label`, but sorts in descending order.
 
@@ -676,7 +676,7 @@ over time and return an instant vector with per-series aggregation results:
 * `last_over_time(range-vector)`: the most recent point value in the specified interval.
 * `present_over_time(range-vector)`: the value 1 for any series in the specified interval.
 
-If the [feature flag](../feature_flags.md)
+If the [feature flag](../feature_flags.md#experimental-promql-functions)
 `--enable-feature=promql-experimental-functions` is set, the following
 additional functions are available:
 

documentation/examples/remote_storage/go.mod

@@ -20,7 +20,7 @@ require (
 	github.com/Azure/azure-sdk-for-go/sdk/internal v1.8.0 // indirect
 	github.com/AzureAD/microsoft-authentication-library-for-go v1.2.2 // indirect
 	github.com/alecthomas/units v0.0.0-20231202071711-9a357b53e9c9 // indirect
-	github.com/aws/aws-sdk-go v1.53.16 // indirect
+	github.com/aws/aws-sdk-go v1.55.5 // indirect
 	github.com/beorn7/perks v1.0.1 // indirect
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
 	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc // indirect
@@ -62,7 +62,7 @@ require (
 	golang.org/x/text v0.16.0 // indirect
 	golang.org/x/time v0.5.0 // indirect
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20240528184218-531527333157 // indirect
-	google.golang.org/grpc v1.64.0 // indirect
+	google.golang.org/grpc v1.65.0 // indirect
 	google.golang.org/protobuf v1.34.2 // indirect
 	gopkg.in/yaml.v2 v2.4.0 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect

documentation/examples/remote_storage/go.sum

@@ -26,8 +26,8 @@ github.com/alecthomas/units v0.0.0-20231202071711-9a357b53e9c9/go.mod h1:OMCwj8V
 github.com/armon/go-metrics v0.4.1 h1:hR91U9KYmb6bLBYLQjyM+3j+rcd/UhE+G78SFnF8gJA=
 github.com/armon/go-metrics v0.4.1/go.mod h1:E6amYzXo6aW1tqzoZGT755KkbgrJsSdpwZ+3JqfkOG4=
 github.com/aws/aws-sdk-go v1.38.35/go.mod h1:hcU610XS61/+aQV88ixoOzUoG7v3b31pl2zKMmprdro=
-github.com/aws/aws-sdk-go v1.53.16 h1:8oZjKQO/ml1WLUZw5hvF7pvYjPf8o9f57Wldoy/q9Qc=
+github.com/aws/aws-sdk-go v1.55.5 h1:KKUZBfBoyqy5d3swXyiC7Q76ic40rYcbqH7qjh59kzU=
-github.com/aws/aws-sdk-go v1.53.16/go.mod h1:LF8svs817+Nz+DmiMQKTO3ubZ/6IaTpq3TjupRn3Eqk=
+github.com/aws/aws-sdk-go v1.55.5/go.mod h1:eRwEWoyTWFMVYVQzKMNHWP5/RV4xIUGMQfXQHfHkpNU=
 github.com/bboreham/go-loser v0.0.0-20230920113527-fcc2c21820a3 h1:6df1vn4bBlDDo4tARvBm7l6KA9iVMnE3NWizDeWSrps=
 github.com/bboreham/go-loser v0.0.0-20230920113527-fcc2c21820a3/go.mod h1:CIWtjkly68+yqLPbvwwR/fjNJA/idrtULjZWh2v1ys0=
 github.com/beorn7/perks v0.0.0-20180321164747-3a771d992973/go.mod h1:Dwedo/Wpr24TaqPxmxbtue+5NUziq4I4S80YR8gNf3Q=
@@ -37,8 +37,8 @@ github.com/beorn7/perks v1.0.1/go.mod h1:G2ZrVWU2WbWT9wwq4/hrbKbnv/1ERSJQ0ibhJ6r
 github.com/cespare/xxhash/v2 v2.1.1/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
 github.com/cespare/xxhash/v2 v2.3.0 h1:UL815xU9SqsFlibzuggzjXhog7bL6oX9BbNZnL2UFvs=
 github.com/cespare/xxhash/v2 v2.3.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
-github.com/cncf/xds/go v0.0.0-20240318125728-8a4994d93e50 h1:DBmgJDC9dTfkVyGgipamEh2BpGYxScCH1TOF1LL1cXc=
+github.com/cncf/xds/go v0.0.0-20240423153145-555b57ec207b h1:ga8SEFjZ60pxLcmhnThWgvH2wg8376yUJmPhEH4H3kw=
-github.com/cncf/xds/go v0.0.0-20240318125728-8a4994d93e50/go.mod h1:5e1+Vvlzido69INQaVO6d87Qn543Xr6nooe9Kz7oBFM=
+github.com/cncf/xds/go v0.0.0-20240423153145-555b57ec207b/go.mod h1:W+zGtBO5Y1IgJhy4+A9GOqVhqLpfZi+vwmdNXUehLA8=
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc h1:U9qPSI2PIWSS1VwoXQT9A3Wy9MM3WgvqSxFWenqJduM=
@@ -402,8 +402,8 @@ google.golang.org/genproto/googleapis/api v0.0.0-20240528184218-531527333157 h1:
 google.golang.org/genproto/googleapis/api v0.0.0-20240528184218-531527333157/go.mod h1:99sLkeliLXfdj2J75X3Ho+rrVCaJze0uwN7zDDkjPVU=
 google.golang.org/genproto/googleapis/rpc v0.0.0-20240528184218-531527333157 h1:Zy9XzmMEflZ/MAaA7vNcoebnRAld7FsPW1EeBB7V0m8=
 google.golang.org/genproto/googleapis/rpc v0.0.0-20240528184218-531527333157/go.mod h1:EfXuqaE1J41VCDicxHzUDm+8rk+7ZdXzHV0IhO/I6s0=
-google.golang.org/grpc v1.64.0 h1:KH3VH9y/MgNQg1dE7b3XfVK0GsPSIzJwdF617gUSbvY=
+google.golang.org/grpc v1.65.0 h1:bs/cUb4lp1G5iImFFd3u5ixQzweKizoZJAwBNLR42lc=
-google.golang.org/grpc v1.64.0/go.mod h1:oxjF8E3FBnjp+/gVFYdWacaLDx9na1aqy9oovLpxQYg=
+google.golang.org/grpc v1.65.0/go.mod h1:WgYC2ypjlB0EiQi6wdKixMqukr6lBc0Vo+oOgjrM5ZQ=
 google.golang.org/protobuf v0.0.0-20200109180630-ec00e32a8dfd/go.mod h1:DFci5gLYBciE7Vtevhsrf46CRTquxDuWsQurQQe4oz8=
 google.golang.org/protobuf v0.0.0-20200221191635-4d8936d0db64/go.mod h1:kwYJMbMJ01Woi6D6+Kah6886xMZcty6N08ah7+eCXa0=
 google.golang.org/protobuf v0.0.0-20200228230310-ab0ca4ff8a60/go.mod h1:cfTl7dwQJ+fmap5saPgwCLgHXTUD7jkjRqWcaiX5VyM=

model/exemplar/exemplar.go

@@ -15,7 +15,9 @@ package exemplar
 
 import "github.com/prometheus/prometheus/model/labels"
 
-// The combined length of the label names and values of an Exemplar's LabelSet MUST NOT exceed 128 UTF-8 characters
+// ExemplarMaxLabelSetLength is defined by OpenMetrics: "The combined length of
+// the label names and values of an Exemplar's LabelSet MUST NOT exceed 128
+// UTF-8 characters."
 // https://github.com/OpenObservability/OpenMetrics/blob/main/specification/OpenMetrics.md#exemplars
 const ExemplarMaxLabelSetLength = 128
 
@@ -49,7 +51,7 @@ func (e Exemplar) Equals(e2 Exemplar) bool {
 	return e.Value == e2.Value
 }
 
-// Sort first by timestamp, then value, then labels.
+// Compare first timestamps, then values, then labels.
 func Compare(a, b Exemplar) int {
 	if a.Ts < b.Ts {
 		return -1

model/labels/labels.go

@@ -315,7 +315,8 @@ func Compare(a, b Labels) int {
 	return len(a) - len(b)
 }
 
-// Copy labels from b on top of whatever was in ls previously, reusing memory or expanding if needed.
+// CopyFrom copies labels from b on top of whatever was in ls previously,
+// reusing memory or expanding if needed.
 func (ls *Labels) CopyFrom(b Labels) {
 	(*ls) = append((*ls)[:0], b...)
 }
@@ -422,7 +423,7 @@ type ScratchBuilder struct {
 	add Labels
 }
 
-// Symbol-table is no-op, just for api parity with dedupelabels.
+// SymbolTable is no-op, just for api parity with dedupelabels.
 type SymbolTable struct{}
 
 func NewSymbolTable() *SymbolTable { return nil }
@@ -458,7 +459,7 @@ func (b *ScratchBuilder) Add(name, value string) {
 	b.add = append(b.add, Label{Name: name, Value: value})
 }
 
-// Add a name/value pair, using []byte instead of string.
+// UnsafeAddBytes adds a name/value pair, using []byte instead of string.
 // The '-tags stringlabels' version of this function is unsafe, hence the name.
 // This version is safe - it copies the strings immediately - but we keep the same name so everything compiles.
 func (b *ScratchBuilder) UnsafeAddBytes(name, value []byte) {
@@ -475,14 +476,14 @@ func (b *ScratchBuilder) Assign(ls Labels) {
 	b.add = append(b.add[:0], ls...) // Copy on top of our slice, so we don't retain the input slice.
 }
 
-// Return the name/value pairs added so far as a Labels object.
+// Labels returns the name/value pairs added so far as a Labels object.
 // Note: if you want them sorted, call Sort() first.
 func (b *ScratchBuilder) Labels() Labels {
 	// Copy the slice, so the next use of ScratchBuilder doesn't overwrite.
 	return append([]Label{}, b.add...)
 }
 
-// Write the newly-built Labels out to ls.
+// Overwrite the newly-built Labels out to ls.
 // Callers must ensure that there are no other references to ls, or any strings fetched from it.
 func (b *ScratchBuilder) Overwrite(ls *Labels) {
 	*ls = append((*ls)[:0], b.add...)

model/textparse/interface.go

@@ -106,8 +106,8 @@ const (
 	EntryInvalid   Entry = -1
 	EntryType      Entry = 0
 	EntryHelp      Entry = 1
-	EntrySeries    Entry = 2 // A series with a simple float64 as value.
+	EntrySeries    Entry = 2 // EntrySeries marks a series with a simple float64 as value.
 	EntryComment   Entry = 3
 	EntryUnit      Entry = 4
-	EntryHistogram Entry = 5 // A series with a native histogram as a value.
+	EntryHistogram Entry = 5 // EntryHistogram marks a series with a native histogram as a value.
 )

notifier/notifier.go

@@ -674,7 +674,6 @@ func (n *Manager) sendOne(ctx context.Context, c *http.Client, url string, b []b
 	}()
 
 	// Any HTTP status 2xx is OK.
-	//nolint:usestdlibvars
 	if resp.StatusCode/100 != 2 {
 		return fmt.Errorf("bad response status %s", resp.Status)
 	}

notifier/notifier_test.go

@@ -711,7 +711,7 @@ func TestHangingNotifier(t *testing.T) {
 	)
 
 	var (
-		sendTimeout = 10 * time.Millisecond
+		sendTimeout = 100 * time.Millisecond
 		sdUpdatert  = sendTimeout / 2
 
 		done = make(chan struct{})

promql/engine.go

@@ -573,7 +573,7 @@ func (ng *Engine) validateOpts(expr parser.Expr) error {
 	return validationErr
 }
 
-// NewTestQuery: inject special behaviour into Query for testing.
+// NewTestQuery injects special behaviour into Query for testing.
 func (ng *Engine) NewTestQuery(f func(context.Context) error) Query {
 	qry := &query{
 		q:           "test statement",
@@ -2789,6 +2789,7 @@ type groupedAggregation struct {
 	seen                   bool // Was this output groups seen in the input at this timestamp.
 	hasFloat               bool // Has at least 1 float64 sample aggregated.
 	hasHistogram           bool // Has at least 1 histogram sample aggregated.
+	incompatibleHistograms bool // If true, group has seen mixed exponential and custom buckets, or incompatible custom buckets.
 	groupAggrComplete      bool // Used by LIMITK to short-cut series loop when we've reached K elem on every group.
 	incrementalMean        bool // True after reverting to incremental calculation of the mean value.
 }
@@ -2817,6 +2818,7 @@ func (ev *evaluator) aggregation(e *parser.AggregateExpr, q float64, inputMatrix
 				seen:                   true,
 				floatValue:             f,
 				floatMean:              f,
+				incompatibleHistograms: false,
 				groupCount:             1,
 			}
 			switch op {
@@ -2838,6 +2840,10 @@ func (ev *evaluator) aggregation(e *parser.AggregateExpr, q float64, inputMatrix
 			continue
 		}
 
+		if group.incompatibleHistograms {
+			continue
+		}
+
 		switch op {
 		case parser.SUM:
 			if h != nil {
@@ -2846,6 +2852,7 @@ func (ev *evaluator) aggregation(e *parser.AggregateExpr, q float64, inputMatrix
 					_, err := group.histogramValue.Add(h)
 					if err != nil {
 						handleAggregationError(err, e, inputMatrix[si].Metric.Get(model.MetricNameLabel), &annos)
+						group.incompatibleHistograms = true
 					}
 				}
 				// Otherwise the aggregation contained floats
@@ -2866,10 +2873,14 @@ func (ev *evaluator) aggregation(e *parser.AggregateExpr, q float64, inputMatrix
 					toAdd, err := left.Sub(right)
 					if err != nil {
 						handleAggregationError(err, e, inputMatrix[si].Metric.Get(model.MetricNameLabel), &annos)
+						group.incompatibleHistograms = true
+						continue
 					}
 					_, err = group.histogramValue.Add(toAdd)
 					if err != nil {
 						handleAggregationError(err, e, inputMatrix[si].Metric.Get(model.MetricNameLabel), &annos)
+						group.incompatibleHistograms = true
+						continue
 					}
 				}
 				// Otherwise the aggregation contained floats
@@ -2966,6 +2977,8 @@ func (ev *evaluator) aggregation(e *parser.AggregateExpr, q float64, inputMatrix
 				continue
 			}
 			switch {
+			case aggr.incompatibleHistograms:
+				continue
 			case aggr.hasHistogram:
 				aggr.histogramValue = aggr.histogramValue.Compact(0)
 			case aggr.incrementalMean:
@@ -2992,9 +3005,12 @@ func (ev *evaluator) aggregation(e *parser.AggregateExpr, q float64, inputMatrix
 				annos.Add(annotations.NewMixedFloatsHistogramsAggWarning(e.Expr.PositionRange()))
 				continue
 			}
-			if aggr.hasHistogram {
+			switch {
+			case aggr.incompatibleHistograms:
+				continue
+			case aggr.hasHistogram:
 				aggr.histogramValue.Compact(0)
-			} else {
+			default:
 				aggr.floatValue += aggr.floatKahanC
 			}
 		default:
@@ -3161,7 +3177,7 @@ seriesLoop:
 	return mat, annos
 }
 
-// aggregationK evaluates count_values on vec.
+// aggregationCountValues evaluates count_values on vec.
 // Outputs as many series per group as there are values in the input.
 func (ev *evaluator) aggregationCountValues(e *parser.AggregateExpr, grouping []string, valueLabel string, vec Vector, enh *EvalNodeHelper) (Vector, annotations.Annotations) {
 	type groupCount struct {
@@ -3515,14 +3531,14 @@ func makeInt64Pointer(val int64) *int64 {
 	return valp
 }
 
-// Add RatioSampler interface to allow unit-testing (previously: Randomizer).
+// RatioSampler allows unit-testing (previously: Randomizer).
 type RatioSampler interface {
 	// Return this sample "offset" between [0.0, 1.0]
 	sampleOffset(ts int64, sample *Sample) float64
 	AddRatioSample(r float64, sample *Sample) bool
 }
 
-// Use Hash(labels.String()) / maxUint64 as a "deterministic"
+// HashRatioSampler uses Hash(labels.String()) / maxUint64 as a "deterministic"
 // value in [0.0, 1.0].
 type HashRatioSampler struct{}
 

promql/functions.go

@@ -406,17 +406,22 @@ func funcSortDesc(vals []parser.Value, args parser.Expressions, enh *EvalNodeHel
 
 // === sort_by_label(vector parser.ValueTypeVector, label parser.ValueTypeString...) (Vector, Annotations) ===
 func funcSortByLabel(vals []parser.Value, args parser.Expressions, enh *EvalNodeHelper) (Vector, annotations.Annotations) {
-	// In case the labels are the same, NaN should sort to the bottom, so take
+	// First, sort by the full label set. This ensures a consistent ordering in case sorting by the
-	// ascending sort with NaN first and reverse it.
+	// labels provided as arguments is not conclusive.
-	var anno annotations.Annotations
-	vals[0], anno = funcSort(vals, args, enh)
-	labels := stringSliceFromArgs(args[1:])
 	slices.SortFunc(vals[0].(Vector), func(a, b Sample) int {
-		// Iterate over each given label
+		return labels.Compare(a.Metric, b.Metric)
+	})
+
+	labels := stringSliceFromArgs(args[1:])
+	// Next, sort by the labels provided as arguments.
+	slices.SortFunc(vals[0].(Vector), func(a, b Sample) int {
+		// Iterate over each given label.
 		for _, label := range labels {
 			lv1 := a.Metric.Get(label)
 			lv2 := b.Metric.Get(label)
 
+			// If we encounter multiple samples with the same label values, the sorting which was
+			// performed in the first step will act as a "tie breaker".
 			if lv1 == lv2 {
 				continue
 			}
@@ -431,22 +436,27 @@ func funcSortByLabel(vals []parser.Value, args parser.Expressions, enh *EvalNode
 		return 0
 	})
 
-	return vals[0].(Vector), anno
+	return vals[0].(Vector), nil
 }
 
 // === sort_by_label_desc(vector parser.ValueTypeVector, label parser.ValueTypeString...) (Vector, Annotations) ===
 func funcSortByLabelDesc(vals []parser.Value, args parser.Expressions, enh *EvalNodeHelper) (Vector, annotations.Annotations) {
-	// In case the labels are the same, NaN should sort to the bottom, so take
+	// First, sort by the full label set. This ensures a consistent ordering in case sorting by the
-	// ascending sort with NaN first and reverse it.
+	// labels provided as arguments is not conclusive.
-	var anno annotations.Annotations
-	vals[0], anno = funcSortDesc(vals, args, enh)
-	labels := stringSliceFromArgs(args[1:])
 	slices.SortFunc(vals[0].(Vector), func(a, b Sample) int {
-		// Iterate over each given label
+		return labels.Compare(b.Metric, a.Metric)
+	})
+
+	labels := stringSliceFromArgs(args[1:])
+	// Next, sort by the labels provided as arguments.
+	slices.SortFunc(vals[0].(Vector), func(a, b Sample) int {
+		// Iterate over each given label.
 		for _, label := range labels {
 			lv1 := a.Metric.Get(label)
 			lv2 := b.Metric.Get(label)
 
+			// If we encounter multiple samples with the same label values, the sorting which was
+			// performed in the first step will act as a "tie breaker".
 			if lv1 == lv2 {
 				continue
 			}
@@ -461,21 +471,21 @@ func funcSortByLabelDesc(vals []parser.Value, args parser.Expressions, enh *Eval
 		return 0
 	})
 
-	return vals[0].(Vector), anno
+	return vals[0].(Vector), nil
 }
 
 // === clamp(Vector parser.ValueTypeVector, min, max Scalar) (Vector, Annotations) ===
 func funcClamp(vals []parser.Value, args parser.Expressions, enh *EvalNodeHelper) (Vector, annotations.Annotations) {
 	vec := vals[0].(Vector)
-	min := vals[1].(Vector)[0].F
+	minVal := vals[1].(Vector)[0].F
-	max := vals[2].(Vector)[0].F
+	maxVal := vals[2].(Vector)[0].F
-	if max < min {
+	if maxVal < minVal {
 		return enh.Out, nil
 	}
 	for _, el := range vec {
 		enh.Out = append(enh.Out, Sample{
 			Metric: el.Metric.DropMetricName(),
-			F:      math.Max(min, math.Min(max, el.F)),
+			F:      math.Max(minVal, math.Min(maxVal, el.F)),
 		})
 	}
 	return enh.Out, nil
@@ -484,11 +494,11 @@ func funcClamp(vals []parser.Value, args parser.Expressions, enh *EvalNodeHelper
 // === clamp_max(Vector parser.ValueTypeVector, max Scalar) (Vector, Annotations) ===
 func funcClampMax(vals []parser.Value, args parser.Expressions, enh *EvalNodeHelper) (Vector, annotations.Annotations) {
 	vec := vals[0].(Vector)
-	max := vals[1].(Vector)[0].F
+	maxVal := vals[1].(Vector)[0].F
 	for _, el := range vec {
 		enh.Out = append(enh.Out, Sample{
 			Metric: el.Metric.DropMetricName(),
-			F:      math.Min(max, el.F),
+			F:      math.Min(maxVal, el.F),
 		})
 	}
 	return enh.Out, nil
@@ -497,11 +507,11 @@ func funcClampMax(vals []parser.Value, args parser.Expressions, enh *EvalNodeHel
 // === clamp_min(Vector parser.ValueTypeVector, min Scalar) (Vector, Annotations) ===
 func funcClampMin(vals []parser.Value, args parser.Expressions, enh *EvalNodeHelper) (Vector, annotations.Annotations) {
 	vec := vals[0].(Vector)
-	min := vals[1].(Vector)[0].F
+	minVal := vals[1].(Vector)[0].F
 	for _, el := range vec {
 		enh.Out = append(enh.Out, Sample{
 			Metric: el.Metric.DropMetricName(),
-			F:      math.Max(min, el.F),
+			F:      math.Max(minVal, el.F),
 		})
 	}
 	return enh.Out, nil
@@ -700,13 +710,13 @@ func funcMaxOverTime(vals []parser.Value, args parser.Expressions, enh *EvalNode
 		return enh.Out, nil
 	}
 	return aggrOverTime(vals, enh, func(s Series) float64 {
-		max := s.Floats[0].F
+		maxVal := s.Floats[0].F
 		for _, f := range s.Floats {
-			if f.F > max || math.IsNaN(max) {
+			if f.F > maxVal || math.IsNaN(maxVal) {
-				max = f.F
+				maxVal = f.F
 			}
 		}
-		return max
+		return maxVal
 	}), nil
 }
 
@@ -720,13 +730,13 @@ func funcMinOverTime(vals []parser.Value, args parser.Expressions, enh *EvalNode
 		return enh.Out, nil
 	}
 	return aggrOverTime(vals, enh, func(s Series) float64 {
-		min := s.Floats[0].F
+		minVal := s.Floats[0].F
 		for _, f := range s.Floats {
-			if f.F < min || math.IsNaN(min) {
+			if f.F < minVal || math.IsNaN(minVal) {
-				min = f.F
+				minVal = f.F
 			}
 		}
-		return min
+		return minVal
 	}), nil
 }
 

promql/parser/ast.go

@@ -352,8 +352,7 @@ func (f inspector) Visit(node Node, path []Node) (Visitor, error) {
 // f(node, path); node must not be nil. If f returns a nil error, Inspect invokes f
 // for all the non-nil children of node, recursively.
 func Inspect(node Node, f inspector) {
-	//nolint: errcheck
+	Walk(f, node, nil) //nolint:errcheck
-	Walk(f, node, nil)
 }
 
 // Children returns a list of all child nodes of a syntax tree node.
@@ -419,7 +418,7 @@ func mergeRanges(first, last Node) posrange.PositionRange {
 	}
 }
 
-// Item implements the Node interface.
+// PositionRange implements the Node interface.
 // This makes it possible to call mergeRanges on them.
 func (i *Item) PositionRange() posrange.PositionRange {
 	return posrange.PositionRange{

promql/parser/lex.go

@@ -727,23 +727,23 @@ func lexValueSequence(l *Lexer) stateFn {
 // was only modified to integrate with our lexer.
 func lexEscape(l *Lexer) stateFn {
 	var n int
-	var base, max uint32
+	var base, maxVal uint32
 
 	ch := l.next()
 	switch ch {
 	case 'a', 'b', 'f', 'n', 'r', 't', 'v', '\\', l.stringOpen:
 		return lexString
 	case '0', '1', '2', '3', '4', '5', '6', '7':
-		n, base, max = 3, 8, 255
+		n, base, maxVal = 3, 8, 255
 	case 'x':
 		ch = l.next()
-		n, base, max = 2, 16, 255
+		n, base, maxVal = 2, 16, 255
 	case 'u':
 		ch = l.next()
-		n, base, max = 4, 16, unicode.MaxRune
+		n, base, maxVal = 4, 16, unicode.MaxRune
 	case 'U':
 		ch = l.next()
-		n, base, max = 8, 16, unicode.MaxRune
+		n, base, maxVal = 8, 16, unicode.MaxRune
 	case eof:
 		l.errorf("escape sequence not terminated")
 		return lexString
@@ -772,7 +772,7 @@ func lexEscape(l *Lexer) stateFn {
 		}
 	}
 
-	if x > max || 0xD800 <= x && x < 0xE000 {
+	if x > maxVal || 0xD800 <= x && x < 0xE000 {
 		l.errorf("escape sequence is an invalid Unicode code point")
 	}
 	return lexString

promql/promqltest/testdata/functions.test

@@ -523,16 +523,16 @@ load 5m
 	node_uname_info{job="node_exporter", instance="4m1000", release="1.111.3"} 0+10x10
 
 eval_ordered instant at 50m sort_by_label(http_requests, "instance")
-	http_requests{group="production", instance="0", job="api-server"} 100
 	http_requests{group="canary", instance="0", job="api-server"} 300
-	http_requests{group="production", instance="0", job="app-server"} 500
 	http_requests{group="canary", instance="0", job="app-server"} 700
-	http_requests{group="production", instance="1", job="api-server"} 200
+	http_requests{group="production", instance="0", job="api-server"} 100
+	http_requests{group="production", instance="0", job="app-server"} 500
 	http_requests{group="canary", instance="1", job="api-server"} 400
-	http_requests{group="production", instance="1", job="app-server"} 600
 	http_requests{group="canary", instance="1", job="app-server"} 800
-	http_requests{group="production", instance="2", job="api-server"} 100
+	http_requests{group="production", instance="1", job="api-server"} 200
+	http_requests{group="production", instance="1", job="app-server"} 600
 	http_requests{group="canary", instance="2", job="api-server"} NaN
+	http_requests{group="production", instance="2", job="api-server"} 100
 
 eval_ordered instant at 50m sort_by_label(http_requests, "instance", "group")
 	http_requests{group="canary", instance="0", job="api-server"} 300
@@ -585,14 +585,14 @@ eval_ordered instant at 50m sort_by_label(http_requests, "job", "instance", "gro
 eval_ordered instant at 50m sort_by_label_desc(http_requests, "instance")
 	http_requests{group="production", instance="2", job="api-server"} 100
 	http_requests{group="canary", instance="2", job="api-server"} NaN
-	http_requests{group="canary", instance="1", job="app-server"} 800
 	http_requests{group="production", instance="1", job="app-server"} 600
-	http_requests{group="canary", instance="1", job="api-server"} 400
 	http_requests{group="production", instance="1", job="api-server"} 200
-	http_requests{group="canary", instance="0", job="app-server"} 700
+	http_requests{group="canary", instance="1", job="app-server"} 800
+	http_requests{group="canary", instance="1", job="api-server"} 400
 	http_requests{group="production", instance="0", job="app-server"} 500
-	http_requests{group="canary", instance="0", job="api-server"} 300
 	http_requests{group="production", instance="0", job="api-server"} 100
+	http_requests{group="canary", instance="0", job="app-server"} 700
+	http_requests{group="canary", instance="0", job="api-server"} 300
 
 eval_ordered instant at 50m sort_by_label_desc(http_requests, "instance", "group")
 	http_requests{group="production", instance="2", job="api-server"} 100

promql/promqltest/testdata/native_histograms.test

@@ -785,6 +785,8 @@ eval_warn instant at 1m rate(some_metric[30s])
 eval_warn instant at 30s rate(some_metric[30s])
     # Should produce no results.
 
+clear
+
 # Histogram with constant buckets.
 load 1m
     const_histogram {{schema:0 sum:1 count:1 buckets:[1 1 1]}} {{schema:0 sum:1 count:1 buckets:[1 1 1]}} {{schema:0 sum:1 count:1 buckets:[1 1 1]}} {{schema:0 sum:1 count:1 buckets:[1 1 1]}} {{schema:0 sum:1 count:1 buckets:[1 1 1]}}
@@ -828,3 +830,59 @@ eval instant at 5m histogram_stddev(rate(const_histogram[5m]))
 # Zero buckets mean no observations, so there is no standard variance.
 eval instant at 5m histogram_stdvar(rate(const_histogram[5m]))
     {} NaN
+
+clear
+
+# Test mixing exponential and custom buckets.
+load 6m
+  metric{series="exponential"}         {{sum:4 count:3 buckets:[1 2 1]}}  _                                                                 {{sum:4 count:3 buckets:[1 2 1]}}
+  metric{series="other-exponential"}   {{sum:3 count:2 buckets:[1 1 1]}}  _                                                                 {{sum:3 count:2 buckets:[1 1 1]}}
+  metric{series="custom"}              _                                  {{schema:-53 sum:1 count:1 custom_values:[5 10] buckets:[1]}}     {{schema:-53 sum:1 count:1 custom_values:[5 10] buckets:[1]}}
+  metric{series="other-custom"}        _                                  {{schema:-53 sum:15 count:2 custom_values:[5 10] buckets:[0 2]}}  {{schema:-53 sum:15 count:2 custom_values:[5 10] buckets:[0 2]}}
+
+# T=0: only exponential
+# T=6: only custom
+# T=12: mixed, should be ignored and emit a warning
+eval_warn range from 0 to 12m step 6m sum(metric)
+  {} {{sum:7 count:5 buckets:[2 3 2]}} {{schema:-53 sum:16 count:3 custom_values:[5 10] buckets:[1 2]}} _
+
+eval_warn range from 0 to 12m step 6m avg(metric)
+  {} {{sum:3.5 count:2.5 buckets:[1 1.5 1]}} {{schema:-53 sum:8 count:1.5 custom_values:[5 10] buckets:[0.5 1]}} _
+
+clear
+
+# Test incompatible custom bucket schemas.
+load 6m
+  metric{series="1"} _                                                             {{schema:-53 sum:1 count:1 custom_values:[5 10] buckets:[1]}} {{schema:-53 sum:1 count:1 custom_values:[5 10] buckets:[1]}}
+  metric{series="2"} {{schema:-53 sum:1 count:1 custom_values:[2] buckets:[1]}}    _                                                             {{schema:-53 sum:1 count:1 custom_values:[2] buckets:[1]}}
+  metric{series="3"} {{schema:-53 sum:1 count:1 custom_values:[5 10] buckets:[1]}} {{schema:-53 sum:1 count:1 custom_values:[5 10] buckets:[1]}} {{schema:-53 sum:1 count:1 custom_values:[5 10] buckets:[1]}}
+
+# T=0: incompatible, should be ignored and emit a warning
+# T=6: compatible
+# T=12: incompatible followed by compatible, should be ignored and emit a warning
+eval_warn range from 0 to 12m step 6m sum(metric)
+  {} _ {{schema:-53 sum:2 count:2 custom_values:[5 10] buckets:[2]}} _
+
+eval_warn range from 0 to 12m step 6m avg(metric)
+  {} _ {{schema:-53 sum:1 count:1 custom_values:[5 10] buckets:[1]}} _
+
+clear
+
+load 1m
+  metric{group="just-floats", series="1"} 2
+  metric{group="just-floats", series="2"} 3
+  metric{group="just-exponential-histograms", series="1"} {{sum:3 count:4 buckets:[1 2 1]}}
+  metric{group="just-exponential-histograms", series="2"} {{sum:2 count:3 buckets:[1 1 1]}}
+  metric{group="just-custom-histograms", series="1"} {{schema:-53 sum:1 count:1 custom_values:[2] buckets:[1]}}
+  metric{group="just-custom-histograms", series="2"} {{schema:-53 sum:3 count:4 custom_values:[2] buckets:[7]}}
+  metric{group="floats-and-histograms", series="1"} 2
+  metric{group="floats-and-histograms", series="2"} {{sum:2 count:3 buckets:[1 1 1]}}
+  metric{group="exponential-and-custom-histograms", series="1"} {{sum:2 count:3 buckets:[1 1 1]}}
+  metric{group="exponential-and-custom-histograms", series="2"} {{schema:-53 sum:1 count:1 custom_values:[5 10] buckets:[1]}}
+  metric{group="incompatible-custom-histograms", series="1"} {{schema:-53 sum:1 count:1 custom_values:[5 10] buckets:[1]}}
+  metric{group="incompatible-custom-histograms", series="2"} {{schema:-53 sum:1 count:1 custom_values:[2] buckets:[1]}}
+
+eval_warn instant at 0 sum by (group) (metric)
+  {group="just-floats"} 5
+  {group="just-exponential-histograms"} {{sum:5 count:7 buckets:[2 3 2]}}
+  {group="just-custom-histograms"} {{schema:-53 sum:4 count:5 custom_values:[2] buckets:[8]}}

scrape/clientprotobuf.go

@@ -23,7 +23,7 @@ import (
 	dto "github.com/prometheus/client_model/go"
 )
 
-// Write a MetricFamily into a protobuf.
+// MetricFamilyToProtobuf writes a MetricFamily into a protobuf.
 // This function is intended for testing scraping by providing protobuf serialized input.
 func MetricFamilyToProtobuf(metricFamily *dto.MetricFamily) ([]byte, error) {
 	buffer := &bytes.Buffer{}
@@ -34,7 +34,7 @@ func MetricFamilyToProtobuf(metricFamily *dto.MetricFamily) ([]byte, error) {
 	return buffer.Bytes(), nil
 }
 
-// Append a MetricFamily protobuf representation to a buffer.
+// AddMetricFamilyToProtobuf appends a MetricFamily protobuf representation to a buffer.
 // This function is intended for testing scraping by providing protobuf serialized input.
 func AddMetricFamilyToProtobuf(buffer *bytes.Buffer, metricFamily *dto.MetricFamily) error {
 	protoBuf, err := proto.Marshal(metricFamily)

scrape/manager.go

@@ -93,6 +93,8 @@ type Options struct {
 	skipOffsetting bool
 }
 
+const DefaultNameEscapingScheme = model.ValueEncodingEscaping
+
 // Manager maintains a set of scrape pools and manages start/stop cycles
 // when receiving new target groups from the discovery manager.
 type Manager struct {

scrape/scrape.go

@@ -303,6 +303,11 @@ func (sp *scrapePool) restartLoops(reuseCache bool) {
 		mrc                      = sp.config.MetricRelabelConfigs
 	)
 
+	validationScheme := model.LegacyValidation
+	if sp.config.MetricNameValidationScheme == config.UTF8ValidationConfig {
+		validationScheme = model.UTF8Validation
+	}
+
 	sp.targetMtx.Lock()
 
 	forcedErr := sp.refreshTargetLimitErr()
@@ -323,7 +328,7 @@ func (sp *scrapePool) restartLoops(reuseCache bool) {
 				client:               sp.client,
 				timeout:              timeout,
 				bodySizeLimit:        bodySizeLimit,
-				acceptHeader:         acceptHeader(sp.config.ScrapeProtocols),
+				acceptHeader:         acceptHeader(sp.config.ScrapeProtocols, validationScheme),
 				acceptEncodingHeader: acceptEncodingHeader(enableCompression),
 			}
 			newLoop = sp.newLoop(scrapeLoopOptions{
@@ -452,6 +457,11 @@ func (sp *scrapePool) sync(targets []*Target) {
 		scrapeClassicHistograms  = sp.config.ScrapeClassicHistograms
 	)
 
+	validationScheme := model.LegacyValidation
+	if sp.config.MetricNameValidationScheme == config.UTF8ValidationConfig {
+		validationScheme = model.UTF8Validation
+	}
+
 	sp.targetMtx.Lock()
 	for _, t := range targets {
 		hash := t.hash()
@@ -467,7 +477,7 @@ func (sp *scrapePool) sync(targets []*Target) {
 				client:               sp.client,
 				timeout:              timeout,
 				bodySizeLimit:        bodySizeLimit,
-				acceptHeader:         acceptHeader(sp.config.ScrapeProtocols),
+				acceptHeader:         acceptHeader(sp.config.ScrapeProtocols, validationScheme),
 				acceptEncodingHeader: acceptEncodingHeader(enableCompression),
 				metrics:              sp.metrics,
 			}
@@ -714,11 +724,16 @@ var errBodySizeLimit = errors.New("body size limit exceeded")
 // acceptHeader transforms preference from the options into specific header values as
 // https://www.rfc-editor.org/rfc/rfc9110.html#name-accept defines.
 // No validation is here, we expect scrape protocols to be validated already.
-func acceptHeader(sps []config.ScrapeProtocol) string {
+func acceptHeader(sps []config.ScrapeProtocol, scheme model.ValidationScheme) string {
 	var vals []string
 	weight := len(config.ScrapeProtocolsHeaders) + 1
 	for _, sp := range sps {
-		vals = append(vals, fmt.Sprintf("%s;q=0.%d", config.ScrapeProtocolsHeaders[sp], weight))
+		val := config.ScrapeProtocolsHeaders[sp]
+		if scheme == model.UTF8Validation {
+			val += ";" + config.UTF8NamesHeader
+		}
+		val += fmt.Sprintf(";q=0.%d", weight)
+		vals = append(vals, val)
 		weight--
 	}
 	// Default match anything.

scrape/scrape_test.go

@@ -2339,11 +2339,15 @@ func TestTargetScraperScrapeOK(t *testing.T) {
 	)
 
 	var protobufParsing bool
+	var allowUTF8 bool
 
 	server := httptest.NewServer(
 		http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-			if protobufParsing {
 			accept := r.Header.Get("Accept")
+			if allowUTF8 {
+				require.Truef(t, strings.Contains(accept, "escaping=allow-utf-8"), "Expected Accept header to allow utf8, got %q", accept)
+			}
+			if protobufParsing {
 				require.True(t, strings.HasPrefix(accept, "application/vnd.google.protobuf;"),
 					"Expected Accept header to prefer application/vnd.google.protobuf.")
 			}
@@ -2351,7 +2355,11 @@ func TestTargetScraperScrapeOK(t *testing.T) {
 			timeout := r.Header.Get("X-Prometheus-Scrape-Timeout-Seconds")
 			require.Equal(t, expectedTimeout, timeout, "Expected scrape timeout header.")
 
+			if allowUTF8 {
+				w.Header().Set("Content-Type", `text/plain; version=1.0.0; escaping=allow-utf-8`)
+			} else {
 				w.Header().Set("Content-Type", `text/plain; version=0.0.4`)
+			}
 			w.Write([]byte("metric_a 1\nmetric_b 2\n"))
 		}),
 	)
@@ -2380,13 +2388,22 @@ func TestTargetScraperScrapeOK(t *testing.T) {
 		require.NoError(t, err)
 		contentType, err := ts.readResponse(context.Background(), resp, &buf)
 		require.NoError(t, err)
+		if allowUTF8 {
+			require.Equal(t, "text/plain; version=1.0.0; escaping=allow-utf-8", contentType)
+		} else {
 			require.Equal(t, "text/plain; version=0.0.4", contentType)
+		}
 		require.Equal(t, "metric_a 1\nmetric_b 2\n", buf.String())
 	}
 
-	runTest(acceptHeader(config.DefaultScrapeProtocols))
+	runTest(acceptHeader(config.DefaultScrapeProtocols, model.LegacyValidation))
 	protobufParsing = true
-	runTest(acceptHeader(config.DefaultProtoFirstScrapeProtocols))
+	runTest(acceptHeader(config.DefaultProtoFirstScrapeProtocols, model.LegacyValidation))
+	protobufParsing = false
+	allowUTF8 = true
+	runTest(acceptHeader(config.DefaultScrapeProtocols, model.UTF8Validation))
+	protobufParsing = true
+	runTest(acceptHeader(config.DefaultProtoFirstScrapeProtocols, model.UTF8Validation))
 }
 
 func TestTargetScrapeScrapeCancel(t *testing.T) {
@@ -2412,7 +2429,7 @@ func TestTargetScrapeScrapeCancel(t *testing.T) {
 			),
 		},
 		client:       http.DefaultClient,
-		acceptHeader: acceptHeader(config.DefaultGlobalConfig.ScrapeProtocols),
+		acceptHeader: acceptHeader(config.DefaultGlobalConfig.ScrapeProtocols, model.LegacyValidation),
 	}
 	ctx, cancel := context.WithCancel(context.Background())
 
@@ -2467,7 +2484,7 @@ func TestTargetScrapeScrapeNotFound(t *testing.T) {
 			),
 		},
 		client:       http.DefaultClient,
-		acceptHeader: acceptHeader(config.DefaultGlobalConfig.ScrapeProtocols),
+		acceptHeader: acceptHeader(config.DefaultGlobalConfig.ScrapeProtocols, model.LegacyValidation),
 	}
 
 	resp, err := ts.scrape(context.Background())
@@ -2511,7 +2528,7 @@ func TestTargetScraperBodySizeLimit(t *testing.T) {
 		},
 		client:        http.DefaultClient,
 		bodySizeLimit: bodySizeLimit,
-		acceptHeader:  acceptHeader(config.DefaultGlobalConfig.ScrapeProtocols),
+		acceptHeader:  acceptHeader(config.DefaultGlobalConfig.ScrapeProtocols, model.LegacyValidation),
 		metrics:       newTestScrapeMetrics(t),
 	}
 	var buf bytes.Buffer

scrape/testdata/ca.cer

@@ -1,8 +1,66 @@
+Certificate:
+    Data:
+        Version: 3 (0x2)
+        Serial Number:
+            93:6c:9e:29:8d:37:7b:66
+        Signature Algorithm: sha256WithRSAEncryption
+        Issuer: C = XX, L = Default City, O = Default Company Ltd, CN = Prometheus Test CA
+        Validity
+            Not Before: Aug 20 11:51:23 2024 GMT
+            Not After : Dec  5 11:51:23 2044 GMT
+        Subject: C = XX, L = Default City, O = Default Company Ltd, CN = Prometheus Test CA
+        Subject Public Key Info:
+            Public Key Algorithm: rsaEncryption
+                Public-Key: (2048 bit)
+                Modulus:
+                    00:e9:52:05:4d:f2:5a:95:04:2d:b8:73:8b:3c:e7:
+                    47:48:67:00:be:dd:6c:41:f3:7c:4b:44:73:a3:77:
+                    3e:84:af:30:d7:2a:ad:45:6a:b7:89:23:05:15:b3:
+                    aa:46:79:b8:95:64:cc:13:c4:44:a1:01:a0:e2:3d:
+                    a5:67:2b:aa:d3:13:06:43:33:1c:96:36:12:9e:c6:
+                    1d:36:9b:d7:47:bd:28:2d:88:15:04:fa:14:a3:ff:
+                    8c:26:22:c5:a2:15:c7:76:b3:11:f6:a3:44:9a:28:
+                    0f:ca:fb:f4:51:a8:6a:05:94:7c:77:47:c8:21:56:
+                    25:bf:e2:2e:df:33:f3:e4:bd:d6:47:a5:49:13:12:
+                    c8:1f:4c:d7:2a:56:a2:6c:c1:cf:55:05:5d:9a:75:
+                    a2:23:4e:e6:8e:ff:76:05:d6:e0:c8:0b:51:f0:b6:
+                    a1:b2:7d:8f:9c:6a:a5:ce:86:92:15:8c:5b:86:45:
+                    c0:4a:ff:54:b8:ee:cf:11:bd:07:cb:4b:7d:0b:a1:
+                    9d:72:86:9f:55:bc:f9:6c:d9:55:60:96:30:3f:ec:
+                    2d:f6:5f:9a:32:9a:5a:5c:1c:5f:32:f9:d6:0f:04:
+                    f8:81:08:04:9a:95:c3:9d:5a:30:8e:a5:0e:47:2f:
+                    00:ce:e0:2e:ad:5a:b8:b6:4c:55:7c:8a:59:22:b0:
+                    ed:73
+                Exponent: 65537 (0x10001)
+        X509v3 extensions:
+            X509v3 Subject Key Identifier: 
+                CC:F5:05:99:E5:AB:12:69:D8:78:89:4A:31:CA:F0:8B:0B:AD:66:1B
+            X509v3 Authority Key Identifier: 
+                CC:F5:05:99:E5:AB:12:69:D8:78:89:4A:31:CA:F0:8B:0B:AD:66:1B
+            X509v3 Basic Constraints: 
+                CA:TRUE
+    Signature Algorithm: sha256WithRSAEncryption
+    Signature Value:
+        4a:a1:b0:bc:c8:87:4f:7c:96:62:e5:09:29:ae:3a:2e:68:d0:
+        d2:c5:68:ed:ea:83:36:b1:86:f3:b9:e9:19:2b:b6:73:10:6f:
+        df:7f:bb:f1:76:81:03:c1:a1:5a:ee:6c:44:b8:7c:10:d1:5a:
+        d7:c1:92:64:59:35:a6:e0:aa:08:41:37:6e:e7:c8:b6:bd:0c:
+        4b:47:78:ec:c4:b4:15:a3:62:76:4a:39:8e:6e:19:ff:f0:c0:
+        8a:7e:1c:cd:87:e5:00:6c:f1:ce:27:26:ff:b8:e9:eb:f7:2f:
+        bd:c2:4b:9c:d6:57:de:74:74:b3:4f:03:98:9a:b5:08:2d:16:
+        ca:7f:b6:c8:76:62:86:1b:7c:f2:3e:6c:78:cc:2c:95:9a:bb:
+        77:25:e8:80:ff:9b:e8:f8:9a:85:3b:85:b7:17:4e:77:a1:cf:
+        4d:b9:d0:25:e8:5d:8c:e6:7c:f1:d9:52:30:3d:ec:2b:37:91:
+        bc:e2:e8:39:31:6f:3d:e9:98:70:80:7c:41:dd:19:13:05:21:
+        94:7b:16:cf:d8:ee:4e:38:34:5e:6a:ff:cd:85:ac:8f:94:9a:
+        dd:4e:77:05:13:a6:b4:80:52:b2:97:64:76:88:f4:dd:42:0a:
+        50:1c:80:fd:4b:6e:a9:62:10:aa:ef:2e:c1:2f:be:0e:c2:2e:
+        b5:28:5f:83
 -----BEGIN CERTIFICATE-----
 MIIDkTCCAnmgAwIBAgIJAJNsnimNN3tmMA0GCSqGSIb3DQEBCwUAMF8xCzAJBgNV
 BAYTAlhYMRUwEwYDVQQHDAxEZWZhdWx0IENpdHkxHDAaBgNVBAoME0RlZmF1bHQg
-Q29tcGFueSBMdGQxGzAZBgNVBAMMElByb21ldGhldXMgVGVzdCBDQTAeFw0xNTA4
+Q29tcGFueSBMdGQxGzAZBgNVBAMMElByb21ldGhldXMgVGVzdCBDQTAeFw0yNDA4
-MDQxNDA5MjFaFw0yNTA4MDExNDA5MjFaMF8xCzAJBgNVBAYTAlhYMRUwEwYDVQQH
+MjAxMTUxMjNaFw00NDEyMDUxMTUxMjNaMF8xCzAJBgNVBAYTAlhYMRUwEwYDVQQH
 DAxEZWZhdWx0IENpdHkxHDAaBgNVBAoME0RlZmF1bHQgQ29tcGFueSBMdGQxGzAZ
 BgNVBAMMElByb21ldGhldXMgVGVzdCBDQTCCASIwDQYJKoZIhvcNAQEBBQADggEP
 ADCCAQoCggEBAOlSBU3yWpUELbhzizznR0hnAL7dbEHzfEtEc6N3PoSvMNcqrUVq
@@ -12,11 +70,11 @@ yB9M1ypWomzBz1UFXZp1oiNO5o7/dgXW4MgLUfC2obJ9j5xqpc6GkhWMW4ZFwEr/
 VLjuzxG9B8tLfQuhnXKGn1W8+WzZVWCWMD/sLfZfmjKaWlwcXzL51g8E+IEIBJqV
 w51aMI6lDkcvAM7gLq1auLZMVXyKWSKw7XMCAwEAAaNQME4wHQYDVR0OBBYEFMz1
 BZnlqxJp2HiJSjHK8IsLrWYbMB8GA1UdIwQYMBaAFMz1BZnlqxJp2HiJSjHK8IsL
-rWYbMAwGA1UdEwQFMAMBAf8wDQYJKoZIhvcNAQELBQADggEBAI2iA3w3TK5J15Pu
+rWYbMAwGA1UdEwQFMAMBAf8wDQYJKoZIhvcNAQELBQADggEBAEqhsLzIh098lmLl
-e4fPFB4jxQqsbUwuyXbCCv/jKLeFNCD4BjM181WZEYjPMumeTBVzU3aF45LWQIG1
+CSmuOi5o0NLFaO3qgzaxhvO56RkrtnMQb99/u/F2gQPBoVrubES4fBDRWtfBkmRZ
-0DJcrCL4mjMz9qgAoGqA7aDDXiJGbukMgYYsn7vrnVmrZH8T3E8ySlltr7+W578k
+NabgqghBN27nyLa9DEtHeOzEtBWjYnZKOY5uGf/wwIp+HM2H5QBs8c4nJv+46ev3
-pJ5FxnbCroQwn0zLyVB3sFbS8E3vpBr3L8oy8PwPHhIScexcNVc3V6/m4vTZsXTH
+L73CS5zWV950dLNPA5iatQgtFsp/tsh2YoYbfPI+bHjMLJWau3cl6ID/m+j4moU7
-U+vUm1XhDgpDcFMTg2QQiJbfpOYUkwIgnRDAT7t282t2KQWtnlqc3zwPQ1F/6Cpx
+hbcXTnehz0250CXoXYzmfPHZUjA97Cs3kbzi6Dkxbz3pmHCAfEHdGRMFIZR7Fs/Y
-j19JeNsaF1DArkD7YlyKj/GhZLtHwFHG5cxznH0mLDJTW7bQvqqh2iQTeXmBk1lU
+7k44NF5q/82FrI+Umt1OdwUTprSAUrKXZHaI9N1CClAcgP1LbqliEKrvLsEvvg7C
-mM5lH/s=
+LrUoX4M=
 -----END CERTIFICATE-----

scripts/golangci-lint.yml

@@ -36,4 +36,4 @@ jobs:
         uses: golangci/golangci-lint-action@aaa42aa0628b4ae2578232a66b541047968fac86 # v6.1.0
         with:
           args: --verbose
-          version: v1.59.1
+          version: v1.60.2

storage/buffer_test.go

@@ -96,10 +96,10 @@ func TestSampleRingMixed(t *testing.T) {
 
 	// With ValNone as the preferred type, nothing should be initialized.
 	r := newSampleRing(10, 2, chunkenc.ValNone)
-	require.Zero(t, len(r.fBuf))
+	require.Empty(t, r.fBuf)
-	require.Zero(t, len(r.hBuf))
+	require.Empty(t, r.hBuf)
-	require.Zero(t, len(r.fhBuf))
+	require.Empty(t, r.fhBuf)
-	require.Zero(t, len(r.iBuf))
+	require.Empty(t, r.iBuf)
 
 	// But then mixed adds should work as expected.
 	r.addF(fSample{t: 1, f: 3.14})
@@ -146,10 +146,10 @@ func TestSampleRingAtFloatHistogram(t *testing.T) {
 
 	// With ValNone as the preferred type, nothing should be initialized.
 	r := newSampleRing(10, 2, chunkenc.ValNone)
-	require.Zero(t, len(r.fBuf))
+	require.Empty(t, r.fBuf)
-	require.Zero(t, len(r.hBuf))
+	require.Empty(t, r.hBuf)
-	require.Zero(t, len(r.fhBuf))
+	require.Empty(t, r.fhBuf)
-	require.Zero(t, len(r.iBuf))
+	require.Empty(t, r.iBuf)
 
 	var (
 		h  *histogram.Histogram

storage/interface.go

@@ -228,9 +228,9 @@ type LabelHints struct {
 	Limit int
 }
 
-// TODO(bwplotka): Move to promql/engine_test.go?
 // QueryableFunc is an adapter to allow the use of ordinary functions as
 // Queryables. It follows the idea of http.HandlerFunc.
+// TODO(bwplotka): Move to promql/engine_test.go?
 type QueryableFunc func(mint, maxt int64) (Querier, error)
 
 // Querier calls f() with the given parameters.

storage/remote/azuread/azuread.go

@@ -31,13 +31,15 @@ import (
 	"github.com/google/uuid"
 )
 
+// Clouds.
 const (
-	// Clouds.
 	AzureChina      = "AzureChina"
 	AzureGovernment = "AzureGovernment"
 	AzurePublic     = "AzurePublic"
+)
 
-	// Audiences.
+// Audiences.
+const (
 	IngestionChinaAudience      = "https://monitor.azure.cn//.default"
 	IngestionGovernmentAudience = "https://monitor.azure.us//.default"
 	IngestionPublicAudience     = "https://monitor.azure.com//.default"

storage/remote/client.go

@@ -287,7 +287,6 @@ func (c *Client) Store(ctx context.Context, req []byte, attempt int) (WriteRespo
 	// we can continue handling.
 	rs, _ := ParseWriteResponseStats(httpResp)
 
-	//nolint:usestdlibvars
 	if httpResp.StatusCode/100 == 2 {
 		return rs, nil
 	}
@@ -297,7 +296,6 @@ func (c *Client) Store(ctx context.Context, req []byte, attempt int) (WriteRespo
 	body, _ := io.ReadAll(io.LimitReader(httpResp.Body, maxErrMsgLen))
 	err = fmt.Errorf("server returned HTTP status %s: %s", httpResp.Status, body)
 
-	//nolint:usestdlibvars
 	if httpResp.StatusCode/100 == 5 ||
 		(c.retryOnRateLimit && httpResp.StatusCode == http.StatusTooManyRequests) {
 		return rs, RecoverableError{err, retryAfterDuration(httpResp.Header.Get("Retry-After"))}
@@ -382,7 +380,6 @@ func (c *Client) Read(ctx context.Context, query *prompb.Query) (*prompb.QueryRe
 		return nil, fmt.Errorf("error reading response. HTTP status code: %s: %w", httpResp.Status, err)
 	}
 
-	//nolint:usestdlibvars
 	if httpResp.StatusCode/100 != 2 {
 		return nil, fmt.Errorf("remote server %s returned HTTP status %s: %s", c.urlString, httpResp.Status, strings.TrimSpace(string(compressed)))
 	}

storage/remote/otlptranslator/prometheusremotewrite/histograms.go

@@ -26,6 +26,7 @@ import (
 
 	"github.com/prometheus/prometheus/model/value"
 	"github.com/prometheus/prometheus/prompb"
+	"github.com/prometheus/prometheus/util/annotations"
 )
 
 const defaultZeroThreshold = 1e-128
@@ -33,13 +34,15 @@ const defaultZeroThreshold = 1e-128
 // addExponentialHistogramDataPoints adds OTel exponential histogram data points to the corresponding time series
 // as native histogram samples.
 func (c *PrometheusConverter) addExponentialHistogramDataPoints(dataPoints pmetric.ExponentialHistogramDataPointSlice,
-	resource pcommon.Resource, settings Settings, promName string) error {
+	resource pcommon.Resource, settings Settings, promName string) (annotations.Annotations, error) {
+	var annots annotations.Annotations
 	for x := 0; x < dataPoints.Len(); x++ {
 		pt := dataPoints.At(x)
 
-		histogram, err := exponentialToNativeHistogram(pt)
+		histogram, ws, err := exponentialToNativeHistogram(pt)
+		annots.Merge(ws)
 		if err != nil {
-			return err
+			return annots, err
 		}
 
 		lbls := createAttributes(
@@ -58,15 +61,16 @@ func (c *PrometheusConverter) addExponentialHistogramDataPoints(dataPoints pmetr
 		ts.Exemplars = append(ts.Exemplars, exemplars...)
 	}
 
-	return nil
+	return annots, nil
 }
 
 // exponentialToNativeHistogram translates OTel Exponential Histogram data point
 // to Prometheus Native Histogram.
-func exponentialToNativeHistogram(p pmetric.ExponentialHistogramDataPoint) (prompb.Histogram, error) {
+func exponentialToNativeHistogram(p pmetric.ExponentialHistogramDataPoint) (prompb.Histogram, annotations.Annotations, error) {
+	var annots annotations.Annotations
 	scale := p.Scale()
 	if scale < -4 {
-		return prompb.Histogram{},
+		return prompb.Histogram{}, annots,
 			fmt.Errorf("cannot convert exponential to native histogram."+
 				" Scale must be >= -4, was %d", scale)
 	}
@@ -114,8 +118,11 @@ func exponentialToNativeHistogram(p pmetric.ExponentialHistogramDataPoint) (prom
 			h.Sum = p.Sum()
 		}
 		h.Count = &prompb.Histogram_CountInt{CountInt: p.Count()}
+		if p.Count() == 0 && h.Sum != 0 {
+			annots.Add(fmt.Errorf("exponential histogram data point has zero count, but non-zero sum: %f", h.Sum))
 		}
-	return h, nil
+	}
+	return h, annots, nil
 }
 
 // convertBucketsLayout translates OTel Exponential Histogram dense buckets

storage/remote/otlptranslator/prometheusremotewrite/metrics_to_prw.go

@@ -27,6 +27,7 @@ import (
 
 	"github.com/prometheus/prometheus/prompb"
 	prometheustranslator "github.com/prometheus/prometheus/storage/remote/otlptranslator/prometheus"
+	"github.com/prometheus/prometheus/util/annotations"
 )
 
 type Settings struct {
@@ -53,7 +54,7 @@ func NewPrometheusConverter() *PrometheusConverter {
 }
 
 // FromMetrics converts pmetric.Metrics to Prometheus remote write format.
-func (c *PrometheusConverter) FromMetrics(md pmetric.Metrics, settings Settings) (errs error) {
+func (c *PrometheusConverter) FromMetrics(md pmetric.Metrics, settings Settings) (annots annotations.Annotations, errs error) {
 	resourceMetricsSlice := md.ResourceMetrics()
 	for i := 0; i < resourceMetricsSlice.Len(); i++ {
 		resourceMetrics := resourceMetricsSlice.At(i)
@@ -107,12 +108,14 @@ func (c *PrometheusConverter) FromMetrics(md pmetric.Metrics, settings Settings)
 						errs = multierr.Append(errs, fmt.Errorf("empty data points. %s is dropped", metric.Name()))
 						break
 					}
-					errs = multierr.Append(errs, c.addExponentialHistogramDataPoints(
+					ws, err := c.addExponentialHistogramDataPoints(
 						dataPoints,
 						resource,
 						settings,
 						promName,
-					))
+					)
+					annots.Merge(ws)
+					errs = multierr.Append(errs, err)
 				case pmetric.MetricTypeSummary:
 					dataPoints := metric.Summary().DataPoints()
 					if dataPoints.Len() == 0 {
@@ -128,7 +131,7 @@ func (c *PrometheusConverter) FromMetrics(md pmetric.Metrics, settings Settings)
 		addResourceTargetInfo(resource, settings, mostRecentTimestamp, c)
 	}
 
-	return
+	return annots, errs
 }
 
 func isSameMetric(ts *prompb.TimeSeries, lbls []prompb.Label) bool {

storage/remote/otlptranslator/prometheusremotewrite/metrics_to_prw_test.go

@@ -27,6 +27,41 @@ import (
 	"go.opentelemetry.io/collector/pdata/pmetric/pmetricotlp"
 )
 
+func TestFromMetrics(t *testing.T) {
+	t.Run("exponential histogram warnings for zero count and non-zero sum", func(t *testing.T) {
+		request := pmetricotlp.NewExportRequest()
+		rm := request.Metrics().ResourceMetrics().AppendEmpty()
+		generateAttributes(rm.Resource().Attributes(), "resource", 10)
+
+		metrics := rm.ScopeMetrics().AppendEmpty().Metrics()
+		ts := pcommon.NewTimestampFromTime(time.Now())
+
+		for i := 1; i <= 10; i++ {
+			m := metrics.AppendEmpty()
+			m.SetEmptyExponentialHistogram()
+			m.SetName(fmt.Sprintf("histogram-%d", i))
+			m.ExponentialHistogram().SetAggregationTemporality(pmetric.AggregationTemporalityCumulative)
+			h := m.ExponentialHistogram().DataPoints().AppendEmpty()
+			h.SetTimestamp(ts)
+
+			h.SetCount(0)
+			h.SetSum(155)
+
+			generateAttributes(h.Attributes(), "series", 10)
+		}
+
+		converter := NewPrometheusConverter()
+		annots, err := converter.FromMetrics(request.Metrics(), Settings{})
+		require.NoError(t, err)
+		require.NotEmpty(t, annots)
+		ws, infos := annots.AsStrings("", 0, 0)
+		require.Empty(t, infos)
+		require.Equal(t, []string{
+			"exponential histogram data point has zero count, but non-zero sum: 155.000000",
+		}, ws)
+	})
+}
+
 func BenchmarkPrometheusConverter_FromMetrics(b *testing.B) {
 	for _, resourceAttributeCount := range []int{0, 5, 50} {
 		b.Run(fmt.Sprintf("resource attribute count: %v", resourceAttributeCount), func(b *testing.B) {
@@ -49,7 +84,9 @@ func BenchmarkPrometheusConverter_FromMetrics(b *testing.B) {
 
 											for i := 0; i < b.N; i++ {
 												converter := NewPrometheusConverter()
-												require.NoError(b, converter.FromMetrics(payload.Metrics(), Settings{}))
+												annots, err := converter.FromMetrics(payload.Metrics(), Settings{})
+												require.NoError(b, err)
+												require.Empty(b, annots)
 												require.NotNil(b, converter.TimeSeries())
 											}
 										})

storage/remote/queue_manager.go

@@ -1522,7 +1522,7 @@ func (s *shards) runShard(ctx context.Context, shardID int, queue *queue) {
 	// Send batches of at most MaxSamplesPerSend samples to the remote storage.
 	// If we have fewer samples than that, flush them out after a deadline anyways.
 	var (
-		max = s.qm.cfg.MaxSamplesPerSend
+		maxCount = s.qm.cfg.MaxSamplesPerSend
 
 		pBuf    = proto.NewBuffer(nil)
 		pBufRaw []byte
@@ -1530,19 +1530,19 @@ func (s *shards) runShard(ctx context.Context, shardID int, queue *queue) {
 	)
 	// TODO(@tpaschalis) Should we also raise the max if we have WAL metadata?
 	if s.qm.sendExemplars {
-		max += int(float64(max) * 0.1)
+		maxCount += int(float64(maxCount) * 0.1)
 	}
 
 	// TODO: Dry all of this, we should make an interface/generic for the timeseries type.
 	batchQueue := queue.Chan()
-	pendingData := make([]prompb.TimeSeries, max)
+	pendingData := make([]prompb.TimeSeries, maxCount)
 	for i := range pendingData {
 		pendingData[i].Samples = []prompb.Sample{{}}
 		if s.qm.sendExemplars {
 			pendingData[i].Exemplars = []prompb.Exemplar{{}}
 		}
 	}
-	pendingDataV2 := make([]writev2.TimeSeries, max)
+	pendingDataV2 := make([]writev2.TimeSeries, maxCount)
 	for i := range pendingDataV2 {
 		pendingDataV2[i].Samples = []writev2.Sample{{}}
 	}

storage/remote/queue_manager_test.go

@@ -930,7 +930,7 @@ func createHistograms(numSamples, numSeries int, floatHistogram bool) ([]record.
 }
 
 func createSeriesMetadata(series []record.RefSeries) []record.RefMetadata {
-	metas := make([]record.RefMetadata, len(series))
+	metas := make([]record.RefMetadata, 0, len(series))
 
 	for _, s := range series {
 		metas = append(metas, record.RefMetadata{

storage/remote/write_handler.go

@@ -502,12 +502,17 @@ func (h *otlpWriteHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 	otlpCfg := h.configFunc().OTLPConfig
 
 	converter := otlptranslator.NewPrometheusConverter()
-	if err := converter.FromMetrics(req.Metrics(), otlptranslator.Settings{
+	annots, err := converter.FromMetrics(req.Metrics(), otlptranslator.Settings{
 		AddMetricSuffixes:         true,
 		PromoteResourceAttributes: otlpCfg.PromoteResourceAttributes,
-	}); err != nil {
+	})
+	if err != nil {
 		level.Warn(h.logger).Log("msg", "Error translating OTLP metrics to Prometheus write request", "err", err)
 	}
+	ws, _ := annots.AsStrings("", 0, 0)
+	if len(ws) > 0 {
+		level.Warn(h.logger).Log("msg", "Warnings translating OTLP metrics to Prometheus write request", "warnings", ws)
+	}
 
 	err = h.rwHandler.write(r.Context(), &prompb.WriteRequest{
 		Timeseries: converter.TimeSeries(),

storage/remote/write_handler_test.go

@@ -453,10 +453,10 @@ func TestRemoteWriteHandler_V2Message(t *testing.T) {
 				expectHeaderValue(t, 0, resp.Header.Get(rw20WrittenHistogramsHeader))
 				expectHeaderValue(t, 0, resp.Header.Get(rw20WrittenExemplarsHeader))
 
-				require.Empty(t, len(appendable.samples))
+				require.Empty(t, appendable.samples)
-				require.Empty(t, len(appendable.histograms))
+				require.Empty(t, appendable.histograms)
-				require.Empty(t, len(appendable.exemplars))
+				require.Empty(t, appendable.exemplars)
-				require.Empty(t, len(appendable.metadata))
+				require.Empty(t, appendable.metadata)
 				return
 			}
 

storage/series_test.go

@@ -72,7 +72,7 @@ func TestListSeriesIterator(t *testing.T) {
 	require.Equal(t, chunkenc.ValNone, it.Seek(2))
 }
 
-// TestSeriesSetToChunkSet test the property of SeriesSet that says
+// TestChunkSeriesSetToSeriesSet test the property of SeriesSet that says
 // returned series should be iterable even after Next is called.
 func TestChunkSeriesSetToSeriesSet(t *testing.T) {
 	series := []struct {

template/template.go

@@ -166,7 +166,7 @@ func NewTemplateExpander(
 				return html_template.HTML(text)
 			},
 			"match":     regexp.MatchString,
-			"title":     strings.Title, //nolint:staticcheck
+			"title":     strings.Title, //nolint:staticcheck // TODO(beorn7): Need to come up with a replacement using the cases package.
 			"toUpper":   strings.ToUpper,
 			"toLower":   strings.ToLower,
 			"graphLink": strutil.GraphLinkForExpression,

tracing/testdata/ca.cer

@@ -1,8 +1,66 @@
+Certificate:
+    Data:
+        Version: 3 (0x2)
+        Serial Number:
+            93:6c:9e:29:8d:37:7b:66
+        Signature Algorithm: sha256WithRSAEncryption
+        Issuer: C = XX, L = Default City, O = Default Company Ltd, CN = Prometheus Test CA
+        Validity
+            Not Before: Aug 20 11:51:23 2024 GMT
+            Not After : Dec  5 11:51:23 2044 GMT
+        Subject: C = XX, L = Default City, O = Default Company Ltd, CN = Prometheus Test CA
+        Subject Public Key Info:
+            Public Key Algorithm: rsaEncryption
+                Public-Key: (2048 bit)
+                Modulus:
+                    00:e9:52:05:4d:f2:5a:95:04:2d:b8:73:8b:3c:e7:
+                    47:48:67:00:be:dd:6c:41:f3:7c:4b:44:73:a3:77:
+                    3e:84:af:30:d7:2a:ad:45:6a:b7:89:23:05:15:b3:
+                    aa:46:79:b8:95:64:cc:13:c4:44:a1:01:a0:e2:3d:
+                    a5:67:2b:aa:d3:13:06:43:33:1c:96:36:12:9e:c6:
+                    1d:36:9b:d7:47:bd:28:2d:88:15:04:fa:14:a3:ff:
+                    8c:26:22:c5:a2:15:c7:76:b3:11:f6:a3:44:9a:28:
+                    0f:ca:fb:f4:51:a8:6a:05:94:7c:77:47:c8:21:56:
+                    25:bf:e2:2e:df:33:f3:e4:bd:d6:47:a5:49:13:12:
+                    c8:1f:4c:d7:2a:56:a2:6c:c1:cf:55:05:5d:9a:75:
+                    a2:23:4e:e6:8e:ff:76:05:d6:e0:c8:0b:51:f0:b6:
+                    a1:b2:7d:8f:9c:6a:a5:ce:86:92:15:8c:5b:86:45:
+                    c0:4a:ff:54:b8:ee:cf:11:bd:07:cb:4b:7d:0b:a1:
+                    9d:72:86:9f:55:bc:f9:6c:d9:55:60:96:30:3f:ec:
+                    2d:f6:5f:9a:32:9a:5a:5c:1c:5f:32:f9:d6:0f:04:
+                    f8:81:08:04:9a:95:c3:9d:5a:30:8e:a5:0e:47:2f:
+                    00:ce:e0:2e:ad:5a:b8:b6:4c:55:7c:8a:59:22:b0:
+                    ed:73
+                Exponent: 65537 (0x10001)
+        X509v3 extensions:
+            X509v3 Subject Key Identifier: 
+                CC:F5:05:99:E5:AB:12:69:D8:78:89:4A:31:CA:F0:8B:0B:AD:66:1B
+            X509v3 Authority Key Identifier: 
+                CC:F5:05:99:E5:AB:12:69:D8:78:89:4A:31:CA:F0:8B:0B:AD:66:1B
+            X509v3 Basic Constraints: 
+                CA:TRUE
+    Signature Algorithm: sha256WithRSAEncryption
+    Signature Value:
+        4a:a1:b0:bc:c8:87:4f:7c:96:62:e5:09:29:ae:3a:2e:68:d0:
+        d2:c5:68:ed:ea:83:36:b1:86:f3:b9:e9:19:2b:b6:73:10:6f:
+        df:7f:bb:f1:76:81:03:c1:a1:5a:ee:6c:44:b8:7c:10:d1:5a:
+        d7:c1:92:64:59:35:a6:e0:aa:08:41:37:6e:e7:c8:b6:bd:0c:
+        4b:47:78:ec:c4:b4:15:a3:62:76:4a:39:8e:6e:19:ff:f0:c0:
+        8a:7e:1c:cd:87:e5:00:6c:f1:ce:27:26:ff:b8:e9:eb:f7:2f:
+        bd:c2:4b:9c:d6:57:de:74:74:b3:4f:03:98:9a:b5:08:2d:16:
+        ca:7f:b6:c8:76:62:86:1b:7c:f2:3e:6c:78:cc:2c:95:9a:bb:
+        77:25:e8:80:ff:9b:e8:f8:9a:85:3b:85:b7:17:4e:77:a1:cf:
+        4d:b9:d0:25:e8:5d:8c:e6:7c:f1:d9:52:30:3d:ec:2b:37:91:
+        bc:e2:e8:39:31:6f:3d:e9:98:70:80:7c:41:dd:19:13:05:21:
+        94:7b:16:cf:d8:ee:4e:38:34:5e:6a:ff:cd:85:ac:8f:94:9a:
+        dd:4e:77:05:13:a6:b4:80:52:b2:97:64:76:88:f4:dd:42:0a:
+        50:1c:80:fd:4b:6e:a9:62:10:aa:ef:2e:c1:2f:be:0e:c2:2e:
+        b5:28:5f:83
 -----BEGIN CERTIFICATE-----
 MIIDkTCCAnmgAwIBAgIJAJNsnimNN3tmMA0GCSqGSIb3DQEBCwUAMF8xCzAJBgNV
 BAYTAlhYMRUwEwYDVQQHDAxEZWZhdWx0IENpdHkxHDAaBgNVBAoME0RlZmF1bHQg
-Q29tcGFueSBMdGQxGzAZBgNVBAMMElByb21ldGhldXMgVGVzdCBDQTAeFw0xNTA4
+Q29tcGFueSBMdGQxGzAZBgNVBAMMElByb21ldGhldXMgVGVzdCBDQTAeFw0yNDA4
-MDQxNDA5MjFaFw0yNTA4MDExNDA5MjFaMF8xCzAJBgNVBAYTAlhYMRUwEwYDVQQH
+MjAxMTUxMjNaFw00NDEyMDUxMTUxMjNaMF8xCzAJBgNVBAYTAlhYMRUwEwYDVQQH
 DAxEZWZhdWx0IENpdHkxHDAaBgNVBAoME0RlZmF1bHQgQ29tcGFueSBMdGQxGzAZ
 BgNVBAMMElByb21ldGhldXMgVGVzdCBDQTCCASIwDQYJKoZIhvcNAQEBBQADggEP
 ADCCAQoCggEBAOlSBU3yWpUELbhzizznR0hnAL7dbEHzfEtEc6N3PoSvMNcqrUVq
@@ -12,11 +70,11 @@ yB9M1ypWomzBz1UFXZp1oiNO5o7/dgXW4MgLUfC2obJ9j5xqpc6GkhWMW4ZFwEr/
 VLjuzxG9B8tLfQuhnXKGn1W8+WzZVWCWMD/sLfZfmjKaWlwcXzL51g8E+IEIBJqV
 w51aMI6lDkcvAM7gLq1auLZMVXyKWSKw7XMCAwEAAaNQME4wHQYDVR0OBBYEFMz1
 BZnlqxJp2HiJSjHK8IsLrWYbMB8GA1UdIwQYMBaAFMz1BZnlqxJp2HiJSjHK8IsL
-rWYbMAwGA1UdEwQFMAMBAf8wDQYJKoZIhvcNAQELBQADggEBAI2iA3w3TK5J15Pu
+rWYbMAwGA1UdEwQFMAMBAf8wDQYJKoZIhvcNAQELBQADggEBAEqhsLzIh098lmLl
-e4fPFB4jxQqsbUwuyXbCCv/jKLeFNCD4BjM181WZEYjPMumeTBVzU3aF45LWQIG1
+CSmuOi5o0NLFaO3qgzaxhvO56RkrtnMQb99/u/F2gQPBoVrubES4fBDRWtfBkmRZ
-0DJcrCL4mjMz9qgAoGqA7aDDXiJGbukMgYYsn7vrnVmrZH8T3E8ySlltr7+W578k
+NabgqghBN27nyLa9DEtHeOzEtBWjYnZKOY5uGf/wwIp+HM2H5QBs8c4nJv+46ev3
-pJ5FxnbCroQwn0zLyVB3sFbS8E3vpBr3L8oy8PwPHhIScexcNVc3V6/m4vTZsXTH
+L73CS5zWV950dLNPA5iatQgtFsp/tsh2YoYbfPI+bHjMLJWau3cl6ID/m+j4moU7
-U+vUm1XhDgpDcFMTg2QQiJbfpOYUkwIgnRDAT7t282t2KQWtnlqc3zwPQ1F/6Cpx
+hbcXTnehz0250CXoXYzmfPHZUjA97Cs3kbzi6Dkxbz3pmHCAfEHdGRMFIZR7Fs/Y
-j19JeNsaF1DArkD7YlyKj/GhZLtHwFHG5cxznH0mLDJTW7bQvqqh2iQTeXmBk1lU
+7k44NF5q/82FrI+Umt1OdwUTprSAUrKXZHaI9N1CClAcgP1LbqliEKrvLsEvvg7C
-mM5lH/s=
+LrUoX4M=
 -----END CERTIFICATE-----

tsdb/chunks/chunks.go

@@ -133,6 +133,9 @@ type Meta struct {
 	// Time range the data covers.
 	// When MaxTime == math.MaxInt64 the chunk is still open and being appended to.
 	MinTime, MaxTime int64
+
+	// Flag to indicate that this meta needs merge with OOO data.
+	MergeOOO bool
 }
 
 // ChunkFromSamples requires all samples to have the same type.

tsdb/chunks/head_chunks.go

@@ -191,7 +191,7 @@ func (f *chunkPos) bytesToWriteForChunk(chkLen uint64) uint64 {
 // ChunkDiskMapper is for writing the Head block chunks to disk
 // and access chunks via mmapped files.
 type ChunkDiskMapper struct {
-	/// Writer.
+	// Writer.
 	dir             *os.File
 	writeBufferSize int
 
@@ -210,7 +210,7 @@ type ChunkDiskMapper struct {
 	crc32        hash.Hash
 	writePathMtx sync.Mutex
 
-	/// Reader.
+	// Reader.
 	// The int key in the map is the file number on the disk.
 	mmappedChunkFiles map[int]*mmappedChunkFile // Contains the m-mapped files for each chunk file mapped with its index.
 	closers           map[int]io.Closer         // Closers for resources behind the byte slices.

tsdb/chunks/queue_test.go

@@ -69,16 +69,16 @@ func TestQueuePushPopSingleGoroutine(t *testing.T) {
 	const maxSize = 500
 	const maxIters = 50
 
-	for max := 1; max < maxSize; max++ {
+	for maxCount := 1; maxCount < maxSize; maxCount++ {
-		queue := newWriteJobQueue(max, 1+(r.Int()%max))
+		queue := newWriteJobQueue(maxCount, 1+(r.Int()%maxCount))
 
 		elements := 0 // total elements in the queue
 		lastWriteID := 0
 		lastReadID := 0
 
 		for iter := 0; iter < maxIters; iter++ {
-			if elements < max {
+			if elements < maxCount {
-				toWrite := r.Int() % (max - elements)
+				toWrite := r.Int() % (maxCount - elements)
 				if toWrite == 0 {
 					toWrite = 1
 				}

tsdb/db.go

@@ -49,7 +49,7 @@ import (
 )
 
 const (
-	// Default duration of a block in milliseconds.
+	// DefaultBlockDuration in milliseconds.
 	DefaultBlockDuration = int64(2 * time.Hour / time.Millisecond)
 
 	// Block dir suffixes to make deletion and creation operations atomic.
@@ -699,7 +699,7 @@ func (db *DBReadOnly) LastBlockID() (string, error) {
 		return "", err
 	}
 
-	max := uint64(0)
+	maxT := uint64(0)
 
 	lastBlockID := ""
 
@@ -711,8 +711,8 @@ func (db *DBReadOnly) LastBlockID() (string, error) {
 			continue // Not a block dir.
 		}
 		timestamp := ulidObj.Time()
-		if timestamp > max {
+		if timestamp > maxT {
-			max = timestamp
+			maxT = timestamp
 			lastBlockID = dirName
 		}
 	}
@@ -2046,7 +2046,7 @@ func (db *DB) Querier(mint, maxt int64) (_ storage.Querier, err error) {
 		}
 	}
 
-	blockQueriers := make([]storage.Querier, 0, len(blocks)+2) // +2 to allow for possible in-order and OOO head queriers
+	blockQueriers := make([]storage.Querier, 0, len(blocks)+1) // +1 to allow for possible head querier.
 
 	defer func() {
 		if err != nil {
@@ -2058,10 +2058,12 @@ func (db *DB) Querier(mint, maxt int64) (_ storage.Querier, err error) {
 		}
 	}()
 
-	if maxt >= db.head.MinTime() {
+	overlapsOOO := overlapsClosedInterval(mint, maxt, db.head.MinOOOTime(), db.head.MaxOOOTime())
+	var headQuerier storage.Querier
+	if maxt >= db.head.MinTime() || overlapsOOO {
 		rh := NewRangeHead(db.head, mint, maxt)
 		var err error
-		inOrderHeadQuerier, err := db.blockQuerierFunc(rh, mint, maxt)
+		headQuerier, err = db.blockQuerierFunc(rh, mint, maxt)
 		if err != nil {
 			return nil, fmt.Errorf("open block querier for head %s: %w", rh, err)
 		}
@@ -2071,36 +2073,28 @@ func (db *DB) Querier(mint, maxt int64) (_ storage.Querier, err error) {
 		// won't run into a race later since any truncation that comes after will wait on this querier if it overlaps.
 		shouldClose, getNew, newMint := db.head.IsQuerierCollidingWithTruncation(mint, maxt)
 		if shouldClose {
-			if err := inOrderHeadQuerier.Close(); err != nil {
+			if err := headQuerier.Close(); err != nil {
 				return nil, fmt.Errorf("closing head block querier %s: %w", rh, err)
 			}
-			inOrderHeadQuerier = nil
+			headQuerier = nil
 		}
 		if getNew {
 			rh := NewRangeHead(db.head, newMint, maxt)
-			inOrderHeadQuerier, err = db.blockQuerierFunc(rh, newMint, maxt)
+			headQuerier, err = db.blockQuerierFunc(rh, newMint, maxt)
 			if err != nil {
 				return nil, fmt.Errorf("open block querier for head while getting new querier %s: %w", rh, err)
 			}
 		}
-
-		if inOrderHeadQuerier != nil {
-			blockQueriers = append(blockQueriers, inOrderHeadQuerier)
-		}
 	}
 
-	if overlapsClosedInterval(mint, maxt, db.head.MinOOOTime(), db.head.MaxOOOTime()) {
+	if overlapsOOO {
-		rh := NewOOORangeHead(db.head, mint, maxt, db.lastGarbageCollectedMmapRef)
+		// We need to fetch from in-order and out-of-order chunks: wrap the headQuerier.
-		var err error
+		isoState := db.head.oooIso.TrackReadAfter(db.lastGarbageCollectedMmapRef)
-		outOfOrderHeadQuerier, err := db.blockQuerierFunc(rh, mint, maxt)
+		headQuerier = NewHeadAndOOOQuerier(mint, maxt, db.head, isoState, headQuerier)
-		if err != nil {
-			// If BlockQuerierFunc() failed, make sure to clean up the pending read created by NewOOORangeHead.
-			rh.isoState.Close()
-
-			return nil, fmt.Errorf("open block querier for ooo head %s: %w", rh, err)
 	}
 
-		blockQueriers = append(blockQueriers, outOfOrderHeadQuerier)
+	if headQuerier != nil {
+		blockQueriers = append(blockQueriers, headQuerier)
 	}
 
 	for _, b := range blocks {
@@ -2128,7 +2122,7 @@ func (db *DB) blockChunkQuerierForRange(mint, maxt int64) (_ []storage.ChunkQuer
 		}
 	}
 
-	blockQueriers := make([]storage.ChunkQuerier, 0, len(blocks)+2) // +2 to allow for possible in-order and OOO head queriers
+	blockQueriers := make([]storage.ChunkQuerier, 0, len(blocks)+1) // +1 to allow for possible head querier.
 
 	defer func() {
 		if err != nil {
@@ -2140,9 +2134,11 @@ func (db *DB) blockChunkQuerierForRange(mint, maxt int64) (_ []storage.ChunkQuer
 		}
 	}()
 
-	if maxt >= db.head.MinTime() {
+	overlapsOOO := overlapsClosedInterval(mint, maxt, db.head.MinOOOTime(), db.head.MaxOOOTime())
+	var headQuerier storage.ChunkQuerier
+	if maxt >= db.head.MinTime() || overlapsOOO {
 		rh := NewRangeHead(db.head, mint, maxt)
-		inOrderHeadQuerier, err := db.blockChunkQuerierFunc(rh, mint, maxt)
+		headQuerier, err = db.blockChunkQuerierFunc(rh, mint, maxt)
 		if err != nil {
 			return nil, fmt.Errorf("open querier for head %s: %w", rh, err)
 		}
@@ -2152,35 +2148,28 @@ func (db *DB) blockChunkQuerierForRange(mint, maxt int64) (_ []storage.ChunkQuer
 		// won't run into a race later since any truncation that comes after will wait on this querier if it overlaps.
 		shouldClose, getNew, newMint := db.head.IsQuerierCollidingWithTruncation(mint, maxt)
 		if shouldClose {
-			if err := inOrderHeadQuerier.Close(); err != nil {
+			if err := headQuerier.Close(); err != nil {
 				return nil, fmt.Errorf("closing head querier %s: %w", rh, err)
 			}
-			inOrderHeadQuerier = nil
+			headQuerier = nil
 		}
 		if getNew {
 			rh := NewRangeHead(db.head, newMint, maxt)
-			inOrderHeadQuerier, err = db.blockChunkQuerierFunc(rh, newMint, maxt)
+			headQuerier, err = db.blockChunkQuerierFunc(rh, newMint, maxt)
 			if err != nil {
 				return nil, fmt.Errorf("open querier for head while getting new querier %s: %w", rh, err)
 			}
 		}
-
-		if inOrderHeadQuerier != nil {
-			blockQueriers = append(blockQueriers, inOrderHeadQuerier)
-		}
 	}
 
-	if overlapsClosedInterval(mint, maxt, db.head.MinOOOTime(), db.head.MaxOOOTime()) {
+	if overlapsOOO {
-		rh := NewOOORangeHead(db.head, mint, maxt, db.lastGarbageCollectedMmapRef)
+		// We need to fetch from in-order and out-of-order chunks: wrap the headQuerier.
-		outOfOrderHeadQuerier, err := db.blockChunkQuerierFunc(rh, mint, maxt)
+		isoState := db.head.oooIso.TrackReadAfter(db.lastGarbageCollectedMmapRef)
-		if err != nil {
+		headQuerier = NewHeadAndOOOChunkQuerier(mint, maxt, db.head, isoState, headQuerier)
-			// If NewBlockQuerier() failed, make sure to clean up the pending read created by NewOOORangeHead.
-			rh.isoState.Close()
-
-			return nil, fmt.Errorf("open block chunk querier for ooo head %s: %w", rh, err)
 	}
 
-		blockQueriers = append(blockQueriers, outOfOrderHeadQuerier)
+	if headQuerier != nil {
+		blockQueriers = append(blockQueriers, headQuerier)
 	}
 
 	for _, b := range blocks {
@@ -2344,13 +2333,13 @@ func blockDirs(dir string) ([]string, error) {
 	return dirs, nil
 }
 
-func exponential(d, min, max time.Duration) time.Duration {
+func exponential(d, minD, maxD time.Duration) time.Duration {
 	d *= 2
-	if d < min {
+	if d < minD {
-		d = min
+		d = minD
 	}
-	if d > max {
+	if d > maxD {
-		d = max
+		d = maxD
 	}
 	return d
 }

tsdb/db_test.go

@@ -2676,8 +2676,9 @@ func TestDBReadOnly_Querier_NoAlteration(t *testing.T) {
 		require.NoError(t, db.Close())
 
 		// Simulate a corrupted chunk: without a header.
-		_, err := os.Create(path.Join(mmappedChunksDir(db.dir), "000001"))
+		chunk, err := os.Create(path.Join(mmappedChunksDir(db.dir), "000001"))
 		require.NoError(t, err)
+		require.NoError(t, chunk.Close())
 
 		spinUpQuerierAndCheck(db.dir, t.TempDir(), 1)
 
@@ -4887,8 +4888,8 @@ func testOOOCompaction(t *testing.T, scenario sampleTypeScenario, addExtraSample
 
 	addSample := func(fromMins, toMins int64) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, _, err := scenario.appendFunc(app, series1, ts, ts)
 			require.NoError(t, err)
 			_, _, err = scenario.appendFunc(app, series2, ts, 2*ts)
@@ -4924,8 +4925,8 @@ func testOOOCompaction(t *testing.T, scenario sampleTypeScenario, addExtraSample
 		var series1Samples, series2Samples []chunks.Sample
 		for _, r := range [][2]int64{{90, 119}, {120, 239}, {240, highest}} {
 			fromMins, toMins := r[0], r[1]
-			for min := fromMins; min <= toMins; min++ {
+			for m := fromMins; m <= toMins; m++ {
-				ts := min * time.Minute.Milliseconds()
+				ts := m * time.Minute.Milliseconds()
 				series1Samples = append(series1Samples, scenario.sampleFunc(ts, ts))
 				series2Samples = append(series2Samples, scenario.sampleFunc(ts, 2*ts))
 			}
@@ -5003,8 +5004,8 @@ func testOOOCompaction(t *testing.T, scenario sampleTypeScenario, addExtraSample
 	verifySamples := func(block *Block, fromMins, toMins int64) {
 		series1Samples := make([]chunks.Sample, 0, toMins-fromMins+1)
 		series2Samples := make([]chunks.Sample, 0, toMins-fromMins+1)
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			series1Samples = append(series1Samples, scenario.sampleFunc(ts, ts))
 			series2Samples = append(series2Samples, scenario.sampleFunc(ts, 2*ts))
 		}
@@ -5090,8 +5091,8 @@ func testOOOCompactionWithNormalCompaction(t *testing.T, scenario sampleTypeScen
 
 	addSamples := func(fromMins, toMins int64) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, _, err := scenario.appendFunc(app, series1, ts, ts)
 			require.NoError(t, err)
 			_, _, err = scenario.appendFunc(app, series2, ts, 2*ts)
@@ -5145,8 +5146,8 @@ func testOOOCompactionWithNormalCompaction(t *testing.T, scenario sampleTypeScen
 	verifySamples := func(block *Block, fromMins, toMins int64) {
 		series1Samples := make([]chunks.Sample, 0, toMins-fromMins+1)
 		series2Samples := make([]chunks.Sample, 0, toMins-fromMins+1)
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			series1Samples = append(series1Samples, scenario.sampleFunc(ts, ts))
 			series2Samples = append(series2Samples, scenario.sampleFunc(ts, 2*ts))
 		}
@@ -5203,8 +5204,8 @@ func testOOOCompactionWithDisabledWriteLog(t *testing.T, scenario sampleTypeScen
 
 	addSamples := func(fromMins, toMins int64) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, _, err := scenario.appendFunc(app, series1, ts, ts)
 			require.NoError(t, err)
 			_, _, err = scenario.appendFunc(app, series2, ts, 2*ts)
@@ -5258,8 +5259,8 @@ func testOOOCompactionWithDisabledWriteLog(t *testing.T, scenario sampleTypeScen
 	verifySamples := func(block *Block, fromMins, toMins int64) {
 		series1Samples := make([]chunks.Sample, 0, toMins-fromMins+1)
 		series2Samples := make([]chunks.Sample, 0, toMins-fromMins+1)
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			series1Samples = append(series1Samples, scenario.sampleFunc(ts, ts))
 			series2Samples = append(series2Samples, scenario.sampleFunc(ts, 2*ts))
 		}
@@ -5314,8 +5315,8 @@ func testOOOQueryAfterRestartWithSnapshotAndRemovedWBL(t *testing.T, scenario sa
 
 	addSamples := func(fromMins, toMins int64) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, _, err := scenario.appendFunc(app, series1, ts, ts)
 			require.NoError(t, err)
 			_, _, err = scenario.appendFunc(app, series2, ts, 2*ts)
@@ -5362,8 +5363,8 @@ func testOOOQueryAfterRestartWithSnapshotAndRemovedWBL(t *testing.T, scenario sa
 	verifySamples := func(fromMins, toMins int64) {
 		series1Samples := make([]chunks.Sample, 0, toMins-fromMins+1)
 		series2Samples := make([]chunks.Sample, 0, toMins-fromMins+1)
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			series1Samples = append(series1Samples, scenario.sampleFunc(ts, ts))
 			series2Samples = append(series2Samples, scenario.sampleFunc(ts, ts*2))
 		}
@@ -5462,7 +5463,6 @@ func testQuerierOOOQuery(t *testing.T,
 	sampleFunc func(ts int64) chunks.Sample,
 ) {
 	opts := DefaultOptions()
-	opts.OutOfOrderCapMax = 30
 	opts.OutOfOrderTimeWindow = 24 * time.Hour.Milliseconds()
 
 	series1 := labels.FromStrings("foo", "bar1")
@@ -5474,18 +5474,19 @@ func testQuerierOOOQuery(t *testing.T,
 	addSample := func(db *DB, fromMins, toMins, queryMinT, queryMaxT int64, expSamples []chunks.Sample, filter filterFunc, counterReset bool) ([]chunks.Sample, int) {
 		app := db.Appender(context.Background())
 		totalAppended := 0
-		for min := fromMins; min <= toMins; min += time.Minute.Milliseconds() {
+		for m := fromMins; m <= toMins; m += time.Minute.Milliseconds() {
-			if !filter(min) {
+			if !filter(m / time.Minute.Milliseconds()) {
 				continue
 			}
-			_, err := appendFunc(app, min, counterReset)
+			_, err := appendFunc(app, m, counterReset)
-			if min >= queryMinT && min <= queryMaxT {
+			if m >= queryMinT && m <= queryMaxT {
-				expSamples = append(expSamples, sampleFunc(min))
+				expSamples = append(expSamples, sampleFunc(m))
 			}
 			require.NoError(t, err)
 			totalAppended++
 		}
 		require.NoError(t, app.Commit())
+		require.Positive(t, totalAppended, 0) // Sanity check that filter is not too zealous.
 		return expSamples, totalAppended
 	}
 
@@ -5499,12 +5500,14 @@ func testQuerierOOOQuery(t *testing.T,
 
 	tests := []struct {
 		name      string
+		oooCap    int64
 		queryMinT int64
 		queryMaxT int64
 		batches   []sampleBatch
 	}{
 		{
 			name:      "query interval covering ooomint and inordermaxt returns all ingested samples",
+			oooCap:    30,
 			queryMinT: minutes(0),
 			queryMaxT: minutes(200),
 			batches: []sampleBatch{
@@ -5523,6 +5526,7 @@ func testQuerierOOOQuery(t *testing.T,
 		},
 		{
 			name:      "partial query interval returns only samples within interval",
+			oooCap:    30,
 			queryMinT: minutes(20),
 			queryMaxT: minutes(180),
 			batches: []sampleBatch{
@@ -5564,9 +5568,102 @@ func testQuerierOOOQuery(t *testing.T,
 				},
 			},
 		},
+		{
+			name:      "query overlapping inorder and ooo samples returns all ingested samples at the end of the interval",
+			oooCap:    30,
+			queryMinT: minutes(0),
+			queryMaxT: minutes(200),
+			batches: []sampleBatch{
+				{
+					minT:   minutes(100),
+					maxT:   minutes(200),
+					filter: func(t int64) bool { return t%2 == 0 },
+					isOOO:  false,
+				},
+				{
+					minT:   minutes(170),
+					maxT:   minutes(180),
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+			},
+		},
+		{
+			name:      "query overlapping inorder and ooo in-memory samples returns all ingested samples at the beginning of the interval",
+			oooCap:    30,
+			queryMinT: minutes(0),
+			queryMaxT: minutes(200),
+			batches: []sampleBatch{
+				{
+					minT:   minutes(100),
+					maxT:   minutes(200),
+					filter: func(t int64) bool { return t%2 == 0 },
+					isOOO:  false,
+				},
+				{
+					minT:   minutes(100),
+					maxT:   minutes(110),
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+			},
+		},
+		{
+			name:      "query inorder contain ooo mmaped samples returns all ingested samples at the beginning of the interval",
+			oooCap:    5,
+			queryMinT: minutes(0),
+			queryMaxT: minutes(200),
+			batches: []sampleBatch{
+				{
+					minT:   minutes(100),
+					maxT:   minutes(200),
+					filter: func(t int64) bool { return t%2 == 0 },
+					isOOO:  false,
+				},
+				{
+					minT:   minutes(101),
+					maxT:   minutes(101 + (5-1)*2), // Append samples to fit in a single mmmaped OOO chunk and fit inside the first in-order mmaped chunk.
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+				{
+					minT:   minutes(191),
+					maxT:   minutes(193), // Append some more OOO samples to trigger mapping the OOO chunk, but use time 151 to not overlap with in-order head chunk.
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+			},
+		},
+		{
+			name:      "query overlapping inorder and ooo mmaped samples returns all ingested samples at the beginning of the interval",
+			oooCap:    30,
+			queryMinT: minutes(0),
+			queryMaxT: minutes(200),
+			batches: []sampleBatch{
+				{
+					minT:   minutes(100),
+					maxT:   minutes(200),
+					filter: func(t int64) bool { return t%2 == 0 },
+					isOOO:  false,
+				},
+				{
+					minT:   minutes(101),
+					maxT:   minutes(101 + (30-1)*2), // Append samples to fit in a single mmmaped OOO chunk and overlap the first in-order mmaped chunk.
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+				{
+					minT:   minutes(191),
+					maxT:   minutes(193), // Append some more OOO samples to trigger mapping the OOO chunk, but use time 151 to not overlap with in-order head chunk.
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+			},
+		},
 	}
 	for _, tc := range tests {
 		t.Run(fmt.Sprintf("name=%s", tc.name), func(t *testing.T) {
+			opts.OutOfOrderCapMax = tc.oooCap
 			db := openTestDB(t, opts, nil)
 			db.DisableCompactions()
 			db.EnableNativeHistograms()
@@ -5676,18 +5773,19 @@ func testChunkQuerierOOOQuery(t *testing.T,
 	addSample := func(db *DB, fromMins, toMins, queryMinT, queryMaxT int64, expSamples []chunks.Sample, filter filterFunc, counterReset bool) ([]chunks.Sample, int) {
 		app := db.Appender(context.Background())
 		totalAppended := 0
-		for min := fromMins; min <= toMins; min += time.Minute.Milliseconds() {
+		for m := fromMins; m <= toMins; m += time.Minute.Milliseconds() {
-			if !filter(min) {
+			if !filter(m / time.Minute.Milliseconds()) {
 				continue
 			}
-			_, err := appendFunc(app, min, counterReset)
+			_, err := appendFunc(app, m, counterReset)
-			if min >= queryMinT && min <= queryMaxT {
+			if m >= queryMinT && m <= queryMaxT {
-				expSamples = append(expSamples, sampleFunc(min))
+				expSamples = append(expSamples, sampleFunc(m))
 			}
 			require.NoError(t, err)
 			totalAppended++
 		}
 		require.NoError(t, app.Commit())
+		require.Positive(t, totalAppended) // Sanity check that filter is not too zealous.
 		return expSamples, totalAppended
 	}
 
@@ -5701,12 +5799,14 @@ func testChunkQuerierOOOQuery(t *testing.T,
 
 	tests := []struct {
 		name      string
+		oooCap    int64
 		queryMinT int64
 		queryMaxT int64
 		batches   []sampleBatch
 	}{
 		{
 			name:      "query interval covering ooomint and inordermaxt returns all ingested samples",
+			oooCap:    30,
 			queryMinT: minutes(0),
 			queryMaxT: minutes(200),
 			batches: []sampleBatch{
@@ -5725,6 +5825,7 @@ func testChunkQuerierOOOQuery(t *testing.T,
 		},
 		{
 			name:      "partial query interval returns only samples within interval",
+			oooCap:    30,
 			queryMinT: minutes(20),
 			queryMaxT: minutes(180),
 			batches: []sampleBatch{
@@ -5766,9 +5867,102 @@ func testChunkQuerierOOOQuery(t *testing.T,
 				},
 			},
 		},
+		{
+			name:      "query overlapping inorder and ooo samples returns all ingested samples at the end of the interval",
+			oooCap:    30,
+			queryMinT: minutes(0),
+			queryMaxT: minutes(200),
+			batches: []sampleBatch{
+				{
+					minT:   minutes(100),
+					maxT:   minutes(200),
+					filter: func(t int64) bool { return t%2 == 0 },
+					isOOO:  false,
+				},
+				{
+					minT:   minutes(170),
+					maxT:   minutes(180),
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+			},
+		},
+		{
+			name:      "query overlapping inorder and ooo in-memory samples returns all ingested samples at the beginning of the interval",
+			oooCap:    30,
+			queryMinT: minutes(0),
+			queryMaxT: minutes(200),
+			batches: []sampleBatch{
+				{
+					minT:   minutes(100),
+					maxT:   minutes(200),
+					filter: func(t int64) bool { return t%2 == 0 },
+					isOOO:  false,
+				},
+				{
+					minT:   minutes(100),
+					maxT:   minutes(110),
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+			},
+		},
+		{
+			name:      "query inorder contain ooo mmaped samples returns all ingested samples at the beginning of the interval",
+			oooCap:    5,
+			queryMinT: minutes(0),
+			queryMaxT: minutes(200),
+			batches: []sampleBatch{
+				{
+					minT:   minutes(100),
+					maxT:   minutes(200),
+					filter: func(t int64) bool { return t%2 == 0 },
+					isOOO:  false,
+				},
+				{
+					minT:   minutes(101),
+					maxT:   minutes(101 + (5-1)*2), // Append samples to fit in a single mmmaped OOO chunk and fit inside the first in-order mmaped chunk.
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+				{
+					minT:   minutes(191),
+					maxT:   minutes(193), // Append some more OOO samples to trigger mapping the OOO chunk, but use time 151 to not overlap with in-order head chunk.
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+			},
+		},
+		{
+			name:      "query overlapping inorder and ooo mmaped samples returns all ingested samples at the beginning of the interval",
+			oooCap:    30,
+			queryMinT: minutes(0),
+			queryMaxT: minutes(200),
+			batches: []sampleBatch{
+				{
+					minT:   minutes(100),
+					maxT:   minutes(200),
+					filter: func(t int64) bool { return t%2 == 0 },
+					isOOO:  false,
+				},
+				{
+					minT:   minutes(101),
+					maxT:   minutes(101 + (30-1)*2), // Append samples to fit in a single mmmaped OOO chunk and overlap the first in-order mmaped chunk.
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+				{
+					minT:   minutes(191),
+					maxT:   minutes(193), // Append some more OOO samples to trigger mapping the OOO chunk, but use time 151 to not overlap with in-order head chunk.
+					filter: func(t int64) bool { return t%2 == 1 },
+					isOOO:  true,
+				},
+			},
+		},
 	}
 	for _, tc := range tests {
 		t.Run(fmt.Sprintf("name=%s", tc.name), func(t *testing.T) {
+			opts.OutOfOrderCapMax = tc.oooCap
 			db := openTestDB(t, opts, nil)
 			db.DisableCompactions()
 			db.EnableNativeHistograms()
@@ -5911,7 +6105,7 @@ func testOOONativeHistogramsWithCounterResets(t *testing.T, scenario sampleTypeS
 					shouldReset: func(v int64) bool {
 						return v == 44
 					},
-					expCounterResetHints: []histogram.CounterResetHint{histogram.UnknownCounterReset, histogram.NotCounterReset, histogram.NotCounterReset, histogram.NotCounterReset, histogram.CounterReset},
+					expCounterResetHints: []histogram.CounterResetHint{histogram.UnknownCounterReset, histogram.NotCounterReset, histogram.NotCounterReset, histogram.NotCounterReset, histogram.UnknownCounterReset},
 				},
 				{
 					from:                 50,
@@ -6017,9 +6211,9 @@ func testOOOAppendAndQuery(t *testing.T, scenario sampleTypeScenario) {
 		app := db.Appender(context.Background())
 		key := lbls.String()
 		from, to := minutes(fromMins), minutes(toMins)
-		for min := from; min <= to; min += time.Minute.Milliseconds() {
+		for m := from; m <= to; m += time.Minute.Milliseconds() {
 			val := rand.Intn(1000)
-			_, s, err := scenario.appendFunc(app, lbls, min, int64(val))
+			_, s, err := scenario.appendFunc(app, lbls, m, int64(val))
 			if faceError {
 				require.Error(t, err)
 			} else {
@@ -6150,14 +6344,14 @@ func testOOODisabled(t *testing.T, scenario sampleTypeScenario) {
 		app := db.Appender(context.Background())
 		key := lbls.String()
 		from, to := minutes(fromMins), minutes(toMins)
-		for min := from; min <= to; min += time.Minute.Milliseconds() {
+		for m := from; m <= to; m += time.Minute.Milliseconds() {
-			_, _, err := scenario.appendFunc(app, lbls, min, min)
+			_, _, err := scenario.appendFunc(app, lbls, m, m)
 			if faceError {
 				require.Error(t, err)
 				failedSamples++
 			} else {
 				require.NoError(t, err)
-				expSamples[key] = append(expSamples[key], scenario.sampleFunc(min, min))
+				expSamples[key] = append(expSamples[key], scenario.sampleFunc(m, m))
 				totalSamples++
 			}
 		}
@@ -6226,9 +6420,9 @@ func testWBLAndMmapReplay(t *testing.T, scenario sampleTypeScenario) {
 		app := db.Appender(context.Background())
 		key := lbls.String()
 		from, to := minutes(fromMins), minutes(toMins)
-		for min := from; min <= to; min += time.Minute.Milliseconds() {
+		for m := from; m <= to; m += time.Minute.Milliseconds() {
 			val := rand.Intn(1000)
-			_, s, err := scenario.appendFunc(app, lbls, min, int64(val))
+			_, s, err := scenario.appendFunc(app, lbls, m, int64(val))
 			require.NoError(t, err)
 			expSamples[key] = append(expSamples[key], s)
 			totalSamples++
@@ -6791,8 +6985,8 @@ func testOOOCompactionFailure(t *testing.T, scenario sampleTypeScenario) {
 
 	addSample := func(fromMins, toMins int64) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, _, err := scenario.appendFunc(app, series1, ts, ts)
 			require.NoError(t, err)
 		}
@@ -6879,8 +7073,8 @@ func testOOOCompactionFailure(t *testing.T, scenario sampleTypeScenario) {
 
 	verifySamples := func(block *Block, fromMins, toMins int64) {
 		series1Samples := make([]chunks.Sample, 0, toMins-fromMins+1)
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			series1Samples = append(series1Samples, scenario.sampleFunc(ts, ts))
 		}
 		expRes := map[string][]chunks.Sample{
@@ -6928,8 +7122,8 @@ func TestWBLCorruption(t *testing.T) {
 	var allSamples, expAfterRestart []chunks.Sample
 	addSamples := func(fromMins, toMins int64, afterRestart bool) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, err := app.Append(0, series1, ts, float64(ts))
 			require.NoError(t, err)
 			allSamples = append(allSamples, sample{t: ts, f: float64(ts)})
@@ -7084,8 +7278,8 @@ func testOOOMmapCorruption(t *testing.T, scenario sampleTypeScenario) {
 	var allSamples, expInMmapChunks []chunks.Sample
 	addSamples := func(fromMins, toMins int64, inMmapAfterCorruption bool) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, s, err := scenario.appendFunc(app, series1, ts, ts)
 			require.NoError(t, err)
 			allSamples = append(allSamples, s)
@@ -7231,8 +7425,8 @@ func testOutOfOrderRuntimeConfig(t *testing.T, scenario sampleTypeScenario) {
 	series1 := labels.FromStrings("foo", "bar1")
 	addSamples := func(t *testing.T, db *DB, fromMins, toMins int64, success bool, allSamples []chunks.Sample) []chunks.Sample {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, s, err := scenario.appendFunc(app, series1, ts, ts)
 			if success {
 				require.NoError(t, err)
@@ -7265,7 +7459,7 @@ func testOutOfOrderRuntimeConfig(t *testing.T, scenario sampleTypeScenario) {
 		// WBL is not empty.
 		size, err := db.head.wbl.Size()
 		require.NoError(t, err)
-		require.Greater(t, size, int64(0))
+		require.Positive(t, size)
 
 		require.Empty(t, db.Blocks())
 		require.NoError(t, db.compactOOOHead(ctx))
@@ -7442,8 +7636,8 @@ func testNoGapAfterRestartWithOOO(t *testing.T, scenario sampleTypeScenario) {
 	series1 := labels.FromStrings("foo", "bar1")
 	addSamples := func(t *testing.T, db *DB, fromMins, toMins int64, success bool) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, _, err := scenario.appendFunc(app, series1, ts, ts)
 			if success {
 				require.NoError(t, err)
@@ -7456,8 +7650,8 @@ func testNoGapAfterRestartWithOOO(t *testing.T, scenario sampleTypeScenario) {
 
 	verifySamples := func(t *testing.T, db *DB, fromMins, toMins int64) {
 		var expSamples []chunks.Sample
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			expSamples = append(expSamples, scenario.sampleFunc(ts, ts))
 		}
 
@@ -7574,8 +7768,8 @@ func testWblReplayAfterOOODisableAndRestart(t *testing.T, scenario sampleTypeSce
 	var allSamples []chunks.Sample
 	addSamples := func(fromMins, toMins int64) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, s, err := scenario.appendFunc(app, series1, ts, ts)
 			require.NoError(t, err)
 			allSamples = append(allSamples, s)
@@ -7643,8 +7837,8 @@ func testPanicOnApplyConfig(t *testing.T, scenario sampleTypeScenario) {
 	var allSamples []chunks.Sample
 	addSamples := func(fromMins, toMins int64) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, s, err := scenario.appendFunc(app, series1, ts, ts)
 			require.NoError(t, err)
 			allSamples = append(allSamples, s)
@@ -7702,8 +7896,8 @@ func testDiskFillingUpAfterDisablingOOO(t *testing.T, scenario sampleTypeScenari
 	var allSamples []chunks.Sample
 	addSamples := func(fromMins, toMins int64) {
 		app := db.Appender(context.Background())
-		for min := fromMins; min <= toMins; min++ {
+		for m := fromMins; m <= toMins; m++ {
-			ts := min * time.Minute.Milliseconds()
+			ts := m * time.Minute.Milliseconds()
 			_, s, err := scenario.appendFunc(app, series1, ts, ts)
 			require.NoError(t, err)
 			allSamples = append(allSamples, s)

tsdb/encoding/encoding.go

@@ -201,8 +201,8 @@ func (d *Decbuf) UvarintStr() string {
 	return string(d.UvarintBytes())
 }
 
-// The return value becomes invalid if the byte slice goes away.
+// UvarintBytes returns invalid values if the byte slice goes away.
-// Compared to UvarintStr, this avoid allocations.
+// Compared to UvarintStr, it avoid allocations.
 func (d *Decbuf) UvarintBytes() []byte {
 	l := d.Uvarint64()
 	if d.E != nil {

tsdb/head_append.go

@@ -19,6 +19,7 @@ import (
 	"fmt"
 	"math"
 
+	"github.com/go-kit/log"
 	"github.com/go-kit/log/level"
 
 	"github.com/prometheus/prometheus/model/exemplar"
@@ -1024,7 +1025,7 @@ func (a *headAppender) Commit() (err error) {
 			// Sample is OOO and OOO handling is enabled
 			// and the delta is within the OOO tolerance.
 			var mmapRefs []chunks.ChunkDiskMapperRef
-			ok, chunkCreated, mmapRefs = series.insert(s.T, s.V, nil, nil, a.head.chunkDiskMapper, oooCapMax)
+			ok, chunkCreated, mmapRefs = series.insert(s.T, s.V, nil, nil, a.head.chunkDiskMapper, oooCapMax, a.head.logger)
 			if chunkCreated {
 				r, ok := oooMmapMarkers[series.ref]
 				if !ok || r != nil {
@@ -1120,7 +1121,7 @@ func (a *headAppender) Commit() (err error) {
 			// Sample is OOO and OOO handling is enabled
 			// and the delta is within the OOO tolerance.
 			var mmapRefs []chunks.ChunkDiskMapperRef
-			ok, chunkCreated, mmapRefs = series.insert(s.T, 0, s.H, nil, a.head.chunkDiskMapper, oooCapMax)
+			ok, chunkCreated, mmapRefs = series.insert(s.T, 0, s.H, nil, a.head.chunkDiskMapper, oooCapMax, a.head.logger)
 			if chunkCreated {
 				r, ok := oooMmapMarkers[series.ref]
 				if !ok || r != nil {
@@ -1216,7 +1217,7 @@ func (a *headAppender) Commit() (err error) {
 			// Sample is OOO and OOO handling is enabled
 			// and the delta is within the OOO tolerance.
 			var mmapRefs []chunks.ChunkDiskMapperRef
-			ok, chunkCreated, mmapRefs = series.insert(s.T, 0, nil, s.FH, a.head.chunkDiskMapper, oooCapMax)
+			ok, chunkCreated, mmapRefs = series.insert(s.T, 0, nil, s.FH, a.head.chunkDiskMapper, oooCapMax, a.head.logger)
 			if chunkCreated {
 				r, ok := oooMmapMarkers[series.ref]
 				if !ok || r != nil {
@@ -1314,14 +1315,14 @@ func (a *headAppender) Commit() (err error) {
 }
 
 // insert is like append, except it inserts. Used for OOO samples.
-func (s *memSeries) insert(t int64, v float64, h *histogram.Histogram, fh *histogram.FloatHistogram, chunkDiskMapper *chunks.ChunkDiskMapper, oooCapMax int64) (inserted, chunkCreated bool, mmapRefs []chunks.ChunkDiskMapperRef) {
+func (s *memSeries) insert(t int64, v float64, h *histogram.Histogram, fh *histogram.FloatHistogram, chunkDiskMapper *chunks.ChunkDiskMapper, oooCapMax int64, logger log.Logger) (inserted, chunkCreated bool, mmapRefs []chunks.ChunkDiskMapperRef) {
 	if s.ooo == nil {
 		s.ooo = &memSeriesOOOFields{}
 	}
 	c := s.ooo.oooHeadChunk
 	if c == nil || c.chunk.NumSamples() == int(oooCapMax) {
 		// Note: If no new samples come in then we rely on compaction to clean up stale in-memory OOO chunks.
-		c, mmapRefs = s.cutNewOOOHeadChunk(t, chunkDiskMapper)
+		c, mmapRefs = s.cutNewOOOHeadChunk(t, chunkDiskMapper, logger)
 		chunkCreated = true
 	}
 
@@ -1633,12 +1634,12 @@ func (s *memSeries) histogramsAppendPreprocessor(t int64, e chunkenc.Encoding, o
 // It assumes that the time range is 1/ratioToFull full.
 // Assuming that the samples will keep arriving at the same rate, it will make the
 // remaining n chunks within this chunk range (before max) equally sized.
-func computeChunkEndTime(start, cur, max int64, ratioToFull float64) int64 {
+func computeChunkEndTime(start, cur, maxT int64, ratioToFull float64) int64 {
-	n := float64(max-start) / (float64(cur-start+1) * ratioToFull)
+	n := float64(maxT-start) / (float64(cur-start+1) * ratioToFull)
 	if n <= 1 {
-		return max
+		return maxT
 	}
-	return int64(float64(start) + float64(max-start)/math.Floor(n))
+	return int64(float64(start) + float64(maxT-start)/math.Floor(n))
 }
 
 func (s *memSeries) cutNewHeadChunk(mint int64, e chunkenc.Encoding, chunkRange int64) *memChunk {
@@ -1675,9 +1676,9 @@ func (s *memSeries) cutNewHeadChunk(mint int64, e chunkenc.Encoding, chunkRange
 }
 
 // cutNewOOOHeadChunk cuts a new OOO chunk and m-maps the old chunk.
-// The caller must ensure that s.ooo is not nil.
+// The caller must ensure that s is locked and s.ooo is not nil.
-func (s *memSeries) cutNewOOOHeadChunk(mint int64, chunkDiskMapper *chunks.ChunkDiskMapper) (*oooHeadChunk, []chunks.ChunkDiskMapperRef) {
+func (s *memSeries) cutNewOOOHeadChunk(mint int64, chunkDiskMapper *chunks.ChunkDiskMapper, logger log.Logger) (*oooHeadChunk, []chunks.ChunkDiskMapperRef) {
-	ref := s.mmapCurrentOOOHeadChunk(chunkDiskMapper)
+	ref := s.mmapCurrentOOOHeadChunk(chunkDiskMapper, logger)
 
 	s.ooo.oooHeadChunk = &oooHeadChunk{
 		chunk:   NewOOOChunk(),
@@ -1688,7 +1689,8 @@ func (s *memSeries) cutNewOOOHeadChunk(mint int64, chunkDiskMapper *chunks.Chunk
 	return s.ooo.oooHeadChunk, ref
 }
 
-func (s *memSeries) mmapCurrentOOOHeadChunk(chunkDiskMapper *chunks.ChunkDiskMapper) []chunks.ChunkDiskMapperRef {
+// s must be locked when calling.
+func (s *memSeries) mmapCurrentOOOHeadChunk(chunkDiskMapper *chunks.ChunkDiskMapper, logger log.Logger) []chunks.ChunkDiskMapperRef {
 	if s.ooo == nil || s.ooo.oooHeadChunk == nil {
 		// OOO is not enabled or there is no head chunk, so nothing to m-map here.
 		return nil
@@ -1700,6 +1702,10 @@ func (s *memSeries) mmapCurrentOOOHeadChunk(chunkDiskMapper *chunks.ChunkDiskMap
 	}
 	chunkRefs := make([]chunks.ChunkDiskMapperRef, 0, len(chks))
 	for _, memchunk := range chks {
+		if len(s.ooo.oooMmappedChunks) >= (oooChunkIDMask - 1) {
+			level.Error(logger).Log("msg", "Too many OOO chunks, dropping data", "series", s.lset.String())
+			break
+		}
 		chunkRef := chunkDiskMapper.WriteChunk(s.ref, memchunk.minTime, memchunk.maxTime, memchunk.chunk, true, handleChunkWriteError)
 		chunkRefs = append(chunkRefs, chunkRef)
 		s.ooo.oooMmappedChunks = append(s.ooo.oooMmappedChunks, &mmappedChunk{

tsdb/head_other.go

@@ -26,7 +26,7 @@ func (s *memSeries) labels() labels.Labels {
 	return s.lset
 }
 
-// No-op when not using dedupelabels.
+// RebuildSymbolTable is a no-op when not using dedupelabels.
 func (h *Head) RebuildSymbolTable(logger log.Logger) *labels.SymbolTable {
 	return nil
 }

tsdb/head_read.go

@@ -199,13 +199,18 @@ func (h *headIndexReader) Series(ref storage.SeriesRef, builder *labels.ScratchB
 	defer s.Unlock()
 
 	*chks = (*chks)[:0]
+	*chks = appendSeriesChunks(s, h.mint, h.maxt, *chks)
 
+	return nil
+}
+
+func appendSeriesChunks(s *memSeries, mint, maxt int64, chks []chunks.Meta) []chunks.Meta {
 	for i, c := range s.mmappedChunks {
 		// Do not expose chunks that are outside of the specified range.
-		if !c.OverlapsClosedInterval(h.mint, h.maxt) {
+		if !c.OverlapsClosedInterval(mint, maxt) {
 			continue
 		}
-		*chks = append(*chks, chunks.Meta{
+		chks = append(chks, chunks.Meta{
 			MinTime: c.minTime,
 			MaxTime: c.maxTime,
 			Ref:     chunks.ChunkRef(chunks.NewHeadChunkRef(s.ref, s.headChunkID(i))),
@@ -223,8 +228,8 @@ func (h *headIndexReader) Series(ref storage.SeriesRef, builder *labels.ScratchB
 			} else {
 				maxTime = chk.maxTime
 			}
-			if chk.OverlapsClosedInterval(h.mint, h.maxt) {
+			if chk.OverlapsClosedInterval(mint, maxt) {
-				*chks = append(*chks, chunks.Meta{
+				chks = append(chks, chunks.Meta{
 					MinTime: chk.minTime,
 					MaxTime: maxTime,
 					Ref:     chunks.ChunkRef(chunks.NewHeadChunkRef(s.ref, s.headChunkID(len(s.mmappedChunks)+j))),
@@ -233,8 +238,7 @@ func (h *headIndexReader) Series(ref storage.SeriesRef, builder *labels.ScratchB
 			j++
 		}
 	}
-
+	return chks
-	return nil
 }
 
 // headChunkID returns the HeadChunkID referred to by the given position.
@@ -244,12 +248,20 @@ func (s *memSeries) headChunkID(pos int) chunks.HeadChunkID {
 	return chunks.HeadChunkID(pos) + s.firstChunkID
 }
 
+const oooChunkIDMask = 1 << 23
+
 // oooHeadChunkID returns the HeadChunkID referred to by the given position.
+// Only the bottom 24 bits are used. Bit 23 is always 1 for an OOO chunk; for the rest:
 // * 0 <= pos < len(s.oooMmappedChunks) refer to s.oooMmappedChunks[pos]
 // * pos == len(s.oooMmappedChunks) refers to s.oooHeadChunk
 // The caller must ensure that s.ooo is not nil.
 func (s *memSeries) oooHeadChunkID(pos int) chunks.HeadChunkID {
-	return chunks.HeadChunkID(pos) + s.ooo.firstOOOChunkID
+	return (chunks.HeadChunkID(pos) + s.ooo.firstOOOChunkID) | oooChunkIDMask
+}
+
+func unpackHeadChunkRef(ref chunks.ChunkRef) (seriesID chunks.HeadSeriesRef, chunkID chunks.HeadChunkID, isOOO bool) {
+	sid, cid := chunks.HeadChunkRef(ref).Unpack()
+	return sid, (cid & (oooChunkIDMask - 1)), (cid & oooChunkIDMask) != 0
 }
 
 // LabelValueFor returns label value for the given label name in the series referred to by ID.
@@ -339,10 +351,15 @@ func (h *headChunkReader) ChunkOrIterable(meta chunks.Meta) (chunkenc.Chunk, chu
 	return chk, nil, err
 }
 
-// ChunkWithCopy returns the chunk for the reference number.
+type ChunkReaderWithCopy interface {
-// If the chunk is the in-memory chunk, then it makes a copy and returns the copied chunk.
+	ChunkOrIterableWithCopy(meta chunks.Meta) (chunkenc.Chunk, chunkenc.Iterable, int64, error)
-func (h *headChunkReader) ChunkWithCopy(meta chunks.Meta) (chunkenc.Chunk, int64, error) {
+}
-	return h.chunk(meta, true)
+
+// ChunkOrIterableWithCopy returns the chunk for the reference number.
+// If the chunk is the in-memory chunk, then it makes a copy and returns the copied chunk, plus the max time of the chunk.
+func (h *headChunkReader) ChunkOrIterableWithCopy(meta chunks.Meta) (chunkenc.Chunk, chunkenc.Iterable, int64, error) {
+	chk, maxTime, err := h.chunk(meta, true)
+	return chk, nil, maxTime, err
 }
 
 // chunk returns the chunk for the reference number.
@@ -358,9 +375,14 @@ func (h *headChunkReader) chunk(meta chunks.Meta, copyLastChunk bool) (chunkenc.
 	}
 
 	s.Lock()
+	defer s.Unlock()
+	return h.chunkFromSeries(s, cid, copyLastChunk)
+}
+
+// Call with s locked.
+func (h *headChunkReader) chunkFromSeries(s *memSeries, cid chunks.HeadChunkID, copyLastChunk bool) (chunkenc.Chunk, int64, error) {
 	c, headChunk, isOpen, err := s.chunk(cid, h.head.chunkDiskMapper, &h.head.memChunkPool)
 	if err != nil {
-		s.Unlock()
 		return nil, 0, err
 	}
 	defer func() {
@@ -374,7 +396,6 @@ func (h *headChunkReader) chunk(meta chunks.Meta, copyLastChunk bool) (chunkenc.
 
 	// This means that the chunk is outside the specified range.
 	if !c.OverlapsClosedInterval(h.mint, h.maxt) {
-		s.Unlock()
 		return nil, 0, storage.ErrNotFound
 	}
 
@@ -391,7 +412,6 @@ func (h *headChunkReader) chunk(meta chunks.Meta, copyLastChunk bool) (chunkenc.
 			return nil, 0, err
 		}
 	}
-	s.Unlock()
 
 	return &safeHeadChunk{
 		Chunk:    chk,
@@ -461,30 +481,12 @@ func (s *memSeries) chunk(id chunks.HeadChunkID, chunkDiskMapper *chunks.ChunkDi
 	return elem, true, offset == 0, nil
 }
 
-// oooMergedChunks return an iterable over one or more OOO chunks for the given
+// mergedChunks return an iterable over all chunks that overlap the
-// chunks.Meta reference from memory or by m-mapping it from the disk. The
+// time window [mint,maxt], plus meta.Chunk if populated.
-// returned iterable will be a merge of all the overlapping chunks, if any,
+// If hr is non-nil then in-order chunks are included.
-// amongst all the chunks in the OOOHead.
 // This function is not thread safe unless the caller holds a lock.
 // The caller must ensure that s.ooo is not nil.
-func (s *memSeries) oooMergedChunks(meta chunks.Meta, cdm *chunks.ChunkDiskMapper, mint, maxt int64, maxMmapRef chunks.ChunkDiskMapperRef) (*mergedOOOChunks, error) {
+func (s *memSeries) mergedChunks(meta chunks.Meta, cdm *chunks.ChunkDiskMapper, hr *headChunkReader, mint, maxt int64, maxMmapRef chunks.ChunkDiskMapperRef) (chunkenc.Iterable, error) {
-	_, cid := chunks.HeadChunkRef(meta.Ref).Unpack()
-
-	// ix represents the index of chunk in the s.mmappedChunks slice. The chunk meta's are
-	// incremented by 1 when new chunk is created, hence (meta - firstChunkID) gives the slice index.
-	// The max index for the s.mmappedChunks slice can be len(s.mmappedChunks)-1, hence if the ix
-	// is len(s.mmappedChunks), it represents the next chunk, which is the head chunk.
-	ix := int(cid) - int(s.ooo.firstOOOChunkID)
-	if ix < 0 || ix > len(s.ooo.oooMmappedChunks) {
-		return nil, storage.ErrNotFound
-	}
-
-	if ix == len(s.ooo.oooMmappedChunks) {
-		if s.ooo.oooHeadChunk == nil {
-			return nil, errors.New("invalid ooo head chunk")
-		}
-	}
-
 	// We create a temporary slice of chunk metas to hold the information of all
 	// possible chunks that may overlap with the requested chunk.
 	tmpChks := make([]chunkMetaAndChunkDiskMapperRef, 0, len(s.ooo.oooMmappedChunks)+1)
@@ -509,6 +511,16 @@ func (s *memSeries) oooMergedChunks(meta chunks.Meta, cdm *chunks.ChunkDiskMappe
 		tmpChks = append(tmpChks, chunkMetaAndChunkDiskMapperRef{meta: meta})
 	}
 
+	if hr != nil { // Include in-order chunks.
+		metas := appendSeriesChunks(s, max(meta.MinTime, mint), min(meta.MaxTime, maxt), nil)
+		for _, m := range metas {
+			tmpChks = append(tmpChks, chunkMetaAndChunkDiskMapperRef{
+				meta: m,
+				ref:  0, // This tells the loop below it's an in-order head chunk.
+			})
+		}
+	}
+
 	// Next we want to sort all the collected chunks by min time so we can find
 	// those that overlap and stop when we know the rest don't.
 	slices.SortFunc(tmpChks, refLessByMinTimeAndMinRef)
@@ -520,9 +532,17 @@ func (s *memSeries) oooMergedChunks(meta chunks.Meta, cdm *chunks.ChunkDiskMappe
 			continue
 		}
 		var iterable chunkenc.Iterable
-		if c.meta.Chunk != nil {
+		switch {
+		case c.meta.Chunk != nil:
 			iterable = c.meta.Chunk
-		} else {
+		case c.ref == 0: // This is an in-order head chunk.
+			_, cid := chunks.HeadChunkRef(c.meta.Ref).Unpack()
+			var err error
+			iterable, _, err = hr.chunkFromSeries(s, cid, false)
+			if err != nil {
+				return nil, fmt.Errorf("invalid head chunk: %w", err)
+			}
+		default:
 			chk, err := cdm.Chunk(c.ref)
 			if err != nil {
 				var cerr *chunks.CorruptionErr

tsdb/head_test.go

@@ -2759,7 +2759,7 @@ func testOutOfOrderSamplesMetric(t *testing.T, scenario sampleTypeScenario) {
 
 	require.Equal(t, int64(math.MinInt64), db.head.minValidTime.Load())
 	require.NoError(t, db.Compact(ctx))
-	require.Greater(t, db.head.minValidTime.Load(), int64(0))
+	require.Positive(t, db.head.minValidTime.Load())
 
 	app = db.Appender(ctx)
 	_, err = appendSample(app, db.head.minValidTime.Load()-2)
@@ -3679,7 +3679,7 @@ func TestHistogramInWALAndMmapChunk(t *testing.T) {
 	require.Len(t, ms.mmappedChunks, 25)
 	expMmapChunks := make([]*mmappedChunk, 0, 20)
 	for _, mmap := range ms.mmappedChunks {
-		require.Greater(t, mmap.numSamples, uint16(0))
+		require.Positive(t, mmap.numSamples)
 		cpy := *mmap
 		expMmapChunks = append(expMmapChunks, &cpy)
 	}
@@ -5868,7 +5868,7 @@ func TestCuttingNewHeadChunks(t *testing.T) {
 	}
 }
 
-// TestHeadDetectsDuplcateSampleAtSizeLimit tests a regression where a duplicate sample
+// TestHeadDetectsDuplicateSampleAtSizeLimit tests a regression where a duplicate sample
 // is appended to the head, right when the head chunk is at the size limit.
 // The test adds all samples as duplicate, thus expecting that the result has
 // exactly half of the samples.

tsdb/head_wal.go

@@ -1006,7 +1006,7 @@ func (wp *wblSubsetProcessor) processWBLSamples(h *Head) (unknownRefs, unknownHi
 				unknownRefs++
 				continue
 			}
-			ok, chunkCreated, _ := ms.insert(s.T, s.V, nil, nil, h.chunkDiskMapper, oooCapMax)
+			ok, chunkCreated, _ := ms.insert(s.T, s.V, nil, nil, h.chunkDiskMapper, oooCapMax, h.logger)
 			if chunkCreated {
 				h.metrics.chunksCreated.Inc()
 				h.metrics.chunks.Inc()
@@ -1033,9 +1033,9 @@ func (wp *wblSubsetProcessor) processWBLSamples(h *Head) (unknownRefs, unknownHi
 			var chunkCreated bool
 			var ok bool
 			if s.h != nil {
-				ok, chunkCreated, _ = ms.insert(s.t, 0, s.h, nil, h.chunkDiskMapper, oooCapMax)
+				ok, chunkCreated, _ = ms.insert(s.t, 0, s.h, nil, h.chunkDiskMapper, oooCapMax, h.logger)
 			} else {
-				ok, chunkCreated, _ = ms.insert(s.t, 0, nil, s.fh, h.chunkDiskMapper, oooCapMax)
+				ok, chunkCreated, _ = ms.insert(s.t, 0, nil, s.fh, h.chunkDiskMapper, oooCapMax, h.logger)
 			}
 			if chunkCreated {
 				h.metrics.chunksCreated.Inc()

tsdb/index/index.go

@@ -196,8 +196,9 @@ func NewTOCFromByteSlice(bs ByteSlice) (*TOC, error) {
 	return toc, d.Err()
 }
 
-// NewWriter returns a new Writer to the given filename. It serializes data in format version 2.
+// NewWriterWithEncoder returns a new Writer to the given filename. It
-// It uses the given encoder to encode each postings list.
+// serializes data in format version 2. It uses the given encoder to encode each
+// postings list.
 func NewWriterWithEncoder(ctx context.Context, fn string, encoder PostingsEncoder) (*Writer, error) {
 	dir := filepath.Dir(fn)
 

tsdb/index/postingsstats_test.go

@@ -20,10 +20,10 @@ import (
 
 func TestPostingsStats(t *testing.T) {
 	stats := &maxHeap{}
-	max := 3000000
+	const maxCount = 3000000
-	heapLength := 10
+	const heapLength = 10
 	stats.init(heapLength)
-	for i := 0; i < max; i++ {
+	for i := 0; i < maxCount; i++ {
 		item := Stat{
 			Name:  "Label-da",
 			Count: uint64(i),
@@ -35,13 +35,13 @@ func TestPostingsStats(t *testing.T) {
 	data := stats.get()
 	require.Len(t, data, 10)
 	for i := 0; i < heapLength; i++ {
-		require.Equal(t, uint64(max-i), data[i].Count)
+		require.Equal(t, uint64(maxCount-i), data[i].Count)
 	}
 }
 
 func TestPostingsStats2(t *testing.T) {
 	stats := &maxHeap{}
-	heapLength := 10
+	const heapLength = 10
 
 	stats.init(heapLength)
 	stats.push(Stat{Name: "Stuff", Count: 10})
@@ -57,12 +57,12 @@ func TestPostingsStats2(t *testing.T) {
 
 func BenchmarkPostingStatsMaxHep(b *testing.B) {
 	stats := &maxHeap{}
-	max := 9000000
+	const maxCount = 9000000
-	heapLength := 10
+	const heapLength = 10
 	b.ResetTimer()
 	for n := 0; n < b.N; n++ {
 		stats.init(heapLength)
-		for i := 0; i < max; i++ {
+		for i := 0; i < maxCount; i++ {
 			item := Stat{
 				Name:  "Label-da",
 				Count: uint64(i),

tsdb/ooo_head.go

@@ -14,16 +14,10 @@
 package tsdb
 
 import (
-	"fmt"
 	"sort"
 
 	"github.com/prometheus/prometheus/model/histogram"
 	"github.com/prometheus/prometheus/tsdb/chunkenc"
-
-	"github.com/oklog/ulid"
-
-	"github.com/prometheus/prometheus/tsdb/chunks"
-	"github.com/prometheus/prometheus/tsdb/tombstones"
 )
 
 // OOOChunk maintains samples in time-ascending order.
@@ -171,75 +165,3 @@ func (o *OOOChunk) ToEncodedChunks(mint, maxt int64) (chks []memChunk, err error
 	}
 	return chks, nil
 }
-
-var _ BlockReader = &OOORangeHead{}
-
-// OOORangeHead allows querying Head out of order samples via BlockReader
-// interface implementation.
-type OOORangeHead struct {
-	head *Head
-	// mint and maxt are tracked because when a query is handled we only want
-	// the timerange of the query and having preexisting pointers to the first
-	// and last timestamp help with that.
-	mint, maxt int64
-
-	isoState *oooIsolationState
-}
-
-func NewOOORangeHead(head *Head, mint, maxt int64, minRef chunks.ChunkDiskMapperRef) *OOORangeHead {
-	isoState := head.oooIso.TrackReadAfter(minRef)
-
-	return &OOORangeHead{
-		head:     head,
-		mint:     mint,
-		maxt:     maxt,
-		isoState: isoState,
-	}
-}
-
-func (oh *OOORangeHead) Index() (IndexReader, error) {
-	return NewOOOHeadIndexReader(oh.head, oh.mint, oh.maxt, oh.isoState.minRef), nil
-}
-
-func (oh *OOORangeHead) Chunks() (ChunkReader, error) {
-	return NewOOOHeadChunkReader(oh.head, oh.mint, oh.maxt, oh.isoState, 0), nil
-}
-
-func (oh *OOORangeHead) Tombstones() (tombstones.Reader, error) {
-	// As stated in the design doc https://docs.google.com/document/d/1Kppm7qL9C-BJB1j6yb6-9ObG3AbdZnFUBYPNNWwDBYM/edit?usp=sharing
-	// Tombstones are not supported for out of order metrics.
-	return tombstones.NewMemTombstones(), nil
-}
-
-var oooRangeHeadULID = ulid.MustParse("0000000000XXXX000RANGEHEAD")
-
-func (oh *OOORangeHead) Meta() BlockMeta {
-	return BlockMeta{
-		MinTime: oh.mint,
-		MaxTime: oh.maxt,
-		ULID:    oooRangeHeadULID,
-		Stats: BlockStats{
-			NumSeries: oh.head.NumSeries(),
-		},
-	}
-}
-
-// Size returns the size taken by the Head block.
-func (oh *OOORangeHead) Size() int64 {
-	return oh.head.Size()
-}
-
-// String returns an human readable representation of the out of order range
-// head. It's important to keep this function in order to avoid the struct dump
-// when the head is stringified in errors or logs.
-func (oh *OOORangeHead) String() string {
-	return fmt.Sprintf("ooo range head (mint: %d, maxt: %d)", oh.MinTime(), oh.MaxTime())
-}
-
-func (oh *OOORangeHead) MinTime() int64 {
-	return oh.mint
-}
-
-func (oh *OOORangeHead) MaxTime() int64 {
-	return oh.maxt
-}

tsdb/ooo_head_read.go

@@ -27,15 +27,10 @@ import (
 	"github.com/prometheus/prometheus/tsdb/chunks"
 	"github.com/prometheus/prometheus/tsdb/index"
 	"github.com/prometheus/prometheus/tsdb/tombstones"
+	"github.com/prometheus/prometheus/util/annotations"
 )
 
-// OOOHeadIndexReader implements IndexReader so ooo samples in the head can be
+type HeadAndOOOIndexReader struct {
-// accessed.
-// It also has a reference to headIndexReader so we can leverage on its
-// IndexReader implementation for all the methods that remain the same. We
-// decided to do this to avoid code duplication.
-// The only methods that change are the ones about getting Series and Postings.
-type OOOHeadIndexReader struct {
 	*headIndexReader            // A reference to the headIndexReader so we can reuse as many interface implementation as possible.
 	lastGarbageCollectedMmapRef chunks.ChunkDiskMapperRef
 }
@@ -49,17 +44,13 @@ func (o mergedOOOChunks) Iterator(iterator chunkenc.Iterator) chunkenc.Iterator
 	return storage.ChainSampleIteratorFromIterables(iterator, o.chunkIterables)
 }
 
-func NewOOOHeadIndexReader(head *Head, mint, maxt int64, lastGarbageCollectedMmapRef chunks.ChunkDiskMapperRef) *OOOHeadIndexReader {
+func NewHeadAndOOOIndexReader(head *Head, mint, maxt int64, lastGarbageCollectedMmapRef chunks.ChunkDiskMapperRef) *HeadAndOOOIndexReader {
 	hr := &headIndexReader{
 		head: head,
 		mint: mint,
 		maxt: maxt,
 	}
-	return &OOOHeadIndexReader{hr, lastGarbageCollectedMmapRef}
+	return &HeadAndOOOIndexReader{hr, lastGarbageCollectedMmapRef}
-}
-
-func (oh *OOOHeadIndexReader) Series(ref storage.SeriesRef, builder *labels.ScratchBuilder, chks *[]chunks.Meta) error {
-	return oh.series(ref, builder, chks, oh.lastGarbageCollectedMmapRef, 0)
 }
 
 type MultiChunk struct {
@@ -109,12 +100,7 @@ func (c MultiChunk) Reset([]byte) {
 	// no-op
 }
 
-// lastGarbageCollectedMmapRef gives the last mmap chunk that may be being garbage collected and so
+func (oh *HeadAndOOOIndexReader) Series(ref storage.SeriesRef, builder *labels.ScratchBuilder, chks *[]chunks.Meta) error {
-// any chunk at or before this ref will not be considered. 0 disables this check.
-//
-// maxMmapRef tells upto what max m-map chunk that we can consider. If it is non-0, then
-// the oooHeadChunk will not be considered.
-func (oh *OOOHeadIndexReader) series(ref storage.SeriesRef, builder *labels.ScratchBuilder, chks *[]chunks.Meta, lastGarbageCollectedMmapRef, maxMmapRef chunks.ChunkDiskMapperRef) error {
 	s := oh.head.series.getByID(chunks.HeadSeriesRef(ref))
 
 	if s == nil {
@@ -131,10 +117,19 @@ func (oh *OOOHeadIndexReader) series(ref storage.SeriesRef, builder *labels.Scra
 	defer s.Unlock()
 	*chks = (*chks)[:0]
 
-	if s.ooo == nil {
+	if s.ooo != nil {
-		return nil
+		return getOOOSeriesChunks(s, oh.mint, oh.maxt, oh.lastGarbageCollectedMmapRef, 0, true, chks)
 	}
+	*chks = appendSeriesChunks(s, oh.mint, oh.maxt, *chks)
+	return nil
+}
 
+// lastGarbageCollectedMmapRef gives the last mmap chunk that may be being garbage collected and so
+// any chunk at or before this ref will not be considered. 0 disables this check.
+//
+// maxMmapRef tells upto what max m-map chunk that we can consider. If it is non-0, then
+// the oooHeadChunk will not be considered.
+func getOOOSeriesChunks(s *memSeries, mint, maxt int64, lastGarbageCollectedMmapRef, maxMmapRef chunks.ChunkDiskMapperRef, includeInOrder bool, chks *[]chunks.Meta) error {
 	tmpChks := make([]chunks.Meta, 0, len(s.ooo.oooMmappedChunks))
 
 	addChunk := func(minT, maxT int64, ref chunks.ChunkRef, chunk chunkenc.Chunk) {
@@ -143,13 +138,14 @@ func (oh *OOOHeadIndexReader) series(ref storage.SeriesRef, builder *labels.Scra
 			MaxTime:  maxT,
 			Ref:      ref,
 			Chunk:    chunk,
+			MergeOOO: true,
 		})
 	}
 
 	// Collect all chunks that overlap the query range.
 	if s.ooo.oooHeadChunk != nil {
 		c := s.ooo.oooHeadChunk
-		if c.OverlapsClosedInterval(oh.mint, oh.maxt) && maxMmapRef == 0 {
+		if c.OverlapsClosedInterval(mint, maxt) && maxMmapRef == 0 {
 			ref := chunks.ChunkRef(chunks.NewHeadChunkRef(s.ref, s.oooHeadChunkID(len(s.ooo.oooMmappedChunks))))
 			if len(c.chunk.samples) > 0 { // Empty samples happens in tests, at least.
 				headChunks := MultiChunk{}
@@ -170,12 +166,16 @@ func (oh *OOOHeadIndexReader) series(ref storage.SeriesRef, builder *labels.Scra
 	}
 	for i := len(s.ooo.oooMmappedChunks) - 1; i >= 0; i-- {
 		c := s.ooo.oooMmappedChunks[i]
-		if c.OverlapsClosedInterval(oh.mint, oh.maxt) && (maxMmapRef == 0 || maxMmapRef.GreaterThanOrEqualTo(c.ref)) && (lastGarbageCollectedMmapRef == 0 || c.ref.GreaterThan(lastGarbageCollectedMmapRef)) {
+		if c.OverlapsClosedInterval(mint, maxt) && (maxMmapRef == 0 || maxMmapRef.GreaterThanOrEqualTo(c.ref)) && (lastGarbageCollectedMmapRef == 0 || c.ref.GreaterThan(lastGarbageCollectedMmapRef)) {
 			ref := chunks.ChunkRef(chunks.NewHeadChunkRef(s.ref, s.oooHeadChunkID(i)))
 			addChunk(c.minTime, c.maxTime, ref, nil)
 		}
 	}
 
+	if includeInOrder {
+		tmpChks = appendSeriesChunks(s, mint, maxt, tmpChks)
+	}
+
 	// There is nothing to do if we did not collect any chunk.
 	if len(tmpChks) == 0 {
 		return nil
@@ -206,17 +206,17 @@ func (oh *OOOHeadIndexReader) series(ref storage.SeriesRef, builder *labels.Scra
 			if c.Chunk != nil {
 				(*chks)[len(*chks)-1].Chunk = c.Chunk
 			}
+			(*chks)[len(*chks)-1].MergeOOO = (*chks)[len(*chks)-1].MergeOOO || c.MergeOOO
 		}
 	}
 
 	return nil
 }
 
-// LabelValues needs to be overridden from the headIndexReader implementation due
+// LabelValues needs to be overridden from the headIndexReader implementation
-// to the check that happens at the beginning where we make sure that the query
+// so we can return labels within either in-order range or ooo range.
-// interval overlaps with the head minooot and maxooot.
+func (oh *HeadAndOOOIndexReader) LabelValues(ctx context.Context, name string, matchers ...*labels.Matcher) ([]string, error) {
-func (oh *OOOHeadIndexReader) LabelValues(ctx context.Context, name string, matchers ...*labels.Matcher) ([]string, error) {
+	if oh.maxt < oh.head.MinTime() && oh.maxt < oh.head.MinOOOTime() || oh.mint > oh.head.MaxTime() && oh.mint > oh.head.MaxOOOTime() {
-	if oh.maxt < oh.head.MinOOOTime() || oh.mint > oh.head.MaxOOOTime() {
 		return []string{}, nil
 	}
 
@@ -268,41 +268,30 @@ func lessByMinTimeAndMinRef(a, b chunks.Meta) int {
 	}
 }
 
-func (oh *OOOHeadIndexReader) Postings(ctx context.Context, name string, values ...string) (index.Postings, error) {
+type HeadAndOOOChunkReader struct {
-	switch len(values) {
-	case 0:
-		return index.EmptyPostings(), nil
-	case 1:
-		return oh.head.postings.Get(name, values[0]), nil // TODO(ganesh) Also call GetOOOPostings
-	default:
-		// TODO(ganesh) We want to only return postings for out of order series.
-		res := make([]index.Postings, 0, len(values))
-		for _, value := range values {
-			res = append(res, oh.head.postings.Get(name, value)) // TODO(ganesh) Also call GetOOOPostings
-		}
-		return index.Merge(ctx, res...), nil
-	}
-}
-
-type OOOHeadChunkReader struct {
 	head        *Head
 	mint, maxt  int64
-	isoState   *oooIsolationState
+	cr          *headChunkReader // If nil, only read OOO chunks.
 	maxMmapRef  chunks.ChunkDiskMapperRef
+	oooIsoState *oooIsolationState
 }
 
-func NewOOOHeadChunkReader(head *Head, mint, maxt int64, isoState *oooIsolationState, maxMmapRef chunks.ChunkDiskMapperRef) *OOOHeadChunkReader {
+func NewHeadAndOOOChunkReader(head *Head, mint, maxt int64, cr *headChunkReader, oooIsoState *oooIsolationState, maxMmapRef chunks.ChunkDiskMapperRef) *HeadAndOOOChunkReader {
-	return &OOOHeadChunkReader{
+	return &HeadAndOOOChunkReader{
 		head:        head,
 		mint:        mint,
 		maxt:        maxt,
-		isoState:   isoState,
+		cr:          cr,
 		maxMmapRef:  maxMmapRef,
+		oooIsoState: oooIsoState,
 	}
 }
 
-func (cr OOOHeadChunkReader) ChunkOrIterable(meta chunks.Meta) (chunkenc.Chunk, chunkenc.Iterable, error) {
+func (cr *HeadAndOOOChunkReader) ChunkOrIterable(meta chunks.Meta) (chunkenc.Chunk, chunkenc.Iterable, error) {
-	sid, _ := chunks.HeadChunkRef(meta.Ref).Unpack()
+	sid, _, _ := unpackHeadChunkRef(meta.Ref)
+	if !meta.MergeOOO {
+		return cr.cr.ChunkOrIterable(meta)
+	}
 
 	s := cr.head.series.getByID(sid)
 	// This means that the series has been garbage collected.
@@ -311,34 +300,38 @@ func (cr OOOHeadChunkReader) ChunkOrIterable(meta chunks.Meta) (chunkenc.Chunk,
 	}
 
 	s.Lock()
-	if s.ooo == nil {
+	if s.ooo == nil { // Must have s.ooo non-nil to call mergedChunks().
-		// There is no OOO data for this series.
 		s.Unlock()
-		return nil, nil, storage.ErrNotFound
+		return cr.cr.ChunkOrIterable(meta)
 	}
-	mc, err := s.oooMergedChunks(meta, cr.head.chunkDiskMapper, cr.mint, cr.maxt, cr.maxMmapRef)
+	mc, err := s.mergedChunks(meta, cr.head.chunkDiskMapper, cr.cr, cr.mint, cr.maxt, cr.maxMmapRef)
 	s.Unlock()
-	if err != nil {
-		return nil, nil, err
-	}
 
-	// This means that the query range did not overlap with the requested chunk.
+	return nil, mc, err
-	if len(mc.chunkIterables) == 0 {
-		return nil, nil, storage.ErrNotFound
-	}
-
-	return nil, mc, nil
 }
 
-func (cr OOOHeadChunkReader) Close() error {
+// ChunkOrIterableWithCopy implements ChunkReaderWithCopy. The special Copy
-	if cr.isoState != nil {
+// behaviour is only implemented for the in-order head chunk.
-		cr.isoState.Close()
+func (cr *HeadAndOOOChunkReader) ChunkOrIterableWithCopy(meta chunks.Meta) (chunkenc.Chunk, chunkenc.Iterable, int64, error) {
+	if !meta.MergeOOO {
+		return cr.cr.ChunkOrIterableWithCopy(meta)
+	}
+	chk, iter, err := cr.ChunkOrIterable(meta)
+	return chk, iter, 0, err
+}
+
+func (cr *HeadAndOOOChunkReader) Close() error {
+	if cr.cr != nil && cr.cr.isoState != nil {
+		cr.cr.isoState.Close()
+	}
+	if cr.oooIsoState != nil {
+		cr.oooIsoState.Close()
 	}
 	return nil
 }
 
 type OOOCompactionHead struct {
-	oooIR       *OOOHeadIndexReader
+	head        *Head
 	lastMmapRef chunks.ChunkDiskMapperRef
 	lastWBLFile int
 	postings    []storage.SeriesRef
@@ -355,6 +348,7 @@ type OOOCompactionHead struct {
 // on the sample append latency. So call NewOOOCompactionHead only right before compaction.
 func NewOOOCompactionHead(ctx context.Context, head *Head) (*OOOCompactionHead, error) {
 	ch := &OOOCompactionHead{
+		head:       head,
 		chunkRange: head.chunkRange.Load(),
 		mint:       math.MaxInt64,
 		maxt:       math.MinInt64,
@@ -368,15 +362,14 @@ func NewOOOCompactionHead(ctx context.Context, head *Head) (*OOOCompactionHead,
 		ch.lastWBLFile = lastWBLFile
 	}
 
-	ch.oooIR = NewOOOHeadIndexReader(head, math.MinInt64, math.MaxInt64, 0)
+	hr := headIndexReader{head: head, mint: ch.mint, maxt: ch.maxt}
 	n, v := index.AllPostingsKey()
-
+	// TODO: filter to series with OOO samples, before sorting.
-	// TODO: verify this gets only ooo samples.
+	p, err := hr.Postings(ctx, n, v)
-	p, err := ch.oooIR.Postings(ctx, n, v)
 	if err != nil {
 		return nil, err
 	}
-	p = ch.oooIR.SortedPostings(p)
+	p = hr.SortedPostings(p)
 
 	var lastSeq, lastOff int
 	for p.Next() {
@@ -397,7 +390,7 @@ func NewOOOCompactionHead(ctx context.Context, head *Head) (*OOOCompactionHead,
 		}
 
 		var lastMmapRef chunks.ChunkDiskMapperRef
-		mmapRefs := ms.mmapCurrentOOOHeadChunk(head.chunkDiskMapper)
+		mmapRefs := ms.mmapCurrentOOOHeadChunk(head.chunkDiskMapper, head.logger)
 		if len(mmapRefs) == 0 && len(ms.ooo.oooMmappedChunks) > 0 {
 			// Nothing was m-mapped. So take the mmapRef from the existing slice if it exists.
 			mmapRefs = []chunks.ChunkDiskMapperRef{ms.ooo.oooMmappedChunks[len(ms.ooo.oooMmappedChunks)-1].ref}
@@ -433,7 +426,7 @@ func (ch *OOOCompactionHead) Index() (IndexReader, error) {
 }
 
 func (ch *OOOCompactionHead) Chunks() (ChunkReader, error) {
-	return NewOOOHeadChunkReader(ch.oooIR.head, ch.oooIR.mint, ch.oooIR.maxt, nil, ch.lastMmapRef), nil
+	return NewHeadAndOOOChunkReader(ch.head, ch.mint, ch.maxt, nil, nil, ch.lastMmapRef), nil
 }
 
 func (ch *OOOCompactionHead) Tombstones() (tombstones.Reader, error) {
@@ -459,12 +452,12 @@ func (ch *OOOCompactionHead) Meta() BlockMeta {
 // Only the method of BlockReader interface are valid for the cloned OOOCompactionHead.
 func (ch *OOOCompactionHead) CloneForTimeRange(mint, maxt int64) *OOOCompactionHead {
 	return &OOOCompactionHead{
-		oooIR:       NewOOOHeadIndexReader(ch.oooIR.head, mint, maxt, 0),
+		head:        ch.head,
 		lastMmapRef: ch.lastMmapRef,
 		postings:    ch.postings,
 		chunkRange:  ch.chunkRange,
-		mint:        ch.mint,
+		mint:        mint,
-		maxt:        ch.maxt,
+		maxt:        maxt,
 	}
 }
 
@@ -484,7 +477,8 @@ func NewOOOCompactionHeadIndexReader(ch *OOOCompactionHead) IndexReader {
 }
 
 func (ir *OOOCompactionHeadIndexReader) Symbols() index.StringIter {
-	return ir.ch.oooIR.Symbols()
+	hr := headIndexReader{head: ir.ch.head, mint: ir.ch.mint, maxt: ir.ch.maxt}
+	return hr.Symbols()
 }
 
 func (ir *OOOCompactionHeadIndexReader) Postings(_ context.Context, name string, values ...string) (index.Postings, error) {
@@ -505,11 +499,28 @@ func (ir *OOOCompactionHeadIndexReader) SortedPostings(p index.Postings) index.P
 }
 
 func (ir *OOOCompactionHeadIndexReader) ShardedPostings(p index.Postings, shardIndex, shardCount uint64) index.Postings {
-	return ir.ch.oooIR.ShardedPostings(p, shardIndex, shardCount)
+	hr := headIndexReader{head: ir.ch.head, mint: ir.ch.mint, maxt: ir.ch.maxt}
+	return hr.ShardedPostings(p, shardIndex, shardCount)
 }
 
 func (ir *OOOCompactionHeadIndexReader) Series(ref storage.SeriesRef, builder *labels.ScratchBuilder, chks *[]chunks.Meta) error {
-	return ir.ch.oooIR.series(ref, builder, chks, 0, ir.ch.lastMmapRef)
+	s := ir.ch.head.series.getByID(chunks.HeadSeriesRef(ref))
+
+	if s == nil {
+		ir.ch.head.metrics.seriesNotFound.Inc()
+		return storage.ErrNotFound
+	}
+	builder.Assign(s.labels())
+
+	s.Lock()
+	defer s.Unlock()
+	*chks = (*chks)[:0]
+
+	if s.ooo == nil {
+		return nil
+	}
+
+	return getOOOSeriesChunks(s, ir.ch.mint, ir.ch.maxt, 0, ir.ch.lastMmapRef, false, chks)
 }
 
 func (ir *OOOCompactionHeadIndexReader) SortedLabelValues(_ context.Context, name string, matchers ...*labels.Matcher) ([]string, error) {
@@ -537,5 +548,91 @@ func (ir *OOOCompactionHeadIndexReader) LabelNamesFor(ctx context.Context, posti
 }
 
 func (ir *OOOCompactionHeadIndexReader) Close() error {
-	return ir.ch.oooIR.Close()
+	return nil
+}
+
+// HeadAndOOOQuerier queries both the head and the out-of-order head.
+type HeadAndOOOQuerier struct {
+	mint, maxt int64
+	head       *Head
+	index      IndexReader
+	chunkr     ChunkReader
+	querier    storage.Querier
+}
+
+func NewHeadAndOOOQuerier(mint, maxt int64, head *Head, oooIsoState *oooIsolationState, querier storage.Querier) storage.Querier {
+	cr := &headChunkReader{
+		head:     head,
+		mint:     mint,
+		maxt:     maxt,
+		isoState: head.iso.State(mint, maxt),
+	}
+	return &HeadAndOOOQuerier{
+		mint:    mint,
+		maxt:    maxt,
+		head:    head,
+		index:   NewHeadAndOOOIndexReader(head, mint, maxt, oooIsoState.minRef),
+		chunkr:  NewHeadAndOOOChunkReader(head, mint, maxt, cr, oooIsoState, 0),
+		querier: querier,
+	}
+}
+
+func (q *HeadAndOOOQuerier) LabelValues(ctx context.Context, name string, hints *storage.LabelHints, matchers ...*labels.Matcher) ([]string, annotations.Annotations, error) {
+	return q.querier.LabelValues(ctx, name, hints, matchers...)
+}
+
+func (q *HeadAndOOOQuerier) LabelNames(ctx context.Context, hints *storage.LabelHints, matchers ...*labels.Matcher) ([]string, annotations.Annotations, error) {
+	return q.querier.LabelNames(ctx, hints, matchers...)
+}
+
+func (q *HeadAndOOOQuerier) Close() error {
+	q.chunkr.Close()
+	return q.querier.Close()
+}
+
+func (q *HeadAndOOOQuerier) Select(ctx context.Context, sortSeries bool, hints *storage.SelectHints, matchers ...*labels.Matcher) storage.SeriesSet {
+	return selectSeriesSet(ctx, sortSeries, hints, matchers, q.index, q.chunkr, q.head.tombstones, q.mint, q.maxt)
+}
+
+// HeadAndOOOChunkQuerier queries both the head and the out-of-order head.
+type HeadAndOOOChunkQuerier struct {
+	mint, maxt int64
+	head       *Head
+	index      IndexReader
+	chunkr     ChunkReader
+	querier    storage.ChunkQuerier
+}
+
+func NewHeadAndOOOChunkQuerier(mint, maxt int64, head *Head, oooIsoState *oooIsolationState, querier storage.ChunkQuerier) storage.ChunkQuerier {
+	cr := &headChunkReader{
+		head:     head,
+		mint:     mint,
+		maxt:     maxt,
+		isoState: head.iso.State(mint, maxt),
+	}
+	return &HeadAndOOOChunkQuerier{
+		mint:    mint,
+		maxt:    maxt,
+		head:    head,
+		index:   NewHeadAndOOOIndexReader(head, mint, maxt, oooIsoState.minRef),
+		chunkr:  NewHeadAndOOOChunkReader(head, mint, maxt, cr, oooIsoState, 0),
+		querier: querier,
+	}
+}
+
+func (q *HeadAndOOOChunkQuerier) LabelValues(ctx context.Context, name string, hints *storage.LabelHints, matchers ...*labels.Matcher) ([]string, annotations.Annotations, error) {
+	return q.querier.LabelValues(ctx, name, hints, matchers...)
+}
+
+func (q *HeadAndOOOChunkQuerier) LabelNames(ctx context.Context, hints *storage.LabelHints, matchers ...*labels.Matcher) ([]string, annotations.Annotations, error) {
+	return q.querier.LabelNames(ctx, hints, matchers...)
+}
+
+func (q *HeadAndOOOChunkQuerier) Close() error {
+	q.chunkr.Close()
+	return q.querier.Close()
+}
+
+func (q *HeadAndOOOChunkQuerier) Select(ctx context.Context, sortSeries bool, hints *storage.SelectHints, matchers ...*labels.Matcher) storage.ChunkSeriesSet {
+	return selectChunkSeriesSet(ctx, sortSeries, hints, matchers, rangeHeadULID, q.index, q.chunkr, q.head.tombstones, q.mint, q.maxt)
 }

tsdb/ooo_head_read_test.go

@@ -35,7 +35,7 @@ import (
 var (
 	_ chunkenc.Chunk    = &MultiChunk{}
 	_ chunkenc.Iterable = &mergedOOOChunks{}
-	_ IndexReader       = &OOOHeadIndexReader{}
+	_ IndexReader       = &HeadAndOOOIndexReader{}
 )
 
 type chunkInterval struct {
@@ -318,12 +318,13 @@ func TestOOOHeadIndexReader_Series(t *testing.T) {
 							Chunk:    chunkenc.Chunk(nil),
 							MinTime:  e.mint,
 							MaxTime:  e.maxt,
+							MergeOOO: true, // Only OOO chunks are tested here, so we always request merge from OOO head.
 						}
 
 						// Ref to whatever Ref the chunk has, that we refer to by ID
 						for ref, c := range intervals {
 							if c.ID == e.ID {
-								meta.Ref = chunks.ChunkRef(chunks.NewHeadChunkRef(chunks.HeadSeriesRef(s1ID), chunks.HeadChunkID(ref)))
+								meta.Ref = chunks.ChunkRef(chunks.NewHeadChunkRef(chunks.HeadSeriesRef(s1ID), s1.oooHeadChunkID(ref)))
 								break
 							}
 						}
@@ -348,7 +349,7 @@ func TestOOOHeadIndexReader_Series(t *testing.T) {
 						})
 					}
 
-					ir := NewOOOHeadIndexReader(h, tc.queryMinT, tc.queryMaxT, 0)
+					ir := NewHeadAndOOOIndexReader(h, tc.queryMinT, tc.queryMaxT, 0)
 
 					var chks []chunks.Meta
 					var b labels.ScratchBuilder
@@ -429,17 +430,17 @@ func testOOOHeadChunkReader_LabelValues(t *testing.T, scenario sampleTypeScenari
 			name:       "LabelValues calls with ooo head query range not overlapping out-of-order data",
 			queryMinT:  100,
 			queryMaxT:  100,
-			expValues1: []string{},
+			expValues1: []string{"bar1"},
-			expValues2: []string{},
+			expValues2: nil,
-			expValues3: []string{},
+			expValues3: []string{"bar1", "bar2"},
-			expValues4: []string{},
+			expValues4: []string{"bar1", "bar2"},
 		},
 	}
 
 	for _, tc := range cases {
 		t.Run(tc.name, func(t *testing.T) {
 			// We first want to test using a head index reader that covers the biggest query interval
-			oh := NewOOOHeadIndexReader(head, tc.queryMinT, tc.queryMaxT, 0)
+			oh := NewHeadAndOOOIndexReader(head, tc.queryMinT, tc.queryMaxT, 0)
 			matchers := []*labels.Matcher{labels.MustNewMatcher(labels.MatchEqual, "foo", "bar1")}
 			values, err := oh.LabelValues(ctx, "foo", matchers...)
 			sort.Strings(values)
@@ -491,10 +492,10 @@ func testOOOHeadChunkReader_Chunk(t *testing.T, scenario sampleTypeScenario) {
 	t.Run("Getting a non existing chunk fails with not found error", func(t *testing.T) {
 		db := newTestDBWithOpts(t, opts)
 
-		cr := NewOOOHeadChunkReader(db.head, 0, 1000, nil, 0)
+		cr := NewHeadAndOOOChunkReader(db.head, 0, 1000, nil, nil, 0)
 		defer cr.Close()
 		c, iterable, err := cr.ChunkOrIterable(chunks.Meta{
-			Ref: 0x1000000, Chunk: chunkenc.Chunk(nil), MinTime: 100, MaxTime: 300,
+			Ref: 0x1800000, Chunk: chunkenc.Chunk(nil), MinTime: 100, MaxTime: 300, MergeOOO: true,
 		})
 		require.Nil(t, iterable)
 		require.Equal(t, err, fmt.Errorf("not found"))
@@ -842,14 +843,14 @@ func testOOOHeadChunkReader_Chunk(t *testing.T, scenario sampleTypeScenario) {
 
 			// The Series method populates the chunk metas, taking a copy of the
 			// head OOO chunk if necessary. These are then used by the ChunkReader.
-			ir := NewOOOHeadIndexReader(db.head, tc.queryMinT, tc.queryMaxT, 0)
+			ir := NewHeadAndOOOIndexReader(db.head, tc.queryMinT, tc.queryMaxT, 0)
 			var chks []chunks.Meta
 			var b labels.ScratchBuilder
 			err = ir.Series(s1Ref, &b, &chks)
 			require.NoError(t, err)
 			require.Equal(t, len(tc.expChunksSamples), len(chks))
 
-			cr := NewOOOHeadChunkReader(db.head, tc.queryMinT, tc.queryMaxT, nil, 0)
+			cr := NewHeadAndOOOChunkReader(db.head, tc.queryMinT, tc.queryMaxT, nil, nil, 0)
 			defer cr.Close()
 			for i := 0; i < len(chks); i++ {
 				c, iterable, err := cr.ChunkOrIterable(chks[i])
@@ -1009,7 +1010,7 @@ func testOOOHeadChunkReader_Chunk_ConsistentQueryResponseDespiteOfHeadExpanding(
 
 			// The Series method populates the chunk metas, taking a copy of the
 			// head OOO chunk if necessary. These are then used by the ChunkReader.
-			ir := NewOOOHeadIndexReader(db.head, tc.queryMinT, tc.queryMaxT, 0)
+			ir := NewHeadAndOOOIndexReader(db.head, tc.queryMinT, tc.queryMaxT, 0)
 			var chks []chunks.Meta
 			var b labels.ScratchBuilder
 			err = ir.Series(s1Ref, &b, &chks)
@@ -1025,7 +1026,7 @@ func testOOOHeadChunkReader_Chunk_ConsistentQueryResponseDespiteOfHeadExpanding(
 			}
 			require.NoError(t, app.Commit())
 
-			cr := NewOOOHeadChunkReader(db.head, tc.queryMinT, tc.queryMaxT, nil, 0)
+			cr := NewHeadAndOOOChunkReader(db.head, tc.queryMinT, tc.queryMaxT, nil, nil, 0)
 			defer cr.Close()
 			for i := 0; i < len(chks); i++ {
 				c, iterable, err := cr.ChunkOrIterable(chks[i])

tsdb/querier.go

@@ -115,20 +115,24 @@ func NewBlockQuerier(b BlockReader, mint, maxt int64) (storage.Querier, error) {
 }
 
 func (q *blockQuerier) Select(ctx context.Context, sortSeries bool, hints *storage.SelectHints, ms ...*labels.Matcher) storage.SeriesSet {
-	mint := q.mint
+	return selectSeriesSet(ctx, sortSeries, hints, ms, q.index, q.chunks, q.tombstones, q.mint, q.maxt)
-	maxt := q.maxt
+}
+
+func selectSeriesSet(ctx context.Context, sortSeries bool, hints *storage.SelectHints, ms []*labels.Matcher,
+	index IndexReader, chunks ChunkReader, tombstones tombstones.Reader, mint, maxt int64,
+) storage.SeriesSet {
 	disableTrimming := false
 	sharded := hints != nil && hints.ShardCount > 0
 
-	p, err := PostingsForMatchers(ctx, q.index, ms...)
+	p, err := PostingsForMatchers(ctx, index, ms...)
 	if err != nil {
 		return storage.ErrSeriesSet(err)
 	}
 	if sharded {
-		p = q.index.ShardedPostings(p, hints.ShardIndex, hints.ShardCount)
+		p = index.ShardedPostings(p, hints.ShardIndex, hints.ShardCount)
 	}
 	if sortSeries {
-		p = q.index.SortedPostings(p)
+		p = index.SortedPostings(p)
 	}
 
 	if hints != nil {
@@ -137,11 +141,11 @@ func (q *blockQuerier) Select(ctx context.Context, sortSeries bool, hints *stora
 		disableTrimming = hints.DisableTrimming
 		if hints.Func == "series" {
 			// When you're only looking up metadata (for example series API), you don't need to load any chunks.
-			return newBlockSeriesSet(q.index, newNopChunkReader(), q.tombstones, p, mint, maxt, disableTrimming)
+			return newBlockSeriesSet(index, newNopChunkReader(), tombstones, p, mint, maxt, disableTrimming)
 		}
 	}
 
-	return newBlockSeriesSet(q.index, q.chunks, q.tombstones, p, mint, maxt, disableTrimming)
+	return newBlockSeriesSet(index, chunks, tombstones, p, mint, maxt, disableTrimming)
 }
 
 // blockChunkQuerier provides chunk querying access to a single block database.
@@ -159,8 +163,12 @@ func NewBlockChunkQuerier(b BlockReader, mint, maxt int64) (storage.ChunkQuerier
 }
 
 func (q *blockChunkQuerier) Select(ctx context.Context, sortSeries bool, hints *storage.SelectHints, ms ...*labels.Matcher) storage.ChunkSeriesSet {
-	mint := q.mint
+	return selectChunkSeriesSet(ctx, sortSeries, hints, ms, q.blockID, q.index, q.chunks, q.tombstones, q.mint, q.maxt)
-	maxt := q.maxt
+}
+
+func selectChunkSeriesSet(ctx context.Context, sortSeries bool, hints *storage.SelectHints, ms []*labels.Matcher,
+	blockID ulid.ULID, index IndexReader, chunks ChunkReader, tombstones tombstones.Reader, mint, maxt int64,
+) storage.ChunkSeriesSet {
 	disableTrimming := false
 	sharded := hints != nil && hints.ShardCount > 0
 
@@ -169,17 +177,17 @@ func (q *blockChunkQuerier) Select(ctx context.Context, sortSeries bool, hints *
 		maxt = hints.End
 		disableTrimming = hints.DisableTrimming
 	}
-	p, err := PostingsForMatchers(ctx, q.index, ms...)
+	p, err := PostingsForMatchers(ctx, index, ms...)
 	if err != nil {
 		return storage.ErrChunkSeriesSet(err)
 	}
 	if sharded {
-		p = q.index.ShardedPostings(p, hints.ShardIndex, hints.ShardCount)
+		p = index.ShardedPostings(p, hints.ShardIndex, hints.ShardCount)
 	}
 	if sortSeries {
-		p = q.index.SortedPostings(p)
+		p = index.SortedPostings(p)
 	}
-	return NewBlockChunkSeriesSet(q.blockID, q.index, q.chunks, q.tombstones, p, mint, maxt, disableTrimming)
+	return NewBlockChunkSeriesSet(blockID, index, chunks, tombstones, p, mint, maxt, disableTrimming)
 }
 
 // PostingsForMatchers assembles a single postings iterator against the index reader
@@ -633,14 +641,16 @@ func (p *populateWithDelGenericSeriesIterator) next(copyHeadChunk bool) bool {
 		}
 	}
 
-	hcr, ok := p.cr.(*headChunkReader)
+	hcr, ok := p.cr.(ChunkReaderWithCopy)
 	var iterable chunkenc.Iterable
 	if ok && copyHeadChunk && len(p.bufIter.Intervals) == 0 {
-		// ChunkWithCopy will copy the head chunk.
+		// ChunkOrIterableWithCopy will copy the head chunk, if it can.
 		var maxt int64
-		p.currMeta.Chunk, maxt, p.err = hcr.ChunkWithCopy(p.currMeta)
+		p.currMeta.Chunk, iterable, maxt, p.err = hcr.ChunkOrIterableWithCopy(p.currMeta)
+		if p.currMeta.Chunk != nil {
 			// For the in-memory head chunk the index reader sets maxt as MaxInt64. We fix it here.
 			p.currMeta.MaxTime = maxt
+		}
 	} else {
 		p.currMeta.Chunk, iterable, p.err = p.cr.ChunkOrIterable(p.currMeta)
 	}
@@ -962,7 +972,7 @@ func (p *populateWithDelChunkSeriesIterator) populateChunksFromIterable() bool {
 		// Check if the encoding has changed (i.e. we need to create a new
 		// chunk as chunks can't have multiple encoding types).
 		// For the first sample, the following condition will always be true as
-		// ValNoneNone != ValFloat | ValHistogram | ValFloatHistogram.
+		// ValNone != ValFloat | ValHistogram | ValFloatHistogram.
 		if currentValueType != prevValueType {
 			if prevValueType != chunkenc.ValNone {
 				p.chunksFromIterable = append(p.chunksFromIterable, chunks.Meta{Chunk: currentChunk, MinTime: cmint, MaxTime: cmaxt})

tsdb/querier_bench_test.go

@@ -20,6 +20,7 @@ import (
 	"testing"
 
 	"github.com/prometheus/prometheus/model/labels"
+	"github.com/prometheus/prometheus/storage"
 	"github.com/prometheus/prometheus/tsdb/index"
 
 	"github.com/stretchr/testify/require"
@@ -254,25 +255,32 @@ func BenchmarkMergedStringIter(b *testing.B) {
 	b.ReportAllocs()
 }
 
-func BenchmarkQuerierSelect(b *testing.B) {
+func createHeadForBenchmarkSelect(b *testing.B, numSeries int, addSeries func(app storage.Appender, i int)) (*Head, *DB) {
-	opts := DefaultHeadOptions()
+	dir := b.TempDir()
-	opts.ChunkRange = 1000
+	opts := DefaultOptions()
-	opts.ChunkDirRoot = b.TempDir()
+	opts.OutOfOrderCapMax = 255
-	h, err := NewHead(nil, nil, nil, nil, opts, nil)
+	opts.OutOfOrderTimeWindow = 1000
+	db, err := Open(dir, nil, nil, opts, nil)
 	require.NoError(b, err)
-	defer h.Close()
+	b.Cleanup(func() {
+		require.NoError(b, db.Close())
+	})
+	h := db.Head()
+
 	app := h.Appender(context.Background())
-	numSeries := 1000000
 	for i := 0; i < numSeries; i++ {
-		app.Append(0, labels.FromStrings("foo", "bar", "i", fmt.Sprintf("%d%s", i, postingsBenchSuffix)), int64(i), 0)
+		addSeries(app, i)
 	}
 	require.NoError(b, app.Commit())
+	return h, db
+}
 
-	bench := func(b *testing.B, br BlockReader, sorted bool) {
+func benchmarkSelect(b *testing.B, queryable storage.Queryable, numSeries int, sorted bool) {
 	matcher := labels.MustNewMatcher(labels.MatchEqual, "foo", "bar")
+	b.ResetTimer()
 	for s := 1; s <= numSeries; s *= 10 {
 		b.Run(fmt.Sprintf("%dof%d", s, numSeries), func(b *testing.B) {
-				q, err := NewBlockQuerier(br, 0, int64(s-1))
+			q, err := queryable.Querier(0, int64(s-1))
 			require.NoError(b, err)
 
 			b.ResetTimer()
@@ -285,15 +293,25 @@ func BenchmarkQuerierSelect(b *testing.B) {
 			q.Close()
 		})
 	}
+}
+
+func BenchmarkQuerierSelect(b *testing.B) {
+	numSeries := 1000000
+	h, db := createHeadForBenchmarkSelect(b, numSeries, func(app storage.Appender, i int) {
+		_, err := app.Append(0, labels.FromStrings("foo", "bar", "i", fmt.Sprintf("%d%s", i, postingsBenchSuffix)), int64(i), 0)
+		if err != nil {
+			b.Fatal(err)
 		}
+	})
 
 	b.Run("Head", func(b *testing.B) {
-		bench(b, h, false)
+		benchmarkSelect(b, db, numSeries, false)
 	})
 	b.Run("SortedHead", func(b *testing.B) {
-		bench(b, h, true)
+		benchmarkSelect(b, db, numSeries, true)
 	})
 
+	b.Run("Block", func(b *testing.B) {
 		tmpdir := b.TempDir()
 
 		blockdir := createBlockFromHead(b, tmpdir, h)
@@ -303,7 +321,32 @@ func BenchmarkQuerierSelect(b *testing.B) {
 			require.NoError(b, block.Close())
 		}()
 
-	b.Run("Block", func(b *testing.B) {
+		benchmarkSelect(b, (*queryableBlock)(block), numSeries, false)
-		bench(b, block, false)
+	})
+}
+
+// Type wrapper to let a Block be a Queryable in benchmarkSelect().
+type queryableBlock Block
+
+func (pb *queryableBlock) Querier(mint, maxt int64) (storage.Querier, error) {
+	return NewBlockQuerier((*Block)(pb), mint, maxt)
+}
+
+func BenchmarkQuerierSelectWithOutOfOrder(b *testing.B) {
+	numSeries := 1000000
+	_, db := createHeadForBenchmarkSelect(b, numSeries, func(app storage.Appender, i int) {
+		l := labels.FromStrings("foo", "bar", "i", fmt.Sprintf("%d%s", i, postingsBenchSuffix))
+		ref, err := app.Append(0, l, int64(i+1), 0)
+		if err != nil {
+			b.Fatal(err)
+		}
+		_, err = app.Append(ref, l, int64(i), 1) // Out of order sample
+		if err != nil {
+			b.Fatal(err)
+		}
+	})
+
+	b.Run("Head", func(b *testing.B) {
+		benchmarkSelect(b, db, numSeries, false)
 	})
 }

tsdb/querier_test.go

@@ -3169,12 +3169,11 @@ func BenchmarkQueries(b *testing.B) {
 
 					qHead, err := NewBlockQuerier(NewRangeHead(head, 1, nSamples), 1, nSamples)
 					require.NoError(b, err)
-					qOOOHead, err := NewBlockQuerier(NewOOORangeHead(head, 1, nSamples, 0), 1, nSamples)
+					isoState := head.oooIso.TrackReadAfter(0)
-					require.NoError(b, err)
+					qOOOHead := NewHeadAndOOOQuerier(1, nSamples, head, isoState, qHead)
 
 					queryTypes = append(queryTypes, qt{
-						fmt.Sprintf("_Head_oooPercent:%d", oooPercentage),
+						fmt.Sprintf("_Head_oooPercent:%d", oooPercentage), qOOOHead,
-						storage.NewMergeQuerier([]storage.Querier{qHead, qOOOHead}, nil, storage.ChainedSeriesMerge),
 					})
 				}
 

tsdb/wlog/watcher.go

@@ -58,15 +58,16 @@ type WriteTo interface {
 	StoreSeries([]record.RefSeries, int)
 	StoreMetadata([]record.RefMetadata)
 
-	// Next two methods are intended for garbage-collection: first we call
+	// UpdateSeriesSegment and SeriesReset are intended for
-	// UpdateSeriesSegment on all current series
+	// garbage-collection:
+	// First we call UpdateSeriesSegment on all current series.
 	UpdateSeriesSegment([]record.RefSeries, int)
-	// Then SeriesReset is called to allow the deletion
+	// Then SeriesReset is called to allow the deletion of all series
-	// of all series created in a segment lower than the argument.
+	// created in a segment lower than the argument.
 	SeriesReset(int)
 }
 
-// Used to notify the watcher that data has been written so that it can read.
+// WriteNotified notifies the watcher that data has been written so that it can read.
 type WriteNotified interface {
 	Notify()
 }
@@ -572,6 +573,7 @@ func (w *Watcher) readSegment(r *LiveReader, segmentNum int, tail bool) error {
 				w.writer.AppendHistograms(histogramsToSend)
 				histogramsToSend = histogramsToSend[:0]
 			}
+
 		case record.FloatHistogramSamples:
 			// Skip if experimental "histograms over remote write" is not enabled.
 			if !w.sendHistograms {
@@ -610,11 +612,13 @@ func (w *Watcher) readSegment(r *LiveReader, segmentNum int, tail bool) error {
 				return err
 			}
 			w.writer.StoreMetadata(meta)
-		case record.Tombstones:
 
-		default:
+		case record.Unknown:
 			// Could be corruption, or reading from a WAL from a newer Prometheus.
 			w.recordDecodeFailsMetric.Inc()
+
+		default:
+			// We're not interested in other types of records.
 		}
 	}
 	if err := r.Err(); err != nil {
@@ -643,14 +647,12 @@ func (w *Watcher) readSegmentForGC(r *LiveReader, segmentNum int, _ bool) error
 			}
 			w.writer.UpdateSeriesSegment(series, segmentNum)
 
-		// Ignore these; we're only interested in series.
+		case record.Unknown:
-		case record.Samples:
-		case record.Exemplars:
-		case record.Tombstones:
-
-		default:
 			// Could be corruption, or reading from a WAL from a newer Prometheus.
 			w.recordDecodeFailsMetric.Inc()
+
+		default:
+			// We're only interested in series.
 		}
 	}
 	if err := r.Err(); err != nil {

tsdb/wlog/wlog.go

@@ -38,8 +38,8 @@ import (
 )
 
 const (
-	DefaultSegmentSize = 128 * 1024 * 1024 // 128 MB
+	DefaultSegmentSize = 128 * 1024 * 1024 // DefaultSegmentSize is 128 MB.
-	pageSize           = 32 * 1024         // 32KB
+	pageSize           = 32 * 1024         // pageSize is 32KB.
 	recordHeaderSize   = 7
 	WblDirName         = "wbl"
 )
@@ -612,16 +612,16 @@ func (w *WL) setSegment(segment *Segment) error {
 
 // flushPage writes the new contents of the page to disk. If no more records will fit into
 // the page, the remaining bytes will be set to zero and a new page will be started.
-// If clear is true, this is enforced regardless of how many bytes are left in the page.
+// If forceClear is true, this is enforced regardless of how many bytes are left in the page.
-func (w *WL) flushPage(clear bool) error {
+func (w *WL) flushPage(forceClear bool) error {
 	w.metrics.pageFlushes.Inc()
 
 	p := w.page
-	clear = clear || p.full()
+	shouldClear := forceClear || p.full()
 
 	// No more data will fit into the page or an implicit clear.
 	// Enqueue and clear it.
-	if clear {
+	if shouldClear {
 		p.alloc = pageSize // Write till end of page.
 	}
 
@@ -633,7 +633,7 @@ func (w *WL) flushPage(clear bool) error {
 	p.flushed += n
 
 	// We flushed an entire page, prepare a new one.
-	if clear {
+	if shouldClear {
 		p.reset()
 		w.donePages++
 		w.metrics.pageCompletions.Inc()

util/annotations/annotations.go

@@ -174,7 +174,7 @@ func NewInvalidQuantileWarning(q float64, pos posrange.PositionRange) error {
 	}
 }
 
-// NewInvalidQuantileWarning is used when the user specifies an invalid ratio
+// NewInvalidRatioWarning is used when the user specifies an invalid ratio
 // value, i.e. a float that is outside the range [-1, 1] or NaN.
 func NewInvalidRatioWarning(q, to float64, pos posrange.PositionRange) error {
 	return annoErr{

util/testutil/cmp.go

@@ -23,13 +23,14 @@ import (
 	"github.com/prometheus/prometheus/model/labels"
 )
 
-// Replacement for require.Equal using go-cmp adapted for Prometheus data structures, instead of DeepEqual.
+// RequireEqual is a replacement for require.Equal using go-cmp adapted for
+// Prometheus data structures, instead of DeepEqual.
 func RequireEqual(t testing.TB, expected, actual interface{}, msgAndArgs ...interface{}) {
 	t.Helper()
 	RequireEqualWithOptions(t, expected, actual, nil, msgAndArgs...)
 }
 
-// As RequireEqual but allows extra cmp.Options.
+// RequireEqualWithOptions works like RequireEqual but allows extra cmp.Options.
 func RequireEqualWithOptions(t testing.TB, expected, actual interface{}, extra []cmp.Option, msgAndArgs ...interface{}) {
 	t.Helper()
 	options := append([]cmp.Option{cmp.Comparer(labels.Equal)}, extra...)

util/testutil/directory.go

@@ -155,7 +155,7 @@ func DirHash(t *testing.T, path string) []byte {
 		modTime, err := info.ModTime().GobEncode()
 		require.NoError(t, err)
 
-		_, err = io.WriteString(hash, string(modTime))
+		_, err = hash.Write(modTime)
 		require.NoError(t, err)
 		return nil
 	})

web/web.go

@@ -481,14 +481,14 @@ func New(logger log.Logger, o *Options) *Handler {
 
 	router.Get("/-/healthy", func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusOK)
-		fmt.Fprintf(w, o.AppName+" is Healthy.\n")
+		fmt.Fprintf(w, "%s is Healthy.\n", o.AppName)
 	})
 	router.Head("/-/healthy", func(w http.ResponseWriter, _ *http.Request) {
 		w.WriteHeader(http.StatusOK)
 	})
 	router.Get("/-/ready", readyf(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusOK)
-		fmt.Fprintf(w, o.AppName+" is Ready.\n")
+		fmt.Fprintf(w, "%s is Ready.\n", o.AppName)
 	}))
 	router.Head("/-/ready", readyf(func(w http.ResponseWriter, r *http.Request) {
 		w.WriteHeader(http.StatusOK)