diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml index cead7abfdb..a56140c191 100644 --- a/.github/workflows/ci.yml +++ b/.github/workflows/ci.yml @@ -161,21 +161,20 @@ jobs: runs-on: ubuntu-latest steps: - name: Checkout repository - uses: actions/checkout@0ad4b8fadaa221de15dcec353f45205ec38ea70b # v4.1.4 + uses: actions/checkout@a5ac7e51b41094c92402da3b24376905380afc29 # v4.1.6 - name: Install Go uses: actions/setup-go@0c52d547c9bc32b1aa3301fd7a9cb496313a4491 # v5.0.0 with: - cache: false go-version: 1.22.x - name: Install snmp_exporter/generator dependencies run: sudo apt-get update && sudo apt-get -y install libsnmp-dev if: github.repository == 'prometheus/snmp_exporter' - name: Lint - uses: golangci/golangci-lint-action@3cfe3a4abbb849e10058ce4af15d205b6da42804 # v4.0.0 + uses: golangci/golangci-lint-action@a4f60bb28d35aeee14e6880718e0c85ff1882e64 # v6.0.1 with: args: --verbose # Make sure to sync this with Makefile.common and scripts/golangci-lint.yml. - version: v1.56.2 + version: v1.59.0 fuzzing: uses: ./.github/workflows/fuzzing.yml if: github.event_name == 'pull_request' diff --git a/.gitpod.Dockerfile b/.gitpod.Dockerfile index d645db5de1..2370ec5f5c 100644 --- a/.gitpod.Dockerfile +++ b/.gitpod.Dockerfile @@ -1,15 +1,33 @@ FROM gitpod/workspace-full +# Set Node.js version as an environment variable. ENV CUSTOM_NODE_VERSION=16 -ENV CUSTOM_GO_VERSION=1.19 -ENV GOPATH=$HOME/go-packages -ENV GOROOT=$HOME/go -ENV PATH=$GOROOT/bin:$GOPATH/bin:$PATH +# Install and use the specified Node.js version via nvm. RUN bash -c ". .nvm/nvm.sh && nvm install ${CUSTOM_NODE_VERSION} && nvm use ${CUSTOM_NODE_VERSION} && nvm alias default ${CUSTOM_NODE_VERSION}" +# Ensure nvm uses the default Node.js version in all new shells. RUN echo "nvm use default &>/dev/null" >> ~/.bashrc.d/51-nvm-fix -RUN curl -fsSL https://dl.google.com/go/go${GO_VERSION}.linux-amd64.tar.gz | tar xzs \ - && printf '%s\n' 'export GOPATH=/workspace/go' \ - 'export PATH=$GOPATH/bin:$PATH' > $HOME/.bashrc.d/300-go +# Remove any existing Go installation in $HOME path. +RUN rm -rf $HOME/go $HOME/go-packages + +# Export go environment variables. +RUN echo "export GOPATH=/workspace/go" >> ~/.bashrc.d/300-go && \ + echo "export GOBIN=\$GOPATH/bin" >> ~/.bashrc.d/300-go && \ + echo "export GOROOT=${HOME}/go" >> ~/.bashrc.d/300-go && \ + echo "export PATH=\$GOROOT/bin:\$GOBIN:\$PATH" >> ~/.bashrc + +# Reload the environment variables to ensure go environment variables are +# available in subsequent commands. +RUN bash -c "source ~/.bashrc && source ~/.bashrc.d/300-go" + +# Fetch the Go version dynamically from the Prometheus go.mod file and Install Go in $HOME path. +RUN export CUSTOM_GO_VERSION=$(curl -sSL "https://raw.githubusercontent.com/prometheus/prometheus/main/go.mod" | awk '/^go/{print $2".0"}') && \ + curl -fsSL "https://dl.google.com/go/go${CUSTOM_GO_VERSION}.linux-amd64.tar.gz" | \ + tar -xz -C $HOME + +# Fetch the goyacc parser version dynamically from the Prometheus Makefile +# and install it globally in $GOBIN path. +RUN GOYACC_VERSION=$(curl -fsSL "https://raw.githubusercontent.com/prometheus/prometheus/main/Makefile" | awk -F'=' '/GOYACC_VERSION \?=/{gsub(/ /, "", $2); print $2}') && \ + go install "golang.org/x/tools/cmd/goyacc@${GOYACC_VERSION}" diff --git a/.golangci.yml b/.golangci.yml index a85a76cdf1..f81b29ed2d 100644 --- a/.golangci.yml +++ b/.golangci.yml @@ -21,6 +21,7 @@ linters: - goimports - misspell - nolintlint + - perfsprint - predeclared - revive - testifylint @@ -44,7 +45,9 @@ issues: - linters: - godot source: "^// ===" - + - linters: + - perfsprint + text: "fmt.Sprintf can be replaced with string concatenation" linters-settings: depguard: rules: @@ -85,6 +88,9 @@ linters-settings: local-prefixes: github.com/prometheus/prometheus gofumpt: extra-rules: true + perfsprint: + # Optimizes `fmt.Errorf`. + 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. diff --git a/CHANGELOG.md b/CHANGELOG.md index 079dd7595a..941f09da91 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -3,6 +3,7 @@ ## unreleased * [CHANGE] Rules: Execute 1 query instead of N (where N is the number of alerts within alert rule) when restoring alerts. #13980 +* [FEATURE] Rules: Add new option `query_offset` for each rule group via rule group configuration file and `rule_query_offset` as part of the global configuration to have more resilience for remote write delays. #14061 * [ENHANCEMENT] Rules: Add `rule_group_last_restore_duration_seconds` to measure the time it takes to restore a rule group. #13974 * [ENHANCEMENT] OTLP: Improve remote write format translation performance by using label set hashes for metric identifiers instead of string based ones. #14006 #13991 * [ENHANCEMENT] TSDB: Optimize querying with regexp matchers. #13620 diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index 7687826ba4..9b1b286ccf 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -42,7 +42,12 @@ go build ./cmd/prometheus/ make test # Make sure all the tests pass before you commit and push :) ``` -We use [`golangci-lint`](https://github.com/golangci/golangci-lint) for linting the code. If it reports an issue and you think that the warning needs to be disregarded or is a false-positive, you can add a special comment `//nolint:linter1[,linter2,...]` before the offending line. Use this sparingly though, fixing the code to comply with the linter's recommendation is in general the preferred course of action. +To run a collection of Go linters through [`golangci-lint`](https://github.com/golangci/golangci-lint), do: +```bash +make lint +``` + +If it reports an issue and you think that the warning needs to be disregarded or is a false-positive, you can add a special comment `//nolint:linter1[,linter2,...]` before the offending line. Use this sparingly though, fixing the code to comply with the linter's recommendation is in general the preferred course of action. See [this section of the golangci-lint documentation](https://golangci-lint.run/usage/false-positives/#nolint-directive) for more information. All our issues are regularly tagged so that you can also filter down the issues involving the components you want to work on. For our labeling policy refer [the wiki page](https://github.com/prometheus/prometheus/wiki/Label-Names-and-Descriptions). diff --git a/Makefile.common b/Makefile.common index 0e9ace29b4..1617292350 100644 --- a/Makefile.common +++ b/Makefile.common @@ -61,7 +61,7 @@ PROMU_URL := https://github.com/prometheus/promu/releases/download/v$(PROMU_ SKIP_GOLANGCI_LINT := GOLANGCI_LINT := GOLANGCI_LINT_OPTS ?= -GOLANGCI_LINT_VERSION ?= v1.56.2 +GOLANGCI_LINT_VERSION ?= v1.59.0 # golangci-lint only supports linux, darwin and windows platforms on i386/amd64/arm64. # windows isn't included here because of the path separator being different. ifeq ($(GOHOSTOS),$(filter $(GOHOSTOS),linux darwin)) diff --git a/cmd/prometheus/main.go b/cmd/prometheus/main.go index 8218ffb18d..0532bc3809 100644 --- a/cmd/prometheus/main.go +++ b/cmd/prometheus/main.go @@ -42,6 +42,7 @@ import ( "github.com/mwitkow/go-conntrack" "github.com/oklog/run" "github.com/prometheus/client_golang/prometheus" + "github.com/prometheus/client_golang/prometheus/collectors" versioncollector "github.com/prometheus/client_golang/prometheus/collectors/version" "github.com/prometheus/common/model" "github.com/prometheus/common/promlog" @@ -252,6 +253,18 @@ func main() { newFlagRetentionDuration model.Duration ) + // Unregister the default GoCollector, and reregister with our defaults. + if prometheus.Unregister(collectors.NewGoCollector()) { + prometheus.MustRegister( + collectors.NewGoCollector( + collectors.WithGoCollectorRuntimeMetrics( + collectors.MetricsGC, + collectors.MetricsScheduler, + ), + ), + ) + } + cfg := flagConfig{ notifier: notifier.Options{ Registerer: prometheus.DefaultRegisterer, @@ -418,7 +431,7 @@ func main() { serverOnlyFlag(a, "rules.alert.resend-delay", "Minimum amount of time to wait before resending an alert to Alertmanager."). Default("1m").SetValue(&cfg.resendDelay) - serverOnlyFlag(a, "rules.max-concurrent-evals", "Global concurrency limit for independent rules that can run concurrently."). + serverOnlyFlag(a, "rules.max-concurrent-evals", "Global concurrency limit for independent rules that can run concurrently. When set, \"query.max-concurrency\" may need to be adjusted accordingly."). Default("4").Int64Var(&cfg.maxConcurrentEvals) a.Flag("scrape.adjust-timestamps", "Adjust scrape timestamps by up to `scrape.timestamp-tolerance` to align them to the intended schedule. See https://github.com/prometheus/prometheus/issues/7846 for more context. Experimental. This flag will be removed in a future release."). @@ -772,6 +785,9 @@ func main() { ResendDelay: time.Duration(cfg.resendDelay), MaxConcurrentEvals: cfg.maxConcurrentEvals, ConcurrentEvalsEnabled: cfg.enableConcurrentRuleEval, + DefaultRuleQueryOffset: func() time.Duration { + return time.Duration(cfgFile.GlobalConfig.RuleQueryOffset) + }, }) } diff --git a/cmd/prometheus/main_test.go b/cmd/prometheus/main_test.go index 03f3a9bc39..89c171bb5b 100644 --- a/cmd/prometheus/main_test.go +++ b/cmd/prometheus/main_test.go @@ -24,6 +24,7 @@ import ( "os/exec" "path/filepath" "runtime" + "strconv" "strings" "syscall" "testing" @@ -189,7 +190,7 @@ func TestSendAlerts(t *testing.T) { for i, tc := range testCases { tc := tc - t.Run(fmt.Sprintf("%d", i), func(t *testing.T) { + t.Run(strconv.Itoa(i), func(t *testing.T) { senderFunc := senderFunc(func(alerts ...*notifier.Alert) { require.NotEmpty(t, tc.in, "sender called with 0 alert") require.Equal(t, tc.exp, alerts) diff --git a/cmd/prometheus/query_log_test.go b/cmd/prometheus/query_log_test.go index dd6b56672c..9a7a3ed855 100644 --- a/cmd/prometheus/query_log_test.go +++ b/cmd/prometheus/query_log_test.go @@ -296,7 +296,7 @@ func (p *queryLogTest) run(t *testing.T) { if p.exactQueryCount() { require.Equal(t, 1, qc) } else { - require.Greater(t, qc, 0, "no queries logged") + require.Positive(t, qc, "no queries logged") } p.validateLastQuery(t, ql) @@ -366,7 +366,7 @@ func (p *queryLogTest) run(t *testing.T) { if p.exactQueryCount() { require.Equal(t, 1, qc) } else { - require.Greater(t, qc, 0, "no queries logged") + require.Positive(t, qc, "no queries logged") } } diff --git a/cmd/promtool/backfill.go b/cmd/promtool/backfill.go index 601c3ced9f..79db428c71 100644 --- a/cmd/promtool/backfill.go +++ b/cmd/promtool/backfill.go @@ -88,7 +88,7 @@ func createBlocks(input []byte, mint, maxt, maxBlockDuration int64, maxSamplesIn blockDuration := getCompatibleBlockDuration(maxBlockDuration) mint = blockDuration * (mint / blockDuration) - db, err := tsdb.OpenDBReadOnly(outputDir, nil) + db, err := tsdb.OpenDBReadOnly(outputDir, "", nil) if err != nil { return err } diff --git a/cmd/promtool/main.go b/cmd/promtool/main.go index c0484adcc0..e1d275e97e 100644 --- a/cmd/promtool/main.go +++ b/cmd/promtool/main.go @@ -235,12 +235,14 @@ func main() { tsdbDumpCmd := tsdbCmd.Command("dump", "Dump samples from a TSDB.") dumpPath := tsdbDumpCmd.Arg("db path", "Database path (default is "+defaultDBPath+").").Default(defaultDBPath).String() + dumpSandboxDirRoot := tsdbDumpCmd.Flag("sandbox-dir-root", "Root directory where a sandbox directory would be created in case WAL replay generates chunks. The sandbox directory is cleaned up at the end.").Default(defaultDBPath).String() dumpMinTime := tsdbDumpCmd.Flag("min-time", "Minimum timestamp to dump.").Default(strconv.FormatInt(math.MinInt64, 10)).Int64() dumpMaxTime := tsdbDumpCmd.Flag("max-time", "Maximum timestamp to dump.").Default(strconv.FormatInt(math.MaxInt64, 10)).Int64() dumpMatch := tsdbDumpCmd.Flag("match", "Series selector. Can be specified multiple times.").Default("{__name__=~'(?s:.*)'}").Strings() - tsdbDumpOpenMetricsCmd := tsdbCmd.Command("dump-openmetrics", "[Experimental] Dump samples from a TSDB into OpenMetrics format. Native histograms are not dumped.") + tsdbDumpOpenMetricsCmd := tsdbCmd.Command("dump-openmetrics", "[Experimental] Dump samples from a TSDB into OpenMetrics text format, excluding native histograms and staleness markers, which are not representable in OpenMetrics.") dumpOpenMetricsPath := tsdbDumpOpenMetricsCmd.Arg("db path", "Database path (default is "+defaultDBPath+").").Default(defaultDBPath).String() + dumpOpenMetricsSandboxDirRoot := tsdbDumpOpenMetricsCmd.Flag("sandbox-dir-root", "Root directory where a sandbox directory would be created in case WAL replay generates chunks. The sandbox directory is cleaned up at the end.").Default(defaultDBPath).String() dumpOpenMetricsMinTime := tsdbDumpOpenMetricsCmd.Flag("min-time", "Minimum timestamp to dump.").Default(strconv.FormatInt(math.MinInt64, 10)).Int64() dumpOpenMetricsMaxTime := tsdbDumpOpenMetricsCmd.Flag("max-time", "Maximum timestamp to dump.").Default(strconv.FormatInt(math.MaxInt64, 10)).Int64() dumpOpenMetricsMatch := tsdbDumpOpenMetricsCmd.Flag("match", "Series selector. Can be specified multiple times.").Default("{__name__=~'(?s:.*)'}").Strings() @@ -396,9 +398,9 @@ func main() { os.Exit(checkErr(listBlocks(*listPath, *listHumanReadable))) case tsdbDumpCmd.FullCommand(): - os.Exit(checkErr(dumpSamples(ctx, *dumpPath, *dumpMinTime, *dumpMaxTime, *dumpMatch, formatSeriesSet))) + os.Exit(checkErr(dumpSamples(ctx, *dumpPath, *dumpSandboxDirRoot, *dumpMinTime, *dumpMaxTime, *dumpMatch, formatSeriesSet))) case tsdbDumpOpenMetricsCmd.FullCommand(): - os.Exit(checkErr(dumpSamples(ctx, *dumpOpenMetricsPath, *dumpOpenMetricsMinTime, *dumpOpenMetricsMaxTime, *dumpOpenMetricsMatch, formatSeriesSetOpenMetrics))) + os.Exit(checkErr(dumpSamples(ctx, *dumpOpenMetricsPath, *dumpOpenMetricsSandboxDirRoot, *dumpOpenMetricsMinTime, *dumpOpenMetricsMaxTime, *dumpOpenMetricsMatch, formatSeriesSetOpenMetrics))) // TODO(aSquare14): Work on adding support for custom block size. case openMetricsImportCmd.FullCommand(): os.Exit(backfillOpenMetrics(*importFilePath, *importDBPath, *importHumanReadable, *importQuiet, *maxBlockDuration)) diff --git a/cmd/promtool/main_test.go b/cmd/promtool/main_test.go index 7306a3e648..78500fe937 100644 --- a/cmd/promtool/main_test.go +++ b/cmd/promtool/main_test.go @@ -25,6 +25,7 @@ import ( "os/exec" "path/filepath" "runtime" + "strconv" "strings" "syscall" "testing" @@ -410,7 +411,7 @@ func TestExitCodes(t *testing.T) { } { t.Run(c.file, func(t *testing.T) { for _, lintFatal := range []bool{true, false} { - t.Run(fmt.Sprintf("%t", lintFatal), func(t *testing.T) { + t.Run(strconv.FormatBool(lintFatal), func(t *testing.T) { args := []string{"-test.main", "check", "config", "testdata/" + c.file} if lintFatal { args = append(args, "--lint-fatal") diff --git a/cmd/promtool/tsdb.go b/cmd/promtool/tsdb.go index 2ad969438c..2ed7244b1c 100644 --- a/cmd/promtool/tsdb.go +++ b/cmd/promtool/tsdb.go @@ -338,7 +338,7 @@ func readPrometheusLabels(r io.Reader, n int) ([]labels.Labels, error) { } func listBlocks(path string, humanReadable bool) error { - db, err := tsdb.OpenDBReadOnly(path, nil) + db, err := tsdb.OpenDBReadOnly(path, "", nil) if err != nil { return err } @@ -393,7 +393,7 @@ func getFormatedBytes(bytes int64, humanReadable bool) string { } func openBlock(path, blockID string) (*tsdb.DBReadOnly, tsdb.BlockReader, error) { - db, err := tsdb.OpenDBReadOnly(path, nil) + db, err := tsdb.OpenDBReadOnly(path, "", nil) if err != nil { return nil, nil, err } @@ -708,8 +708,8 @@ func analyzeCompaction(ctx context.Context, block tsdb.BlockReader, indexr tsdb. type SeriesSetFormatter func(series storage.SeriesSet) error -func dumpSamples(ctx context.Context, path string, mint, maxt int64, match []string, formatter SeriesSetFormatter) (err error) { - db, err := tsdb.OpenDBReadOnly(path, nil) +func dumpSamples(ctx context.Context, dbDir, sandboxDirRoot string, mint, maxt int64, match []string, formatter SeriesSetFormatter) (err error) { + db, err := tsdb.OpenDBReadOnly(dbDir, sandboxDirRoot, nil) if err != nil { return err } @@ -856,9 +856,9 @@ func displayHistogram(dataType string, datas []int, total int) { } avg := sum / len(datas) fmt.Printf("%s (min/avg/max): %d/%d/%d\n", dataType, datas[0], avg, datas[len(datas)-1]) - maxLeftLen := strconv.Itoa(len(fmt.Sprintf("%d", end))) - maxRightLen := strconv.Itoa(len(fmt.Sprintf("%d", end+step))) - maxCountLen := strconv.Itoa(len(fmt.Sprintf("%d", maxCount))) + maxLeftLen := strconv.Itoa(len(strconv.Itoa(end))) + maxRightLen := strconv.Itoa(len(strconv.Itoa(end + step))) + maxCountLen := strconv.Itoa(len(strconv.Itoa(maxCount))) for bucket, count := range buckets { percentage := 100.0 * count / total fmt.Printf("[%"+maxLeftLen+"d, %"+maxRightLen+"d]: %"+maxCountLen+"d %s\n", bucket*step+start+1, (bucket+1)*step+start, count, strings.Repeat("#", percentage)) diff --git a/cmd/promtool/tsdb_test.go b/cmd/promtool/tsdb_test.go index 70e8877659..75089b168b 100644 --- a/cmd/promtool/tsdb_test.go +++ b/cmd/promtool/tsdb_test.go @@ -64,6 +64,7 @@ func getDumpedSamples(t *testing.T, path string, mint, maxt int64, match []strin err := dumpSamples( context.Background(), path, + t.TempDir(), mint, maxt, match, diff --git a/cmd/promtool/unittest.go b/cmd/promtool/unittest.go index 27d5dd98b9..5451c5296c 100644 --- a/cmd/promtool/unittest.go +++ b/cmd/promtool/unittest.go @@ -573,7 +573,7 @@ func (la labelsAndAnnotations) String() string { } s := "[\n0:" + indentLines("\n"+la[0].String(), " ") for i, l := range la[1:] { - s += ",\n" + fmt.Sprintf("%d", i+1) + ":" + indentLines("\n"+l.String(), " ") + s += ",\n" + strconv.Itoa(i+1) + ":" + indentLines("\n"+l.String(), " ") } s += "\n]" diff --git a/config/config.go b/config/config.go index 1cfd588643..463dbc3571 100644 --- a/config/config.go +++ b/config/config.go @@ -145,6 +145,7 @@ var ( ScrapeInterval: model.Duration(1 * time.Minute), ScrapeTimeout: model.Duration(10 * time.Second), EvaluationInterval: model.Duration(1 * time.Minute), + RuleQueryOffset: model.Duration(0 * time.Minute), // When native histogram feature flag is enabled, ScrapeProtocols default // changes to DefaultNativeHistogramScrapeProtocols. ScrapeProtocols: DefaultScrapeProtocols, @@ -397,6 +398,8 @@ type GlobalConfig struct { ScrapeProtocols []ScrapeProtocol `yaml:"scrape_protocols,omitempty"` // How frequently to evaluate rules by default. EvaluationInterval model.Duration `yaml:"evaluation_interval,omitempty"` + // Offset the rule evaluation timestamp of this particular group by the specified duration into the past to ensure the underlying metrics have been received. + RuleQueryOffset model.Duration `yaml:"rule_query_offset"` // File to which PromQL queries are logged. QueryLogFile string `yaml:"query_log_file,omitempty"` // The labels to add to any timeseries that this Prometheus instance scrapes. @@ -556,6 +559,7 @@ func (c *GlobalConfig) isZero() bool { c.ScrapeInterval == 0 && c.ScrapeTimeout == 0 && c.EvaluationInterval == 0 && + c.RuleQueryOffset == 0 && c.QueryLogFile == "" && c.ScrapeProtocols == nil } diff --git a/discovery/aws/ec2.go b/discovery/aws/ec2.go index aa79fd9c62..a6a0a82577 100644 --- a/discovery/aws/ec2.go +++ b/discovery/aws/ec2.go @@ -18,6 +18,7 @@ import ( "errors" "fmt" "net" + "strconv" "strings" "time" @@ -279,7 +280,7 @@ func (d *EC2Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error if inst.PrivateDnsName != nil { labels[ec2LabelPrivateDNS] = model.LabelValue(*inst.PrivateDnsName) } - addr := net.JoinHostPort(*inst.PrivateIpAddress, fmt.Sprintf("%d", d.cfg.Port)) + addr := net.JoinHostPort(*inst.PrivateIpAddress, strconv.Itoa(d.cfg.Port)) labels[model.AddressLabel] = model.LabelValue(addr) if inst.Platform != nil { diff --git a/discovery/aws/lightsail.go b/discovery/aws/lightsail.go index 86b138be55..0ad7f2d541 100644 --- a/discovery/aws/lightsail.go +++ b/discovery/aws/lightsail.go @@ -18,6 +18,7 @@ import ( "errors" "fmt" "net" + "strconv" "strings" "time" @@ -229,7 +230,7 @@ func (d *LightsailDiscovery) refresh(ctx context.Context) ([]*targetgroup.Group, lightsailLabelRegion: model.LabelValue(d.cfg.Region), } - addr := net.JoinHostPort(*inst.PrivateIpAddress, fmt.Sprintf("%d", d.cfg.Port)) + addr := net.JoinHostPort(*inst.PrivateIpAddress, strconv.Itoa(d.cfg.Port)) labels[model.AddressLabel] = model.LabelValue(addr) if inst.PublicIpAddress != nil { diff --git a/discovery/azure/azure.go b/discovery/azure/azure.go index 7c2ece2c7b..70d95b9f3a 100644 --- a/discovery/azure/azure.go +++ b/discovery/azure/azure.go @@ -20,6 +20,7 @@ import ( "math/rand" "net" "net/http" + "strconv" "strings" "sync" "time" @@ -492,7 +493,7 @@ func (d *Discovery) vmToLabelSet(ctx context.Context, client client, vm virtualM } if ip.Properties != nil && ip.Properties.PrivateIPAddress != nil { labels[azureLabelMachinePrivateIP] = model.LabelValue(*ip.Properties.PrivateIPAddress) - address := net.JoinHostPort(*ip.Properties.PrivateIPAddress, fmt.Sprintf("%d", d.port)) + address := net.JoinHostPort(*ip.Properties.PrivateIPAddress, strconv.Itoa(d.port)) labels[model.AddressLabel] = model.LabelValue(address) return labels, nil } diff --git a/discovery/consul/consul.go b/discovery/consul/consul.go index 40eed7697a..bdc1fc8dce 100644 --- a/discovery/consul/consul.go +++ b/discovery/consul/consul.go @@ -539,9 +539,9 @@ func (srv *consulService) watch(ctx context.Context, ch chan<- []*targetgroup.Gr // since the service may be registered remotely through a different node. var addr string if serviceNode.Service.Address != "" { - addr = net.JoinHostPort(serviceNode.Service.Address, fmt.Sprintf("%d", serviceNode.Service.Port)) + addr = net.JoinHostPort(serviceNode.Service.Address, strconv.Itoa(serviceNode.Service.Port)) } else { - addr = net.JoinHostPort(serviceNode.Node.Address, fmt.Sprintf("%d", serviceNode.Service.Port)) + addr = net.JoinHostPort(serviceNode.Node.Address, strconv.Itoa(serviceNode.Service.Port)) } labels := model.LabelSet{ diff --git a/discovery/digitalocean/digitalocean.go b/discovery/digitalocean/digitalocean.go index 18380b7296..ecee60cb1f 100644 --- a/discovery/digitalocean/digitalocean.go +++ b/discovery/digitalocean/digitalocean.go @@ -177,7 +177,7 @@ func (d *Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error) { } labels := model.LabelSet{ - doLabelID: model.LabelValue(fmt.Sprintf("%d", droplet.ID)), + doLabelID: model.LabelValue(strconv.Itoa(droplet.ID)), doLabelName: model.LabelValue(droplet.Name), doLabelImage: model.LabelValue(droplet.Image.Slug), doLabelImageName: model.LabelValue(droplet.Image.Name), diff --git a/discovery/dns/dns.go b/discovery/dns/dns.go index cf56a2ad02..314c3d38cd 100644 --- a/discovery/dns/dns.go +++ b/discovery/dns/dns.go @@ -18,6 +18,7 @@ import ( "errors" "fmt" "net" + "strconv" "strings" "sync" "time" @@ -200,7 +201,7 @@ func (d *Discovery) refreshOne(ctx context.Context, name string, ch chan<- *targ tg := &targetgroup.Group{} hostPort := func(a string, p int) model.LabelValue { - return model.LabelValue(net.JoinHostPort(a, fmt.Sprintf("%d", p))) + return model.LabelValue(net.JoinHostPort(a, strconv.Itoa(p))) } for _, record := range response.Answer { @@ -209,7 +210,7 @@ func (d *Discovery) refreshOne(ctx context.Context, name string, ch chan<- *targ switch addr := record.(type) { case *dns.SRV: dnsSrvRecordTarget = model.LabelValue(addr.Target) - dnsSrvRecordPort = model.LabelValue(fmt.Sprintf("%d", addr.Port)) + dnsSrvRecordPort = model.LabelValue(strconv.Itoa(int(addr.Port))) // Remove the final dot from rooted DNS names to make them look more usual. addr.Target = strings.TrimRight(addr.Target, ".") diff --git a/discovery/hetzner/hcloud.go b/discovery/hetzner/hcloud.go index 6d0599dfa2..df56f94c5f 100644 --- a/discovery/hetzner/hcloud.go +++ b/discovery/hetzner/hcloud.go @@ -15,7 +15,6 @@ package hetzner import ( "context" - "fmt" "net" "net/http" "strconv" @@ -92,7 +91,7 @@ func (d *hcloudDiscovery) refresh(ctx context.Context) ([]*targetgroup.Group, er for i, server := range servers { labels := model.LabelSet{ hetznerLabelRole: model.LabelValue(HetznerRoleHcloud), - hetznerLabelServerID: model.LabelValue(fmt.Sprintf("%d", server.ID)), + hetznerLabelServerID: model.LabelValue(strconv.FormatInt(server.ID, 10)), hetznerLabelServerName: model.LabelValue(server.Name), hetznerLabelDatacenter: model.LabelValue(server.Datacenter.Name), hetznerLabelPublicIPv4: model.LabelValue(server.PublicNet.IPv4.IP.String()), @@ -102,10 +101,10 @@ func (d *hcloudDiscovery) refresh(ctx context.Context) ([]*targetgroup.Group, er hetznerLabelHcloudDatacenterLocation: model.LabelValue(server.Datacenter.Location.Name), hetznerLabelHcloudDatacenterLocationNetworkZone: model.LabelValue(server.Datacenter.Location.NetworkZone), hetznerLabelHcloudType: model.LabelValue(server.ServerType.Name), - hetznerLabelHcloudCPUCores: model.LabelValue(fmt.Sprintf("%d", server.ServerType.Cores)), + hetznerLabelHcloudCPUCores: model.LabelValue(strconv.Itoa(server.ServerType.Cores)), hetznerLabelHcloudCPUType: model.LabelValue(server.ServerType.CPUType), - hetznerLabelHcloudMemoryGB: model.LabelValue(fmt.Sprintf("%d", int(server.ServerType.Memory))), - hetznerLabelHcloudDiskGB: model.LabelValue(fmt.Sprintf("%d", server.ServerType.Disk)), + hetznerLabelHcloudMemoryGB: model.LabelValue(strconv.Itoa(int(server.ServerType.Memory))), + hetznerLabelHcloudDiskGB: model.LabelValue(strconv.Itoa(server.ServerType.Disk)), model.AddressLabel: model.LabelValue(net.JoinHostPort(server.PublicNet.IPv4.IP.String(), strconv.FormatUint(uint64(d.port), 10))), } diff --git a/discovery/hetzner/robot.go b/discovery/hetzner/robot.go index b862c33f5b..516470b05a 100644 --- a/discovery/hetzner/robot.go +++ b/discovery/hetzner/robot.go @@ -112,7 +112,7 @@ func (d *robotDiscovery) refresh(context.Context) ([]*targetgroup.Group, error) hetznerLabelPublicIPv4: model.LabelValue(server.Server.ServerIP), hetznerLabelServerStatus: model.LabelValue(server.Server.Status), hetznerLabelRobotProduct: model.LabelValue(server.Server.Product), - hetznerLabelRobotCancelled: model.LabelValue(fmt.Sprintf("%t", server.Server.Canceled)), + hetznerLabelRobotCancelled: model.LabelValue(strconv.FormatBool(server.Server.Canceled)), model.AddressLabel: model.LabelValue(net.JoinHostPort(server.Server.ServerIP, strconv.FormatUint(uint64(d.port), 10))), } diff --git a/discovery/legacymanager/manager_test.go b/discovery/legacymanager/manager_test.go index 1ed699645d..a455a8e341 100644 --- a/discovery/legacymanager/manager_test.go +++ b/discovery/legacymanager/manager_test.go @@ -720,7 +720,7 @@ func staticConfig(addrs ...string) discovery.StaticConfig { var cfg discovery.StaticConfig for i, addr := range addrs { cfg = append(cfg, &targetgroup.Group{ - Source: fmt.Sprint(i), + Source: strconv.Itoa(i), Targets: []model.LabelSet{ {model.AddressLabel: model.LabelValue(addr)}, }, diff --git a/discovery/linode/linode.go b/discovery/linode/linode.go index 2a5475b854..f09c1a734c 100644 --- a/discovery/linode/linode.go +++ b/discovery/linode/linode.go @@ -325,7 +325,7 @@ func (d *Discovery) refreshData(ctx context.Context) ([]*targetgroup.Group, erro } labels := model.LabelSet{ - linodeLabelID: model.LabelValue(fmt.Sprintf("%d", instance.ID)), + linodeLabelID: model.LabelValue(strconv.Itoa(instance.ID)), linodeLabelName: model.LabelValue(instance.Label), linodeLabelImage: model.LabelValue(instance.Image), linodeLabelPrivateIPv4: model.LabelValue(privateIPv4), @@ -338,13 +338,13 @@ func (d *Discovery) refreshData(ctx context.Context) ([]*targetgroup.Group, erro linodeLabelType: model.LabelValue(instance.Type), linodeLabelStatus: model.LabelValue(instance.Status), linodeLabelGroup: model.LabelValue(instance.Group), - linodeLabelGPUs: model.LabelValue(fmt.Sprintf("%d", instance.Specs.GPUs)), + linodeLabelGPUs: model.LabelValue(strconv.Itoa(instance.Specs.GPUs)), linodeLabelHypervisor: model.LabelValue(instance.Hypervisor), linodeLabelBackups: model.LabelValue(backupsStatus), - linodeLabelSpecsDiskBytes: model.LabelValue(fmt.Sprintf("%d", int64(instance.Specs.Disk)<<20)), - linodeLabelSpecsMemoryBytes: model.LabelValue(fmt.Sprintf("%d", int64(instance.Specs.Memory)<<20)), - linodeLabelSpecsVCPUs: model.LabelValue(fmt.Sprintf("%d", instance.Specs.VCPUs)), - linodeLabelSpecsTransferBytes: model.LabelValue(fmt.Sprintf("%d", int64(instance.Specs.Transfer)<<20)), + linodeLabelSpecsDiskBytes: model.LabelValue(strconv.FormatInt(int64(instance.Specs.Disk)<<20, 10)), + linodeLabelSpecsMemoryBytes: model.LabelValue(strconv.FormatInt(int64(instance.Specs.Memory)<<20, 10)), + linodeLabelSpecsVCPUs: model.LabelValue(strconv.Itoa(instance.Specs.VCPUs)), + linodeLabelSpecsTransferBytes: model.LabelValue(strconv.FormatInt(int64(instance.Specs.Transfer)<<20, 10)), } addr := net.JoinHostPort(publicIPv4, strconv.FormatUint(uint64(d.port), 10)) diff --git a/discovery/manager_test.go b/discovery/manager_test.go index 656d7c3c66..be07edbdb4 100644 --- a/discovery/manager_test.go +++ b/discovery/manager_test.go @@ -720,7 +720,7 @@ func staticConfig(addrs ...string) StaticConfig { var cfg StaticConfig for i, addr := range addrs { cfg = append(cfg, &targetgroup.Group{ - Source: fmt.Sprint(i), + Source: strconv.Itoa(i), Targets: []model.LabelSet{ {model.AddressLabel: model.LabelValue(addr)}, }, diff --git a/discovery/marathon/marathon.go b/discovery/marathon/marathon.go index 3e9e15967f..38b47accff 100644 --- a/discovery/marathon/marathon.go +++ b/discovery/marathon/marathon.go @@ -505,7 +505,7 @@ func targetEndpoint(task *task, port uint32, containerNet bool) string { host = task.Host } - return net.JoinHostPort(host, fmt.Sprintf("%d", port)) + return net.JoinHostPort(host, strconv.Itoa(int(port))) } // Get a list of ports and a list of labels from a PortMapping. diff --git a/discovery/moby/network.go b/discovery/moby/network.go index 0e0d0041de..794d2e607d 100644 --- a/discovery/moby/network.go +++ b/discovery/moby/network.go @@ -15,7 +15,7 @@ package moby import ( "context" - "fmt" + "strconv" "github.com/docker/docker/api/types" "github.com/docker/docker/client" @@ -44,8 +44,8 @@ func getNetworksLabels(ctx context.Context, client *client.Client, labelPrefix s labelPrefix + labelNetworkID: network.ID, labelPrefix + labelNetworkName: network.Name, labelPrefix + labelNetworkScope: network.Scope, - labelPrefix + labelNetworkInternal: fmt.Sprintf("%t", network.Internal), - labelPrefix + labelNetworkIngress: fmt.Sprintf("%t", network.Ingress), + labelPrefix + labelNetworkInternal: strconv.FormatBool(network.Internal), + labelPrefix + labelNetworkIngress: strconv.FormatBool(network.Ingress), } for k, v := range network.Labels { ln := strutil.SanitizeLabelName(k) diff --git a/discovery/moby/nodes.go b/discovery/moby/nodes.go index a7c5551c02..b5be844eda 100644 --- a/discovery/moby/nodes.go +++ b/discovery/moby/nodes.go @@ -66,7 +66,7 @@ func (d *Discovery) refreshNodes(ctx context.Context) ([]*targetgroup.Group, err swarmLabelNodeAddress: model.LabelValue(n.Status.Addr), } if n.ManagerStatus != nil { - labels[swarmLabelNodeManagerLeader] = model.LabelValue(fmt.Sprintf("%t", n.ManagerStatus.Leader)) + labels[swarmLabelNodeManagerLeader] = model.LabelValue(strconv.FormatBool(n.ManagerStatus.Leader)) labels[swarmLabelNodeManagerReachability] = model.LabelValue(n.ManagerStatus.Reachability) labels[swarmLabelNodeManagerAddr] = model.LabelValue(n.ManagerStatus.Addr) } diff --git a/discovery/moby/services.go b/discovery/moby/services.go index 1d472b5c00..c61b499259 100644 --- a/discovery/moby/services.go +++ b/discovery/moby/services.go @@ -116,7 +116,7 @@ func (d *Discovery) refreshServices(ctx context.Context) ([]*targetgroup.Group, labels[model.LabelName(k)] = model.LabelValue(v) } - addr := net.JoinHostPort(ip.String(), fmt.Sprintf("%d", d.port)) + addr := net.JoinHostPort(ip.String(), strconv.Itoa(d.port)) labels[model.AddressLabel] = model.LabelValue(addr) tg.Targets = append(tg.Targets, labels) diff --git a/discovery/moby/tasks.go b/discovery/moby/tasks.go index 2505a7b07a..38b9d33de2 100644 --- a/discovery/moby/tasks.go +++ b/discovery/moby/tasks.go @@ -150,7 +150,7 @@ func (d *Discovery) refreshTasks(ctx context.Context) ([]*targetgroup.Group, err labels[model.LabelName(k)] = model.LabelValue(v) } - addr := net.JoinHostPort(ip.String(), fmt.Sprintf("%d", d.port)) + addr := net.JoinHostPort(ip.String(), strconv.Itoa(d.port)) labels[model.AddressLabel] = model.LabelValue(addr) tg.Targets = append(tg.Targets, labels) diff --git a/discovery/openstack/hypervisor.go b/discovery/openstack/hypervisor.go index 16964cfb62..8964da9294 100644 --- a/discovery/openstack/hypervisor.go +++ b/discovery/openstack/hypervisor.go @@ -17,6 +17,7 @@ import ( "context" "fmt" "net" + "strconv" "github.com/go-kit/log" "github.com/gophercloud/gophercloud" @@ -72,7 +73,7 @@ func (h *HypervisorDiscovery) refresh(ctx context.Context) ([]*targetgroup.Group } tg := &targetgroup.Group{ - Source: fmt.Sprintf("OS_" + h.region), + Source: "OS_" + h.region, } // OpenStack API reference // https://developer.openstack.org/api-ref/compute/#list-hypervisors-details @@ -84,7 +85,7 @@ func (h *HypervisorDiscovery) refresh(ctx context.Context) ([]*targetgroup.Group } for _, hypervisor := range hypervisorList { labels := model.LabelSet{} - addr := net.JoinHostPort(hypervisor.HostIP, fmt.Sprintf("%d", h.port)) + addr := net.JoinHostPort(hypervisor.HostIP, strconv.Itoa(h.port)) labels[model.AddressLabel] = model.LabelValue(addr) labels[openstackLabelHypervisorID] = model.LabelValue(hypervisor.ID) labels[openstackLabelHypervisorHostName] = model.LabelValue(hypervisor.HypervisorHostname) diff --git a/discovery/openstack/instance.go b/discovery/openstack/instance.go index 9b28c1d6e1..750d414a2b 100644 --- a/discovery/openstack/instance.go +++ b/discovery/openstack/instance.go @@ -17,6 +17,7 @@ import ( "context" "fmt" "net" + "strconv" "github.com/go-kit/log" "github.com/go-kit/log/level" @@ -120,7 +121,7 @@ func (i *InstanceDiscovery) refresh(ctx context.Context) ([]*targetgroup.Group, } pager := servers.List(client, opts) tg := &targetgroup.Group{ - Source: fmt.Sprintf("OS_" + i.region), + Source: "OS_" + i.region, } err = pager.EachPage(func(page pagination.Page) (bool, error) { if ctx.Err() != nil { @@ -194,7 +195,7 @@ func (i *InstanceDiscovery) refresh(ctx context.Context) ([]*targetgroup.Group, if val, ok := floatingIPList[floatingIPKey{id: s.ID, fixed: addr}]; ok { lbls[openstackLabelPublicIP] = model.LabelValue(val) } - addr = net.JoinHostPort(addr, fmt.Sprintf("%d", i.port)) + addr = net.JoinHostPort(addr, strconv.Itoa(i.port)) lbls[model.AddressLabel] = model.LabelValue(addr) tg.Targets = append(tg.Targets, lbls) diff --git a/discovery/ovhcloud/dedicated_server.go b/discovery/ovhcloud/dedicated_server.go index bb5dadcd7b..a70857a08b 100644 --- a/discovery/ovhcloud/dedicated_server.go +++ b/discovery/ovhcloud/dedicated_server.go @@ -144,12 +144,12 @@ func (d *dedicatedServerDiscovery) refresh(context.Context) ([]*targetgroup.Grou model.InstanceLabel: model.LabelValue(server.Name), dedicatedServerLabelPrefix + "state": model.LabelValue(server.State), dedicatedServerLabelPrefix + "commercial_range": model.LabelValue(server.CommercialRange), - dedicatedServerLabelPrefix + "link_speed": model.LabelValue(fmt.Sprintf("%d", server.LinkSpeed)), + dedicatedServerLabelPrefix + "link_speed": model.LabelValue(strconv.Itoa(server.LinkSpeed)), dedicatedServerLabelPrefix + "rack": model.LabelValue(server.Rack), dedicatedServerLabelPrefix + "no_intervention": model.LabelValue(strconv.FormatBool(server.NoIntervention)), dedicatedServerLabelPrefix + "os": model.LabelValue(server.Os), dedicatedServerLabelPrefix + "support_level": model.LabelValue(server.SupportLevel), - dedicatedServerLabelPrefix + "server_id": model.LabelValue(fmt.Sprintf("%d", server.ServerID)), + dedicatedServerLabelPrefix + "server_id": model.LabelValue(strconv.FormatInt(server.ServerID, 10)), dedicatedServerLabelPrefix + "reverse": model.LabelValue(server.Reverse), dedicatedServerLabelPrefix + "datacenter": model.LabelValue(server.Datacenter), dedicatedServerLabelPrefix + "name": model.LabelValue(server.Name), diff --git a/discovery/ovhcloud/vps.go b/discovery/ovhcloud/vps.go index e2d1dee364..58ceeabd87 100644 --- a/discovery/ovhcloud/vps.go +++ b/discovery/ovhcloud/vps.go @@ -19,6 +19,7 @@ import ( "net/netip" "net/url" "path" + "strconv" "github.com/go-kit/log" "github.com/go-kit/log/level" @@ -161,21 +162,21 @@ func (d *vpsDiscovery) refresh(context.Context) ([]*targetgroup.Group, error) { model.InstanceLabel: model.LabelValue(server.Name), vpsLabelPrefix + "offer": model.LabelValue(server.Model.Offer), vpsLabelPrefix + "datacenter": model.LabelValue(fmt.Sprintf("%+v", server.Model.Datacenter)), - vpsLabelPrefix + "model_vcore": model.LabelValue(fmt.Sprintf("%d", server.Model.Vcore)), - vpsLabelPrefix + "maximum_additional_ip": model.LabelValue(fmt.Sprintf("%d", server.Model.MaximumAdditionalIP)), + vpsLabelPrefix + "model_vcore": model.LabelValue(strconv.Itoa(server.Model.Vcore)), + vpsLabelPrefix + "maximum_additional_ip": model.LabelValue(strconv.Itoa(server.Model.MaximumAdditionalIP)), vpsLabelPrefix + "version": model.LabelValue(server.Model.Version), vpsLabelPrefix + "model_name": model.LabelValue(server.Model.Name), - vpsLabelPrefix + "disk": model.LabelValue(fmt.Sprintf("%d", server.Model.Disk)), - vpsLabelPrefix + "memory": model.LabelValue(fmt.Sprintf("%d", server.Model.Memory)), + vpsLabelPrefix + "disk": model.LabelValue(strconv.Itoa(server.Model.Disk)), + vpsLabelPrefix + "memory": model.LabelValue(strconv.Itoa(server.Model.Memory)), vpsLabelPrefix + "zone": model.LabelValue(server.Zone), vpsLabelPrefix + "display_name": model.LabelValue(server.DisplayName), vpsLabelPrefix + "cluster": model.LabelValue(server.Cluster), vpsLabelPrefix + "state": model.LabelValue(server.State), vpsLabelPrefix + "name": model.LabelValue(server.Name), vpsLabelPrefix + "netboot_mode": model.LabelValue(server.NetbootMode), - vpsLabelPrefix + "memory_limit": model.LabelValue(fmt.Sprintf("%d", server.MemoryLimit)), + vpsLabelPrefix + "memory_limit": model.LabelValue(strconv.Itoa(server.MemoryLimit)), vpsLabelPrefix + "offer_type": model.LabelValue(server.OfferType), - vpsLabelPrefix + "vcore": model.LabelValue(fmt.Sprintf("%d", server.Vcore)), + vpsLabelPrefix + "vcore": model.LabelValue(strconv.Itoa(server.Vcore)), vpsLabelPrefix + "ipv4": model.LabelValue(ipv4), vpsLabelPrefix + "ipv6": model.LabelValue(ipv6), } diff --git a/discovery/puppetdb/puppetdb.go b/discovery/puppetdb/puppetdb.go index 8c9ccde0a4..8f89acbf93 100644 --- a/discovery/puppetdb/puppetdb.go +++ b/discovery/puppetdb/puppetdb.go @@ -237,7 +237,7 @@ func (d *Discovery) refresh(ctx context.Context) ([]*targetgroup.Group, error) { pdbLabelResource: model.LabelValue(resource.Resource), pdbLabelType: model.LabelValue(resource.Type), pdbLabelTitle: model.LabelValue(resource.Title), - pdbLabelExported: model.LabelValue(fmt.Sprintf("%t", resource.Exported)), + pdbLabelExported: model.LabelValue(strconv.FormatBool(resource.Exported)), pdbLabelFile: model.LabelValue(resource.File), pdbLabelEnvironment: model.LabelValue(resource.Environment), } diff --git a/discovery/uyuni/uyuni.go b/discovery/uyuni/uyuni.go index e885ef2e80..c8af2f1587 100644 --- a/discovery/uyuni/uyuni.go +++ b/discovery/uyuni/uyuni.go @@ -20,6 +20,7 @@ import ( "net/http" "net/url" "path" + "strconv" "strings" "time" @@ -269,7 +270,7 @@ func (d *Discovery) getEndpointLabels( model.AddressLabel: model.LabelValue(addr), uyuniLabelMinionHostname: model.LabelValue(networkInfo.Hostname), uyuniLabelPrimaryFQDN: model.LabelValue(networkInfo.PrimaryFQDN), - uyuniLablelSystemID: model.LabelValue(fmt.Sprintf("%d", endpoint.SystemID)), + uyuniLablelSystemID: model.LabelValue(strconv.Itoa(endpoint.SystemID)), uyuniLablelGroups: model.LabelValue(strings.Join(managedGroupNames, d.separator)), uyuniLablelEndpointName: model.LabelValue(endpoint.EndpointName), uyuniLablelExporter: model.LabelValue(endpoint.ExporterName), diff --git a/discovery/zookeeper/zookeeper.go b/discovery/zookeeper/zookeeper.go index 303c7ca6d0..92904dd71c 100644 --- a/discovery/zookeeper/zookeeper.go +++ b/discovery/zookeeper/zookeeper.go @@ -280,17 +280,17 @@ func parseServersetMember(data []byte, path string) (model.LabelSet, error) { labels := model.LabelSet{} labels[serversetPathLabel] = model.LabelValue(path) labels[model.AddressLabel] = model.LabelValue( - net.JoinHostPort(member.ServiceEndpoint.Host, fmt.Sprintf("%d", member.ServiceEndpoint.Port))) + net.JoinHostPort(member.ServiceEndpoint.Host, strconv.Itoa(member.ServiceEndpoint.Port))) labels[serversetEndpointLabelPrefix+"_host"] = model.LabelValue(member.ServiceEndpoint.Host) - labels[serversetEndpointLabelPrefix+"_port"] = model.LabelValue(fmt.Sprintf("%d", member.ServiceEndpoint.Port)) + labels[serversetEndpointLabelPrefix+"_port"] = model.LabelValue(strconv.Itoa(member.ServiceEndpoint.Port)) for name, endpoint := range member.AdditionalEndpoints { cleanName := model.LabelName(strutil.SanitizeLabelName(name)) labels[serversetEndpointLabelPrefix+"_host_"+cleanName] = model.LabelValue( endpoint.Host) labels[serversetEndpointLabelPrefix+"_port_"+cleanName] = model.LabelValue( - fmt.Sprintf("%d", endpoint.Port)) + strconv.Itoa(endpoint.Port)) } labels[serversetStatusLabel] = model.LabelValue(member.Status) @@ -321,10 +321,10 @@ func parseNerveMember(data []byte, path string) (model.LabelSet, error) { labels := model.LabelSet{} labels[nervePathLabel] = model.LabelValue(path) labels[model.AddressLabel] = model.LabelValue( - net.JoinHostPort(member.Host, fmt.Sprintf("%d", member.Port))) + net.JoinHostPort(member.Host, strconv.Itoa(member.Port))) labels[nerveEndpointLabelPrefix+"_host"] = model.LabelValue(member.Host) - labels[nerveEndpointLabelPrefix+"_port"] = model.LabelValue(fmt.Sprintf("%d", member.Port)) + labels[nerveEndpointLabelPrefix+"_port"] = model.LabelValue(strconv.Itoa(member.Port)) labels[nerveEndpointLabelPrefix+"_name"] = model.LabelValue(member.Name) return labels, nil diff --git a/docs/command-line/prometheus.md b/docs/command-line/prometheus.md index 93eaf251d0..aa9bf3bfb0 100644 --- a/docs/command-line/prometheus.md +++ b/docs/command-line/prometheus.md @@ -48,7 +48,7 @@ The Prometheus monitoring server | --rules.alert.for-outage-tolerance | Max time to tolerate prometheus outage for restoring "for" state of alert. Use with server mode only. | `1h` | | --rules.alert.for-grace-period | Minimum duration between alert and restored "for" state. This is maintained only for alerts with configured "for" time greater than grace period. Use with server mode only. | `10m` | | --rules.alert.resend-delay | Minimum amount of time to wait before resending an alert to Alertmanager. Use with server mode only. | `1m` | -| --rules.max-concurrent-evals | Global concurrency limit for independent rules that can run concurrently. Use with server mode only. | `4` | +| --rules.max-concurrent-evals | Global concurrency limit for independent rules that can run concurrently. When set, "query.max-concurrency" may need to be adjusted accordingly. Use with server mode only. | `4` | | --alertmanager.notification-queue-capacity | The capacity of the queue for pending Alertmanager notifications. Use with server mode only. | `10000` | | --query.lookback-delta | The maximum lookback duration for retrieving metrics during expression evaluations and federation. Use with server mode only. | `5m` | | --query.timeout | Maximum time a query may take before being aborted. Use with server mode only. | `2m` | diff --git a/docs/command-line/promtool.md b/docs/command-line/promtool.md index 3eceed48f2..443cd3f0cb 100644 --- a/docs/command-line/promtool.md +++ b/docs/command-line/promtool.md @@ -566,6 +566,7 @@ Dump samples from a TSDB. | Flag | Description | Default | | --- | --- | --- | +| --sandbox-dir-root | Root directory where a sandbox directory would be created in case WAL replay generates chunks. The sandbox directory is cleaned up at the end. | `data/` | | --min-time | Minimum timestamp to dump. | `-9223372036854775808` | | --max-time | Maximum timestamp to dump. | `9223372036854775807` | | --match | Series selector. Can be specified multiple times. | `{__name__=~'(?s:.*)'}` | @@ -584,7 +585,7 @@ Dump samples from a TSDB. ##### `promtool tsdb dump-openmetrics` -[Experimental] Dump samples from a TSDB into OpenMetrics format. Native histograms are not dumped. +[Experimental] Dump samples from a TSDB into OpenMetrics text format, excluding native histograms and staleness markers, which are not representable in OpenMetrics. @@ -592,6 +593,7 @@ Dump samples from a TSDB. | Flag | Description | Default | | --- | --- | --- | +| --sandbox-dir-root | Root directory where a sandbox directory would be created in case WAL replay generates chunks. The sandbox directory is cleaned up at the end. | `data/` | | --min-time | Minimum timestamp to dump. | `-9223372036854775808` | | --max-time | Maximum timestamp to dump. | `9223372036854775807` | | --match | Series selector. Can be specified multiple times. | `{__name__=~'(?s:.*)'}` | diff --git a/docs/configuration/configuration.md b/docs/configuration/configuration.md index a90defc785..a8fc9c626e 100644 --- a/docs/configuration/configuration.md +++ b/docs/configuration/configuration.md @@ -70,6 +70,10 @@ global: # How frequently to evaluate rules. [ evaluation_interval: | default = 1m ] + + # Offset the rule evaluation timestamp of this particular group by the specified duration into the past to ensure the underlying metrics have been received. + # Metric availability delays are more likely to occur when Prometheus is running as a remote write target, but can also occur when there's anomalies with scraping. + [ rule_query_offset: | default = 0s ] # The labels to add to any time series or alerts when communicating with # external systems (federation, remote storage, Alertmanager). @@ -1349,7 +1353,7 @@ interface. The following meta labels are available on targets during [relabeling](#relabel_config): * `__meta_openstack_address_pool`: the pool of the private IP. -* `__meta_openstack_instance_flavor`: the flavor of the OpenStack instance. +* `__meta_openstack_instance_flavor`: the flavor ID of the OpenStack instance. * `__meta_openstack_instance_id`: the OpenStack instance ID. * `__meta_openstack_instance_image`: the ID of the image the OpenStack instance is using. * `__meta_openstack_instance_name`: the OpenStack instance name. @@ -1357,7 +1361,7 @@ The following meta labels are available on targets during [relabeling](#relabel_ * `__meta_openstack_private_ip`: the private IP of the OpenStack instance. * `__meta_openstack_project_id`: the project (tenant) owning this instance. * `__meta_openstack_public_ip`: the public IP of the OpenStack instance. -* `__meta_openstack_tag_`: each tag value of the instance. +* `__meta_openstack_tag_`: each metadata item of the instance, with any unsupported characters converted to an underscore. * `__meta_openstack_user_id`: the user account owning the tenant. See below for the configuration options for OpenStack discovery: @@ -1467,6 +1471,7 @@ For OVHcloud's [public cloud instances](https://www.ovhcloud.com/en/public-cloud * `__meta_ovhcloud_dedicated_server_ipv6`: the IPv6 of the server * `__meta_ovhcloud_dedicated_server_link_speed`: the link speed of the server * `__meta_ovhcloud_dedicated_server_name`: the name of the server +* `__meta_ovhcloud_dedicated_server_no_intervention`: whether datacenter intervention is disabled for the server * `__meta_ovhcloud_dedicated_server_os`: the operating system of the server * `__meta_ovhcloud_dedicated_server_rack`: the rack of the server * `__meta_ovhcloud_dedicated_server_reverse`: the reverse DNS name of the server @@ -3673,7 +3678,8 @@ queue_config: [ min_shards: | default = 1 ] # Maximum number of samples per send. [ max_samples_per_send: | default = 2000] - # Maximum time a sample will wait in buffer. + # Maximum time a sample will wait for a send. The sample might wait less + # if the buffer is full. Further time might pass due to potential retries. [ batch_send_deadline: | default = 5s ] # Initial retry delay. Gets doubled for every retry. [ min_backoff: | default = 30ms ] diff --git a/docs/configuration/recording_rules.md b/docs/configuration/recording_rules.md index 48ab951f94..9aa226bbc0 100644 --- a/docs/configuration/recording_rules.md +++ b/docs/configuration/recording_rules.md @@ -86,6 +86,9 @@ name: # rule can produce. 0 is no limit. [ limit: | default = 0 ] +# Offset the rule evaluation timestamp of this particular group by the specified duration into the past. +[ query_offset: | default = global.rule_query_offset ] + rules: [ - ... ] ``` @@ -148,6 +151,9 @@ the rule, active, pending, or inactive, are cleared as well. The event will be recorded as an error in the evaluation, and as such no stale markers are written. +# Rule query offset +This is useful to ensure the underlying metrics have been received and stored in Prometheus. Metric availability delays are more likely to occur when Prometheus is running as a remote write target due to the nature of distributed systems, but can also occur when there's anomalies with scraping and/or short evaluation intervals. + # Failed rule evaluations due to slow evaluation If a rule group hasn't finished evaluating before its next evaluation is supposed to start (as defined by the `evaluation_interval`), the next evaluation will be skipped. Subsequent evaluations of the rule group will continue to be skipped until the initial evaluation either completes or times out. When this happens, there will be a gap in the metric produced by the recording rule. The `rule_group_iterations_missed_total` metric will be incremented for each missed iteration of the rule group. diff --git a/docs/storage.md b/docs/storage.md index 46bb7210e0..b66f2062af 100644 --- a/docs/storage.md +++ b/docs/storage.md @@ -197,6 +197,9 @@ or time-series database to Prometheus. To do so, the user must first convert the source data into [OpenMetrics](https://openmetrics.io/) format, which is the input format for the backfilling as described below. +Note that native histograms and staleness markers are not supported by this +procedure, as they cannot be represented in the OpenMetrics format. + ### Usage Backfilling can be used via the Promtool command line. Promtool will write the blocks diff --git a/documentation/examples/custom-sd/adapter-usage/main.go b/documentation/examples/custom-sd/adapter-usage/main.go index bfbca7b70c..8ccbafe6f1 100644 --- a/documentation/examples/custom-sd/adapter-usage/main.go +++ b/documentation/examples/custom-sd/adapter-usage/main.go @@ -127,9 +127,9 @@ func (d *discovery) parseServiceNodes(resp *http.Response, name string) (*target // since the service may be registered remotely through a different node. var addr string if node.ServiceAddress != "" { - addr = net.JoinHostPort(node.ServiceAddress, fmt.Sprintf("%d", node.ServicePort)) + addr = net.JoinHostPort(node.ServiceAddress, strconv.Itoa(node.ServicePort)) } else { - addr = net.JoinHostPort(node.Address, fmt.Sprintf("%d", node.ServicePort)) + addr = net.JoinHostPort(node.Address, strconv.Itoa(node.ServicePort)) } target := model.LabelSet{model.AddressLabel: model.LabelValue(addr)} diff --git a/documentation/prometheus-mixin/config.libsonnet b/documentation/prometheus-mixin/config.libsonnet index ab9079a5e3..70d46a2212 100644 --- a/documentation/prometheus-mixin/config.libsonnet +++ b/documentation/prometheus-mixin/config.libsonnet @@ -44,5 +44,10 @@ // The default refresh time for all dashboards, default to 60s refresh: '60s', }, + + // Opt-out of multi-cluster dashboards by overriding this. + showMultiCluster: true, + // The cluster label to infer the cluster name from. + clusterLabel: 'cluster', }, } diff --git a/documentation/prometheus-mixin/dashboards.libsonnet b/documentation/prometheus-mixin/dashboards.libsonnet index efe53dbac9..2bdd168cc9 100644 --- a/documentation/prometheus-mixin/dashboards.libsonnet +++ b/documentation/prometheus-mixin/dashboards.libsonnet @@ -10,21 +10,32 @@ local template = grafana.template; { grafanaDashboards+:: { 'prometheus.json': - g.dashboard( + local showMultiCluster = $._config.showMultiCluster; + local dashboard = g.dashboard( '%(prefix)sOverview' % $._config.grafanaPrometheus - ) - .addMultiTemplate('cluster', 'prometheus_build_info{%(prometheusSelector)s}' % $._config, 'cluster') - .addMultiTemplate('job', 'prometheus_build_info{cluster=~"$cluster"}', 'job') - .addMultiTemplate('instance', 'prometheus_build_info{cluster=~"$cluster", job=~"$job"}', 'instance') + ); + local templatedDashboard = if showMultiCluster then + dashboard + .addMultiTemplate('cluster', 'prometheus_build_info{%(prometheusSelector)s}' % $._config, $._config.clusterLabel) + .addMultiTemplate('job', 'prometheus_build_info{cluster=~"$cluster"}', 'job') + .addMultiTemplate('instance', 'prometheus_build_info{cluster=~"$cluster", job=~"$job"}', 'instance') + else + dashboard + .addMultiTemplate('job', 'prometheus_build_info{%(prometheusSelector)s}' % $._config, 'job') + .addMultiTemplate('instance', 'prometheus_build_info{job=~"$job"}', 'instance'); + templatedDashboard .addRow( g.row('Prometheus Stats') .addPanel( g.panel('Prometheus Stats') + - g.tablePanel([ + g.tablePanel(if showMultiCluster then [ 'count by (cluster, job, instance, version) (prometheus_build_info{cluster=~"$cluster", job=~"$job", instance=~"$instance"})', 'max by (cluster, job, instance) (time() - process_start_time_seconds{cluster=~"$cluster", job=~"$job", instance=~"$instance"})', + ] else [ + 'count by (job, instance, version) (prometheus_build_info{job=~"$job", instance=~"$instance"})', + 'max by (job, instance) (time() - process_start_time_seconds{job=~"$job", instance=~"$instance"})', ], { - cluster: { alias: 'Cluster' }, + cluster: { alias: if showMultiCluster then 'Cluster' else '' }, job: { alias: 'Job' }, instance: { alias: 'Instance' }, version: { alias: 'Version' }, @@ -37,12 +48,18 @@ local template = grafana.template; g.row('Discovery') .addPanel( g.panel('Target Sync') + - g.queryPanel('sum(rate(prometheus_target_sync_length_seconds_sum{cluster=~"$cluster",job=~"$job",instance=~"$instance"}[5m])) by (cluster, job, scrape_job, instance) * 1e3', '{{cluster}}:{{job}}:{{instance}}:{{scrape_job}}') + + g.queryPanel(if showMultiCluster then 'sum(rate(prometheus_target_sync_length_seconds_sum{cluster=~"$cluster",job=~"$job",instance=~"$instance"}[5m])) by (cluster, job, scrape_job, instance) * 1e3' + else 'sum(rate(prometheus_target_sync_length_seconds_sum{job=~"$job",instance=~"$instance"}[5m])) by (scrape_job) * 1e3', + if showMultiCluster then '{{cluster}}:{{job}}:{{instance}}:{{scrape_job}}' + else '{{scrape_job}}') + { yaxes: g.yaxes('ms') } ) .addPanel( g.panel('Targets') + - g.queryPanel('sum by (cluster, job, instance) (prometheus_sd_discovered_targets{cluster=~"$cluster", job=~"$job",instance=~"$instance"})', '{{cluster}}:{{job}}:{{instance}}') + + g.queryPanel(if showMultiCluster then 'sum by (cluster, job, instance) (prometheus_sd_discovered_targets{cluster=~"$cluster", job=~"$job",instance=~"$instance"})' + else 'sum(prometheus_sd_discovered_targets{job=~"$job",instance=~"$instance"})', + if showMultiCluster then '{{cluster}}:{{job}}:{{instance}}' + else 'Targets') + g.stack ) ) @@ -50,29 +67,47 @@ local template = grafana.template; g.row('Retrieval') .addPanel( g.panel('Average Scrape Interval Duration') + - g.queryPanel('rate(prometheus_target_interval_length_seconds_sum{cluster=~"$cluster", job=~"$job",instance=~"$instance"}[5m]) / rate(prometheus_target_interval_length_seconds_count{cluster=~"$cluster", job=~"$job",instance=~"$instance"}[5m]) * 1e3', '{{cluster}}:{{job}}:{{instance}} {{interval}} configured') + + g.queryPanel(if showMultiCluster then 'rate(prometheus_target_interval_length_seconds_sum{cluster=~"$cluster", job=~"$job",instance=~"$instance"}[5m]) / rate(prometheus_target_interval_length_seconds_count{cluster=~"$cluster", job=~"$job",instance=~"$instance"}[5m]) * 1e3' + else 'rate(prometheus_target_interval_length_seconds_sum{job=~"$job",instance=~"$instance"}[5m]) / rate(prometheus_target_interval_length_seconds_count{job=~"$job",instance=~"$instance"}[5m]) * 1e3', + if showMultiCluster then '{{cluster}}:{{job}}:{{instance}} {{interval}} configured' + else '{{interval}} configured') + { yaxes: g.yaxes('ms') } ) .addPanel( g.panel('Scrape failures') + - g.queryPanel([ + g.queryPanel(if showMultiCluster then [ 'sum by (cluster, job, instance) (rate(prometheus_target_scrapes_exceeded_body_size_limit_total{cluster=~"$cluster",job=~"$job",instance=~"$instance"}[1m]))', 'sum by (cluster, job, instance) (rate(prometheus_target_scrapes_exceeded_sample_limit_total{cluster=~"$cluster",job=~"$job",instance=~"$instance"}[1m]))', 'sum by (cluster, job, instance) (rate(prometheus_target_scrapes_sample_duplicate_timestamp_total{cluster=~"$cluster",job=~"$job",instance=~"$instance"}[1m]))', 'sum by (cluster, job, instance) (rate(prometheus_target_scrapes_sample_out_of_bounds_total{cluster=~"$cluster",job=~"$job",instance=~"$instance"}[1m]))', 'sum by (cluster, job, instance) (rate(prometheus_target_scrapes_sample_out_of_order_total{cluster=~"$cluster",job=~"$job",instance=~"$instance"}[1m]))', - ], [ + ] else [ + 'sum by (job) (rate(prometheus_target_scrapes_exceeded_body_size_limit_total[1m]))', + 'sum by (job) (rate(prometheus_target_scrapes_exceeded_sample_limit_total[1m]))', + 'sum by (job) (rate(prometheus_target_scrapes_sample_duplicate_timestamp_total[1m]))', + 'sum by (job) (rate(prometheus_target_scrapes_sample_out_of_bounds_total[1m]))', + 'sum by (job) (rate(prometheus_target_scrapes_sample_out_of_order_total[1m]))', + ], if showMultiCluster then [ 'exceeded body size limit: {{cluster}} {{job}} {{instance}}', 'exceeded sample limit: {{cluster}} {{job}} {{instance}}', 'duplicate timestamp: {{cluster}} {{job}} {{instance}}', 'out of bounds: {{cluster}} {{job}} {{instance}}', 'out of order: {{cluster}} {{job}} {{instance}}', + ] else [ + 'exceeded body size limit: {{job}}', + 'exceeded sample limit: {{job}}', + 'duplicate timestamp: {{job}}', + 'out of bounds: {{job}}', + 'out of order: {{job}}', ]) + g.stack ) .addPanel( g.panel('Appended Samples') + - g.queryPanel('rate(prometheus_tsdb_head_samples_appended_total{cluster=~"$cluster", job=~"$job",instance=~"$instance"}[5m])', '{{cluster}} {{job}} {{instance}}') + + g.queryPanel(if showMultiCluster then 'rate(prometheus_tsdb_head_samples_appended_total{cluster=~"$cluster", job=~"$job",instance=~"$instance"}[5m])' + else 'rate(prometheus_tsdb_head_samples_appended_total{job=~"$job",instance=~"$instance"}[5m])', + if showMultiCluster then '{{cluster}} {{job}} {{instance}}' + else '{{job}} {{instance}}') + g.stack ) ) @@ -80,12 +115,18 @@ local template = grafana.template; g.row('Storage') .addPanel( g.panel('Head Series') + - g.queryPanel('prometheus_tsdb_head_series{cluster=~"$cluster",job=~"$job",instance=~"$instance"}', '{{cluster}} {{job}} {{instance}} head series') + + g.queryPanel(if showMultiCluster then 'prometheus_tsdb_head_series{cluster=~"$cluster",job=~"$job",instance=~"$instance"}' + else 'prometheus_tsdb_head_series{job=~"$job",instance=~"$instance"}', + if showMultiCluster then '{{cluster}} {{job}} {{instance}} head series' + else '{{job}} {{instance}} head series') + g.stack ) .addPanel( g.panel('Head Chunks') + - g.queryPanel('prometheus_tsdb_head_chunks{cluster=~"$cluster",job=~"$job",instance=~"$instance"}', '{{cluster}} {{job}} {{instance}} head chunks') + + g.queryPanel(if showMultiCluster then 'prometheus_tsdb_head_chunks{cluster=~"$cluster",job=~"$job",instance=~"$instance"}' + else 'prometheus_tsdb_head_chunks{job=~"$job",instance=~"$instance"}', + if showMultiCluster then '{{cluster}} {{job}} {{instance}} head chunks' + else '{{job}} {{instance}} head chunks') + g.stack ) ) @@ -93,12 +134,18 @@ local template = grafana.template; g.row('Query') .addPanel( g.panel('Query Rate') + - g.queryPanel('rate(prometheus_engine_query_duration_seconds_count{cluster=~"$cluster",job=~"$job",instance=~"$instance",slice="inner_eval"}[5m])', '{{cluster}} {{job}} {{instance}}') + + g.queryPanel(if showMultiCluster then 'rate(prometheus_engine_query_duration_seconds_count{cluster=~"$cluster",job=~"$job",instance=~"$instance",slice="inner_eval"}[5m])' + else 'rate(prometheus_engine_query_duration_seconds_count{job=~"$job",instance=~"$instance",slice="inner_eval"}[5m])', + if showMultiCluster then '{{cluster}} {{job}} {{instance}}' + else '{{job}} {{instance}}') + g.stack, ) .addPanel( g.panel('Stage Duration') + - g.queryPanel('max by (slice) (prometheus_engine_query_duration_seconds{quantile="0.9",cluster=~"$cluster", job=~"$job",instance=~"$instance"}) * 1e3', '{{slice}}') + + g.queryPanel(if showMultiCluster then 'max by (slice) (prometheus_engine_query_duration_seconds{quantile="0.9",cluster=~"$cluster", job=~"$job",instance=~"$instance"}) * 1e3' + else 'max by (slice) (prometheus_engine_query_duration_seconds{quantile="0.9",job=~"$job",instance=~"$instance"}) * 1e3', + if showMultiCluster then '{{slice}}' + else '{{slice}}') + { yaxes: g.yaxes('ms') } + g.stack, ) diff --git a/model/histogram/float_histogram_test.go b/model/histogram/float_histogram_test.go index 49fb77ab0b..759da6540c 100644 --- a/model/histogram/float_histogram_test.go +++ b/model/histogram/float_histogram_test.go @@ -14,9 +14,9 @@ package histogram import ( - "fmt" "math" "math/rand" + "strconv" "testing" "github.com/stretchr/testify/require" @@ -2134,7 +2134,7 @@ func TestAllFloatBucketIterator(t *testing.T) { } for i, c := range cases { - t.Run(fmt.Sprintf("%d", i), func(t *testing.T) { + t.Run(strconv.Itoa(i), func(t *testing.T) { var expBuckets, actBuckets []Bucket[float64] if c.includeNeg { @@ -2360,7 +2360,7 @@ func TestAllReverseFloatBucketIterator(t *testing.T) { } for i, c := range cases { - t.Run(fmt.Sprintf("%d", i), func(t *testing.T) { + t.Run(strconv.Itoa(i), func(t *testing.T) { var expBuckets, actBuckets []Bucket[float64] if c.includePos { diff --git a/model/histogram/histogram_test.go b/model/histogram/histogram_test.go index 14a948e644..d1a074135c 100644 --- a/model/histogram/histogram_test.go +++ b/model/histogram/histogram_test.go @@ -14,8 +14,8 @@ package histogram import ( - "fmt" "math" + "strconv" "testing" "github.com/stretchr/testify/require" @@ -72,7 +72,7 @@ func TestHistogramString(t *testing.T) { } for i, c := range cases { - t.Run(fmt.Sprintf("%d", i), func(t *testing.T) { + t.Run(strconv.Itoa(i), func(t *testing.T) { actualString := c.histogram.String() require.Equal(t, c.expectedString, actualString) }) @@ -211,7 +211,7 @@ func TestCumulativeBucketIterator(t *testing.T) { } for i, c := range cases { - t.Run(fmt.Sprintf("%d", i), func(t *testing.T) { + t.Run(strconv.Itoa(i), func(t *testing.T) { it := c.histogram.CumulativeBucketIterator() actualBuckets := make([]Bucket[uint64], 0, len(c.expectedBuckets)) for it.Next() { @@ -371,7 +371,7 @@ func TestRegularBucketIterator(t *testing.T) { } for i, c := range cases { - t.Run(fmt.Sprintf("%d", i), func(t *testing.T) { + t.Run(strconv.Itoa(i), func(t *testing.T) { it := c.histogram.PositiveBucketIterator() actualPositiveBuckets := make([]Bucket[uint64], 0, len(c.expectedPositiveBuckets)) for it.Next() { diff --git a/model/labels/labels_test.go b/model/labels/labels_test.go index 3d6e7659f4..6464d007d3 100644 --- a/model/labels/labels_test.go +++ b/model/labels/labels_test.go @@ -17,6 +17,7 @@ import ( "encoding/json" "fmt" "net/http" + "strconv" "strings" "testing" @@ -732,7 +733,7 @@ func TestScratchBuilder(t *testing.T) { want: FromStrings("ddd", "444"), }, } { - t.Run(fmt.Sprint(i), func(t *testing.T) { + t.Run(strconv.Itoa(i), func(t *testing.T) { b := NewScratchBuilder(len(tcase.add)) for _, lbl := range tcase.add { b.Add(lbl.Name, lbl.Value) diff --git a/model/labels/matcher.go b/model/labels/matcher.go index 1282f80d63..8e220e392d 100644 --- a/model/labels/matcher.go +++ b/model/labels/matcher.go @@ -14,7 +14,8 @@ package labels import ( - "fmt" + "bytes" + "strconv" ) // MatchType is an enum for label matching types. @@ -78,7 +79,29 @@ func MustNewMatcher(mt MatchType, name, val string) *Matcher { } func (m *Matcher) String() string { - return fmt.Sprintf("%s%s%q", m.Name, m.Type, m.Value) + // Start a buffer with a pre-allocated size on stack to cover most needs. + var bytea [1024]byte + b := bytes.NewBuffer(bytea[:0]) + + if m.shouldQuoteName() { + b.Write(strconv.AppendQuote(b.AvailableBuffer(), m.Name)) + } else { + b.WriteString(m.Name) + } + b.WriteString(m.Type.String()) + b.Write(strconv.AppendQuote(b.AvailableBuffer(), m.Value)) + + return b.String() +} + +func (m *Matcher) shouldQuoteName() bool { + for i, c := range m.Name { + if c == '_' || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || (i > 0 && c >= '0' && c <= '9') { + continue + } + return true + } + return false } // Matches returns whether the matcher matches the given string value. diff --git a/model/labels/matcher_test.go b/model/labels/matcher_test.go index c23deafe61..ff39d40d0f 100644 --- a/model/labels/matcher_test.go +++ b/model/labels/matcher_test.go @@ -15,6 +15,7 @@ package labels import ( "fmt" + "math/rand" "testing" "github.com/stretchr/testify/require" @@ -225,3 +226,128 @@ func BenchmarkNewMatcher(b *testing.B) { } }) } + +func BenchmarkMatcher_String(b *testing.B) { + type benchCase struct { + name string + matchers []*Matcher + } + cases := []benchCase{ + { + name: "short name equal", + matchers: []*Matcher{ + MustNewMatcher(MatchEqual, "foo", "bar"), + MustNewMatcher(MatchEqual, "bar", "baz"), + MustNewMatcher(MatchEqual, "abc", "def"), + MustNewMatcher(MatchEqual, "ghi", "klm"), + MustNewMatcher(MatchEqual, "nop", "qrs"), + }, + }, + { + name: "short quoted name not equal", + matchers: []*Matcher{ + MustNewMatcher(MatchEqual, "f.o", "bar"), + MustNewMatcher(MatchEqual, "b.r", "baz"), + MustNewMatcher(MatchEqual, "a.c", "def"), + MustNewMatcher(MatchEqual, "g.i", "klm"), + MustNewMatcher(MatchEqual, "n.p", "qrs"), + }, + }, + { + name: "short quoted name with quotes not equal", + matchers: []*Matcher{ + MustNewMatcher(MatchEqual, `"foo"`, "bar"), + MustNewMatcher(MatchEqual, `"foo"`, "baz"), + MustNewMatcher(MatchEqual, `"foo"`, "def"), + MustNewMatcher(MatchEqual, `"foo"`, "klm"), + MustNewMatcher(MatchEqual, `"foo"`, "qrs"), + }, + }, + { + name: "short name value with quotes equal", + matchers: []*Matcher{ + MustNewMatcher(MatchEqual, "foo", `"bar"`), + MustNewMatcher(MatchEqual, "bar", `"baz"`), + MustNewMatcher(MatchEqual, "abc", `"def"`), + MustNewMatcher(MatchEqual, "ghi", `"klm"`), + MustNewMatcher(MatchEqual, "nop", `"qrs"`), + }, + }, + { + name: "short name and long value regexp", + matchers: []*Matcher{ + MustNewMatcher(MatchRegexp, "foo", "five_six_seven_eight_nine_ten_one_two_three_four"), + MustNewMatcher(MatchRegexp, "bar", "one_two_three_four_five_six_seven_eight_nine_ten"), + MustNewMatcher(MatchRegexp, "abc", "two_three_four_five_six_seven_eight_nine_ten_one"), + MustNewMatcher(MatchRegexp, "ghi", "three_four_five_six_seven_eight_nine_ten_one_two"), + MustNewMatcher(MatchRegexp, "nop", "four_five_six_seven_eight_nine_ten_one_two_three"), + }, + }, + { + name: "short name and long value with quotes equal", + matchers: []*Matcher{ + MustNewMatcher(MatchEqual, "foo", `five_six_seven_eight_nine_ten_"one"_two_three_four`), + MustNewMatcher(MatchEqual, "bar", `one_two_three_four_five_six_"seven"_eight_nine_ten`), + MustNewMatcher(MatchEqual, "abc", `two_three_four_five_six_seven_"eight"_nine_ten_one`), + MustNewMatcher(MatchEqual, "ghi", `three_four_five_six_seven_eight_"nine"_ten_one_two`), + MustNewMatcher(MatchEqual, "nop", `four_five_six_seven_eight_nine_"ten"_one_two_three`), + }, + }, + { + name: "long name regexp", + matchers: []*Matcher{ + MustNewMatcher(MatchRegexp, "one_two_three_four_five_six_seven_eight_nine_ten", "val"), + MustNewMatcher(MatchRegexp, "two_three_four_five_six_seven_eight_nine_ten_one", "val"), + MustNewMatcher(MatchRegexp, "three_four_five_six_seven_eight_nine_ten_one_two", "val"), + MustNewMatcher(MatchRegexp, "four_five_six_seven_eight_nine_ten_one_two_three", "val"), + MustNewMatcher(MatchRegexp, "five_six_seven_eight_nine_ten_one_two_three_four", "val"), + }, + }, + { + name: "long quoted name regexp", + matchers: []*Matcher{ + MustNewMatcher(MatchRegexp, "one.two.three.four.five.six.seven.eight.nine.ten", "val"), + MustNewMatcher(MatchRegexp, "two.three.four.five.six.seven.eight.nine.ten.one", "val"), + MustNewMatcher(MatchRegexp, "three.four.five.six.seven.eight.nine.ten.one.two", "val"), + MustNewMatcher(MatchRegexp, "four.five.six.seven.eight.nine.ten.one.two.three", "val"), + MustNewMatcher(MatchRegexp, "five.six.seven.eight.nine.ten.one.two.three.four", "val"), + }, + }, + { + name: "long name and long value regexp", + matchers: []*Matcher{ + MustNewMatcher(MatchRegexp, "one_two_three_four_five_six_seven_eight_nine_ten", "five_six_seven_eight_nine_ten_one_two_three_four"), + MustNewMatcher(MatchRegexp, "two_three_four_five_six_seven_eight_nine_ten_one", "one_two_three_four_five_six_seven_eight_nine_ten"), + MustNewMatcher(MatchRegexp, "three_four_five_six_seven_eight_nine_ten_one_two", "two_three_four_five_six_seven_eight_nine_ten_one"), + MustNewMatcher(MatchRegexp, "four_five_six_seven_eight_nine_ten_one_two_three", "three_four_five_six_seven_eight_nine_ten_one_two"), + MustNewMatcher(MatchRegexp, "five_six_seven_eight_nine_ten_one_two_three_four", "four_five_six_seven_eight_nine_ten_one_two_three"), + }, + }, + { + name: "long quoted name and long value regexp", + matchers: []*Matcher{ + MustNewMatcher(MatchRegexp, "one.two.three.four.five.six.seven.eight.nine.ten", "five.six.seven.eight.nine.ten.one.two.three.four"), + MustNewMatcher(MatchRegexp, "two.three.four.five.six.seven.eight.nine.ten.one", "one.two.three.four.five.six.seven.eight.nine.ten"), + MustNewMatcher(MatchRegexp, "three.four.five.six.seven.eight.nine.ten.one.two", "two.three.four.five.six.seven.eight.nine.ten.one"), + MustNewMatcher(MatchRegexp, "four.five.six.seven.eight.nine.ten.one.two.three", "three.four.five.six.seven.eight.nine.ten.one.two"), + MustNewMatcher(MatchRegexp, "five.six.seven.eight.nine.ten.one.two.three.four", "four.five.six.seven.eight.nine.ten.one.two.three"), + }, + }, + } + + var mixed []*Matcher + for _, bc := range cases { + mixed = append(mixed, bc.matchers...) + } + rand.Shuffle(len(mixed), func(i, j int) { mixed[i], mixed[j] = mixed[j], mixed[i] }) + cases = append(cases, benchCase{name: "mixed", matchers: mixed}) + + for _, bc := range cases { + b.Run(bc.name, func(b *testing.B) { + for i := 0; i <= b.N; i++ { + m := bc.matchers[i%len(bc.matchers)] + _ = m.String() + } + }) + } +} diff --git a/model/labels/regexp.go b/model/labels/regexp.go index 79e340984a..b484e27168 100644 --- a/model/labels/regexp.go +++ b/model/labels/regexp.go @@ -828,7 +828,12 @@ type zeroOrOneCharacterStringMatcher struct { } func (m *zeroOrOneCharacterStringMatcher) Matches(s string) bool { - if moreThanOneRune(s) { + // If there's more than one rune in the string, then it can't match. + if r, size := utf8.DecodeRuneInString(s); r == utf8.RuneError { + // Size is 0 for empty strings, 1 for invalid rune. + // Empty string matches, invalid rune matches if there isn't anything else. + return size == len(s) + } else if size < len(s) { return false } @@ -840,27 +845,6 @@ func (m *zeroOrOneCharacterStringMatcher) Matches(s string) bool { return s[0] != '\n' } -// moreThanOneRune returns true if there are more than one runes in the string. -// It doesn't check whether the string is valid UTF-8. -// The return value should be always equal to utf8.RuneCountInString(s) > 1, -// but the function is optimized for the common case where the string prefix is ASCII. -func moreThanOneRune(s string) bool { - // If len(s) is exactly one or zero, there can't be more than one rune. - // Exit through this path quickly. - if len(s) <= 1 { - return false - } - - // There's one or more bytes: - // If first byte is ASCII then there are multiple runes if there are more bytes after that. - if s[0] < utf8.RuneSelf { - return len(s) > 1 - } - - // Less common case: first is a multibyte rune. - return utf8.RuneCountInString(s) > 1 -} - // trueMatcher is a stringMatcher which matches any string (always returns true). type trueMatcher struct{} diff --git a/model/labels/regexp_test.go b/model/labels/regexp_test.go index 47d3eeb4a2..1db90a473d 100644 --- a/model/labels/regexp_test.go +++ b/model/labels/regexp_test.go @@ -19,6 +19,7 @@ import ( "strings" "testing" "time" + "unicode/utf8" "github.com/grafana/regexp" "github.com/grafana/regexp/syntax" @@ -36,6 +37,7 @@ var ( ".*foo", "^.*foo$", "^.+foo$", + ".?", ".*", ".+", "foo.+", @@ -88,6 +90,12 @@ var ( // Values matching / not matching the test regexps on long alternations. "zQPbMkNO", "zQPbMkNo", "jyyfj00j0061", "jyyfj00j006", "jyyfj00j00612", "NNSPdvMi", "NNSPdvMiXXX", "NNSPdvMixxx", "nnSPdvMi", "nnSPdvMiXXX", + + // Invalid utf8 + "\xfefoo", + "foo\xfe", + "\xfd", + "\xff\xff", } ) @@ -926,19 +934,91 @@ func BenchmarkOptimizeEqualStringMatchers(b *testing.B) { } func TestZeroOrOneCharacterStringMatcher(t *testing.T) { - matcher := &zeroOrOneCharacterStringMatcher{matchNL: true} - require.True(t, matcher.Matches("")) - require.True(t, matcher.Matches("x")) - require.True(t, matcher.Matches("\n")) - require.False(t, matcher.Matches("xx")) - require.False(t, matcher.Matches("\n\n")) + t.Run("match newline", func(t *testing.T) { + matcher := &zeroOrOneCharacterStringMatcher{matchNL: true} + require.True(t, matcher.Matches("")) + require.True(t, matcher.Matches("x")) + require.True(t, matcher.Matches("\n")) + require.False(t, matcher.Matches("xx")) + require.False(t, matcher.Matches("\n\n")) + }) - matcher = &zeroOrOneCharacterStringMatcher{matchNL: false} - require.True(t, matcher.Matches("")) - require.True(t, matcher.Matches("x")) - require.False(t, matcher.Matches("\n")) - require.False(t, matcher.Matches("xx")) - require.False(t, matcher.Matches("\n\n")) + t.Run("do not match newline", func(t *testing.T) { + matcher := &zeroOrOneCharacterStringMatcher{matchNL: false} + require.True(t, matcher.Matches("")) + require.True(t, matcher.Matches("x")) + require.False(t, matcher.Matches("\n")) + require.False(t, matcher.Matches("xx")) + require.False(t, matcher.Matches("\n\n")) + }) + + t.Run("unicode", func(t *testing.T) { + // Just for documentation purposes, emoji1 is 1 rune, emoji2 is 2 runes. + // Having this in mind, will make future readers fixing tests easier. + emoji1 := "😀" + emoji2 := "❤️" + require.Equal(t, 1, utf8.RuneCountInString(emoji1)) + require.Equal(t, 2, utf8.RuneCountInString(emoji2)) + + matcher := &zeroOrOneCharacterStringMatcher{matchNL: true} + require.True(t, matcher.Matches(emoji1)) + require.False(t, matcher.Matches(emoji2)) + require.False(t, matcher.Matches(emoji1+emoji1)) + require.False(t, matcher.Matches("x"+emoji1)) + require.False(t, matcher.Matches(emoji1+"x")) + require.False(t, matcher.Matches(emoji1+emoji2)) + }) + + t.Run("invalid unicode", func(t *testing.T) { + // Just for reference, we also compare to what `^.?$` regular expression matches. + re := regexp.MustCompile("^.?$") + matcher := &zeroOrOneCharacterStringMatcher{matchNL: true} + + requireMatches := func(s string, expected bool) { + t.Helper() + require.Equal(t, expected, matcher.Matches(s)) + require.Equal(t, re.MatchString(s), matcher.Matches(s)) + } + + requireMatches("\xff", true) + requireMatches("x\xff", false) + requireMatches("\xffx", false) + requireMatches("\xff\xfe", false) + }) +} + +func BenchmarkZeroOrOneCharacterStringMatcher(b *testing.B) { + type benchCase struct { + str string + matches bool + } + + emoji1 := "😀" + emoji2 := "❤️" + cases := []benchCase{ + {"", true}, + {"x", true}, + {"\n", true}, + {"xx", false}, + {"\n\n", false}, + {emoji1, true}, + {emoji2, false}, + {emoji1 + emoji1, false}, + {strings.Repeat("x", 100), false}, + {strings.Repeat(emoji1, 100), false}, + {strings.Repeat(emoji2, 100), false}, + } + + matcher := &zeroOrOneCharacterStringMatcher{matchNL: true} + b.ResetTimer() + + for n := 0; n < b.N; n++ { + c := cases[n%len(cases)] + got := matcher.Matches(c.str) + if got != c.matches { + b.Fatalf("unexpected result for %q: got %t, want %t", c.str, got, c.matches) + } + } } func TestLiteralPrefixStringMatcher(t *testing.T) { diff --git a/model/relabel/relabel.go b/model/relabel/relabel.go index 4ddfa26c71..db2e6ce2ff 100644 --- a/model/relabel/relabel.go +++ b/model/relabel/relabel.go @@ -17,6 +17,7 @@ import ( "crypto/md5" "encoding/binary" "fmt" + "strconv" "strings" "github.com/grafana/regexp" @@ -290,7 +291,7 @@ func relabel(cfg *Config, lb *labels.Builder) (keep bool) { hash := md5.Sum([]byte(val)) // Use only the last 8 bytes of the hash to give the same result as earlier versions of this code. mod := binary.BigEndian.Uint64(hash[8:]) % cfg.Modulus - lb.Set(cfg.TargetLabel, fmt.Sprintf("%d", mod)) + lb.Set(cfg.TargetLabel, strconv.FormatUint(mod, 10)) case LabelMap: lb.Range(func(l labels.Label) { if cfg.Regex.MatchString(l.Name) { diff --git a/model/relabel/relabel_test.go b/model/relabel/relabel_test.go index 6798fb02a5..507ea101dd 100644 --- a/model/relabel/relabel_test.go +++ b/model/relabel/relabel_test.go @@ -14,7 +14,7 @@ package relabel import ( - "fmt" + "strconv" "testing" "github.com/prometheus/common/model" @@ -657,7 +657,7 @@ func TestRelabelValidate(t *testing.T) { }, } for i, test := range tests { - t.Run(fmt.Sprint(i), func(t *testing.T) { + t.Run(strconv.Itoa(i), func(t *testing.T) { err := test.config.Validate() if test.expected == "" { require.NoError(t, err) diff --git a/model/rulefmt/rulefmt.go b/model/rulefmt/rulefmt.go index 4ed1619d64..bfb85ce740 100644 --- a/model/rulefmt/rulefmt.go +++ b/model/rulefmt/rulefmt.go @@ -136,10 +136,11 @@ func (g *RuleGroups) Validate(node ruleGroups) (errs []error) { // RuleGroup is a list of sequentially evaluated recording and alerting rules. type RuleGroup struct { - Name string `yaml:"name"` - Interval model.Duration `yaml:"interval,omitempty"` - Limit int `yaml:"limit,omitempty"` - Rules []RuleNode `yaml:"rules"` + Name string `yaml:"name"` + Interval model.Duration `yaml:"interval,omitempty"` + QueryOffset *model.Duration `yaml:"query_offset,omitempty"` + Limit int `yaml:"limit,omitempty"` + Rules []RuleNode `yaml:"rules"` } // Rule describes an alerting or recording rule. diff --git a/notifier/notifier_test.go b/notifier/notifier_test.go index e7a9243bc2..d2e72ca33b 100644 --- a/notifier/notifier_test.go +++ b/notifier/notifier_test.go @@ -74,7 +74,7 @@ func TestHandlerNextBatch(t *testing.T) { for i := range make([]struct{}, 2*maxBatchSize+1) { h.queue = append(h.queue, &Alert{ - Labels: labels.FromStrings("alertname", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("alertname", strconv.Itoa(i)), }) } @@ -186,10 +186,10 @@ func TestHandlerSendAll(t *testing.T) { for i := range make([]struct{}, maxBatchSize) { h.queue = append(h.queue, &Alert{ - Labels: labels.FromStrings("alertname", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("alertname", strconv.Itoa(i)), }) expected = append(expected, &Alert{ - Labels: labels.FromStrings("alertname", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("alertname", strconv.Itoa(i)), }) } @@ -297,23 +297,23 @@ func TestHandlerSendAllRemapPerAm(t *testing.T) { for i := range make([]struct{}, maxBatchSize/2) { h.queue = append(h.queue, &Alert{ - Labels: labels.FromStrings("alertname", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("alertname", strconv.Itoa(i)), }, &Alert{ - Labels: labels.FromStrings("alertname", "test", "alertnamedrop", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("alertname", "test", "alertnamedrop", strconv.Itoa(i)), }, ) expected1 = append(expected1, &Alert{ - Labels: labels.FromStrings("alertname", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("alertname", strconv.Itoa(i)), }, &Alert{ - Labels: labels.FromStrings("alertname", "test", "alertnamedrop", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("alertname", "test", "alertnamedrop", strconv.Itoa(i)), }, ) expected2 = append(expected2, &Alert{ - Labels: labels.FromStrings("alertname", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("alertname", strconv.Itoa(i)), }) } @@ -502,7 +502,7 @@ func TestHandlerQueuing(t *testing.T) { var alerts []*Alert for i := range make([]struct{}, 20*maxBatchSize) { alerts = append(alerts, &Alert{ - Labels: labels.FromStrings("alertname", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("alertname", strconv.Itoa(i)), }) } @@ -762,7 +762,7 @@ func TestHangingNotifier(t *testing.T) { var alerts []*Alert for i := range make([]struct{}, 20*maxBatchSize) { alerts = append(alerts, &Alert{ - Labels: labels.FromStrings("alertname", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("alertname", strconv.Itoa(i)), }) } diff --git a/promql/engine_test.go b/promql/engine_test.go index 7b8deed935..69d9ea0361 100644 --- a/promql/engine_test.go +++ b/promql/engine_test.go @@ -20,6 +20,8 @@ import ( "math" "os" "sort" + "strconv" + "sync" "testing" "time" @@ -34,8 +36,6 @@ import ( "github.com/prometheus/prometheus/promql/parser/posrange" "github.com/prometheus/prometheus/promql/promqltest" "github.com/prometheus/prometheus/storage" - "github.com/prometheus/prometheus/tsdb/tsdbutil" - "github.com/prometheus/prometheus/util/almost" "github.com/prometheus/prometheus/util/annotations" "github.com/prometheus/prometheus/util/stats" "github.com/prometheus/prometheus/util/teststorage" @@ -59,7 +59,9 @@ func TestQueryConcurrency(t *testing.T) { require.NoError(t, err) defer os.RemoveAll(dir) queryTracker := promql.NewActiveQueryTracker(dir, maxConcurrency, nil) - t.Cleanup(queryTracker.Close) + t.Cleanup(func() { + require.NoError(t, queryTracker.Close()) + }) opts := promql.EngineOpts{ Logger: nil, @@ -91,9 +93,14 @@ func TestQueryConcurrency(t *testing.T) { return nil } + var wg sync.WaitGroup for i := 0; i < maxConcurrency; i++ { q := engine.NewTestQuery(f) - go q.Exec(ctx) + wg.Add(1) + go func() { + q.Exec(ctx) + wg.Done() + }() select { case <-processing: // Expected. @@ -103,7 +110,11 @@ func TestQueryConcurrency(t *testing.T) { } q := engine.NewTestQuery(f) - go q.Exec(ctx) + wg.Add(1) + go func() { + q.Exec(ctx) + wg.Done() + }() select { case <-processing: @@ -126,6 +137,8 @@ func TestQueryConcurrency(t *testing.T) { for i := 0; i < maxConcurrency; i++ { block <- struct{}{} } + + wg.Wait() } // contextDone returns an error if the context was canceled or timed out. @@ -3285,1076 +3298,6 @@ func TestInstantQueryWithRangeVectorSelector(t *testing.T) { } } -func TestNativeHistogramRate(t *testing.T) { - // TODO(beorn7): Integrate histograms into the PromQL testing framework - // and write more tests there. - engine := newTestEngine() - storage := teststorage.New(t) - t.Cleanup(func() { storage.Close() }) - - seriesName := "sparse_histogram_series" - lbls := labels.FromStrings("__name__", seriesName) - - app := storage.Appender(context.Background()) - for i, h := range tsdbutil.GenerateTestHistograms(100) { - _, err := app.AppendHistogram(0, lbls, int64(i)*int64(15*time.Second/time.Millisecond), h, nil) - require.NoError(t, err) - } - require.NoError(t, app.Commit()) - - queryString := fmt.Sprintf("rate(%s[45s])", seriesName) - t.Run("instant_query", func(t *testing.T) { - qry, err := engine.NewInstantQuery(context.Background(), storage, nil, queryString, timestamp.Time(int64(5*time.Minute/time.Millisecond))) - require.NoError(t, err) - res := qry.Exec(context.Background()) - require.NoError(t, res.Err) - vector, err := res.Vector() - require.NoError(t, err) - require.Len(t, vector, 1) - actualHistogram := vector[0].H - expectedHistogram := &histogram.FloatHistogram{ - CounterResetHint: histogram.GaugeType, - Schema: 1, - ZeroThreshold: 0.001, - ZeroCount: 1. / 15., - Count: 9. / 15., - Sum: 1.2266666666666663, - PositiveSpans: []histogram.Span{{Offset: 0, Length: 2}, {Offset: 1, Length: 2}}, - PositiveBuckets: []float64{1. / 15., 1. / 15., 1. / 15., 1. / 15.}, - NegativeSpans: []histogram.Span{{Offset: 0, Length: 2}, {Offset: 1, Length: 2}}, - NegativeBuckets: []float64{1. / 15., 1. / 15., 1. / 15., 1. / 15.}, - } - require.Equal(t, expectedHistogram, actualHistogram) - }) - - t.Run("range_query", func(t *testing.T) { - step := 30 * time.Second - start := timestamp.Time(int64(5 * time.Minute / time.Millisecond)) - end := start.Add(step) - qry, err := engine.NewRangeQuery(context.Background(), storage, nil, queryString, start, end, step) - require.NoError(t, err) - res := qry.Exec(context.Background()) - require.NoError(t, res.Err) - matrix, err := res.Matrix() - require.NoError(t, err) - require.Len(t, matrix, 1) - require.Len(t, matrix[0].Histograms, 2) - actualHistograms := matrix[0].Histograms - expectedHistograms := []promql.HPoint{{ - T: 300000, - H: &histogram.FloatHistogram{ - CounterResetHint: histogram.GaugeType, - Schema: 1, - ZeroThreshold: 0.001, - ZeroCount: 1. / 15., - Count: 9. / 15., - Sum: 1.2266666666666663, - PositiveSpans: []histogram.Span{{Offset: 0, Length: 2}, {Offset: 1, Length: 2}}, - PositiveBuckets: []float64{1. / 15., 1. / 15., 1. / 15., 1. / 15.}, - NegativeSpans: []histogram.Span{{Offset: 0, Length: 2}, {Offset: 1, Length: 2}}, - NegativeBuckets: []float64{1. / 15., 1. / 15., 1. / 15., 1. / 15.}, - }, - }, { - T: 330000, - H: &histogram.FloatHistogram{ - CounterResetHint: histogram.GaugeType, - Schema: 1, - ZeroThreshold: 0.001, - ZeroCount: 1. / 15., - Count: 9. / 15., - Sum: 1.2266666666666663, - PositiveSpans: []histogram.Span{{Offset: 0, Length: 2}, {Offset: 1, Length: 2}}, - PositiveBuckets: []float64{1. / 15., 1. / 15., 1. / 15., 1. / 15.}, - NegativeSpans: []histogram.Span{{Offset: 0, Length: 2}, {Offset: 1, Length: 2}}, - NegativeBuckets: []float64{1. / 15., 1. / 15., 1. / 15., 1. / 15.}, - }, - }} - require.Equal(t, expectedHistograms, actualHistograms) - }) -} - -func TestNativeFloatHistogramRate(t *testing.T) { - // TODO(beorn7): Integrate histograms into the PromQL testing framework - // and write more tests there. - engine := newTestEngine() - storage := teststorage.New(t) - t.Cleanup(func() { storage.Close() }) - - seriesName := "sparse_histogram_series" - lbls := labels.FromStrings("__name__", seriesName) - - app := storage.Appender(context.Background()) - for i, fh := range tsdbutil.GenerateTestFloatHistograms(100) { - _, err := app.AppendHistogram(0, lbls, int64(i)*int64(15*time.Second/time.Millisecond), nil, fh) - require.NoError(t, err) - } - require.NoError(t, app.Commit()) - - queryString := fmt.Sprintf("rate(%s[1m])", seriesName) - qry, err := engine.NewInstantQuery(context.Background(), storage, nil, queryString, timestamp.Time(int64(5*time.Minute/time.Millisecond))) - require.NoError(t, err) - res := qry.Exec(context.Background()) - require.NoError(t, res.Err) - vector, err := res.Vector() - require.NoError(t, err) - require.Len(t, vector, 1) - actualHistogram := vector[0].H - expectedHistogram := &histogram.FloatHistogram{ - CounterResetHint: histogram.GaugeType, - Schema: 1, - ZeroThreshold: 0.001, - ZeroCount: 1. / 15., - Count: 9. / 15., - Sum: 1.226666666666667, - PositiveSpans: []histogram.Span{{Offset: 0, Length: 2}, {Offset: 1, Length: 2}}, - PositiveBuckets: []float64{1. / 15., 1. / 15., 1. / 15., 1. / 15.}, - NegativeSpans: []histogram.Span{{Offset: 0, Length: 2}, {Offset: 1, Length: 2}}, - NegativeBuckets: []float64{1. / 15., 1. / 15., 1. / 15., 1. / 15.}, - } - require.Equal(t, expectedHistogram, actualHistogram) -} - -func TestNativeHistogram_HistogramCountAndSum(t *testing.T) { - // TODO(codesome): Integrate histograms into the PromQL testing framework - // and write more tests there. - h := &histogram.Histogram{ - Count: 24, - ZeroCount: 4, - ZeroThreshold: 0.001, - Sum: 100, - Schema: 0, - PositiveSpans: []histogram.Span{ - {Offset: 0, Length: 2}, - {Offset: 1, Length: 2}, - }, - PositiveBuckets: []int64{2, 1, -2, 3}, - NegativeSpans: []histogram.Span{ - {Offset: 0, Length: 2}, - {Offset: 1, Length: 2}, - }, - NegativeBuckets: []int64{2, 1, -2, 3}, - } - for _, floatHisto := range []bool{true, false} { - t.Run(fmt.Sprintf("floatHistogram=%t", floatHisto), func(t *testing.T) { - engine := newTestEngine() - storage := teststorage.New(t) - t.Cleanup(func() { storage.Close() }) - - seriesName := "sparse_histogram_series" - lbls := labels.FromStrings("__name__", seriesName) - - ts := int64(10 * time.Minute / time.Millisecond) - app := storage.Appender(context.Background()) - var err error - if floatHisto { - _, err = app.AppendHistogram(0, lbls, ts, nil, h.ToFloat(nil)) - } else { - _, err = app.AppendHistogram(0, lbls, ts, h, nil) - } - require.NoError(t, err) - require.NoError(t, app.Commit()) - - queryString := fmt.Sprintf("histogram_count(%s)", seriesName) - qry, err := engine.NewInstantQuery(context.Background(), storage, nil, queryString, timestamp.Time(ts)) - require.NoError(t, err) - - res := qry.Exec(context.Background()) - require.NoError(t, res.Err) - - vector, err := res.Vector() - require.NoError(t, err) - - require.Len(t, vector, 1) - require.Nil(t, vector[0].H) - if floatHisto { - require.Equal(t, h.ToFloat(nil).Count, vector[0].F) - } else { - require.Equal(t, float64(h.Count), vector[0].F) - } - - queryString = fmt.Sprintf("histogram_sum(%s)", seriesName) - qry, err = engine.NewInstantQuery(context.Background(), storage, nil, queryString, timestamp.Time(ts)) - require.NoError(t, err) - - res = qry.Exec(context.Background()) - require.NoError(t, res.Err) - - vector, err = res.Vector() - require.NoError(t, err) - - require.Len(t, vector, 1) - require.Nil(t, vector[0].H) - if floatHisto { - require.Equal(t, h.ToFloat(nil).Sum, vector[0].F) - } else { - require.Equal(t, h.Sum, vector[0].F) - } - }) - } -} - -func TestNativeHistogram_HistogramStdDevVar(t *testing.T) { - // TODO(codesome): Integrate histograms into the PromQL testing framework - // and write more tests there. - testCases := []struct { - name string - h *histogram.Histogram - stdVar float64 - }{ - { - name: "1, 2, 3, 4 low-res", - h: &histogram.Histogram{ - Count: 4, - Sum: 10, - Schema: 2, - PositiveSpans: []histogram.Span{ - {Offset: 0, Length: 1}, - {Offset: 3, Length: 1}, - {Offset: 2, Length: 2}, - }, - PositiveBuckets: []int64{1, 0, 0, 0}, - }, - stdVar: 1.163807968526718, // actual variance: 1.25 - }, - { - name: "1, 2, 3, 4 hi-res", - h: &histogram.Histogram{ - Count: 4, - Sum: 10, - Schema: 8, - PositiveSpans: []histogram.Span{ - {Offset: 0, Length: 1}, - {Offset: 255, Length: 1}, - {Offset: 149, Length: 1}, - {Offset: 105, Length: 1}, - }, - PositiveBuckets: []int64{1, 0, 0, 0}, - }, - stdVar: 1.2471347737158793, // actual variance: 1.25 - }, - { - name: "-50, -8, 0, 3, 8, 9, 100", - h: &histogram.Histogram{ - Count: 7, - ZeroCount: 1, - Sum: 62, - Schema: 3, - PositiveSpans: []histogram.Span{ - {Offset: 13, Length: 1}, - {Offset: 10, Length: 1}, - {Offset: 1, Length: 1}, - {Offset: 27, Length: 1}, - }, - PositiveBuckets: []int64{1, 0, 0, 0}, - NegativeSpans: []histogram.Span{ - {Offset: 24, Length: 1}, - {Offset: 21, Length: 1}, - }, - NegativeBuckets: []int64{1, 0}, - }, - stdVar: 1844.4651144196398, // actual variance: 1738.4082 - }, - { - name: "-100000, -10000, -1000, -888, -888, -100, -50, -9, -8, -3", - h: &histogram.Histogram{ - Count: 10, - ZeroCount: 0, - Sum: -112946, - Schema: 0, - NegativeSpans: []histogram.Span{ - {Offset: 2, Length: 3}, - {Offset: 1, Length: 2}, - {Offset: 2, Length: 1}, - {Offset: 3, Length: 1}, - {Offset: 2, Length: 1}, - }, - NegativeBuckets: []int64{1, 0, 0, 0, 0, 2, -2, 0}, - }, - stdVar: 759352122.1939945, // actual variance: 882690990 - }, - { - name: "-10 x10", - h: &histogram.Histogram{ - Count: 10, - ZeroCount: 0, - Sum: -100, - Schema: 0, - NegativeSpans: []histogram.Span{ - {Offset: 4, Length: 1}, - }, - NegativeBuckets: []int64{10}, - }, - stdVar: 1.725830020304794, // actual variance: 0 - }, - { - name: "-50, -8, 0, 3, 8, 9, 100, NaN", - h: &histogram.Histogram{ - Count: 8, - ZeroCount: 1, - Sum: math.NaN(), - Schema: 3, - PositiveSpans: []histogram.Span{ - {Offset: 13, Length: 1}, - {Offset: 10, Length: 1}, - {Offset: 1, Length: 1}, - {Offset: 27, Length: 1}, - }, - PositiveBuckets: []int64{1, 0, 0, 0}, - NegativeSpans: []histogram.Span{ - {Offset: 24, Length: 1}, - {Offset: 21, Length: 1}, - }, - NegativeBuckets: []int64{1, 0}, - }, - stdVar: math.NaN(), - }, - { - name: "-50, -8, 0, 3, 8, 9, 100, +Inf", - h: &histogram.Histogram{ - Count: 7, - ZeroCount: 1, - Sum: math.Inf(1), - Schema: 3, - PositiveSpans: []histogram.Span{ - {Offset: 13, Length: 1}, - {Offset: 10, Length: 1}, - {Offset: 1, Length: 1}, - {Offset: 27, Length: 1}, - }, - PositiveBuckets: []int64{1, 0, 0, 0}, - NegativeSpans: []histogram.Span{ - {Offset: 24, Length: 1}, - {Offset: 21, Length: 1}, - }, - NegativeBuckets: []int64{1, 0}, - }, - stdVar: math.NaN(), - }, - } - for _, tc := range testCases { - for _, floatHisto := range []bool{true, false} { - t.Run(fmt.Sprintf("%s floatHistogram=%t", tc.name, floatHisto), func(t *testing.T) { - engine := newTestEngine() - storage := teststorage.New(t) - t.Cleanup(func() { storage.Close() }) - - seriesName := "sparse_histogram_series" - lbls := labels.FromStrings("__name__", seriesName) - - ts := int64(10 * time.Minute / time.Millisecond) - app := storage.Appender(context.Background()) - var err error - if floatHisto { - _, err = app.AppendHistogram(0, lbls, ts, nil, tc.h.ToFloat(nil)) - } else { - _, err = app.AppendHistogram(0, lbls, ts, tc.h, nil) - } - require.NoError(t, err) - require.NoError(t, app.Commit()) - - queryString := fmt.Sprintf("histogram_stdvar(%s)", seriesName) - qry, err := engine.NewInstantQuery(context.Background(), storage, nil, queryString, timestamp.Time(ts)) - require.NoError(t, err) - - res := qry.Exec(context.Background()) - require.NoError(t, res.Err) - - vector, err := res.Vector() - require.NoError(t, err) - - require.Len(t, vector, 1) - require.Nil(t, vector[0].H) - require.InEpsilon(t, tc.stdVar, vector[0].F, 1e-12) - - queryString = fmt.Sprintf("histogram_stddev(%s)", seriesName) - qry, err = engine.NewInstantQuery(context.Background(), storage, nil, queryString, timestamp.Time(ts)) - require.NoError(t, err) - - res = qry.Exec(context.Background()) - require.NoError(t, res.Err) - - vector, err = res.Vector() - require.NoError(t, err) - - require.Len(t, vector, 1) - require.Nil(t, vector[0].H) - require.InEpsilon(t, math.Sqrt(tc.stdVar), vector[0].F, 1e-12) - }) - } - } -} - -func TestNativeHistogram_HistogramQuantile(t *testing.T) { - // TODO(codesome): Integrate histograms into the PromQL testing framework - // and write more tests there. - type subCase struct { - quantile string - value float64 - } - - cases := []struct { - text string - // Histogram to test. - h *histogram.Histogram - // Different quantiles to test for this histogram. - subCases []subCase - }{ - { - text: "all positive buckets with zero bucket", - h: &histogram.Histogram{ - Count: 12, - ZeroCount: 2, - ZeroThreshold: 0.001, - Sum: 100, // Does not matter. - Schema: 0, - PositiveSpans: []histogram.Span{ - {Offset: 0, Length: 2}, - {Offset: 1, Length: 2}, - }, - PositiveBuckets: []int64{2, 1, -2, 3}, - }, - subCases: []subCase{ - { - quantile: "1.0001", - value: math.Inf(1), - }, - { - quantile: "1", - value: 16, - }, - { - quantile: "0.99", - value: 15.759999999999998, - }, - { - quantile: "0.9", - value: 13.600000000000001, - }, - { - quantile: "0.6", - value: 4.799999999999997, - }, - { - quantile: "0.5", - value: 1.6666666666666665, - }, - { // Zero bucket. - quantile: "0.1", - value: 0.0006000000000000001, - }, - { - quantile: "0", - value: 0, - }, - { - quantile: "-1", - value: math.Inf(-1), - }, - }, - }, - { - text: "all negative buckets with zero bucket", - h: &histogram.Histogram{ - Count: 12, - ZeroCount: 2, - ZeroThreshold: 0.001, - Sum: 100, // Does not matter. - Schema: 0, - NegativeSpans: []histogram.Span{ - {Offset: 0, Length: 2}, - {Offset: 1, Length: 2}, - }, - NegativeBuckets: []int64{2, 1, -2, 3}, - }, - subCases: []subCase{ - { - quantile: "1.0001", - value: math.Inf(1), - }, - { // Zero bucket. - quantile: "1", - value: 0, - }, - { // Zero bucket. - quantile: "0.99", - value: -6.000000000000048e-05, - }, - { // Zero bucket. - quantile: "0.9", - value: -0.0005999999999999996, - }, - { - quantile: "0.5", - value: -1.6666666666666667, - }, - { - quantile: "0.1", - value: -13.6, - }, - { - quantile: "0", - value: -16, - }, - { - quantile: "-1", - value: math.Inf(-1), - }, - }, - }, - { - text: "both positive and negative buckets with zero bucket", - h: &histogram.Histogram{ - Count: 24, - ZeroCount: 4, - ZeroThreshold: 0.001, - Sum: 100, // Does not matter. - Schema: 0, - PositiveSpans: []histogram.Span{ - {Offset: 0, Length: 2}, - {Offset: 1, Length: 2}, - }, - PositiveBuckets: []int64{2, 1, -2, 3}, - NegativeSpans: []histogram.Span{ - {Offset: 0, Length: 2}, - {Offset: 1, Length: 2}, - }, - NegativeBuckets: []int64{2, 1, -2, 3}, - }, - subCases: []subCase{ - { - quantile: "1.0001", - value: math.Inf(1), - }, - { - quantile: "1", - value: 16, - }, - { - quantile: "0.99", - value: 15.519999999999996, - }, - { - quantile: "0.9", - value: 11.200000000000003, - }, - { - quantile: "0.7", - value: 1.2666666666666657, - }, - { // Zero bucket. - quantile: "0.55", - value: 0.0006000000000000005, - }, - { // Zero bucket. - quantile: "0.5", - value: 0, - }, - { // Zero bucket. - quantile: "0.45", - value: -0.0005999999999999996, - }, - { - quantile: "0.3", - value: -1.266666666666667, - }, - { - quantile: "0.1", - value: -11.2, - }, - { - quantile: "0.01", - value: -15.52, - }, - { - quantile: "0", - value: -16, - }, - { - quantile: "-1", - value: math.Inf(-1), - }, - }, - }, - } - - engine := newTestEngine() - storage := teststorage.New(t) - t.Cleanup(func() { storage.Close() }) - idx := int64(0) - for _, floatHisto := range []bool{true, false} { - for _, c := range cases { - t.Run(fmt.Sprintf("%s floatHistogram=%t", c.text, floatHisto), func(t *testing.T) { - seriesName := "sparse_histogram_series" - lbls := labels.FromStrings("__name__", seriesName) - ts := idx * int64(10*time.Minute/time.Millisecond) - app := storage.Appender(context.Background()) - var err error - if floatHisto { - _, err = app.AppendHistogram(0, lbls, ts, nil, c.h.ToFloat(nil)) - } else { - _, err = app.AppendHistogram(0, lbls, ts, c.h, nil) - } - require.NoError(t, err) - require.NoError(t, app.Commit()) - - for j, sc := range c.subCases { - t.Run(fmt.Sprintf("%d %s", j, sc.quantile), func(t *testing.T) { - queryString := fmt.Sprintf("histogram_quantile(%s, %s)", sc.quantile, seriesName) - qry, err := engine.NewInstantQuery(context.Background(), storage, nil, queryString, timestamp.Time(ts)) - require.NoError(t, err) - - res := qry.Exec(context.Background()) - require.NoError(t, res.Err) - - vector, err := res.Vector() - require.NoError(t, err) - - require.Len(t, vector, 1) - require.Nil(t, vector[0].H) - require.True(t, almost.Equal(sc.value, vector[0].F, defaultEpsilon)) - }) - } - idx++ - }) - } - } -} - -func TestNativeHistogram_HistogramFraction(t *testing.T) { - // TODO(codesome): Integrate histograms into the PromQL testing framework - // and write more tests there. - type subCase struct { - lower, upper string - value float64 - } - - invariantCases := []subCase{ - { - lower: "42", - upper: "3.1415", - value: 0, - }, - { - lower: "0", - upper: "0", - value: 0, - }, - { - lower: "0.000001", - upper: "0.000001", - value: 0, - }, - { - lower: "42", - upper: "42", - value: 0, - }, - { - lower: "-3.1", - upper: "-3.1", - value: 0, - }, - { - lower: "3.1415", - upper: "NaN", - value: math.NaN(), - }, - { - lower: "NaN", - upper: "42", - value: math.NaN(), - }, - { - lower: "NaN", - upper: "NaN", - value: math.NaN(), - }, - { - lower: "-Inf", - upper: "+Inf", - value: 1, - }, - } - - cases := []struct { - text string - // Histogram to test. - h *histogram.Histogram - // Different ranges to test for this histogram. - subCases []subCase - }{ - { - text: "empty histogram", - h: &histogram.Histogram{}, - subCases: []subCase{ - { - lower: "3.1415", - upper: "42", - value: math.NaN(), - }, - }, - }, - { - text: "all positive buckets with zero bucket", - h: &histogram.Histogram{ - Count: 12, - ZeroCount: 2, - ZeroThreshold: 0.001, - Sum: 100, // Does not matter. - Schema: 0, - PositiveSpans: []histogram.Span{ - {Offset: 0, Length: 2}, - {Offset: 1, Length: 2}, - }, - PositiveBuckets: []int64{2, 1, -2, 3}, // Abs: 2, 3, 1, 4 - }, - subCases: append([]subCase{ - { - lower: "0", - upper: "+Inf", - value: 1, - }, - { - lower: "-Inf", - upper: "0", - value: 0, - }, - { - lower: "-0.001", - upper: "0", - value: 0, - }, - { - lower: "0", - upper: "0.001", - value: 2. / 12., - }, - { - lower: "0", - upper: "0.0005", - value: 1. / 12., - }, - { - lower: "0.001", - upper: "inf", - value: 10. / 12., - }, - { - lower: "-inf", - upper: "-0.001", - value: 0, - }, - { - lower: "1", - upper: "2", - value: 3. / 12., - }, - { - lower: "1.5", - upper: "2", - value: 1.5 / 12., - }, - { - lower: "1", - upper: "8", - value: 4. / 12., - }, - { - lower: "1", - upper: "6", - value: 3.5 / 12., - }, - { - lower: "1.5", - upper: "6", - value: 2. / 12., - }, - { - lower: "-2", - upper: "-1", - value: 0, - }, - { - lower: "-2", - upper: "-1.5", - value: 0, - }, - { - lower: "-8", - upper: "-1", - value: 0, - }, - { - lower: "-6", - upper: "-1", - value: 0, - }, - { - lower: "-6", - upper: "-1.5", - value: 0, - }, - }, invariantCases...), - }, - { - text: "all negative buckets with zero bucket", - h: &histogram.Histogram{ - Count: 12, - ZeroCount: 2, - ZeroThreshold: 0.001, - Sum: 100, // Does not matter. - Schema: 0, - NegativeSpans: []histogram.Span{ - {Offset: 0, Length: 2}, - {Offset: 1, Length: 2}, - }, - NegativeBuckets: []int64{2, 1, -2, 3}, - }, - subCases: append([]subCase{ - { - lower: "0", - upper: "+Inf", - value: 0, - }, - { - lower: "-Inf", - upper: "0", - value: 1, - }, - { - lower: "-0.001", - upper: "0", - value: 2. / 12., - }, - { - lower: "0", - upper: "0.001", - value: 0, - }, - { - lower: "-0.0005", - upper: "0", - value: 1. / 12., - }, - { - lower: "0.001", - upper: "inf", - value: 0, - }, - { - lower: "-inf", - upper: "-0.001", - value: 10. / 12., - }, - { - lower: "1", - upper: "2", - value: 0, - }, - { - lower: "1.5", - upper: "2", - value: 0, - }, - { - lower: "1", - upper: "8", - value: 0, - }, - { - lower: "1", - upper: "6", - value: 0, - }, - { - lower: "1.5", - upper: "6", - value: 0, - }, - { - lower: "-2", - upper: "-1", - value: 3. / 12., - }, - { - lower: "-2", - upper: "-1.5", - value: 1.5 / 12., - }, - { - lower: "-8", - upper: "-1", - value: 4. / 12., - }, - { - lower: "-6", - upper: "-1", - value: 3.5 / 12., - }, - { - lower: "-6", - upper: "-1.5", - value: 2. / 12., - }, - }, invariantCases...), - }, - { - text: "both positive and negative buckets with zero bucket", - h: &histogram.Histogram{ - Count: 24, - ZeroCount: 4, - ZeroThreshold: 0.001, - Sum: 100, // Does not matter. - Schema: 0, - PositiveSpans: []histogram.Span{ - {Offset: 0, Length: 2}, - {Offset: 1, Length: 2}, - }, - PositiveBuckets: []int64{2, 1, -2, 3}, - NegativeSpans: []histogram.Span{ - {Offset: 0, Length: 2}, - {Offset: 1, Length: 2}, - }, - NegativeBuckets: []int64{2, 1, -2, 3}, - }, - subCases: append([]subCase{ - { - lower: "0", - upper: "+Inf", - value: 0.5, - }, - { - lower: "-Inf", - upper: "0", - value: 0.5, - }, - { - lower: "-0.001", - upper: "0", - value: 2. / 24, - }, - { - lower: "0", - upper: "0.001", - value: 2. / 24., - }, - { - lower: "-0.0005", - upper: "0.0005", - value: 2. / 24., - }, - { - lower: "0.001", - upper: "inf", - value: 10. / 24., - }, - { - lower: "-inf", - upper: "-0.001", - value: 10. / 24., - }, - { - lower: "1", - upper: "2", - value: 3. / 24., - }, - { - lower: "1.5", - upper: "2", - value: 1.5 / 24., - }, - { - lower: "1", - upper: "8", - value: 4. / 24., - }, - { - lower: "1", - upper: "6", - value: 3.5 / 24., - }, - { - lower: "1.5", - upper: "6", - value: 2. / 24., - }, - { - lower: "-2", - upper: "-1", - value: 3. / 24., - }, - { - lower: "-2", - upper: "-1.5", - value: 1.5 / 24., - }, - { - lower: "-8", - upper: "-1", - value: 4. / 24., - }, - { - lower: "-6", - upper: "-1", - value: 3.5 / 24., - }, - { - lower: "-6", - upper: "-1.5", - value: 2. / 24., - }, - }, invariantCases...), - }, - } - idx := int64(0) - for _, floatHisto := range []bool{true, false} { - for _, c := range cases { - t.Run(fmt.Sprintf("%s floatHistogram=%t", c.text, floatHisto), func(t *testing.T) { - engine := newTestEngine() - storage := teststorage.New(t) - t.Cleanup(func() { storage.Close() }) - - seriesName := "sparse_histogram_series" - lbls := labels.FromStrings("__name__", seriesName) - - ts := idx * int64(10*time.Minute/time.Millisecond) - app := storage.Appender(context.Background()) - var err error - if floatHisto { - _, err = app.AppendHistogram(0, lbls, ts, nil, c.h.ToFloat(nil)) - } else { - _, err = app.AppendHistogram(0, lbls, ts, c.h, nil) - } - require.NoError(t, err) - require.NoError(t, app.Commit()) - - for j, sc := range c.subCases { - t.Run(fmt.Sprintf("%d %s %s", j, sc.lower, sc.upper), func(t *testing.T) { - queryString := fmt.Sprintf("histogram_fraction(%s, %s, %s)", sc.lower, sc.upper, seriesName) - qry, err := engine.NewInstantQuery(context.Background(), storage, nil, queryString, timestamp.Time(ts)) - require.NoError(t, err) - - res := qry.Exec(context.Background()) - require.NoError(t, res.Err) - - vector, err := res.Vector() - require.NoError(t, err) - - require.Len(t, vector, 1) - require.Nil(t, vector[0].H) - if math.IsNaN(sc.value) { - require.True(t, math.IsNaN(vector[0].F)) - return - } - require.Equal(t, sc.value, vector[0].F) - }) - } - idx++ - }) - } - } -} - func TestNativeHistogram_Sum_Count_Add_AvgOperator(t *testing.T) { // TODO(codesome): Integrate histograms into the PromQL testing framework // and write more tests there. @@ -4482,7 +3425,7 @@ func TestNativeHistogram_Sum_Count_Add_AvgOperator(t *testing.T) { _, err := app.Append(0, labels.FromStrings("__name__", "float_series", "idx", "0"), ts, 42) require.NoError(t, err) for idx1, h := range c.histograms { - lbls := labels.FromStrings("__name__", seriesName, "idx", fmt.Sprintf("%d", idx1)) + lbls := labels.FromStrings("__name__", seriesName, "idx", strconv.Itoa(idx1)) // Since we mutate h later, we need to create a copy here. var err error if floatHisto { @@ -4754,7 +3697,7 @@ func TestNativeHistogram_SubOperator(t *testing.T) { ts := idx0 * int64(10*time.Minute/time.Millisecond) app := storage.Appender(context.Background()) for idx1, h := range c.histograms { - lbls := labels.FromStrings("__name__", seriesName, "idx", fmt.Sprintf("%d", idx1)) + lbls := labels.FromStrings("__name__", seriesName, "idx", strconv.Itoa(idx1)) // Since we mutate h later, we need to create a copy here. var err error if floatHisto { diff --git a/promql/parser/printer_test.go b/promql/parser/printer_test.go index 554fb69626..f224d841d0 100644 --- a/promql/parser/printer_test.go +++ b/promql/parser/printer_test.go @@ -138,6 +138,16 @@ func TestExprString(t *testing.T) { { in: `{__name__="",a="x"}`, }, + { + in: `{"a.b"="c"}`, + }, + { + in: `{"0"="1"}`, + }, + { + in: `{"_0"="1"}`, + out: `{_0="1"}`, + }, } for _, test := range inputs { diff --git a/promql/promqltest/testdata/native_histograms.test b/promql/promqltest/testdata/native_histograms.test index 1da68a385f..37818e4f88 100644 --- a/promql/promqltest/testdata/native_histograms.test +++ b/promql/promqltest/testdata/native_histograms.test @@ -269,3 +269,448 @@ eval instant at 50m histogram_sum(sum(incr_sum_histogram)) eval instant at 50m histogram_sum(sum(last_over_time(incr_sum_histogram[5m]))) {} 30 + +# Apply rate function to histogram. +load 15s + histogram_rate {{schema:1 count:12 sum:18.4 z_bucket:2 z_bucket_w:0.001 buckets:[1 2 0 1 1] n_buckets:[1 2 0 1 1]}}+{{schema:1 count:9 sum:18.4 z_bucket:1 z_bucket_w:0.001 buckets:[1 1 0 1 1] n_buckets:[1 1 0 1 1]}}x100 + +eval instant at 5m rate(histogram_rate[45s]) + {} {{schema:1 count:0.6 sum:1.2266666666666652 z_bucket:0.06666666666666667 z_bucket_w:0.001 buckets:[0.06666666666666667 0.06666666666666667 0 0.06666666666666667 0.06666666666666667] n_buckets:[0.06666666666666667 0.06666666666666667 0 0.06666666666666667 0.06666666666666667]}} + +eval range from 5m to 5m30s step 30s rate(histogram_rate[45s]) + {} {{schema:1 count:0.6 sum:1.2266666666666652 z_bucket:0.06666666666666667 z_bucket_w:0.001 buckets:[0.06666666666666667 0.06666666666666667 0 0.06666666666666667 0.06666666666666667] n_buckets:[0.06666666666666667 0.06666666666666667 0 0.06666666666666667 0.06666666666666667]}}x1 + +# Apply count and sum function to histogram. +load 10m + histogram_count_sum_2 {{schema:0 count:24 sum:100 z_bucket:4 z_bucket_w:0.001 buckets:[2 3 0 1 4] n_buckets:[2 3 0 1 4]}}x1 + +eval instant at 10m histogram_count(histogram_count_sum_2) + {} 24 + +eval instant at 10m histogram_sum(histogram_count_sum_2) + {} 100 + +# Apply stddev and stdvar function to histogram with {1, 2, 3, 4} (low res). +load 10m + histogram_stddev_stdvar_1 {{schema:2 count:4 sum:10 buckets:[1 0 0 0 1 0 0 1 1]}}x1 + +eval instant at 10m histogram_stddev(histogram_stddev_stdvar_1) + {} 1.0787993180043811 + +eval instant at 10m histogram_stdvar(histogram_stddev_stdvar_1) + {} 1.163807968526718 + +# Apply stddev and stdvar function to histogram with {1, 1, 1, 1} (high res). +load 10m + histogram_stddev_stdvar_2 {{schema:8 count:10 sum:10 buckets:[1 2 3 4]}}x1 + +eval instant at 10m histogram_stddev(histogram_stddev_stdvar_2) + {} 0.0048960313898237465 + +eval instant at 10m histogram_stdvar(histogram_stddev_stdvar_2) + {} 2.3971123370139447e-05 + +# Apply stddev and stdvar function to histogram with {-50, -8, 0, 3, 8, 9}. +load 10m + histogram_stddev_stdvar_3 {{schema:3 count:7 sum:62 z_bucket:1 buckets:[0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 ] n_buckets:[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 ]}}x1 + +eval instant at 10m histogram_stddev(histogram_stddev_stdvar_3) + {} 42.947236400258 + +eval instant at 10m histogram_stdvar(histogram_stddev_stdvar_3) + {} 1844.4651144196398 + +# Apply stddev and stdvar function to histogram with {-100000, -10000, -1000, -888, -888, -100, -50, -9, -8, -3}. +load 10m + histogram_stddev_stdvar_4 {{schema:0 count:10 sum:-112946 z_bucket:0 n_buckets:[0 0 1 1 1 0 1 1 0 0 3 0 0 0 1 0 0 1]}}x1 + +eval instant at 10m histogram_stddev(histogram_stddev_stdvar_4) + {} 27556.344499842 + +eval instant at 10m histogram_stdvar(histogram_stddev_stdvar_4) + {} 759352122.1939945 + +# Apply stddev and stdvar function to histogram with {-10x10}. +load 10m + histogram_stddev_stdvar_5 {{schema:0 count:10 sum:-100 z_bucket:0 n_buckets:[0 0 0 0 10]}}x1 + +eval instant at 10m histogram_stddev(histogram_stddev_stdvar_5) + {} 1.3137084989848 + +eval instant at 10m histogram_stdvar(histogram_stddev_stdvar_5) + {} 1.725830020304794 + +# Apply stddev and stdvar function to histogram with {-50, -8, 0, 3, 8, 9, NaN}. +load 10m + histogram_stddev_stdvar_6 {{schema:3 count:7 sum:NaN z_bucket:1 buckets:[0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 ] n_buckets:[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 ]}}x1 + +eval instant at 10m histogram_stddev(histogram_stddev_stdvar_6) + {} NaN + +eval instant at 10m histogram_stdvar(histogram_stddev_stdvar_6) + {} NaN + +# Apply stddev and stdvar function to histogram with {-50, -8, 0, 3, 8, 9, Inf}. +load 10m + histogram_stddev_stdvar_7 {{schema:3 count:7 sum:Inf z_bucket:1 buckets:[0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 ] n_buckets:[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 ]}}x1 + +eval instant at 10m histogram_stddev(histogram_stddev_stdvar_7) + {} NaN + +eval instant at 10m histogram_stdvar(histogram_stddev_stdvar_7) + {} NaN + +# Apply quantile function to histogram with all positive buckets with zero bucket. +load 10m + histogram_quantile_1 {{schema:0 count:12 sum:100 z_bucket:2 z_bucket_w:0.001 buckets:[2 3 0 1 4]}}x1 + +eval instant at 10m histogram_quantile(1.001, histogram_quantile_1) + {} Inf + +eval instant at 10m histogram_quantile(1, histogram_quantile_1) + {} 16 + +eval instant at 10m histogram_quantile(0.99, histogram_quantile_1) + {} 15.759999999999998 + +eval instant at 10m histogram_quantile(0.9, histogram_quantile_1) + {} 13.600000000000001 + +eval instant at 10m histogram_quantile(0.6, histogram_quantile_1) + {} 4.799999999999997 + +eval instant at 10m histogram_quantile(0.5, histogram_quantile_1) + {} 1.6666666666666665 + +eval instant at 10m histogram_quantile(0.1, histogram_quantile_1) + {} 0.0006000000000000001 + +eval instant at 10m histogram_quantile(0, histogram_quantile_1) + {} 0 + +eval instant at 10m histogram_quantile(-1, histogram_quantile_1) + {} -Inf + +# Apply quantile function to histogram with all negative buckets with zero bucket. +load 10m + histogram_quantile_2 {{schema:0 count:12 sum:100 z_bucket:2 z_bucket_w:0.001 n_buckets:[2 3 0 1 4]}}x1 + +eval instant at 10m histogram_quantile(1.001, histogram_quantile_2) + {} Inf + +eval instant at 10m histogram_quantile(1, histogram_quantile_2) + {} 0 + +eval instant at 10m histogram_quantile(0.99, histogram_quantile_2) + {} -6.000000000000048e-05 + +eval instant at 10m histogram_quantile(0.9, histogram_quantile_2) + {} -0.0005999999999999996 + +eval instant at 10m histogram_quantile(0.5, histogram_quantile_2) + {} -1.6666666666666667 + +eval instant at 10m histogram_quantile(0.1, histogram_quantile_2) + {} -13.6 + +eval instant at 10m histogram_quantile(0, histogram_quantile_2) + {} -16 + +eval instant at 10m histogram_quantile(-1, histogram_quantile_2) + {} -Inf + +# Apply quantile function to histogram with both positive and negative buckets with zero bucket. +load 10m + histogram_quantile_3 {{schema:0 count:24 sum:100 z_bucket:4 z_bucket_w:0.001 buckets:[2 3 0 1 4] n_buckets:[2 3 0 1 4]}}x1 + +eval instant at 10m histogram_quantile(1.001, histogram_quantile_3) + {} Inf + +eval instant at 10m histogram_quantile(1, histogram_quantile_3) + {} 16 + +eval instant at 10m histogram_quantile(0.99, histogram_quantile_3) + {} 15.519999999999996 + +eval instant at 10m histogram_quantile(0.9, histogram_quantile_3) + {} 11.200000000000003 + +eval instant at 10m histogram_quantile(0.7, histogram_quantile_3) + {} 1.2666666666666657 + +eval instant at 10m histogram_quantile(0.55, histogram_quantile_3) + {} 0.0006000000000000005 + +eval instant at 10m histogram_quantile(0.5, histogram_quantile_3) + {} 0 + +eval instant at 10m histogram_quantile(0.45, histogram_quantile_3) + {} -0.0005999999999999996 + +eval instant at 10m histogram_quantile(0.3, histogram_quantile_3) + {} -1.266666666666667 + +eval instant at 10m histogram_quantile(0.1, histogram_quantile_3) + {} -11.2 + +eval instant at 10m histogram_quantile(0.01, histogram_quantile_3) + {} -15.52 + +eval instant at 10m histogram_quantile(0, histogram_quantile_3) + {} -16 + +eval instant at 10m histogram_quantile(-1, histogram_quantile_3) + {} -Inf + +# Apply fraction function to empty histogram. +load 10m + histogram_fraction_1 {{}}x1 + +eval instant at 10m histogram_fraction(3.1415, 42, histogram_fraction_1) + {} NaN + +# Apply fraction function to histogram with positive and zero buckets. +load 10m + histogram_fraction_2 {{schema:0 count:12 sum:100 z_bucket:2 z_bucket_w:0.001 buckets:[2 3 0 1 4]}}x1 + +eval instant at 10m histogram_fraction(0, +Inf, histogram_fraction_2) + {} 1 + +eval instant at 10m histogram_fraction(-Inf, 0, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(-0.001, 0, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(0, 0.001, histogram_fraction_2) + {} 0.16666666666666666 + +eval instant at 10m histogram_fraction(0, 0.0005, histogram_fraction_2) + {} 0.08333333333333333 + +eval instant at 10m histogram_fraction(0.001, inf, histogram_fraction_2) + {} 0.8333333333333334 + +eval instant at 10m histogram_fraction(-inf, -0.001, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(1, 2, histogram_fraction_2) + {} 0.25 + +eval instant at 10m histogram_fraction(1.5, 2, histogram_fraction_2) + {} 0.125 + +eval instant at 10m histogram_fraction(1, 8, histogram_fraction_2) + {} 0.3333333333333333 + +eval instant at 10m histogram_fraction(1, 6, histogram_fraction_2) + {} 0.2916666666666667 + +eval instant at 10m histogram_fraction(1.5, 6, histogram_fraction_2) + {} 0.16666666666666666 + +eval instant at 10m histogram_fraction(-2, -1, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(-2, -1.5, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(-8, -1, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(-6, -1, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(-6, -1.5, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(42, 3.1415, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(0, 0, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(0.000001, 0.000001, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(42, 42, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(-3.1, -3.1, histogram_fraction_2) + {} 0 + +eval instant at 10m histogram_fraction(3.1415, NaN, histogram_fraction_2) + {} NaN + +eval instant at 10m histogram_fraction(NaN, 42, histogram_fraction_2) + {} NaN + +eval instant at 10m histogram_fraction(NaN, NaN, histogram_fraction_2) + {} NaN + +eval instant at 10m histogram_fraction(-Inf, +Inf, histogram_fraction_2) + {} 1 + +# Apply fraction function to histogram with negative and zero buckets. +load 10m + histogram_fraction_3 {{schema:0 count:12 sum:100 z_bucket:2 z_bucket_w:0.001 n_buckets:[2 3 0 1 4]}}x1 + +eval instant at 10m histogram_fraction(0, +Inf, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(-Inf, 0, histogram_fraction_3) + {} 1 + +eval instant at 10m histogram_fraction(-0.001, 0, histogram_fraction_3) + {} 0.16666666666666666 + +eval instant at 10m histogram_fraction(0, 0.001, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(-0.0005, 0, histogram_fraction_3) + {} 0.08333333333333333 + +eval instant at 10m histogram_fraction(0.001, inf, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(-inf, -0.001, histogram_fraction_3) + {} 0.8333333333333334 + +eval instant at 10m histogram_fraction(1, 2, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(1.5, 2, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(1, 8, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(1, 6, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(1.5, 6, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(-2, -1, histogram_fraction_3) + {} 0.25 + +eval instant at 10m histogram_fraction(-2, -1.5, histogram_fraction_3) + {} 0.125 + +eval instant at 10m histogram_fraction(-8, -1, histogram_fraction_3) + {} 0.3333333333333333 + +eval instant at 10m histogram_fraction(-6, -1, histogram_fraction_3) + {} 0.2916666666666667 + +eval instant at 10m histogram_fraction(-6, -1.5, histogram_fraction_3) + {} 0.16666666666666666 + +eval instant at 10m histogram_fraction(42, 3.1415, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(0, 0, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(0.000001, 0.000001, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(42, 42, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(-3.1, -3.1, histogram_fraction_3) + {} 0 + +eval instant at 10m histogram_fraction(3.1415, NaN, histogram_fraction_3) + {} NaN + +eval instant at 10m histogram_fraction(NaN, 42, histogram_fraction_3) + {} NaN + +eval instant at 10m histogram_fraction(NaN, NaN, histogram_fraction_3) + {} NaN + +eval instant at 10m histogram_fraction(-Inf, +Inf, histogram_fraction_3) + {} 1 + +# Apply fraction function to histogram with both positive, negative and zero buckets. +load 10m + histogram_fraction_4 {{schema:0 count:24 sum:100 z_bucket:4 z_bucket_w:0.001 buckets:[2 3 0 1 4] n_buckets:[2 3 0 1 4]}}x1 + +eval instant at 10m histogram_fraction(0, +Inf, histogram_fraction_4) + {} 0.5 + +eval instant at 10m histogram_fraction(-Inf, 0, histogram_fraction_4) + {} 0.5 + +eval instant at 10m histogram_fraction(-0.001, 0, histogram_fraction_4) + {} 0.08333333333333333 + +eval instant at 10m histogram_fraction(0, 0.001, histogram_fraction_4) + {} 0.08333333333333333 + +eval instant at 10m histogram_fraction(-0.0005, 0.0005, histogram_fraction_4) + {} 0.08333333333333333 + +eval instant at 10m histogram_fraction(0.001, inf, histogram_fraction_4) + {} 0.4166666666666667 + +eval instant at 10m histogram_fraction(-inf, -0.001, histogram_fraction_4) + {} 0.4166666666666667 + +eval instant at 10m histogram_fraction(1, 2, histogram_fraction_4) + {} 0.125 + +eval instant at 10m histogram_fraction(1.5, 2, histogram_fraction_4) + {} 0.0625 + +eval instant at 10m histogram_fraction(1, 8, histogram_fraction_4) + {} 0.16666666666666666 + +eval instant at 10m histogram_fraction(1, 6, histogram_fraction_4) + {} 0.14583333333333334 + +eval instant at 10m histogram_fraction(1.5, 6, histogram_fraction_4) + {} 0.08333333333333333 + +eval instant at 10m histogram_fraction(-2, -1, histogram_fraction_4) + {} 0.125 + +eval instant at 10m histogram_fraction(-2, -1.5, histogram_fraction_4) + {} 0.0625 + +eval instant at 10m histogram_fraction(-8, -1, histogram_fraction_4) + {} 0.16666666666666666 + +eval instant at 10m histogram_fraction(-6, -1, histogram_fraction_4) + {} 0.14583333333333334 + +eval instant at 10m histogram_fraction(-6, -1.5, histogram_fraction_4) + {} 0.08333333333333333 + +eval instant at 10m histogram_fraction(42, 3.1415, histogram_fraction_4) + {} 0 + +eval instant at 10m histogram_fraction(0, 0, histogram_fraction_4) + {} 0 + +eval instant at 10m histogram_fraction(0.000001, 0.000001, histogram_fraction_4) + {} 0 + +eval instant at 10m histogram_fraction(42, 42, histogram_fraction_4) + {} 0 + +eval instant at 10m histogram_fraction(-3.1, -3.1, histogram_fraction_4) + {} 0 + +eval instant at 10m histogram_fraction(3.1415, NaN, histogram_fraction_4) + {} NaN + +eval instant at 10m histogram_fraction(NaN, 42, histogram_fraction_4) + {} NaN + +eval instant at 10m histogram_fraction(NaN, NaN, histogram_fraction_4) + {} NaN + +eval instant at 10m histogram_fraction(-Inf, +Inf, histogram_fraction_4) + {} 1 diff --git a/promql/query_logger.go b/promql/query_logger.go index 7ddd8c2d5a..7e06ebb97f 100644 --- a/promql/query_logger.go +++ b/promql/query_logger.go @@ -16,6 +16,8 @@ package promql import ( "context" "encoding/json" + "errors" + "fmt" "io" "os" "path/filepath" @@ -36,6 +38,8 @@ type ActiveQueryTracker struct { maxConcurrent int } +var _ io.Closer = &ActiveQueryTracker{} + type Entry struct { Query string `json:"query"` Timestamp int64 `json:"timestamp_sec"` @@ -83,6 +87,23 @@ func logUnfinishedQueries(filename string, filesize int, logger log.Logger) { } } +type mmapedFile struct { + f io.Closer + m mmap.MMap +} + +func (f *mmapedFile) Close() error { + err := f.m.Unmap() + if err != nil { + err = fmt.Errorf("mmapedFile: unmapping: %w", err) + } + if fErr := f.f.Close(); fErr != nil { + return errors.Join(fmt.Errorf("close mmapedFile.f: %w", fErr), err) + } + + return err +} + func getMMapedFile(filename string, filesize int, logger log.Logger) ([]byte, io.Closer, error) { file, err := os.OpenFile(filename, os.O_CREATE|os.O_RDWR|os.O_TRUNC, 0o666) if err != nil { @@ -108,7 +129,7 @@ func getMMapedFile(filename string, filesize int, logger log.Logger) ([]byte, io return nil, nil, err } - return fileAsBytes, file, err + return fileAsBytes, &mmapedFile{f: file, m: fileAsBytes}, err } func NewActiveQueryTracker(localStoragePath string, maxConcurrent int, logger log.Logger) *ActiveQueryTracker { @@ -204,9 +225,13 @@ func (tracker ActiveQueryTracker) Insert(ctx context.Context, query string) (int } } -func (tracker *ActiveQueryTracker) Close() { +// Close closes tracker. +func (tracker *ActiveQueryTracker) Close() error { if tracker == nil || tracker.closer == nil { - return + return nil } - tracker.closer.Close() + if err := tracker.closer.Close(); err != nil { + return fmt.Errorf("close ActiveQueryTracker.closer: %w", err) + } + return nil } diff --git a/promql/query_logger_test.go b/promql/query_logger_test.go index 376d61b641..7bd93781ec 100644 --- a/promql/query_logger_test.go +++ b/promql/query_logger_test.go @@ -16,6 +16,7 @@ package promql import ( "context" "os" + "path/filepath" "testing" "github.com/grafana/regexp" @@ -104,26 +105,26 @@ func TestIndexReuse(t *testing.T) { } func TestMMapFile(t *testing.T) { - file, err := os.CreateTemp("", "mmapedFile") + dir := t.TempDir() + fpath := filepath.Join(dir, "mmapedFile") + const data = "ab" + + fileAsBytes, closer, err := getMMapedFile(fpath, 2, nil) require.NoError(t, err) + copy(fileAsBytes, data) + require.NoError(t, closer.Close()) - filename := file.Name() - defer os.Remove(filename) - - fileAsBytes, _, err := getMMapedFile(filename, 2, nil) - - require.NoError(t, err) - copy(fileAsBytes, "ab") - - f, err := os.Open(filename) + f, err := os.Open(fpath) require.NoError(t, err) + t.Cleanup(func() { + _ = f.Close() + }) bytes := make([]byte, 4) n, err := f.Read(bytes) - require.Equal(t, 2, n) require.NoError(t, err, "Unexpected error while reading file.") - - require.Equal(t, fileAsBytes, bytes[:2], "Mmap failed") + require.Equal(t, 2, n) + require.Equal(t, []byte(data), bytes[:2], "Mmap failed") } func TestParseBrokenJSON(t *testing.T) { diff --git a/rules/alerting.go b/rules/alerting.go index 2d2b19efe7..2dc0917dce 100644 --- a/rules/alerting.go +++ b/rules/alerting.go @@ -338,10 +338,9 @@ const resolvedRetention = 15 * time.Minute // Eval evaluates the rule expression and then creates pending alerts and fires // or removes previously pending alerts accordingly. -func (r *AlertingRule) Eval(ctx context.Context, ts time.Time, query QueryFunc, externalURL *url.URL, limit int) (promql.Vector, error) { +func (r *AlertingRule) Eval(ctx context.Context, queryOffset time.Duration, ts time.Time, query QueryFunc, externalURL *url.URL, limit int) (promql.Vector, error) { ctx = NewOriginContext(ctx, NewRuleDetail(r)) - - res, err := query(ctx, r.vector.String(), ts) + res, err := query(ctx, r.vector.String(), ts.Add(-queryOffset)) if err != nil { return nil, err } @@ -484,8 +483,8 @@ func (r *AlertingRule) Eval(ctx context.Context, ts time.Time, query QueryFunc, } if r.restored.Load() { - vec = append(vec, r.sample(a, ts)) - vec = append(vec, r.forStateSample(a, ts, float64(a.ActiveAt.Unix()))) + vec = append(vec, r.sample(a, ts.Add(-queryOffset))) + vec = append(vec, r.forStateSample(a, ts.Add(-queryOffset), float64(a.ActiveAt.Unix()))) } } diff --git a/rules/alerting_test.go b/rules/alerting_test.go index a9315b47ee..5ebd049f66 100644 --- a/rules/alerting_test.go +++ b/rules/alerting_test.go @@ -123,7 +123,7 @@ func TestAlertingRuleTemplateWithHistogram(t *testing.T) { ) evalTime := time.Now() - res, err := rule.Eval(context.TODO(), evalTime, q, nil, 0) + res, err := rule.Eval(context.TODO(), 0, evalTime, q, nil, 0) require.NoError(t, err) require.Len(t, res, 2) @@ -230,7 +230,7 @@ func TestAlertingRuleLabelsUpdate(t *testing.T) { t.Logf("case %d", i) evalTime := baseTime.Add(time.Duration(i) * time.Minute) result[0].T = timestamp.FromTime(evalTime) - res, err := rule.Eval(context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0) + res, err := rule.Eval(context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0) require.NoError(t, err) var filteredRes promql.Vector // After removing 'ALERTS_FOR_STATE' samples. @@ -247,7 +247,7 @@ func TestAlertingRuleLabelsUpdate(t *testing.T) { testutil.RequireEqual(t, result, filteredRes) } evalTime := baseTime.Add(time.Duration(len(results)) * time.Minute) - res, err := rule.Eval(context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0) + res, err := rule.Eval(context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0) require.NoError(t, err) require.Empty(t, res) } @@ -315,7 +315,7 @@ func TestAlertingRuleExternalLabelsInTemplate(t *testing.T) { var filteredRes promql.Vector // After removing 'ALERTS_FOR_STATE' samples. res, err := ruleWithoutExternalLabels.Eval( - context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0, + context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0, ) require.NoError(t, err) for _, smpl := range res { @@ -329,7 +329,7 @@ func TestAlertingRuleExternalLabelsInTemplate(t *testing.T) { } res, err = ruleWithExternalLabels.Eval( - context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0, + context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0, ) require.NoError(t, err) for _, smpl := range res { @@ -408,7 +408,7 @@ func TestAlertingRuleExternalURLInTemplate(t *testing.T) { var filteredRes promql.Vector // After removing 'ALERTS_FOR_STATE' samples. res, err := ruleWithoutExternalURL.Eval( - context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0, + context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0, ) require.NoError(t, err) for _, smpl := range res { @@ -422,7 +422,7 @@ func TestAlertingRuleExternalURLInTemplate(t *testing.T) { } res, err = ruleWithExternalURL.Eval( - context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0, + context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0, ) require.NoError(t, err) for _, smpl := range res { @@ -477,7 +477,7 @@ func TestAlertingRuleEmptyLabelFromTemplate(t *testing.T) { var filteredRes promql.Vector // After removing 'ALERTS_FOR_STATE' samples. res, err := rule.Eval( - context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0, + context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0, ) require.NoError(t, err) for _, smpl := range res { @@ -544,7 +544,7 @@ instance: {{ $v.Labels.instance }}, value: {{ printf "%.0f" $v.Value }}; close(getDoneCh) }() _, err = ruleWithQueryInTemplate.Eval( - context.TODO(), evalTime, slowQueryFunc, nil, 0, + context.TODO(), 0, evalTime, slowQueryFunc, nil, 0, ) require.NoError(t, err) } @@ -596,7 +596,7 @@ func TestAlertingRuleDuplicate(t *testing.T) { "", true, log.NewNopLogger(), ) - _, err := rule.Eval(ctx, now, EngineQueryFunc(engine, storage), nil, 0) + _, err := rule.Eval(ctx, 0, now, EngineQueryFunc(engine, storage), nil, 0) require.Error(t, err) require.EqualError(t, err, "vector contains metrics with the same labelset after applying alert labels") } @@ -644,7 +644,7 @@ func TestAlertingRuleLimit(t *testing.T) { evalTime := time.Unix(0, 0) for _, test := range tests { - switch _, err := rule.Eval(context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, test.limit); { + switch _, err := rule.Eval(context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, test.limit); { case err != nil: require.EqualError(t, err, test.err) case test.err != "": @@ -871,7 +871,7 @@ func TestKeepFiringFor(t *testing.T) { t.Logf("case %d", i) evalTime := baseTime.Add(time.Duration(i) * time.Minute) result[0].T = timestamp.FromTime(evalTime) - res, err := rule.Eval(context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0) + res, err := rule.Eval(context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0) require.NoError(t, err) var filteredRes promql.Vector // After removing 'ALERTS_FOR_STATE' samples. @@ -888,7 +888,7 @@ func TestKeepFiringFor(t *testing.T) { testutil.RequireEqual(t, result, filteredRes) } evalTime := baseTime.Add(time.Duration(len(results)) * time.Minute) - res, err := rule.Eval(context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0) + res, err := rule.Eval(context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0) require.NoError(t, err) require.Empty(t, res) } @@ -925,7 +925,7 @@ func TestPendingAndKeepFiringFor(t *testing.T) { baseTime := time.Unix(0, 0) result.T = timestamp.FromTime(baseTime) - res, err := rule.Eval(context.TODO(), baseTime, EngineQueryFunc(testEngine, storage), nil, 0) + res, err := rule.Eval(context.TODO(), 0, baseTime, EngineQueryFunc(testEngine, storage), nil, 0) require.NoError(t, err) require.Len(t, res, 2) @@ -940,7 +940,7 @@ func TestPendingAndKeepFiringFor(t *testing.T) { } evalTime := baseTime.Add(time.Minute) - res, err = rule.Eval(context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0) + res, err = rule.Eval(context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0) require.NoError(t, err) require.Empty(t, res) } @@ -974,7 +974,7 @@ func TestAlertingEvalWithOrigin(t *testing.T) { true, log.NewNopLogger(), ) - _, err = rule.Eval(ctx, now, func(ctx context.Context, qs string, _ time.Time) (promql.Vector, error) { + _, err = rule.Eval(ctx, 0, now, func(ctx context.Context, qs string, _ time.Time) (promql.Vector, error) { detail = FromOriginContext(ctx) return nil, nil }, nil, 0) diff --git a/rules/group.go b/rules/group.go index 1f4757de3c..9ae89789d0 100644 --- a/rules/group.go +++ b/rules/group.go @@ -47,6 +47,7 @@ type Group struct { name string file string interval time.Duration + queryOffset *time.Duration limit int rules []Rule seriesInPreviousEval []map[string]labels.Labels // One per Rule. @@ -90,6 +91,7 @@ type GroupOptions struct { Rules []Rule ShouldRestore bool Opts *ManagerOptions + QueryOffset *time.Duration done chan struct{} EvalIterationFunc GroupEvalIterationFunc } @@ -126,6 +128,7 @@ func NewGroup(o GroupOptions) *Group { name: o.Name, file: o.File, interval: o.Interval, + queryOffset: o.QueryOffset, limit: o.Limit, rules: o.Rules, shouldRestore: o.ShouldRestore, @@ -443,6 +446,8 @@ func (g *Group) Eval(ctx context.Context, ts time.Time) { wg sync.WaitGroup ) + ruleQueryOffset := g.QueryOffset() + for i, rule := range g.rules { select { case <-g.done: @@ -473,7 +478,7 @@ func (g *Group) Eval(ctx context.Context, ts time.Time) { g.metrics.EvalTotal.WithLabelValues(GroupKey(g.File(), g.Name())).Inc() - vector, err := rule.Eval(ctx, ts, g.opts.QueryFunc, g.opts.ExternalURL, g.Limit()) + vector, err := rule.Eval(ctx, ruleQueryOffset, ts, g.opts.QueryFunc, g.opts.ExternalURL, g.Limit()) if err != nil { rule.SetHealth(HealthBad) rule.SetLastError(err) @@ -562,7 +567,7 @@ func (g *Group) Eval(ctx context.Context, ts time.Time) { for metric, lset := range g.seriesInPreviousEval[i] { if _, ok := seriesReturned[metric]; !ok { // Series no longer exposed, mark it stale. - _, err = app.Append(0, lset, timestamp.FromTime(ts), math.Float64frombits(value.StaleNaN)) + _, err = app.Append(0, lset, timestamp.FromTime(ts.Add(-ruleQueryOffset)), math.Float64frombits(value.StaleNaN)) unwrappedErr := errors.Unwrap(err) if unwrappedErr == nil { unwrappedErr = err @@ -601,14 +606,27 @@ func (g *Group) Eval(ctx context.Context, ts time.Time) { g.cleanupStaleSeries(ctx, ts) } +func (g *Group) QueryOffset() time.Duration { + if g.queryOffset != nil { + return *g.queryOffset + } + + if g.opts.DefaultRuleQueryOffset != nil { + return g.opts.DefaultRuleQueryOffset() + } + + return time.Duration(0) +} + func (g *Group) cleanupStaleSeries(ctx context.Context, ts time.Time) { if len(g.staleSeries) == 0 { return } app := g.opts.Appendable.Appender(ctx) + queryOffset := g.QueryOffset() for _, s := range g.staleSeries { // Rule that produced series no longer configured, mark it stale. - _, err := app.Append(0, s, timestamp.FromTime(ts), math.Float64frombits(value.StaleNaN)) + _, err := app.Append(0, s, timestamp.FromTime(ts.Add(-queryOffset)), math.Float64frombits(value.StaleNaN)) unwrappedErr := errors.Unwrap(err) if unwrappedErr == nil { unwrappedErr = err diff --git a/rules/manager.go b/rules/manager.go index 165dca144e..063189e0ab 100644 --- a/rules/manager.go +++ b/rules/manager.go @@ -116,6 +116,7 @@ type ManagerOptions struct { ForGracePeriod time.Duration ResendDelay time.Duration GroupLoader GroupLoader + DefaultRuleQueryOffset func() time.Duration MaxConcurrentEvals int64 ConcurrentEvalsEnabled bool RuleConcurrencyController RuleConcurrencyController @@ -336,6 +337,7 @@ func (m *Manager) LoadGroups( Rules: rules, ShouldRestore: shouldRestore, Opts: m.opts, + QueryOffset: (*time.Duration)(rg.QueryOffset), done: m.done, EvalIterationFunc: groupEvalIterationFunc, }) diff --git a/rules/manager_test.go b/rules/manager_test.go index 1862b58075..11d1282bd3 100644 --- a/rules/manager_test.go +++ b/rules/manager_test.go @@ -16,9 +16,12 @@ package rules import ( "context" "fmt" + "io/fs" "math" "os" + "path" "sort" + "strconv" "sync" "testing" "time" @@ -161,7 +164,7 @@ func TestAlertingRule(t *testing.T) { evalTime := baseTime.Add(test.time) - res, err := rule.Eval(context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0) + res, err := rule.Eval(context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, 0) require.NoError(t, err) var filteredRes promql.Vector // After removing 'ALERTS_FOR_STATE' samples. @@ -191,152 +194,156 @@ func TestAlertingRule(t *testing.T) { } func TestForStateAddSamples(t *testing.T) { - storage := promqltest.LoadedStorage(t, ` + for _, queryOffset := range []time.Duration{0, time.Minute} { + t.Run(fmt.Sprintf("queryOffset %s", queryOffset.String()), func(t *testing.T) { + storage := promqltest.LoadedStorage(t, ` load 5m http_requests{job="app-server", instance="0", group="canary", severity="overwrite-me"} 75 85 95 105 105 95 85 http_requests{job="app-server", instance="1", group="canary", severity="overwrite-me"} 80 90 100 110 120 130 140 `) - t.Cleanup(func() { storage.Close() }) + t.Cleanup(func() { storage.Close() }) - expr, err := parser.ParseExpr(`http_requests{group="canary", job="app-server"} < 100`) - require.NoError(t, err) + expr, err := parser.ParseExpr(`http_requests{group="canary", job="app-server"} < 100`) + require.NoError(t, err) - rule := NewAlertingRule( - "HTTPRequestRateLow", - expr, - time.Minute, - 0, - labels.FromStrings("severity", "{{\"c\"}}ritical"), - labels.EmptyLabels(), labels.EmptyLabels(), "", true, nil, - ) - result := promql.Vector{ - promql.Sample{ - Metric: labels.FromStrings( - "__name__", "ALERTS_FOR_STATE", - "alertname", "HTTPRequestRateLow", - "group", "canary", - "instance", "0", - "job", "app-server", - "severity", "critical", - ), - F: 1, - }, - promql.Sample{ - Metric: labels.FromStrings( - "__name__", "ALERTS_FOR_STATE", - "alertname", "HTTPRequestRateLow", - "group", "canary", - "instance", "1", - "job", "app-server", - "severity", "critical", - ), - F: 1, - }, - promql.Sample{ - Metric: labels.FromStrings( - "__name__", "ALERTS_FOR_STATE", - "alertname", "HTTPRequestRateLow", - "group", "canary", - "instance", "0", - "job", "app-server", - "severity", "critical", - ), - F: 1, - }, - promql.Sample{ - Metric: labels.FromStrings( - "__name__", "ALERTS_FOR_STATE", - "alertname", "HTTPRequestRateLow", - "group", "canary", - "instance", "1", - "job", "app-server", - "severity", "critical", - ), - F: 1, - }, - } - - baseTime := time.Unix(0, 0) - - tests := []struct { - time time.Duration - result promql.Vector - persistThisTime bool // If true, it means this 'time' is persisted for 'for'. - }{ - { - time: 0, - result: append(promql.Vector{}, result[:2]...), - persistThisTime: true, - }, - { - time: 5 * time.Minute, - result: append(promql.Vector{}, result[2:]...), - }, - { - time: 10 * time.Minute, - result: append(promql.Vector{}, result[2:3]...), - }, - { - time: 15 * time.Minute, - result: nil, - }, - { - time: 20 * time.Minute, - result: nil, - }, - { - time: 25 * time.Minute, - result: append(promql.Vector{}, result[:1]...), - persistThisTime: true, - }, - { - time: 30 * time.Minute, - result: append(promql.Vector{}, result[2:3]...), - }, - } - - var forState float64 - for i, test := range tests { - t.Logf("case %d", i) - evalTime := baseTime.Add(test.time) - - if test.persistThisTime { - forState = float64(evalTime.Unix()) - } - if test.result == nil { - forState = float64(value.StaleNaN) - } - - res, err := rule.Eval(context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, 0) - require.NoError(t, err) - - var filteredRes promql.Vector // After removing 'ALERTS' samples. - for _, smpl := range res { - smplName := smpl.Metric.Get("__name__") - if smplName == "ALERTS_FOR_STATE" { - filteredRes = append(filteredRes, smpl) - } else { - // If not 'ALERTS_FOR_STATE', it has to be 'ALERTS'. - require.Equal(t, "ALERTS", smplName) + rule := NewAlertingRule( + "HTTPRequestRateLow", + expr, + time.Minute, + 0, + labels.FromStrings("severity", "{{\"c\"}}ritical"), + labels.EmptyLabels(), labels.EmptyLabels(), "", true, nil, + ) + result := promql.Vector{ + promql.Sample{ + Metric: labels.FromStrings( + "__name__", "ALERTS_FOR_STATE", + "alertname", "HTTPRequestRateLow", + "group", "canary", + "instance", "0", + "job", "app-server", + "severity", "critical", + ), + F: 1, + }, + promql.Sample{ + Metric: labels.FromStrings( + "__name__", "ALERTS_FOR_STATE", + "alertname", "HTTPRequestRateLow", + "group", "canary", + "instance", "1", + "job", "app-server", + "severity", "critical", + ), + F: 1, + }, + promql.Sample{ + Metric: labels.FromStrings( + "__name__", "ALERTS_FOR_STATE", + "alertname", "HTTPRequestRateLow", + "group", "canary", + "instance", "0", + "job", "app-server", + "severity", "critical", + ), + F: 1, + }, + promql.Sample{ + Metric: labels.FromStrings( + "__name__", "ALERTS_FOR_STATE", + "alertname", "HTTPRequestRateLow", + "group", "canary", + "instance", "1", + "job", "app-server", + "severity", "critical", + ), + F: 1, + }, } - } - for i := range test.result { - test.result[i].T = timestamp.FromTime(evalTime) - // Updating the expected 'for' state. - if test.result[i].F >= 0 { - test.result[i].F = forState - } - } - require.Equal(t, len(test.result), len(filteredRes), "%d. Number of samples in expected and actual output don't match (%d vs. %d)", i, len(test.result), len(res)) - sort.Slice(filteredRes, func(i, j int) bool { - return labels.Compare(filteredRes[i].Metric, filteredRes[j].Metric) < 0 + baseTime := time.Unix(0, 0) + + tests := []struct { + time time.Duration + result promql.Vector + persistThisTime bool // If true, it means this 'time' is persisted for 'for'. + }{ + { + time: 0, + result: append(promql.Vector{}, result[:2]...), + persistThisTime: true, + }, + { + time: 5 * time.Minute, + result: append(promql.Vector{}, result[2:]...), + }, + { + time: 10 * time.Minute, + result: append(promql.Vector{}, result[2:3]...), + }, + { + time: 15 * time.Minute, + result: nil, + }, + { + time: 20 * time.Minute, + result: nil, + }, + { + time: 25 * time.Minute, + result: append(promql.Vector{}, result[:1]...), + persistThisTime: true, + }, + { + time: 30 * time.Minute, + result: append(promql.Vector{}, result[2:3]...), + }, + } + + var forState float64 + for i, test := range tests { + t.Logf("case %d", i) + evalTime := baseTime.Add(test.time).Add(queryOffset) + + if test.persistThisTime { + forState = float64(evalTime.Unix()) + } + if test.result == nil { + forState = float64(value.StaleNaN) + } + + res, err := rule.Eval(context.TODO(), queryOffset, evalTime, EngineQueryFunc(testEngine, storage), nil, 0) + require.NoError(t, err) + + var filteredRes promql.Vector // After removing 'ALERTS' samples. + for _, smpl := range res { + smplName := smpl.Metric.Get("__name__") + if smplName == "ALERTS_FOR_STATE" { + filteredRes = append(filteredRes, smpl) + } else { + // If not 'ALERTS_FOR_STATE', it has to be 'ALERTS'. + require.Equal(t, "ALERTS", smplName) + } + } + for i := range test.result { + test.result[i].T = timestamp.FromTime(evalTime.Add(-queryOffset)) + // Updating the expected 'for' state. + if test.result[i].F >= 0 { + test.result[i].F = forState + } + } + require.Equal(t, len(test.result), len(filteredRes), "%d. Number of samples in expected and actual output don't match (%d vs. %d)", i, len(test.result), len(res)) + + sort.Slice(filteredRes, func(i, j int) bool { + return labels.Compare(filteredRes[i].Metric, filteredRes[j].Metric) < 0 + }) + prom_testutil.RequireEqual(t, test.result, filteredRes) + + for _, aa := range rule.ActiveAlerts() { + require.Zero(t, aa.Labels.Get(model.MetricNameLabel), "%s label set on active alert: %s", model.MetricNameLabel, aa.Labels) + } + } }) - prom_testutil.RequireEqual(t, test.result, filteredRes) - - for _, aa := range rule.ActiveAlerts() { - require.Zero(t, aa.Labels.Get(model.MetricNameLabel), "%s label set on active alert: %s", model.MetricNameLabel, aa.Labels) - } } } @@ -348,243 +355,251 @@ func sortAlerts(items []*Alert) { } func TestForStateRestore(t *testing.T) { - storage := promqltest.LoadedStorage(t, ` + for _, queryOffset := range []time.Duration{0, time.Minute} { + t.Run(fmt.Sprintf("queryOffset %s", queryOffset.String()), func(t *testing.T) { + storage := promqltest.LoadedStorage(t, ` load 5m http_requests{job="app-server", instance="0", group="canary", severity="overwrite-me"} 75 85 50 0 0 25 0 0 40 0 120 http_requests{job="app-server", instance="1", group="canary", severity="overwrite-me"} 125 90 60 0 0 25 0 0 40 0 130 `) - t.Cleanup(func() { storage.Close() }) + t.Cleanup(func() { storage.Close() }) - expr, err := parser.ParseExpr(`http_requests{group="canary", job="app-server"} < 100`) - require.NoError(t, err) + expr, err := parser.ParseExpr(`http_requests{group="canary", job="app-server"} < 100`) + require.NoError(t, err) - opts := &ManagerOptions{ - QueryFunc: EngineQueryFunc(testEngine, storage), - Appendable: storage, - Queryable: storage, - Context: context.Background(), - Logger: log.NewNopLogger(), - NotifyFunc: func(ctx context.Context, expr string, alerts ...*Alert) {}, - OutageTolerance: 30 * time.Minute, - ForGracePeriod: 10 * time.Minute, - } - - alertForDuration := 25 * time.Minute - // Initial run before prometheus goes down. - rule := NewAlertingRule( - "HTTPRequestRateLow", - expr, - alertForDuration, - 0, - labels.FromStrings("severity", "critical"), - labels.EmptyLabels(), labels.EmptyLabels(), "", true, nil, - ) - - group := NewGroup(GroupOptions{ - Name: "default", - Interval: time.Second, - Rules: []Rule{rule}, - ShouldRestore: true, - Opts: opts, - }) - groups := make(map[string]*Group) - groups["default;"] = group - - initialRuns := []time.Duration{0, 5 * time.Minute} - - baseTime := time.Unix(0, 0) - for _, duration := range initialRuns { - evalTime := baseTime.Add(duration) - group.Eval(context.TODO(), evalTime) - } - - // Prometheus goes down here. We create new rules and groups. - type testInput struct { - name string - restoreDuration time.Duration - expectedAlerts []*Alert - - num int - noRestore bool - gracePeriod bool - downDuration time.Duration - before func() - } - - tests := []testInput{ - { - name: "normal restore (alerts were not firing)", - restoreDuration: 15 * time.Minute, - expectedAlerts: rule.ActiveAlerts(), - downDuration: 10 * time.Minute, - }, - { - name: "outage tolerance", - restoreDuration: 40 * time.Minute, - noRestore: true, - num: 2, - }, - { - name: "no active alerts", - restoreDuration: 50 * time.Minute, - expectedAlerts: []*Alert{}, - }, - { - name: "test the grace period", - restoreDuration: 25 * time.Minute, - expectedAlerts: []*Alert{}, - gracePeriod: true, - before: func() { - for _, duration := range []time.Duration{10 * time.Minute, 15 * time.Minute, 20 * time.Minute} { - evalTime := baseTime.Add(duration) - group.Eval(context.TODO(), evalTime) - } - }, - num: 2, - }, - } - - for _, tt := range tests { - t.Run(tt.name, func(t *testing.T) { - if tt.before != nil { - tt.before() + opts := &ManagerOptions{ + QueryFunc: EngineQueryFunc(testEngine, storage), + Appendable: storage, + Queryable: storage, + Context: context.Background(), + Logger: log.NewNopLogger(), + NotifyFunc: func(ctx context.Context, expr string, alerts ...*Alert) {}, + OutageTolerance: 30 * time.Minute, + ForGracePeriod: 10 * time.Minute, } - newRule := NewAlertingRule( + alertForDuration := 25 * time.Minute + // Initial run before prometheus goes down. + rule := NewAlertingRule( "HTTPRequestRateLow", expr, alertForDuration, 0, labels.FromStrings("severity", "critical"), - labels.EmptyLabels(), labels.EmptyLabels(), "", false, nil, + labels.EmptyLabels(), labels.EmptyLabels(), "", true, nil, ) - newGroup := NewGroup(GroupOptions{ + + group := NewGroup(GroupOptions{ Name: "default", Interval: time.Second, - Rules: []Rule{newRule}, + Rules: []Rule{rule}, ShouldRestore: true, Opts: opts, }) + groups := make(map[string]*Group) + groups["default;"] = group - newGroups := make(map[string]*Group) - newGroups["default;"] = newGroup + initialRuns := []time.Duration{0, 5 * time.Minute} - restoreTime := baseTime.Add(tt.restoreDuration) - // First eval before restoration. - newGroup.Eval(context.TODO(), restoreTime) - // Restore happens here. - newGroup.RestoreForState(restoreTime) - - got := newRule.ActiveAlerts() - for _, aa := range got { - require.Zero(t, aa.Labels.Get(model.MetricNameLabel), "%s label set on active alert: %s", model.MetricNameLabel, aa.Labels) + baseTime := time.Unix(0, 0) + for _, duration := range initialRuns { + evalTime := baseTime.Add(duration) + group.Eval(context.TODO(), evalTime) } - sort.Slice(got, func(i, j int) bool { - return labels.Compare(got[i].Labels, got[j].Labels) < 0 - }) - // In all cases, we expect the restoration process to have completed. - require.Truef(t, newRule.Restored(), "expected the rule restoration process to have completed") + // Prometheus goes down here. We create new rules and groups. + type testInput struct { + name string + restoreDuration time.Duration + expectedAlerts []*Alert - // Checking if we have restored it correctly. - switch { - case tt.noRestore: - require.Len(t, got, tt.num) - for _, e := range got { - require.Equal(t, e.ActiveAt, restoreTime) - } - case tt.gracePeriod: + num int + noRestore bool + gracePeriod bool + downDuration time.Duration + before func() + } - require.Len(t, got, tt.num) - for _, e := range got { - require.Equal(t, opts.ForGracePeriod, e.ActiveAt.Add(alertForDuration).Sub(restoreTime)) - } - default: - exp := tt.expectedAlerts - require.Equal(t, len(exp), len(got)) - sortAlerts(exp) - sortAlerts(got) - for i, e := range exp { - require.Equal(t, e.Labels, got[i].Labels) + tests := []testInput{ + { + name: "normal restore (alerts were not firing)", + restoreDuration: 15 * time.Minute, + expectedAlerts: rule.ActiveAlerts(), + downDuration: 10 * time.Minute, + }, + { + name: "outage tolerance", + restoreDuration: 40 * time.Minute, + noRestore: true, + num: 2, + }, + { + name: "no active alerts", + restoreDuration: 50 * time.Minute, + expectedAlerts: []*Alert{}, + }, + { + name: "test the grace period", + restoreDuration: 25 * time.Minute, + expectedAlerts: []*Alert{}, + gracePeriod: true, + before: func() { + for _, duration := range []time.Duration{10 * time.Minute, 15 * time.Minute, 20 * time.Minute} { + evalTime := baseTime.Add(duration) + group.Eval(context.TODO(), evalTime) + } + }, + num: 2, + }, + } - // Difference in time should be within 1e6 ns, i.e. 1ms - // (due to conversion between ns & ms, float64 & int64). - activeAtDiff := float64(e.ActiveAt.Unix() + int64(tt.downDuration/time.Second) - got[i].ActiveAt.Unix()) - require.Equal(t, 0.0, math.Abs(activeAtDiff), "'for' state restored time is wrong") - } + for _, tt := range tests { + t.Run(tt.name, func(t *testing.T) { + if tt.before != nil { + tt.before() + } + + newRule := NewAlertingRule( + "HTTPRequestRateLow", + expr, + alertForDuration, + 0, + labels.FromStrings("severity", "critical"), + labels.EmptyLabels(), labels.EmptyLabels(), "", false, nil, + ) + newGroup := NewGroup(GroupOptions{ + Name: "default", + Interval: time.Second, + Rules: []Rule{newRule}, + ShouldRestore: true, + Opts: opts, + QueryOffset: &queryOffset, + }) + + newGroups := make(map[string]*Group) + newGroups["default;"] = newGroup + + restoreTime := baseTime.Add(tt.restoreDuration).Add(queryOffset) + // First eval before restoration. + newGroup.Eval(context.TODO(), restoreTime) + // Restore happens here. + newGroup.RestoreForState(restoreTime) + + got := newRule.ActiveAlerts() + for _, aa := range got { + require.Zero(t, aa.Labels.Get(model.MetricNameLabel), "%s label set on active alert: %s", model.MetricNameLabel, aa.Labels) + } + sort.Slice(got, func(i, j int) bool { + return labels.Compare(got[i].Labels, got[j].Labels) < 0 + }) + + // In all cases, we expect the restoration process to have completed. + require.Truef(t, newRule.Restored(), "expected the rule restoration process to have completed") + + // Checking if we have restored it correctly. + switch { + case tt.noRestore: + require.Len(t, got, tt.num) + for _, e := range got { + require.Equal(t, e.ActiveAt, restoreTime) + } + case tt.gracePeriod: + + require.Len(t, got, tt.num) + for _, e := range got { + require.Equal(t, opts.ForGracePeriod, e.ActiveAt.Add(alertForDuration).Sub(restoreTime)) + } + default: + exp := tt.expectedAlerts + require.Equal(t, len(exp), len(got)) + sortAlerts(exp) + sortAlerts(got) + for i, e := range exp { + require.Equal(t, e.Labels, got[i].Labels) + + // Difference in time should be within 1e6 ns, i.e. 1ms + // (due to conversion between ns & ms, float64 & int64). + activeAtDiff := queryOffset.Seconds() + float64(e.ActiveAt.Unix()+int64(tt.downDuration/time.Second)-got[i].ActiveAt.Unix()) + require.Equal(t, 0.0, math.Abs(activeAtDiff), "'for' state restored time is wrong") + } + } + }) } }) } } func TestStaleness(t *testing.T) { - st := teststorage.New(t) - defer st.Close() - engineOpts := promql.EngineOpts{ - Logger: nil, - Reg: nil, - MaxSamples: 10, - Timeout: 10 * time.Second, + for _, queryOffset := range []time.Duration{0, time.Minute} { + st := teststorage.New(t) + defer st.Close() + engineOpts := promql.EngineOpts{ + Logger: nil, + Reg: nil, + MaxSamples: 10, + Timeout: 10 * time.Second, + } + engine := promql.NewEngine(engineOpts) + opts := &ManagerOptions{ + QueryFunc: EngineQueryFunc(engine, st), + Appendable: st, + Queryable: st, + Context: context.Background(), + Logger: log.NewNopLogger(), + } + + expr, err := parser.ParseExpr("a + 1") + require.NoError(t, err) + rule := NewRecordingRule("a_plus_one", expr, labels.Labels{}) + group := NewGroup(GroupOptions{ + Name: "default", + Interval: time.Second, + Rules: []Rule{rule}, + ShouldRestore: true, + Opts: opts, + QueryOffset: &queryOffset, + }) + + // A time series that has two samples and then goes stale. + app := st.Appender(context.Background()) + app.Append(0, labels.FromStrings(model.MetricNameLabel, "a"), 0, 1) + app.Append(0, labels.FromStrings(model.MetricNameLabel, "a"), 1000, 2) + app.Append(0, labels.FromStrings(model.MetricNameLabel, "a"), 2000, math.Float64frombits(value.StaleNaN)) + + err = app.Commit() + require.NoError(t, err) + + ctx := context.Background() + + // Execute 3 times, 1 second apart. + group.Eval(ctx, time.Unix(0, 0).Add(queryOffset)) + group.Eval(ctx, time.Unix(1, 0).Add(queryOffset)) + group.Eval(ctx, time.Unix(2, 0).Add(queryOffset)) + + querier, err := st.Querier(0, 2000) + require.NoError(t, err) + defer querier.Close() + + matcher, err := labels.NewMatcher(labels.MatchEqual, model.MetricNameLabel, "a_plus_one") + require.NoError(t, err) + + set := querier.Select(ctx, false, nil, matcher) + samples, err := readSeriesSet(set) + require.NoError(t, err) + + metric := labels.FromStrings(model.MetricNameLabel, "a_plus_one").String() + metricSample, ok := samples[metric] + + require.True(t, ok, "Series %s not returned.", metric) + require.True(t, value.IsStaleNaN(metricSample[2].F), "Appended second sample not as expected. Wanted: stale NaN Got: %x", math.Float64bits(metricSample[2].F)) + metricSample[2].F = 42 // require.Equal cannot handle NaN. + + want := map[string][]promql.FPoint{ + metric: {{T: 0, F: 2}, {T: 1000, F: 3}, {T: 2000, F: 42}}, + } + + require.Equal(t, want, samples) } - engine := promql.NewEngine(engineOpts) - opts := &ManagerOptions{ - QueryFunc: EngineQueryFunc(engine, st), - Appendable: st, - Queryable: st, - Context: context.Background(), - Logger: log.NewNopLogger(), - } - - expr, err := parser.ParseExpr("a + 1") - require.NoError(t, err) - rule := NewRecordingRule("a_plus_one", expr, labels.Labels{}) - group := NewGroup(GroupOptions{ - Name: "default", - Interval: time.Second, - Rules: []Rule{rule}, - ShouldRestore: true, - Opts: opts, - }) - - // A time series that has two samples and then goes stale. - app := st.Appender(context.Background()) - app.Append(0, labels.FromStrings(model.MetricNameLabel, "a"), 0, 1) - app.Append(0, labels.FromStrings(model.MetricNameLabel, "a"), 1000, 2) - app.Append(0, labels.FromStrings(model.MetricNameLabel, "a"), 2000, math.Float64frombits(value.StaleNaN)) - - err = app.Commit() - require.NoError(t, err) - - ctx := context.Background() - - // Execute 3 times, 1 second apart. - group.Eval(ctx, time.Unix(0, 0)) - group.Eval(ctx, time.Unix(1, 0)) - group.Eval(ctx, time.Unix(2, 0)) - - querier, err := st.Querier(0, 2000) - require.NoError(t, err) - defer querier.Close() - - matcher, err := labels.NewMatcher(labels.MatchEqual, model.MetricNameLabel, "a_plus_one") - require.NoError(t, err) - - set := querier.Select(ctx, false, nil, matcher) - samples, err := readSeriesSet(set) - require.NoError(t, err) - - metric := labels.FromStrings(model.MetricNameLabel, "a_plus_one").String() - metricSample, ok := samples[metric] - - require.True(t, ok, "Series %s not returned.", metric) - require.True(t, value.IsStaleNaN(metricSample[2].F), "Appended second sample not as expected. Wanted: stale NaN Got: %x", math.Float64bits(metricSample[2].F)) - metricSample[2].F = 42 // require.Equal cannot handle NaN. - - want := map[string][]promql.FPoint{ - metric: {{T: 0, F: 2}, {T: 1000, F: 3}, {T: 2000, F: 42}}, - } - - require.Equal(t, want, samples) } // Convert a SeriesSet into a form usable with require.Equal. @@ -608,6 +623,46 @@ func readSeriesSet(ss storage.SeriesSet) (map[string][]promql.FPoint, error) { return result, ss.Err() } +func TestGroup_QueryOffset(t *testing.T) { + config := ` +groups: + - name: group1 + query_offset: 2m + - name: group2 + query_offset: 0s + - name: group3 +` + + dir := t.TempDir() + fname := path.Join(dir, "rules.yaml") + err := os.WriteFile(fname, []byte(config), fs.ModePerm) + require.NoError(t, err) + + m := NewManager(&ManagerOptions{ + Logger: log.NewNopLogger(), + DefaultRuleQueryOffset: func() time.Duration { + return time.Minute + }, + }) + m.start() + err = m.Update(time.Second, []string{fname}, labels.EmptyLabels(), "", nil) + require.NoError(t, err) + + rgs := m.RuleGroups() + sort.Slice(rgs, func(i, j int) bool { + return rgs[i].Name() < rgs[j].Name() + }) + + // From config. + require.Equal(t, 2*time.Minute, rgs[0].QueryOffset()) + // Setting 0 in config is detected. + require.Equal(t, time.Duration(0), rgs[1].QueryOffset()) + // Default when nothing is set. + require.Equal(t, time.Minute, rgs[2].QueryOffset()) + + m.Stop() +} + func TestCopyState(t *testing.T) { oldGroup := &Group{ rules: []Rule{ @@ -1361,7 +1416,7 @@ func TestNativeHistogramsInRecordingRules(t *testing.T) { ts := time.Now() app := db.Appender(context.Background()) for i, h := range hists { - l := labels.FromStrings("__name__", "histogram_metric", "idx", fmt.Sprintf("%d", i)) + l := labels.FromStrings("__name__", "histogram_metric", "idx", strconv.Itoa(i)) _, err := app.AppendHistogram(0, l, ts.UnixMilli(), h.Copy(), nil) require.NoError(t, err) } @@ -2043,7 +2098,7 @@ func TestBoundedRuleEvalConcurrency(t *testing.T) { require.EqualValues(t, maxInflight.Load(), int32(maxConcurrency)+int32(groupCount)) } -const artificialDelay = 10 * time.Millisecond +const artificialDelay = 15 * time.Millisecond func optsFactory(storage storage.Storage, maxInflight, inflightQueries *atomic.Int32, maxConcurrent int64) *ManagerOptions { var inflightMu sync.Mutex diff --git a/rules/origin_test.go b/rules/origin_test.go index ca466301dd..75c83f9a4e 100644 --- a/rules/origin_test.go +++ b/rules/origin_test.go @@ -31,7 +31,7 @@ type unknownRule struct{} func (u unknownRule) Name() string { return "" } func (u unknownRule) Labels() labels.Labels { return labels.EmptyLabels() } -func (u unknownRule) Eval(context.Context, time.Time, QueryFunc, *url.URL, int) (promql.Vector, error) { +func (u unknownRule) Eval(context.Context, time.Duration, time.Time, QueryFunc, *url.URL, int) (promql.Vector, error) { return nil, nil } func (u unknownRule) String() string { return "" } diff --git a/rules/recording.go b/rules/recording.go index e2b0a31a03..17a75fdd1a 100644 --- a/rules/recording.go +++ b/rules/recording.go @@ -77,10 +77,9 @@ func (rule *RecordingRule) Labels() labels.Labels { } // Eval evaluates the rule and then overrides the metric names and labels accordingly. -func (rule *RecordingRule) Eval(ctx context.Context, ts time.Time, query QueryFunc, _ *url.URL, limit int) (promql.Vector, error) { +func (rule *RecordingRule) Eval(ctx context.Context, queryOffset time.Duration, ts time.Time, query QueryFunc, _ *url.URL, limit int) (promql.Vector, error) { ctx = NewOriginContext(ctx, NewRuleDetail(rule)) - - vector, err := query(ctx, rule.vector.String(), ts) + vector, err := query(ctx, rule.vector.String(), ts.Add(-queryOffset)) if err != nil { return nil, err } diff --git a/rules/recording_test.go b/rules/recording_test.go index 49f37b1ac9..fdddd4e02e 100644 --- a/rules/recording_test.go +++ b/rules/recording_test.go @@ -126,7 +126,7 @@ func TestRuleEval(t *testing.T) { for _, scenario := range ruleEvalTestScenarios { t.Run(scenario.name, func(t *testing.T) { rule := NewRecordingRule("test_rule", scenario.expr, scenario.ruleLabels) - result, err := rule.Eval(context.TODO(), ruleEvaluationTime, EngineQueryFunc(testEngine, storage), nil, 0) + result, err := rule.Eval(context.TODO(), 0, ruleEvaluationTime, EngineQueryFunc(testEngine, storage), nil, 0) require.NoError(t, err) testutil.RequireEqual(t, scenario.expected, result) }) @@ -144,7 +144,7 @@ func BenchmarkRuleEval(b *testing.B) { b.ResetTimer() for i := 0; i < b.N; i++ { - _, err := rule.Eval(context.TODO(), ruleEvaluationTime, EngineQueryFunc(testEngine, storage), nil, 0) + _, err := rule.Eval(context.TODO(), 0, ruleEvaluationTime, EngineQueryFunc(testEngine, storage), nil, 0) if err != nil { require.NoError(b, err) } @@ -173,7 +173,7 @@ func TestRuleEvalDuplicate(t *testing.T) { expr, _ := parser.ParseExpr(`vector(0) or label_replace(vector(0),"test","x","","")`) rule := NewRecordingRule("foo", expr, labels.FromStrings("test", "test")) - _, err := rule.Eval(ctx, now, EngineQueryFunc(engine, storage), nil, 0) + _, err := rule.Eval(ctx, 0, now, EngineQueryFunc(engine, storage), nil, 0) require.Error(t, err) require.EqualError(t, err, "vector contains metrics with the same labelset after applying rule labels") } @@ -215,7 +215,7 @@ func TestRecordingRuleLimit(t *testing.T) { evalTime := time.Unix(0, 0) for _, test := range tests { - switch _, err := rule.Eval(context.TODO(), evalTime, EngineQueryFunc(testEngine, storage), nil, test.limit); { + switch _, err := rule.Eval(context.TODO(), 0, evalTime, EngineQueryFunc(testEngine, storage), nil, test.limit); { case err != nil: require.EqualError(t, err, test.err) case test.err != "": @@ -243,7 +243,7 @@ func TestRecordingEvalWithOrigin(t *testing.T) { require.NoError(t, err) rule := NewRecordingRule(name, expr, lbs) - _, err = rule.Eval(ctx, now, func(ctx context.Context, qs string, _ time.Time) (promql.Vector, error) { + _, err = rule.Eval(ctx, 0, now, func(ctx context.Context, qs string, _ time.Time) (promql.Vector, error) { detail = FromOriginContext(ctx) return nil, nil }, nil, 0) diff --git a/rules/rule.go b/rules/rule.go index 59af3e0bba..687c03d000 100644 --- a/rules/rule.go +++ b/rules/rule.go @@ -40,7 +40,7 @@ type Rule interface { // Labels of the rule. Labels() labels.Labels // Eval evaluates the rule, including any associated recording or alerting actions. - Eval(context.Context, time.Time, QueryFunc, *url.URL, int) (promql.Vector, error) + Eval(ctx context.Context, queryOffset time.Duration, evaluationTime time.Time, queryFunc QueryFunc, externalURL *url.URL, limit int) (promql.Vector, error) // String returns a human-readable string representation of the rule. String() string // Query returns the rule query expression. diff --git a/scrape/scrape_test.go b/scrape/scrape_test.go index 51bd377e47..b5a31cb650 100644 --- a/scrape/scrape_test.go +++ b/scrape/scrape_test.go @@ -1285,7 +1285,7 @@ func TestScrapeLoopCacheMemoryExhaustionProtection(t *testing.T) { for i := 0; i < 500; i++ { s = fmt.Sprintf("%smetric_%d_%d 42\n", s, i, numScrapes) } - w.Write([]byte(fmt.Sprintf(s + "&"))) + w.Write([]byte(s + "&")) } else { cancel() } diff --git a/scrape/target_test.go b/scrape/target_test.go index f91e31050a..b43ff24063 100644 --- a/scrape/target_test.go +++ b/scrape/target_test.go @@ -21,6 +21,7 @@ import ( "net/http/httptest" "net/url" "os" + "strconv" "strings" "testing" "time" @@ -67,7 +68,7 @@ func TestTargetOffset(t *testing.T) { // Calculate offsets for 10000 different targets. for i := range offsets { target := newTestTarget("example.com:80", 0, labels.FromStrings( - "label", fmt.Sprintf("%d", i), + "label", strconv.Itoa(i), )) offsets[i] = target.offset(interval, offsetSeed) } diff --git a/scripts/golangci-lint.yml b/scripts/golangci-lint.yml index b36f71c3c1..5ceb590991 100644 --- a/scripts/golangci-lint.yml +++ b/scripts/golangci-lint.yml @@ -24,8 +24,8 @@ jobs: runs-on: ubuntu-latest steps: - name: Checkout repository - uses: actions/checkout@9bb56186c3b09b4f86b1c65136769dd318469633 # v4.1.2 - - name: install Go + uses: actions/checkout@a5ac7e51b41094c92402da3b24376905380afc29 # v4.1.6 + - name: Install Go uses: actions/setup-go@0c52d547c9bc32b1aa3301fd7a9cb496313a4491 # v5.0.0 with: go-version: 1.22.x @@ -33,6 +33,7 @@ jobs: run: sudo apt-get update && sudo apt-get -y install libsnmp-dev if: github.repository == 'prometheus/snmp_exporter' - name: Lint - uses: golangci/golangci-lint-action@9d1e0624a798bb64f6c3cea93db47765312263dc # v5.1.0 + uses: golangci/golangci-lint-action@a4f60bb28d35aeee14e6880718e0c85ff1882e64 # v6.0.1 with: - version: v1.56.2 + args: --verbose + version: v1.59.0 diff --git a/storage/remote/azuread/azuread.go b/storage/remote/azuread/azuread.go index e2058fb54d..58520c6a5d 100644 --- a/storage/remote/azuread/azuread.go +++ b/storage/remote/azuread/azuread.go @@ -75,7 +75,7 @@ type AzureADConfig struct { //nolint:revive // exported. // OAuth is the oauth config that is being used to authenticate. OAuth *OAuthConfig `yaml:"oauth,omitempty"` - // OAuth is the oauth config that is being used to authenticate. + // SDK is the SDK config that is being used to authenticate. SDK *SDKConfig `yaml:"sdk,omitempty"` // Cloud is the Azure cloud in which the service is running. Example: AzurePublic/AzureGovernment/AzureChina. diff --git a/storage/remote/queue_manager_test.go b/storage/remote/queue_manager_test.go index e32a3ace05..6121fb6c03 100644 --- a/storage/remote/queue_manager_test.go +++ b/storage/remote/queue_manager_test.go @@ -468,7 +468,7 @@ func TestReleaseNoninternedString(t *testing.T) { m.StoreSeries([]record.RefSeries{ { Ref: chunks.HeadSeriesRef(i), - Labels: labels.FromStrings("asdf", fmt.Sprintf("%d", i)), + Labels: labels.FromStrings("asdf", strconv.Itoa(i)), }, }, 0) m.SeriesReset(1) diff --git a/tsdb/agent/series_test.go b/tsdb/agent/series_test.go index ae327d8582..bc5a4af5d3 100644 --- a/tsdb/agent/series_test.go +++ b/tsdb/agent/series_test.go @@ -14,8 +14,8 @@ package agent import ( - "fmt" "math" + "strconv" "sync" "testing" "time" @@ -53,7 +53,7 @@ func TestNoDeadlock(t *testing.T) { series := &memSeries{ ref: chunks.HeadSeriesRef(i), lset: labels.FromMap(map[string]string{ - "id": fmt.Sprintf("%d", i), + "id": strconv.Itoa(i), }), } stripeSeries.Set(series.lset.Hash(), series) diff --git a/tsdb/chunks/head_chunks.go b/tsdb/chunks/head_chunks.go index 66dbb07b71..6c8707c57b 100644 --- a/tsdb/chunks/head_chunks.go +++ b/tsdb/chunks/head_chunks.go @@ -381,6 +381,33 @@ func listChunkFiles(dir string) (map[int]string, error) { return res, nil } +// HardLinkChunkFiles creates hardlinks for chunk files from src to dst. +// It does nothing if src doesn't exist and ensures dst is created if not. +func HardLinkChunkFiles(src, dst string) error { + _, err := os.Stat(src) + if os.IsNotExist(err) { + return nil + } + if err != nil { + return fmt.Errorf("check source chunks dir: %w", err) + } + if err := os.MkdirAll(dst, 0o777); err != nil { + return fmt.Errorf("set up destination chunks dir: %w", err) + } + files, err := listChunkFiles(src) + if err != nil { + return fmt.Errorf("list chunks: %w", err) + } + for _, filePath := range files { + _, fileName := filepath.Split(filePath) + err := os.Link(filepath.Join(src, fileName), filepath.Join(dst, fileName)) + if err != nil { + return fmt.Errorf("hardlink a chunk: %w", err) + } + } + return nil +} + // repairLastChunkFile deletes the last file if it's empty. // Because we don't fsync when creating these files, we could end // up with an empty file at the end during an abrupt shutdown. diff --git a/tsdb/chunks/queue_test.go b/tsdb/chunks/queue_test.go index 5756e45856..9f761a5f3b 100644 --- a/tsdb/chunks/queue_test.go +++ b/tsdb/chunks/queue_test.go @@ -55,7 +55,7 @@ func (q *writeJobQueue) assertInvariants(t *testing.T) { require.Len(t, s.segment, s.nextWrite) } // Last segment must have at least one element, or we wouldn't have created it. - require.Greater(t, s.nextWrite, 0) + require.Positive(t, s.nextWrite) } require.Equal(t, q.size, totalSize) diff --git a/tsdb/compact_test.go b/tsdb/compact_test.go index b2d2ea6e7f..7a353a556a 100644 --- a/tsdb/compact_test.go +++ b/tsdb/compact_test.go @@ -22,6 +22,7 @@ import ( "os" "path" "path/filepath" + "strconv" "sync" "testing" "time" @@ -1129,7 +1130,7 @@ func BenchmarkCompactionFromHead(b *testing.B) { for ln := 0; ln < labelNames; ln++ { app := h.Appender(context.Background()) for lv := 0; lv < labelValues; lv++ { - app.Append(0, labels.FromStrings(fmt.Sprintf("%d", ln), fmt.Sprintf("%d%s%d", lv, postingsBenchSuffix, ln)), 0, 0) + app.Append(0, labels.FromStrings(strconv.Itoa(ln), fmt.Sprintf("%d%s%d", lv, postingsBenchSuffix, ln)), 0, 0) } require.NoError(b, app.Commit()) } @@ -1161,7 +1162,7 @@ func BenchmarkCompactionFromOOOHead(b *testing.B) { for ln := 0; ln < labelNames; ln++ { app := h.Appender(context.Background()) for lv := 0; lv < labelValues; lv++ { - lbls := labels.FromStrings(fmt.Sprintf("%d", ln), fmt.Sprintf("%d%s%d", lv, postingsBenchSuffix, ln)) + lbls := labels.FromStrings(strconv.Itoa(ln), fmt.Sprintf("%d%s%d", lv, postingsBenchSuffix, ln)) _, err = app.Append(0, lbls, int64(totalSamples), 0) require.NoError(b, err) for ts := 0; ts < totalSamples; ts++ { @@ -1297,7 +1298,7 @@ func TestCancelCompactions(t *testing.T) { // This checks that the `context.Canceled` error is properly checked at all levels: // - tsdb_errors.NewMulti() should have the Is() method implemented for correct checks. // - callers should check with errors.Is() instead of ==. - readOnlyDB, err := OpenDBReadOnly(tmpdirCopy, log.NewNopLogger()) + readOnlyDB, err := OpenDBReadOnly(tmpdirCopy, "", log.NewNopLogger()) require.NoError(t, err) blocks, err := readOnlyDB.Blocks() require.NoError(t, err) diff --git a/tsdb/db.go b/tsdb/db.go index c2e8904a25..bca3c99480 100644 --- a/tsdb/db.go +++ b/tsdb/db.go @@ -383,26 +383,36 @@ var ErrClosed = errors.New("db already closed") // Current implementation doesn't support concurrency so // all API calls should happen in the same go routine. type DBReadOnly struct { - logger log.Logger - dir string - closers []io.Closer - closed chan struct{} + logger log.Logger + dir string + sandboxDir string + closers []io.Closer + closed chan struct{} } // OpenDBReadOnly opens DB in the given directory for read only operations. -func OpenDBReadOnly(dir string, l log.Logger) (*DBReadOnly, error) { +func OpenDBReadOnly(dir, sandboxDirRoot string, l log.Logger) (*DBReadOnly, error) { if _, err := os.Stat(dir); err != nil { return nil, fmt.Errorf("opening the db dir: %w", err) } + if sandboxDirRoot == "" { + sandboxDirRoot = dir + } + sandboxDir, err := os.MkdirTemp(sandboxDirRoot, "tmp_dbro_sandbox") + if err != nil { + return nil, fmt.Errorf("setting up sandbox dir: %w", err) + } + if l == nil { l = log.NewNopLogger() } return &DBReadOnly{ - logger: l, - dir: dir, - closed: make(chan struct{}), + logger: l, + dir: dir, + sandboxDir: sandboxDir, + closed: make(chan struct{}), }, nil } @@ -491,7 +501,14 @@ func (db *DBReadOnly) loadDataAsQueryable(maxt int64) (storage.SampleAndChunkQue } opts := DefaultHeadOptions() - opts.ChunkDirRoot = db.dir + // Hard link the chunk files to a dir in db.sandboxDir in case the Head needs to truncate some of them + // or cut new ones while replaying the WAL. + // See https://github.com/prometheus/prometheus/issues/11618. + err = chunks.HardLinkChunkFiles(mmappedChunksDir(db.dir), mmappedChunksDir(db.sandboxDir)) + if err != nil { + return nil, err + } + opts.ChunkDirRoot = db.sandboxDir head, err := NewHead(nil, db.logger, nil, nil, opts, NewHeadStats()) if err != nil { return nil, err @@ -519,7 +536,7 @@ func (db *DBReadOnly) loadDataAsQueryable(maxt int64) (storage.SampleAndChunkQue } } opts := DefaultHeadOptions() - opts.ChunkDirRoot = db.dir + opts.ChunkDirRoot = db.sandboxDir head, err = NewHead(nil, db.logger, w, wbl, opts, NewHeadStats()) if err != nil { return nil, err @@ -690,8 +707,14 @@ func (db *DBReadOnly) Block(blockID string) (BlockReader, error) { return block, nil } -// Close all block readers. +// Close all block readers and delete the sandbox dir. func (db *DBReadOnly) Close() error { + defer func() { + // Delete the temporary sandbox directory that was created when opening the DB. + if err := os.RemoveAll(db.sandboxDir); err != nil { + level.Error(db.logger).Log("msg", "delete sandbox dir", "err", err) + } + }() select { case <-db.closed: return ErrClosed diff --git a/tsdb/db_test.go b/tsdb/db_test.go index 71b2f05ac7..5965e53179 100644 --- a/tsdb/db_test.go +++ b/tsdb/db_test.go @@ -25,6 +25,7 @@ import ( "os" "path" "path/filepath" + "runtime" "sort" "strconv" "sync" @@ -1065,7 +1066,7 @@ func TestWALSegmentSizeOptions(t *testing.T) { for i := int64(0); i < 155; i++ { app := db.Appender(context.Background()) - ref, err := app.Append(0, labels.FromStrings("wal"+fmt.Sprintf("%d", i), "size"), i, rand.Float64()) + ref, err := app.Append(0, labels.FromStrings("wal"+strconv.Itoa(int(i)), "size"), i, rand.Float64()) require.NoError(t, err) for j := int64(1); j <= 78; j++ { _, err := app.Append(ref, labels.EmptyLabels(), i+j, rand.Float64()) @@ -2494,7 +2495,7 @@ func TestDBReadOnly(t *testing.T) { } // Open a read only db and ensure that the API returns the same result as the normal DB. - dbReadOnly, err := OpenDBReadOnly(dbDir, logger) + dbReadOnly, err := OpenDBReadOnly(dbDir, "", logger) require.NoError(t, err) defer func() { require.NoError(t, dbReadOnly.Close()) }() @@ -2548,10 +2549,14 @@ func TestDBReadOnly(t *testing.T) { // TestDBReadOnlyClosing ensures that after closing the db // all api methods return an ErrClosed. func TestDBReadOnlyClosing(t *testing.T) { - dbDir := t.TempDir() - db, err := OpenDBReadOnly(dbDir, log.NewLogfmtLogger(log.NewSyncWriter(os.Stderr))) + sandboxDir := t.TempDir() + db, err := OpenDBReadOnly(t.TempDir(), sandboxDir, log.NewLogfmtLogger(log.NewSyncWriter(os.Stderr))) require.NoError(t, err) + // The sandboxDir was there. + require.DirExists(t, db.sandboxDir) require.NoError(t, db.Close()) + // The sandboxDir was deleted when closing. + require.NoDirExists(t, db.sandboxDir) require.Equal(t, db.Close(), ErrClosed) _, err = db.Blocks() require.Equal(t, err, ErrClosed) @@ -2587,7 +2592,7 @@ func TestDBReadOnly_FlushWAL(t *testing.T) { } // Flush WAL. - db, err := OpenDBReadOnly(dbDir, logger) + db, err := OpenDBReadOnly(dbDir, "", logger) require.NoError(t, err) flush := t.TempDir() @@ -2595,7 +2600,7 @@ func TestDBReadOnly_FlushWAL(t *testing.T) { require.NoError(t, db.Close()) // Reopen the DB from the flushed WAL block. - db, err = OpenDBReadOnly(flush, logger) + db, err = OpenDBReadOnly(flush, "", logger) require.NoError(t, err) defer func() { require.NoError(t, db.Close()) }() blocks, err := db.Blocks() @@ -2624,6 +2629,80 @@ func TestDBReadOnly_FlushWAL(t *testing.T) { require.Equal(t, 1000.0, sum) } +func TestDBReadOnly_Querier_NoAlteration(t *testing.T) { + countChunks := func(dir string) int { + files, err := os.ReadDir(mmappedChunksDir(dir)) + require.NoError(t, err) + return len(files) + } + + dirHash := func(dir string) (hash []byte) { + // Windows requires the DB to be closed: "xxx\lock: The process cannot access the file because it is being used by another process." + // But closing the DB alters the directory in this case (it'll cut a new chunk). + if runtime.GOOS != "windows" { + hash = testutil.DirHash(t, dir) + } + return + } + + spinUpQuerierAndCheck := func(dir, sandboxDir string, chunksCount int) { + dBDirHash := dirHash(dir) + // Bootsrap a RO db from the same dir and set up a querier. + dbReadOnly, err := OpenDBReadOnly(dir, sandboxDir, nil) + require.NoError(t, err) + require.Equal(t, chunksCount, countChunks(dir)) + q, err := dbReadOnly.Querier(math.MinInt, math.MaxInt) + require.NoError(t, err) + require.NoError(t, q.Close()) + require.NoError(t, dbReadOnly.Close()) + // The RO Head doesn't alter RW db chunks_head/. + require.Equal(t, chunksCount, countChunks(dir)) + require.Equal(t, dirHash(dir), dBDirHash) + } + + t.Run("doesn't cut chunks while replaying WAL", func(t *testing.T) { + db := openTestDB(t, nil, nil) + defer func() { + require.NoError(t, db.Close()) + }() + + // Append until the first mmaped head chunk. + for i := 0; i < 121; i++ { + app := db.Appender(context.Background()) + _, err := app.Append(0, labels.FromStrings("foo", "bar"), int64(i), 0) + require.NoError(t, err) + require.NoError(t, app.Commit()) + } + + spinUpQuerierAndCheck(db.dir, t.TempDir(), 0) + + // The RW Head should have no problem cutting its own chunk, + // this also proves that a chunk needed to be cut. + require.NotPanics(t, func() { db.ForceHeadMMap() }) + require.Equal(t, 1, countChunks(db.dir)) + }) + + t.Run("doesn't truncate corrupted chunks", func(t *testing.T) { + db := openTestDB(t, nil, nil) + require.NoError(t, db.Close()) + + // Simulate a corrupted chunk: without a header. + _, err := os.Create(path.Join(mmappedChunksDir(db.dir), "000001")) + require.NoError(t, err) + + spinUpQuerierAndCheck(db.dir, t.TempDir(), 1) + + // The RW Head should have no problem truncating its corrupted file: + // this proves that the chunk needed to be truncated. + db, err = Open(db.dir, nil, nil, nil, nil) + defer func() { + require.NoError(t, db.Close()) + }() + require.NoError(t, err) + require.Equal(t, 0, countChunks(db.dir)) + }) +} + func TestDBCannotSeePartialCommits(t *testing.T) { if defaultIsolationDisabled { t.Skip("skipping test since tsdb isolation is disabled") @@ -4495,7 +4574,7 @@ func TestOOOCompaction(t *testing.T) { ms, created, err := db.head.getOrCreate(lbls.Hash(), lbls) require.NoError(t, err) require.False(t, created) - require.Greater(t, ms.ooo.oooHeadChunk.chunk.NumSamples(), 0) + require.Positive(t, ms.ooo.oooHeadChunk.chunk.NumSamples()) require.Len(t, ms.ooo.oooMmappedChunks, 14) // 7 original, 7 duplicate. } checkNonEmptyOOOChunk(series1) @@ -4636,7 +4715,7 @@ func TestOOOCompactionWithNormalCompaction(t *testing.T) { ms, created, err := db.head.getOrCreate(lbls.Hash(), lbls) require.NoError(t, err) require.False(t, created) - require.Greater(t, ms.ooo.oooHeadChunk.chunk.NumSamples(), 0) + require.Positive(t, ms.ooo.oooHeadChunk.chunk.NumSamples()) } // If the normal Head is not compacted, the OOO head compaction does not take place. @@ -4737,7 +4816,7 @@ func TestOOOCompactionWithDisabledWriteLog(t *testing.T) { ms, created, err := db.head.getOrCreate(lbls.Hash(), lbls) require.NoError(t, err) require.False(t, created) - require.Greater(t, ms.ooo.oooHeadChunk.chunk.NumSamples(), 0) + require.Positive(t, ms.ooo.oooHeadChunk.chunk.NumSamples()) } // If the normal Head is not compacted, the OOO head compaction does not take place. @@ -5438,8 +5517,8 @@ func TestWBLAndMmapReplay(t *testing.T) { addedRecs++ require.NoError(t, newWbl.Log(rec)) } - require.Greater(t, markers, 0) - require.Greater(t, addedRecs, 0) + require.Positive(t, markers) + require.Positive(t, addedRecs) require.NoError(t, newWbl.Close()) require.NoError(t, sr.Close()) require.NoError(t, os.RemoveAll(wblDir)) diff --git a/tsdb/exemplar.go b/tsdb/exemplar.go index 213fff5c25..7545ab9a60 100644 --- a/tsdb/exemplar.go +++ b/tsdb/exemplar.go @@ -37,7 +37,7 @@ const ( type CircularExemplarStorage struct { lock sync.RWMutex - exemplars []*circularBufferEntry + exemplars []circularBufferEntry nextIndex int metrics *ExemplarMetrics @@ -121,7 +121,7 @@ func NewCircularExemplarStorage(length int64, m *ExemplarMetrics) (ExemplarStora length = 0 } c := &CircularExemplarStorage{ - exemplars: make([]*circularBufferEntry, length), + exemplars: make([]circularBufferEntry, length), index: make(map[string]*indexEntry, length/estimatedExemplarsPerSeries), metrics: m, } @@ -214,12 +214,12 @@ func (ce *CircularExemplarStorage) ValidateExemplar(l labels.Labels, e exemplar. // Optimize by moving the lock to be per series (& benchmark it). ce.lock.RLock() defer ce.lock.RUnlock() - return ce.validateExemplar(seriesLabels, e, false) + return ce.validateExemplar(ce.index[string(seriesLabels)], e, false) } // Not thread safe. The appended parameters tells us whether this is an external validation, or internal // as a result of an AddExemplar call, in which case we should update any relevant metrics. -func (ce *CircularExemplarStorage) validateExemplar(key []byte, e exemplar.Exemplar, appended bool) error { +func (ce *CircularExemplarStorage) validateExemplar(idx *indexEntry, e exemplar.Exemplar, appended bool) error { if len(ce.exemplars) == 0 { return storage.ErrExemplarsDisabled } @@ -239,8 +239,7 @@ func (ce *CircularExemplarStorage) validateExemplar(key []byte, e exemplar.Exemp return err } - idx, ok := ce.index[string(key)] - if !ok { + if idx == nil { return nil } @@ -292,7 +291,7 @@ func (ce *CircularExemplarStorage) Resize(l int64) int { oldBuffer := ce.exemplars oldNextIndex := int64(ce.nextIndex) - ce.exemplars = make([]*circularBufferEntry, l) + ce.exemplars = make([]circularBufferEntry, l) ce.index = make(map[string]*indexEntry, l/estimatedExemplarsPerSeries) ce.nextIndex = 0 @@ -311,10 +310,11 @@ func (ce *CircularExemplarStorage) Resize(l int64) int { // This way we don't migrate exemplars that would just be overwritten when migrating later exemplars. startIndex := (oldNextIndex - count + int64(len(oldBuffer))) % int64(len(oldBuffer)) + var buf [1024]byte for i := int64(0); i < count; i++ { idx := (startIndex + i) % int64(len(oldBuffer)) - if entry := oldBuffer[idx]; entry != nil { - ce.migrate(entry) + if oldBuffer[idx].ref != nil { + ce.migrate(&oldBuffer[idx], buf[:]) migrated++ } } @@ -328,9 +328,8 @@ func (ce *CircularExemplarStorage) Resize(l int64) int { // migrate is like AddExemplar but reuses existing structs. Expected to be called in batch and requires // external lock and does not compute metrics. -func (ce *CircularExemplarStorage) migrate(entry *circularBufferEntry) { - var buf [1024]byte - seriesLabels := entry.ref.seriesLabels.Bytes(buf[:]) +func (ce *CircularExemplarStorage) migrate(entry *circularBufferEntry, buf []byte) { + seriesLabels := entry.ref.seriesLabels.Bytes(buf[:0]) idx, ok := ce.index[string(seriesLabels)] if !ok { @@ -344,7 +343,7 @@ func (ce *CircularExemplarStorage) migrate(entry *circularBufferEntry) { idx.newest = ce.nextIndex entry.next = noExemplar - ce.exemplars[ce.nextIndex] = entry + ce.exemplars[ce.nextIndex] = *entry ce.nextIndex = (ce.nextIndex + 1) % len(ce.exemplars) } @@ -362,7 +361,8 @@ func (ce *CircularExemplarStorage) AddExemplar(l labels.Labels, e exemplar.Exemp ce.lock.Lock() defer ce.lock.Unlock() - err := ce.validateExemplar(seriesLabels, e, true) + idx, ok := ce.index[string(seriesLabels)] + err := ce.validateExemplar(idx, e, true) if err != nil { if errors.Is(err, storage.ErrDuplicateExemplar) { // Duplicate exemplar, noop. @@ -371,25 +371,23 @@ func (ce *CircularExemplarStorage) AddExemplar(l labels.Labels, e exemplar.Exemp return err } - _, ok := ce.index[string(seriesLabels)] if !ok { - ce.index[string(seriesLabels)] = &indexEntry{oldest: ce.nextIndex, seriesLabels: l} + idx = &indexEntry{oldest: ce.nextIndex, seriesLabels: l} + ce.index[string(seriesLabels)] = idx } else { - ce.exemplars[ce.index[string(seriesLabels)].newest].next = ce.nextIndex + ce.exemplars[idx.newest].next = ce.nextIndex } - if prev := ce.exemplars[ce.nextIndex]; prev == nil { - ce.exemplars[ce.nextIndex] = &circularBufferEntry{} - } else { + if prev := &ce.exemplars[ce.nextIndex]; prev.ref != nil { // There exists an exemplar already on this ce.nextIndex entry, // drop it, to make place for others. - var buf [1024]byte - prevLabels := prev.ref.seriesLabels.Bytes(buf[:]) if prev.next == noExemplar { // Last item for this series, remove index entry. + var buf [1024]byte + prevLabels := prev.ref.seriesLabels.Bytes(buf[:]) delete(ce.index, string(prevLabels)) } else { - ce.index[string(prevLabels)].oldest = prev.next + prev.ref.oldest = prev.next } } @@ -397,8 +395,8 @@ func (ce *CircularExemplarStorage) AddExemplar(l labels.Labels, e exemplar.Exemp // since this is the first exemplar stored for this series. ce.exemplars[ce.nextIndex].next = noExemplar ce.exemplars[ce.nextIndex].exemplar = e - ce.exemplars[ce.nextIndex].ref = ce.index[string(seriesLabels)] - ce.index[string(seriesLabels)].newest = ce.nextIndex + ce.exemplars[ce.nextIndex].ref = idx + idx.newest = ce.nextIndex ce.nextIndex = (ce.nextIndex + 1) % len(ce.exemplars) @@ -416,15 +414,15 @@ func (ce *CircularExemplarStorage) computeMetrics() { return } - if next := ce.exemplars[ce.nextIndex]; next != nil { + if ce.exemplars[ce.nextIndex].ref != nil { ce.metrics.exemplarsInStorage.Set(float64(len(ce.exemplars))) - ce.metrics.lastExemplarsTs.Set(float64(next.exemplar.Ts) / 1000) + ce.metrics.lastExemplarsTs.Set(float64(ce.exemplars[ce.nextIndex].exemplar.Ts) / 1000) return } // We did not yet fill the buffer. ce.metrics.exemplarsInStorage.Set(float64(ce.nextIndex)) - if ce.exemplars[0] != nil { + if ce.exemplars[0].ref != nil { ce.metrics.lastExemplarsTs.Set(float64(ce.exemplars[0].exemplar.Ts) / 1000) } } @@ -438,7 +436,7 @@ func (ce *CircularExemplarStorage) IterateExemplars(f func(seriesLabels labels.L idx := ce.nextIndex l := len(ce.exemplars) for i := 0; i < l; i, idx = i+1, (idx+1)%l { - if ce.exemplars[idx] == nil { + if ce.exemplars[idx].ref == nil { continue } err := f(ce.exemplars[idx].ref.seriesLabels, ce.exemplars[idx].exemplar) diff --git a/tsdb/exemplar_test.go b/tsdb/exemplar_test.go index 21030e4577..7723ec3894 100644 --- a/tsdb/exemplar_test.go +++ b/tsdb/exemplar_test.go @@ -415,27 +415,29 @@ func BenchmarkAddExemplar(b *testing.B) { // before adding. exLabels := labels.FromStrings("trace_id", "89620921") - for _, n := range []int{10000, 100000, 1000000} { - b.Run(fmt.Sprintf("%d", n), func(b *testing.B) { - for j := 0; j < b.N; j++ { - b.StopTimer() - exs, err := NewCircularExemplarStorage(int64(n), eMetrics) - require.NoError(b, err) - es := exs.(*CircularExemplarStorage) - var l labels.Labels - b.StartTimer() + for _, capacity := range []int{1000, 10000, 100000} { + for _, n := range []int{10000, 100000, 1000000} { + b.Run(fmt.Sprintf("%d/%d", n, capacity), func(b *testing.B) { + for j := 0; j < b.N; j++ { + b.StopTimer() + exs, err := NewCircularExemplarStorage(int64(capacity), eMetrics) + require.NoError(b, err) + es := exs.(*CircularExemplarStorage) + var l labels.Labels + b.StartTimer() - for i := 0; i < n; i++ { - if i%100 == 0 { - l = labels.FromStrings("service", strconv.Itoa(i)) - } - err = es.AddExemplar(l, exemplar.Exemplar{Value: float64(i), Ts: int64(i), Labels: exLabels}) - if err != nil { - require.NoError(b, err) + for i := 0; i < n; i++ { + if i%100 == 0 { + l = labels.FromStrings("service", strconv.Itoa(i)) + } + err = es.AddExemplar(l, exemplar.Exemplar{Value: float64(i), Ts: int64(i), Labels: exLabels}) + if err != nil { + require.NoError(b, err) + } } } - } - }) + }) + } } } @@ -480,8 +482,11 @@ func BenchmarkResizeExemplars(b *testing.B) { require.NoError(b, err) es := exs.(*CircularExemplarStorage) + var l labels.Labels for i := 0; i < int(float64(tc.startSize)*float64(1.5)); i++ { - l := labels.FromStrings("service", strconv.Itoa(i)) + if i%100 == 0 { + l = labels.FromStrings("service", strconv.Itoa(i)) + } err = es.AddExemplar(l, exemplar.Exemplar{Value: float64(i), Ts: int64(i)}) if err != nil { diff --git a/tsdb/head.go b/tsdb/head.go index 8b3d9787ca..d5f7144fdb 100644 --- a/tsdb/head.go +++ b/tsdb/head.go @@ -310,12 +310,22 @@ func (h *Head) resetInMemoryState() error { return err } + if h.series != nil { + // reset the existing series to make sure we call the appropriated hooks + // and increment the series removed metrics + fs := h.series.iterForDeletion(func(_ int, _ uint64, s *memSeries, flushedForCallback map[chunks.HeadSeriesRef]labels.Labels) { + // All series should be flushed + flushedForCallback[s.ref] = s.lset + }) + h.metrics.seriesRemoved.Add(float64(fs)) + } + + h.series = newStripeSeries(h.opts.StripeSize, h.opts.SeriesCallback) h.iso = newIsolation(h.opts.IsolationDisabled) h.oooIso = newOOOIsolation() - + h.numSeries.Store(0) h.exemplarMetrics = em h.exemplars = es - h.series = newStripeSeries(h.opts.StripeSize, h.opts.SeriesCallback) h.postings = index.NewUnorderedMemPostings() h.tombstones = tombstones.NewMemTombstones() h.deleted = map[chunks.HeadSeriesRef]int{} @@ -1861,11 +1871,10 @@ func newStripeSeries(stripeSize int, seriesCallback SeriesLifecycleCallback) *st // minMmapFile is the min mmap file number seen in the series (in-order and out-of-order) after gc'ing the series. func (s *stripeSeries) gc(mint int64, minOOOMmapRef chunks.ChunkDiskMapperRef) (_ map[storage.SeriesRef]struct{}, _ int, _, _ int64, minMmapFile int) { var ( - deleted = map[storage.SeriesRef]struct{}{} - rmChunks = 0 - actualMint int64 = math.MaxInt64 - minOOOTime int64 = math.MaxInt64 - deletedFromPrevStripe = 0 + deleted = map[storage.SeriesRef]struct{}{} + rmChunks = 0 + actualMint int64 = math.MaxInt64 + minOOOTime int64 = math.MaxInt64 ) minMmapFile = math.MaxInt32 @@ -1923,27 +1932,7 @@ func (s *stripeSeries) gc(mint int64, minOOOMmapRef chunks.ChunkDiskMapperRef) ( deletedForCallback[series.ref] = series.lset } - // Run through all series shard by shard, checking which should be deleted. - for i := 0; i < s.size; i++ { - deletedForCallback := make(map[chunks.HeadSeriesRef]labels.Labels, deletedFromPrevStripe) - s.locks[i].Lock() - - // Delete conflicts first so seriesHashmap.del doesn't move them to the `unique` field, - // after deleting `unique`. - for hash, all := range s.hashes[i].conflicts { - for _, series := range all { - check(i, hash, series, deletedForCallback) - } - } - for hash, series := range s.hashes[i].unique { - check(i, hash, series, deletedForCallback) - } - - s.locks[i].Unlock() - - s.seriesLifecycleCallback.PostDeletion(deletedForCallback) - deletedFromPrevStripe = len(deletedForCallback) - } + s.iterForDeletion(check) if actualMint == math.MaxInt64 { actualMint = mint @@ -1952,6 +1941,35 @@ func (s *stripeSeries) gc(mint int64, minOOOMmapRef chunks.ChunkDiskMapperRef) ( return deleted, rmChunks, actualMint, minOOOTime, minMmapFile } +// The iterForDeletion function iterates through all series, invoking the checkDeletedFunc for each. +// The checkDeletedFunc takes a map as input and should add to it all series that were deleted and should be included +// when invoking the PostDeletion hook. +func (s *stripeSeries) iterForDeletion(checkDeletedFunc func(int, uint64, *memSeries, map[chunks.HeadSeriesRef]labels.Labels)) int { + seriesSetFromPrevStripe := 0 + totalDeletedSeries := 0 + // Run through all series shard by shard + for i := 0; i < s.size; i++ { + seriesSet := make(map[chunks.HeadSeriesRef]labels.Labels, seriesSetFromPrevStripe) + s.locks[i].Lock() + // Iterate conflicts first so f doesn't move them to the `unique` field, + // after deleting `unique`. + for hash, all := range s.hashes[i].conflicts { + for _, series := range all { + checkDeletedFunc(i, hash, series, seriesSet) + } + } + + for hash, series := range s.hashes[i].unique { + checkDeletedFunc(i, hash, series, seriesSet) + } + s.locks[i].Unlock() + s.seriesLifecycleCallback.PostDeletion(seriesSet) + totalDeletedSeries += len(seriesSet) + seriesSetFromPrevStripe = len(seriesSet) + } + return totalDeletedSeries +} + func (s *stripeSeries) getByID(id chunks.HeadSeriesRef) *memSeries { i := uint64(id) & uint64(s.size-1) diff --git a/tsdb/head_test.go b/tsdb/head_test.go index d9631b3b91..bb437ab598 100644 --- a/tsdb/head_test.go +++ b/tsdb/head_test.go @@ -3383,7 +3383,7 @@ func TestWaitForPendingReadersInTimeRange(t *testing.T) { func TestAppendHistogram(t *testing.T) { l := labels.FromStrings("a", "b") for _, numHistograms := range []int{1, 10, 150, 200, 250, 300} { - t.Run(fmt.Sprintf("%d", numHistograms), func(t *testing.T) { + t.Run(strconv.Itoa(numHistograms), func(t *testing.T) { head, _ := newTestHead(t, 1000, wlog.CompressionNone, false) t.Cleanup(func() { require.NoError(t, head.Close()) @@ -3557,7 +3557,7 @@ func TestHistogramInWALAndMmapChunk(t *testing.T) { expMmapChunks = append(expMmapChunks, &cpy) } expHeadChunkSamples := ms.headChunks.chunk.NumSamples() - require.Greater(t, expHeadChunkSamples, 0) + require.Positive(t, expHeadChunkSamples) // Series with mix of histograms and float. s2 := labels.FromStrings("a", "b2") @@ -3692,7 +3692,7 @@ func TestChunkSnapshot(t *testing.T) { e := ex{ seriesLabels: lbls, e: exemplar.Exemplar{ - Labels: labels.FromStrings("trace_id", fmt.Sprintf("%d", rand.Int())), + Labels: labels.FromStrings("trace_id", strconv.Itoa(rand.Int())), Value: rand.Float64(), Ts: ts, }, @@ -4007,6 +4007,9 @@ func TestSnapshotError(t *testing.T) { require.NoError(t, err) f, err := os.OpenFile(path.Join(snapDir, files[0].Name()), os.O_RDWR, 0) require.NoError(t, err) + // Create snapshot backup to be restored on future test cases. + snapshotBackup, err := io.ReadAll(f) + require.NoError(t, err) _, err = f.WriteAt([]byte{0b11111111}, 18) require.NoError(t, err) require.NoError(t, f.Close()) @@ -4021,10 +4024,44 @@ func TestSnapshotError(t *testing.T) { // There should be no series in the memory after snapshot error since WAL was removed. require.Equal(t, 1.0, prom_testutil.ToFloat64(head.metrics.snapshotReplayErrorTotal)) + require.Equal(t, uint64(0), head.NumSeries()) require.Nil(t, head.series.getByHash(lbls.Hash(), lbls)) tm, err = head.tombstones.Get(1) require.NoError(t, err) require.Empty(t, tm) + require.NoError(t, head.Close()) + + // Test corruption in the middle of the snapshot. + f, err = os.OpenFile(path.Join(snapDir, files[0].Name()), os.O_RDWR, 0) + require.NoError(t, err) + _, err = f.WriteAt(snapshotBackup, 0) + require.NoError(t, err) + _, err = f.WriteAt([]byte{0b11111111}, 300) + require.NoError(t, err) + require.NoError(t, f.Close()) + + c := &countSeriesLifecycleCallback{} + opts := head.opts + opts.SeriesCallback = c + + w, err = wlog.NewSize(nil, nil, head.wal.Dir(), 32768, wlog.CompressionNone) + require.NoError(t, err) + head, err = NewHead(prometheus.NewRegistry(), nil, w, nil, head.opts, nil) + require.NoError(t, err) + require.NoError(t, head.Init(math.MinInt64)) + + // There should be no series in the memory after snapshot error since WAL was removed. + require.Equal(t, 1.0, prom_testutil.ToFloat64(head.metrics.snapshotReplayErrorTotal)) + require.Nil(t, head.series.getByHash(lbls.Hash(), lbls)) + require.Equal(t, uint64(0), head.NumSeries()) + + // Since the snapshot could replay certain series, we continue invoking the create hooks. + // In such instances, we need to ensure that we also trigger the delete hooks when resetting the memory. + require.Equal(t, int64(2), c.created.Load()) + require.Equal(t, int64(2), c.deleted.Load()) + + require.Equal(t, 2.0, prom_testutil.ToFloat64(head.metrics.seriesRemoved)) + require.Equal(t, 2.0, prom_testutil.ToFloat64(head.metrics.seriesCreated)) } func TestHistogramMetrics(t *testing.T) { @@ -4601,7 +4638,7 @@ func TestChunkSnapshotTakenAfterIncompleteSnapshot(t *testing.T) { require.NoError(t, err) require.NotEqual(t, "", name) require.Equal(t, 0, idx) - require.Greater(t, offset, 0) + require.Positive(t, offset) } // TestWBLReplay checks the replay at a low level. @@ -5032,7 +5069,7 @@ func TestOOOAppendWithNoSeries(t *testing.T) { require.Equal(t, expSamples, ms.headChunks.chunk.NumSamples()) } - newLabels := func(idx int) labels.Labels { return labels.FromStrings("foo", fmt.Sprintf("%d", idx)) } + newLabels := func(idx int) labels.Labels { return labels.FromStrings("foo", strconv.Itoa(idx)) } s1 := newLabels(1) appendSample(s1, 300) // At 300m. @@ -5829,3 +5866,14 @@ func TestHeadCompactableDoesNotCompactEmptyHead(t *testing.T) { require.False(t, head.compactable()) } + +type countSeriesLifecycleCallback struct { + created atomic.Int64 + deleted atomic.Int64 +} + +func (c *countSeriesLifecycleCallback) PreCreation(labels.Labels) error { return nil } +func (c *countSeriesLifecycleCallback) PostCreation(labels.Labels) { c.created.Inc() } +func (c *countSeriesLifecycleCallback) PostDeletion(s map[chunks.HeadSeriesRef]labels.Labels) { + c.deleted.Add(int64(len(s))) +} diff --git a/tsdb/index/index.go b/tsdb/index/index.go index a36c33c4fb..480e6a8fc7 100644 --- a/tsdb/index/index.go +++ b/tsdb/index/index.go @@ -51,6 +51,9 @@ const ( indexFilename = "index" seriesByteAlign = 16 + + // checkContextEveryNIterations is used in some tight loops to check if the context is done. + checkContextEveryNIterations = 128 ) type indexWriterSeries struct { @@ -1797,7 +1800,12 @@ func (r *Reader) postingsForLabelMatchingV1(ctx context.Context, name string, ma } var its []Postings + count := 1 for val, offset := range e { + if count%checkContextEveryNIterations == 0 && ctx.Err() != nil { + return ErrPostings(ctx.Err()) + } + count++ if !match(val) { continue } diff --git a/tsdb/index/index_test.go b/tsdb/index/index_test.go index c451c38dd2..5c6d64e076 100644 --- a/tsdb/index/index_test.go +++ b/tsdb/index/index_test.go @@ -20,7 +20,9 @@ import ( "hash/crc32" "os" "path/filepath" + "slices" "sort" + "strconv" "testing" "github.com/stretchr/testify/require" @@ -160,39 +162,14 @@ func TestIndexRW_Create_Open(t *testing.T) { } func TestIndexRW_Postings(t *testing.T) { - dir := t.TempDir() ctx := context.Background() - - fn := filepath.Join(dir, indexFilename) - - iw, err := NewWriter(context.Background(), fn) - require.NoError(t, err) - - series := []labels.Labels{ - labels.FromStrings("a", "1", "b", "1"), - labels.FromStrings("a", "1", "b", "2"), - labels.FromStrings("a", "1", "b", "3"), - labels.FromStrings("a", "1", "b", "4"), + var input indexWriterSeriesSlice + for i := 1; i < 5; i++ { + input = append(input, &indexWriterSeries{ + labels: labels.FromStrings("a", "1", "b", strconv.Itoa(i)), + }) } - - require.NoError(t, iw.AddSymbol("1")) - require.NoError(t, iw.AddSymbol("2")) - require.NoError(t, iw.AddSymbol("3")) - require.NoError(t, iw.AddSymbol("4")) - require.NoError(t, iw.AddSymbol("a")) - require.NoError(t, iw.AddSymbol("b")) - - // Postings lists are only written if a series with the respective - // reference was added before. - require.NoError(t, iw.AddSeries(1, series[0])) - require.NoError(t, iw.AddSeries(2, series[1])) - require.NoError(t, iw.AddSeries(3, series[2])) - require.NoError(t, iw.AddSeries(4, series[3])) - - require.NoError(t, iw.Close()) - - ir, err := NewFileReader(fn) - require.NoError(t, err) + ir, fn, _ := createFileReader(ctx, t, input) p, err := ir.Postings(ctx, "a", "1") require.NoError(t, err) @@ -205,7 +182,7 @@ func TestIndexRW_Postings(t *testing.T) { require.NoError(t, err) require.Empty(t, c) - testutil.RequireEqual(t, series[i], builder.Labels()) + testutil.RequireEqual(t, input[i].labels, builder.Labels()) } require.NoError(t, p.Err()) @@ -240,8 +217,6 @@ func TestIndexRW_Postings(t *testing.T) { "b": {"1", "2", "3", "4"}, }, labelIndices) - require.NoError(t, ir.Close()) - t.Run("ShardedPostings()", func(t *testing.T) { ir, err := NewFileReader(fn) require.NoError(t, err) @@ -296,42 +271,16 @@ func TestIndexRW_Postings(t *testing.T) { } func TestPostingsMany(t *testing.T) { - dir := t.TempDir() ctx := context.Background() - - fn := filepath.Join(dir, indexFilename) - - iw, err := NewWriter(context.Background(), fn) - require.NoError(t, err) - // Create a label in the index which has 999 values. - symbols := map[string]struct{}{} - series := []labels.Labels{} + var input indexWriterSeriesSlice for i := 1; i < 1000; i++ { v := fmt.Sprintf("%03d", i) - series = append(series, labels.FromStrings("i", v, "foo", "bar")) - symbols[v] = struct{}{} + input = append(input, &indexWriterSeries{ + labels: labels.FromStrings("i", v, "foo", "bar"), + }) } - symbols["i"] = struct{}{} - symbols["foo"] = struct{}{} - symbols["bar"] = struct{}{} - syms := []string{} - for s := range symbols { - syms = append(syms, s) - } - sort.Strings(syms) - for _, s := range syms { - require.NoError(t, iw.AddSymbol(s)) - } - - for i, s := range series { - require.NoError(t, iw.AddSeries(storage.SeriesRef(i), s)) - } - require.NoError(t, iw.Close()) - - ir, err := NewFileReader(fn) - require.NoError(t, err) - defer func() { require.NoError(t, ir.Close()) }() + ir, _, symbols := createFileReader(ctx, t, input) cases := []struct { in []string @@ -387,25 +336,13 @@ func TestPostingsMany(t *testing.T) { } func TestPersistence_index_e2e(t *testing.T) { - dir := t.TempDir() ctx := context.Background() - lbls, err := labels.ReadLabels(filepath.Join("..", "testdata", "20kseries.json"), 20000) require.NoError(t, err) - // Sort labels as the index writer expects series in sorted order. sort.Sort(labels.Slice(lbls)) - symbols := map[string]struct{}{} - for _, lset := range lbls { - lset.Range(func(l labels.Label) { - symbols[l.Name] = struct{}{} - symbols[l.Value] = struct{}{} - }) - } - var input indexWriterSeriesSlice - ref := uint64(0) // Generate ChunkMetas for every label set. for i, lset := range lbls { @@ -426,17 +363,7 @@ func TestPersistence_index_e2e(t *testing.T) { }) } - iw, err := NewWriter(context.Background(), filepath.Join(dir, indexFilename)) - require.NoError(t, err) - - syms := []string{} - for s := range symbols { - syms = append(syms, s) - } - sort.Strings(syms) - for _, s := range syms { - require.NoError(t, iw.AddSymbol(s)) - } + ir, _, _ := createFileReader(ctx, t, input) // Population procedure as done by compaction. var ( @@ -447,8 +374,6 @@ func TestPersistence_index_e2e(t *testing.T) { mi := newMockIndex() for i, s := range input { - err = iw.AddSeries(storage.SeriesRef(i), s.labels, s.chunks...) - require.NoError(t, err) require.NoError(t, mi.AddSeries(storage.SeriesRef(i), s.labels, s.chunks...)) s.labels.Range(func(l labels.Label) { @@ -462,12 +387,6 @@ func TestPersistence_index_e2e(t *testing.T) { postings.Add(storage.SeriesRef(i), s.labels) } - err = iw.Close() - require.NoError(t, err) - - ir, err := NewFileReader(filepath.Join(dir, indexFilename)) - require.NoError(t, err) - for p := range mi.postings { gotp, err := ir.Postings(ctx, p.Name, p.Value) require.NoError(t, err) @@ -523,8 +442,6 @@ func TestPersistence_index_e2e(t *testing.T) { } sort.Strings(expSymbols) require.Equal(t, expSymbols, gotSymbols) - - require.NoError(t, ir.Close()) } func TestWriter_ShouldReturnErrorOnSeriesWithDuplicatedLabelNames(t *testing.T) { @@ -624,39 +541,14 @@ func BenchmarkReader_ShardedPostings(b *testing.B) { numShards = 16 ) - dir, err := os.MkdirTemp("", "benchmark_reader_sharded_postings") - require.NoError(b, err) - defer func() { - require.NoError(b, os.RemoveAll(dir)) - }() - ctx := context.Background() - - // Generate an index. - fn := filepath.Join(dir, indexFilename) - - iw, err := NewWriter(ctx, fn) - require.NoError(b, err) - + var input indexWriterSeriesSlice for i := 1; i <= numSeries; i++ { - require.NoError(b, iw.AddSymbol(fmt.Sprintf("%10d", i))) + input = append(input, &indexWriterSeries{ + labels: labels.FromStrings("const", fmt.Sprintf("%10d", 1), "unique", fmt.Sprintf("%10d", i)), + }) } - require.NoError(b, iw.AddSymbol("const")) - require.NoError(b, iw.AddSymbol("unique")) - - for i := 1; i <= numSeries; i++ { - require.NoError(b, iw.AddSeries(storage.SeriesRef(i), - labels.FromStrings("const", fmt.Sprintf("%10d", 1), "unique", fmt.Sprintf("%10d", i)))) - } - - require.NoError(b, iw.Close()) - - b.ResetTimer() - - // Create a reader to read back all postings from the index. - ir, err := NewFileReader(fn) - require.NoError(b, err) - + ir, _, _ := createFileReader(ctx, b, input) b.ResetTimer() for n := 0; n < b.N; n++ { @@ -719,3 +611,64 @@ func TestChunksTimeOrdering(t *testing.T) { require.NoError(t, idx.Close()) } + +func TestReader_PostingsForLabelMatchingHonorsContextCancel(t *testing.T) { + const seriesCount = 1000 + var input indexWriterSeriesSlice + for i := 1; i < seriesCount; i++ { + input = append(input, &indexWriterSeries{ + labels: labels.FromStrings("__name__", fmt.Sprintf("%4d", i)), + chunks: []chunks.Meta{ + {Ref: 1, MinTime: 0, MaxTime: 10}, + }, + }) + } + ir, _, _ := createFileReader(context.Background(), t, input) + + failAfter := uint64(seriesCount / 2) // Fail after processing half of the series. + ctx := &testutil.MockContextErrAfter{FailAfter: failAfter} + p := ir.PostingsForLabelMatching(ctx, "__name__", func(string) bool { + return true + }) + require.Error(t, p.Err()) + require.Equal(t, failAfter, ctx.Count()) +} + +// createFileReader creates a temporary index file. It writes the provided input to this file. +// It returns a Reader for this file, the file's name, and the symbol map. +func createFileReader(ctx context.Context, tb testing.TB, input indexWriterSeriesSlice) (*Reader, string, map[string]struct{}) { + tb.Helper() + + fn := filepath.Join(tb.TempDir(), indexFilename) + + iw, err := NewWriter(ctx, fn) + require.NoError(tb, err) + + symbols := map[string]struct{}{} + for _, s := range input { + s.labels.Range(func(l labels.Label) { + symbols[l.Name] = struct{}{} + symbols[l.Value] = struct{}{} + }) + } + + syms := []string{} + for s := range symbols { + syms = append(syms, s) + } + slices.Sort(syms) + for _, s := range syms { + require.NoError(tb, iw.AddSymbol(s)) + } + for i, s := range input { + require.NoError(tb, iw.AddSeries(storage.SeriesRef(i), s.labels, s.chunks...)) + } + require.NoError(tb, iw.Close()) + + ir, err := NewFileReader(fn) + require.NoError(tb, err) + tb.Cleanup(func() { + require.NoError(tb, ir.Close()) + }) + return ir, fn, symbols +} diff --git a/tsdb/index/postings.go b/tsdb/index/postings.go index 136b3441eb..159f6416e2 100644 --- a/tsdb/index/postings.go +++ b/tsdb/index/postings.go @@ -416,7 +416,13 @@ func (p *MemPostings) PostingsForLabelMatching(ctx context.Context, name string, } var its []Postings + count := 1 for _, v := range vals { + if count%checkContextEveryNIterations == 0 && ctx.Err() != nil { + p.mtx.RUnlock() + return ErrPostings(ctx.Err()) + } + count++ if match(v) { its = append(its, NewListPostings(e[v])) } diff --git a/tsdb/index/postings_test.go b/tsdb/index/postings_test.go index 9e6bd23f8c..2cbc14ac64 100644 --- a/tsdb/index/postings_test.go +++ b/tsdb/index/postings_test.go @@ -22,12 +22,15 @@ import ( "math/rand" "sort" "strconv" + "strings" "testing" + "github.com/grafana/regexp" "github.com/stretchr/testify/require" "github.com/prometheus/prometheus/model/labels" "github.com/prometheus/prometheus/storage" + "github.com/prometheus/prometheus/util/testutil" ) func TestMemPostings_addFor(t *testing.T) { @@ -49,7 +52,7 @@ func TestMemPostings_ensureOrder(t *testing.T) { for j := range l { l[j] = storage.SeriesRef(rand.Uint64()) } - v := fmt.Sprintf("%d", i) + v := strconv.Itoa(i) p.m["a"][v] = l } @@ -390,7 +393,7 @@ func BenchmarkMerge(t *testing.B) { its := make([]Postings, len(refs)) for _, nSeries := range []int{1, 10, 100, 1000, 10000, 100000} { - t.Run(fmt.Sprint(nSeries), func(bench *testing.B) { + t.Run(strconv.Itoa(nSeries), func(bench *testing.B) { ctx := context.Background() for i := 0; i < bench.N; i++ { // Reset the ListPostings to their original values each time round the loop. @@ -1282,3 +1285,71 @@ func BenchmarkListPostings(b *testing.B) { }) } } + +func slowRegexpString() string { + nums := map[int]struct{}{} + for i := 10_000; i < 20_000; i++ { + if i%3 == 0 { + nums[i] = struct{}{} + } + } + + var sb strings.Builder + sb.WriteString(".*(9999") + for i := range nums { + sb.WriteString("|") + sb.WriteString(strconv.Itoa(i)) + } + sb.WriteString(").*") + return sb.String() +} + +func BenchmarkMemPostings_PostingsForLabelMatching(b *testing.B) { + fast := regexp.MustCompile("^(100|200)$") + slowRegexp := "^" + slowRegexpString() + "$" + b.Logf("Slow regexp length = %d", len(slowRegexp)) + slow := regexp.MustCompile(slowRegexp) + + for _, labelValueCount := range []int{1_000, 10_000, 100_000} { + b.Run(fmt.Sprintf("labels=%d", labelValueCount), func(b *testing.B) { + mp := NewMemPostings() + for i := 0; i < labelValueCount; i++ { + mp.Add(storage.SeriesRef(i), labels.FromStrings("label", strconv.Itoa(i))) + } + + fp, err := ExpandPostings(mp.PostingsForLabelMatching(context.Background(), "label", fast.MatchString)) + require.NoError(b, err) + b.Logf("Fast matcher matches %d series", len(fp)) + b.Run("matcher=fast", func(b *testing.B) { + for i := 0; i < b.N; i++ { + mp.PostingsForLabelMatching(context.Background(), "label", fast.MatchString).Next() + } + }) + + sp, err := ExpandPostings(mp.PostingsForLabelMatching(context.Background(), "label", slow.MatchString)) + require.NoError(b, err) + b.Logf("Slow matcher matches %d series", len(sp)) + b.Run("matcher=slow", func(b *testing.B) { + for i := 0; i < b.N; i++ { + mp.PostingsForLabelMatching(context.Background(), "label", slow.MatchString).Next() + } + }) + }) + } +} + +func TestMemPostings_PostingsForLabelMatchingHonorsContextCancel(t *testing.T) { + memP := NewMemPostings() + seriesCount := 10 * checkContextEveryNIterations + for i := 1; i <= seriesCount; i++ { + memP.Add(storage.SeriesRef(i), labels.FromStrings("__name__", fmt.Sprintf("%4d", i))) + } + + failAfter := uint64(seriesCount / 2 / checkContextEveryNIterations) + ctx := &testutil.MockContextErrAfter{FailAfter: failAfter} + p := memP.PostingsForLabelMatching(ctx, "__name__", func(string) bool { + return true + }) + require.Error(t, p.Err()) + require.Equal(t, failAfter+1, ctx.Count()) // Plus one for the Err() call that puts the error in the result. +} diff --git a/tsdb/querier.go b/tsdb/querier.go index 1170493beb..1071c4a716 100644 --- a/tsdb/querier.go +++ b/tsdb/querier.go @@ -33,6 +33,9 @@ import ( "github.com/prometheus/prometheus/util/annotations" ) +// checkContextEveryNIterations is used in some tight loops to check if the context is done. +const checkContextEveryNIterations = 100 + type blockBaseQuerier struct { blockID ulid.ULID index IndexReader @@ -354,11 +357,16 @@ func inversePostingsForMatcher(ctx context.Context, ix IndexReader, m *labels.Ma } res := vals[:0] - // If the inverse match is ="", we just want all the values. - if m.Type == labels.MatchEqual && m.Value == "" { + // If the match before inversion was !="" or !~"", we just want all the values. + if m.Value == "" && (m.Type == labels.MatchRegexp || m.Type == labels.MatchEqual) { res = vals } else { + count := 1 for _, val := range vals { + if count%checkContextEveryNIterations == 0 && ctx.Err() != nil { + return nil, ctx.Err() + } + count++ if !m.Matches(val) { res = append(res, val) } @@ -387,7 +395,12 @@ func labelValuesWithMatchers(ctx context.Context, r IndexReader, name string, ma // re-use the allValues slice to avoid allocations // this is safe because the iteration is always ahead of the append filteredValues := allValues[:0] + count := 1 for _, v := range allValues { + if count%checkContextEveryNIterations == 0 && ctx.Err() != nil { + return nil, ctx.Err() + } + count++ if m.Matches(v) { filteredValues = append(filteredValues, v) } diff --git a/tsdb/querier_test.go b/tsdb/querier_test.go index 16de6373d0..c7e60a0e16 100644 --- a/tsdb/querier_test.go +++ b/tsdb/querier_test.go @@ -38,6 +38,7 @@ import ( "github.com/prometheus/prometheus/tsdb/tombstones" "github.com/prometheus/prometheus/tsdb/tsdbutil" "github.com/prometheus/prometheus/util/annotations" + "github.com/prometheus/prometheus/util/testutil" ) // TODO(bwplotka): Replace those mocks with remote.concreteSeriesSet. @@ -2807,6 +2808,13 @@ func TestPostingsForMatchers(t *testing.T) { }, }, // Not regex. + { + matchers: []*labels.Matcher{labels.MustNewMatcher(labels.MatchNotRegexp, "i", "")}, + exp: []labels.Labels{ + labels.FromStrings("n", "1", "i", "a"), + labels.FromStrings("n", "1", "i", "b"), + }, + }, { matchers: []*labels.Matcher{labels.MustNewMatcher(labels.MatchNotRegexp, "n", "^1$")}, exp: []labels.Labels{ @@ -3638,3 +3646,77 @@ func TestQueryWithOneChunkCompletelyDeleted(t *testing.T) { require.NoError(t, css.Err()) require.Equal(t, 1, seriesCount) } + +func TestReader_PostingsForLabelMatchingHonorsContextCancel(t *testing.T) { + ir := mockReaderOfLabels{} + + failAfter := uint64(mockReaderOfLabelsSeriesCount / 2 / checkContextEveryNIterations) + ctx := &testutil.MockContextErrAfter{FailAfter: failAfter} + _, err := labelValuesWithMatchers(ctx, ir, "__name__", labels.MustNewMatcher(labels.MatchRegexp, "__name__", ".+")) + + require.Error(t, err) + require.Equal(t, failAfter+1, ctx.Count()) // Plus one for the Err() call that puts the error in the result. +} + +func TestReader_InversePostingsForMatcherHonorsContextCancel(t *testing.T) { + ir := mockReaderOfLabels{} + + failAfter := uint64(mockReaderOfLabelsSeriesCount / 2 / checkContextEveryNIterations) + ctx := &testutil.MockContextErrAfter{FailAfter: failAfter} + _, err := inversePostingsForMatcher(ctx, ir, labels.MustNewMatcher(labels.MatchRegexp, "__name__", ".*")) + + require.Error(t, err) + require.Equal(t, failAfter+1, ctx.Count()) // Plus one for the Err() call that puts the error in the result. +} + +type mockReaderOfLabels struct{} + +const mockReaderOfLabelsSeriesCount = checkContextEveryNIterations * 10 + +func (m mockReaderOfLabels) LabelValues(context.Context, string, ...*labels.Matcher) ([]string, error) { + return make([]string, mockReaderOfLabelsSeriesCount), nil +} + +func (m mockReaderOfLabels) LabelValueFor(context.Context, storage.SeriesRef, string) (string, error) { + panic("LabelValueFor called") +} + +func (m mockReaderOfLabels) SortedLabelValues(context.Context, string, ...*labels.Matcher) ([]string, error) { + panic("SortedLabelValues called") +} + +func (m mockReaderOfLabels) Close() error { + return nil +} + +func (m mockReaderOfLabels) LabelNames(context.Context, ...*labels.Matcher) ([]string, error) { + panic("LabelNames called") +} + +func (m mockReaderOfLabels) LabelNamesFor(context.Context, ...storage.SeriesRef) ([]string, error) { + panic("LabelNamesFor called") +} + +func (m mockReaderOfLabels) PostingsForLabelMatching(context.Context, string, func(string) bool) index.Postings { + panic("PostingsForLabelMatching called") +} + +func (m mockReaderOfLabels) Postings(context.Context, string, ...string) (index.Postings, error) { + panic("Postings called") +} + +func (m mockReaderOfLabels) ShardedPostings(index.Postings, uint64, uint64) index.Postings { + panic("Postings called") +} + +func (m mockReaderOfLabels) SortedPostings(index.Postings) index.Postings { + panic("SortedPostings called") +} + +func (m mockReaderOfLabels) Series(storage.SeriesRef, *labels.ScratchBuilder, *[]chunks.Meta) error { + panic("Series called") +} + +func (m mockReaderOfLabels) Symbols() index.StringIter { + panic("Series called") +} diff --git a/tsdb/wlog/checkpoint_test.go b/tsdb/wlog/checkpoint_test.go index 279f7c4356..a9786454de 100644 --- a/tsdb/wlog/checkpoint_test.go +++ b/tsdb/wlog/checkpoint_test.go @@ -19,6 +19,7 @@ import ( "os" "path/filepath" "sort" + "strconv" "strings" "testing" @@ -232,10 +233,10 @@ func TestCheckpoint(t *testing.T) { // Write changing metadata for each series. In the end, only the latest // version should end up in the checkpoint. b = enc.Metadata([]record.RefMetadata{ - {Ref: 0, Unit: fmt.Sprintf("%d", last), Help: fmt.Sprintf("%d", last)}, - {Ref: 1, Unit: fmt.Sprintf("%d", last), Help: fmt.Sprintf("%d", last)}, - {Ref: 2, Unit: fmt.Sprintf("%d", last), Help: fmt.Sprintf("%d", last)}, - {Ref: 3, Unit: fmt.Sprintf("%d", last), Help: fmt.Sprintf("%d", last)}, + {Ref: 0, Unit: strconv.FormatInt(last, 10), Help: strconv.FormatInt(last, 10)}, + {Ref: 1, Unit: strconv.FormatInt(last, 10), Help: strconv.FormatInt(last, 10)}, + {Ref: 2, Unit: strconv.FormatInt(last, 10), Help: strconv.FormatInt(last, 10)}, + {Ref: 3, Unit: strconv.FormatInt(last, 10), Help: strconv.FormatInt(last, 10)}, }, nil) require.NoError(t, w.Log(b)) @@ -250,7 +251,7 @@ func TestCheckpoint(t *testing.T) { require.NoError(t, w.Truncate(107)) require.NoError(t, DeleteCheckpoints(w.Dir(), 106)) require.Equal(t, histogramsInWAL+floatHistogramsInWAL+samplesInWAL, stats.TotalSamples) - require.Greater(t, stats.DroppedSamples, 0) + require.Positive(t, stats.DroppedSamples) // Only the new checkpoint should be left. files, err := os.ReadDir(dir) @@ -324,8 +325,8 @@ func TestCheckpoint(t *testing.T) { testutil.RequireEqual(t, expectedRefSeries, series) expectedRefMetadata := []record.RefMetadata{ - {Ref: 0, Unit: fmt.Sprintf("%d", last-100), Help: fmt.Sprintf("%d", last-100)}, - {Ref: 2, Unit: fmt.Sprintf("%d", last-100), Help: fmt.Sprintf("%d", last-100)}, + {Ref: 0, Unit: strconv.FormatInt(last-100, 10), Help: strconv.FormatInt(last-100, 10)}, + {Ref: 2, Unit: strconv.FormatInt(last-100, 10), Help: strconv.FormatInt(last-100, 10)}, {Ref: 4, Unit: "unit", Help: "help"}, } sort.Slice(metadata, func(i, j int) bool { return metadata[i].Ref < metadata[j].Ref }) diff --git a/util/testutil/context.go b/util/testutil/context.go index 3f63b030d7..0c9e0f6f64 100644 --- a/util/testutil/context.go +++ b/util/testutil/context.go @@ -13,7 +13,12 @@ package testutil -import "time" +import ( + "context" + "time" + + "go.uber.org/atomic" +) // A MockContext provides a simple stub implementation of a Context. type MockContext struct { @@ -40,3 +45,23 @@ func (c *MockContext) Err() error { func (c *MockContext) Value(interface{}) interface{} { return nil } + +// MockContextErrAfter is a MockContext that will return an error after a certain +// number of calls to Err(). +type MockContextErrAfter struct { + MockContext + count atomic.Uint64 + FailAfter uint64 +} + +func (c *MockContextErrAfter) Err() error { + c.count.Inc() + if c.count.Load() >= c.FailAfter { + return context.Canceled + } + return c.MockContext.Err() +} + +func (c *MockContextErrAfter) Count() uint64 { + return c.count.Load() +} diff --git a/web/api/v1/api.go b/web/api/v1/api.go index dc22365073..f0884926e1 100644 --- a/web/api/v1/api.go +++ b/web/api/v1/api.go @@ -116,9 +116,11 @@ type RulesRetriever interface { AlertingRules() []*rules.AlertingRule } +// StatsRenderer converts engine statistics into a format suitable for the API. type StatsRenderer func(context.Context, *stats.Statistics, string) stats.QueryStats -func defaultStatsRenderer(_ context.Context, s *stats.Statistics, param string) stats.QueryStats { +// DefaultStatsRenderer is the default stats renderer for the API. +func DefaultStatsRenderer(_ context.Context, s *stats.Statistics, param string) stats.QueryStats { if param != "" { return stats.NewQueryStats(s) } @@ -272,7 +274,7 @@ func NewAPI( buildInfo: buildInfo, gatherer: gatherer, isAgent: isAgent, - statsRenderer: defaultStatsRenderer, + statsRenderer: DefaultStatsRenderer, remoteReadHandler: remote.NewReadHandler(logger, registerer, q, configFunc, remoteReadSampleLimit, remoteReadConcurrencyLimit, remoteReadMaxBytesInFrame), } @@ -461,7 +463,7 @@ func (api *API) query(r *http.Request) (result apiFuncResult) { // Optional stats field in response if parameter "stats" is not empty. sr := api.statsRenderer if sr == nil { - sr = defaultStatsRenderer + sr = DefaultStatsRenderer } qs := sr(ctx, qry.Stats(), r.FormValue("stats")) @@ -563,7 +565,7 @@ func (api *API) queryRange(r *http.Request) (result apiFuncResult) { // Optional stats field in response if parameter "stats" is not empty. sr := api.statsRenderer if sr == nil { - sr = defaultStatsRenderer + sr = DefaultStatsRenderer } qs := sr(ctx, qry.Stats(), r.FormValue("stats")) @@ -702,7 +704,7 @@ func (api *API) labelNames(r *http.Request) apiFuncResult { names = []string{} } - if len(names) >= limit { + if len(names) > limit { names = names[:limit] warnings = warnings.Add(errors.New("results truncated due to limit")) } @@ -791,7 +793,7 @@ func (api *API) labelValues(r *http.Request) (result apiFuncResult) { slices.Sort(vals) - if len(vals) >= limit { + if len(vals) > limit { vals = vals[:limit] warnings = warnings.Add(errors.New("results truncated due to limit")) } @@ -887,7 +889,8 @@ func (api *API) series(r *http.Request) (result apiFuncResult) { } metrics = append(metrics, set.At().Labels()) - if len(metrics) >= limit { + if len(metrics) > limit { + metrics = metrics[:limit] warnings.Add(errors.New("results truncated due to limit")) return apiFuncResult{metrics, nil, warnings, closer} } diff --git a/web/api/v1/api_test.go b/web/api/v1/api_test.go index 9d7d1d502a..b30890893b 100644 --- a/web/api/v1/api_test.go +++ b/web/api/v1/api_test.go @@ -25,6 +25,7 @@ import ( "reflect" "runtime" "sort" + "strconv" "strings" "testing" "time" @@ -1059,6 +1060,7 @@ func testEndpoints(t *testing.T, api *API, tr *testTargetRetriever, es storage.E responseLen int // If nonzero, check only the length; `response` is ignored. responseMetadataTotal int responseAsJSON string + warningsCount int errType errorType sorter func(interface{}) metadata []targetMetadata @@ -1416,7 +1418,17 @@ func testEndpoints(t *testing.T, api *API, tr *testTargetRetriever, es storage.E "match[]": []string{"test_metric1"}, "limit": []string{"1"}, }, - responseLen: 1, // API does not specify which particular value will come back. + responseLen: 1, // API does not specify which particular value will come back. + warningsCount: 1, + }, + { + endpoint: api.series, + query: url.Values{ + "match[]": []string{"test_metric1"}, + "limit": []string{"2"}, + }, + responseLen: 2, // API does not specify which particular value will come back. + warningsCount: 0, // No warnings if limit isn't exceeded. }, // Missing match[] query params in series requests. { @@ -2699,7 +2711,19 @@ func testEndpoints(t *testing.T, api *API, tr *testTargetRetriever, es storage.E query: url.Values{ "limit": []string{"2"}, }, - responseLen: 2, // API does not specify which particular values will come back. + responseLen: 2, // API does not specify which particular values will come back. + warningsCount: 1, + }, + { + endpoint: api.labelValues, + params: map[string]string{ + "name": "__name__", + }, + query: url.Values{ + "limit": []string{"4"}, + }, + responseLen: 4, // API does not specify which particular values will come back. + warningsCount: 0, // No warnings if limit isn't exceeded. }, // Label names. { @@ -2846,7 +2870,16 @@ func testEndpoints(t *testing.T, api *API, tr *testTargetRetriever, es storage.E query: url.Values{ "limit": []string{"2"}, }, - responseLen: 2, // API does not specify which particular values will come back. + responseLen: 2, // API does not specify which particular values will come back. + warningsCount: 1, + }, + { + endpoint: api.labelNames, + query: url.Values{ + "limit": []string{"3"}, + }, + responseLen: 3, // API does not specify which particular values will come back. + warningsCount: 0, // No warnings if limit isn't exceeded. }, }...) } @@ -2923,6 +2956,8 @@ func testEndpoints(t *testing.T, api *API, tr *testTargetRetriever, es storage.E require.NoError(t, err) require.JSONEq(t, test.responseAsJSON, string(s)) } + + require.Len(t, res.warnings, test.warningsCount) }) } }) @@ -2938,8 +2973,10 @@ func assertAPIError(t *testing.T, got *apiError, exp errorType) { t.Helper() if exp == errorNone { + //nolint:testifylint require.Nil(t, got) } else { + //nolint:testifylint require.NotNil(t, got) require.Equal(t, exp, got.typ, "(%q)", got) } @@ -3544,7 +3581,7 @@ func TestTSDBStatus(t *testing.T) { }, } { tc := tc - t.Run(fmt.Sprintf("%d", i), func(t *testing.T) { + t.Run(strconv.Itoa(i), func(t *testing.T) { api := &API{db: tc.db, gatherer: prometheus.DefaultGatherer} endpoint := tc.endpoint(api) req, err := http.NewRequest(tc.method, fmt.Sprintf("?%s", tc.values.Encode()), nil) diff --git a/web/federate_test.go b/web/federate_test.go index e5adb97eb5..056a95d67f 100644 --- a/web/federate_test.go +++ b/web/federate_test.go @@ -22,6 +22,7 @@ import ( "net/http" "net/http/httptest" "sort" + "strconv" "strings" "testing" "time" @@ -341,8 +342,8 @@ func TestFederationWithNativeHistograms(t *testing.T) { } app := db.Appender(context.Background()) for i := 0; i < 6; i++ { - l := labels.FromStrings("__name__", "test_metric", "foo", fmt.Sprintf("%d", i)) - expL := labels.FromStrings("__name__", "test_metric", "instance", "", "foo", fmt.Sprintf("%d", i)) + l := labels.FromStrings("__name__", "test_metric", "foo", strconv.Itoa(i)) + expL := labels.FromStrings("__name__", "test_metric", "instance", "", "foo", strconv.Itoa(i)) var err error switch i { case 0, 3: