Label Volume Endpoint (#9588)

For a given set of matchers, returns the top N associated label/value pairs by volume. A query for `{cluster=prod}` will return ``` cluster=prod: size (total logs matching this matcher) . . . nth-label=nth-value ``` This is to service use cases where users want to understand where their log volume has come from by label without making multiple requests to the stats endpoint. Note: This PR is a monster but it's mostly plumbing. I've pointed out the most interesting bits that actually get the volumes from ingesters/indexs
grafana · Jun 12, 2023 · 065bee7 · 065bee7
1 parent 4d997a5
commit 065bee7
Show file tree

Hide file tree

Showing 85 changed files with 3,868 additions and 447 deletions.
diff --git a/clients/pkg/promtail/client/client_writeto.go b/clients/pkg/promtail/client/client_writeto.go
@@ -11,6 +11,7 @@ import (
 	"github.com/prometheus/prometheus/tsdb/record"
 
 	"github.com/grafana/loki/clients/pkg/promtail/api"
+
 	"github.com/grafana/loki/pkg/ingester/wal"
 	"github.com/grafana/loki/pkg/util"
 )

diff --git a/clients/pkg/promtail/client/client_writeto_test.go b/clients/pkg/promtail/client/client_writeto_test.go
@@ -2,13 +2,14 @@ package client
 
 import (
 	"fmt"
-	"go.uber.org/atomic"
 	"math/rand"
 	"os"
 	"sync"
 	"testing"
 	"time"
 
+	"go.uber.org/atomic"
+
 	"github.com/go-kit/log"
 	"github.com/go-kit/log/level"
 	"github.com/prometheus/common/model"
@@ -18,6 +19,7 @@ import (
 	"github.com/stretchr/testify/require"
 
 	"github.com/grafana/loki/clients/pkg/promtail/api"
+
 	"github.com/grafana/loki/pkg/ingester/wal"
 	"github.com/grafana/loki/pkg/logproto"
 )

diff --git a/clients/pkg/promtail/promtail_wal_test.go b/clients/pkg/promtail/promtail_wal_test.go
@@ -26,6 +26,7 @@ import (
 	"github.com/grafana/loki/clients/pkg/promtail/scrapeconfig"
 	"github.com/grafana/loki/clients/pkg/promtail/utils"
 	"github.com/grafana/loki/clients/pkg/promtail/wal"
+
 	"github.com/grafana/loki/pkg/push"
 	util_log "github.com/grafana/loki/pkg/util/log"
 )

diff --git a/clients/pkg/promtail/server/ui/assets_vfsdata.go b/clients/pkg/promtail/server/ui/assets_vfsdata.go
@@ -1,5 +1,6 @@
 // Code generated by vfsgen; DO NOT EDIT.
 
+//go:build !dev
 // +build !dev
 
 package ui

diff --git a/clients/pkg/promtail/targets/azureeventhubs/parser.go b/clients/pkg/promtail/targets/azureeventhubs/parser.go
@@ -13,6 +13,7 @@ import (
 	"github.com/prometheus/prometheus/model/relabel"
 
 	"github.com/grafana/loki/clients/pkg/promtail/api"
+
 	"github.com/grafana/loki/pkg/logproto"
 )
 

diff --git a/clients/pkg/promtail/targets/azureeventhubs/target_syncer_test.go b/clients/pkg/promtail/targets/azureeventhubs/target_syncer_test.go
@@ -2,13 +2,15 @@ package azureeventhubs
 
 import (
 	"errors"
+	"testing"
+	"time"
+
 	"github.com/Shopify/sarama"
 	"github.com/go-kit/log"
-	"github.com/grafana/loki/clients/pkg/promtail/scrapeconfig"
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/stretchr/testify/assert"
-	"testing"
-	"time"
+
+	"github.com/grafana/loki/clients/pkg/promtail/scrapeconfig"
 )
 
 func Test_validateConfig(t *testing.T) {

diff --git a/clients/pkg/promtail/targets/kafka/parser.go b/clients/pkg/promtail/targets/kafka/parser.go
@@ -6,6 +6,7 @@ import (
 	"github.com/prometheus/prometheus/model/relabel"
 
 	"github.com/grafana/loki/clients/pkg/promtail/api"
+
 	"github.com/grafana/loki/pkg/logproto"
 )
 

diff --git a/clients/pkg/promtail/targets/kafka/target_syncer.go b/clients/pkg/promtail/targets/kafka/target_syncer.go
@@ -20,6 +20,7 @@ import (
 	"github.com/grafana/loki/clients/pkg/promtail/api"
 	"github.com/grafana/loki/clients/pkg/promtail/scrapeconfig"
 	"github.com/grafana/loki/clients/pkg/promtail/targets/target"
+
 	"github.com/grafana/loki/pkg/util"
 )
 

diff --git a/clients/pkg/promtail/targets/windows/win_eventlog/event.go b/clients/pkg/promtail/targets/windows/win_eventlog/event.go
@@ -23,8 +23,9 @@
 //go:build windows
 // +build windows
 
-//revive:disable-next-line:var-naming
 // Package win_eventlog Input plugin to collect Windows Event Log messages
+//
+//revive:disable-next-line:var-naming
 package win_eventlog
 
 // Event is the event entry representation

diff --git a/clients/pkg/promtail/targets/windows/win_eventlog/syscall_windows.go b/clients/pkg/promtail/targets/windows/win_eventlog/syscall_windows.go
@@ -22,8 +22,9 @@
 //go:build windows
 // +build windows
 
-//revive:disable-next-line:var-naming
 // Package win_eventlog Input plugin to collect Windows Event Log messages
+//
+//revive:disable-next-line:var-naming
 package win_eventlog
 
 import (

diff --git a/clients/pkg/promtail/targets/windows/win_eventlog/util.go b/clients/pkg/promtail/targets/windows/win_eventlog/util.go
@@ -22,8 +22,9 @@
 //go:build windows
 // +build windows
 
-//revive:disable-next-line:var-naming
 // Package win_eventlog Input plugin to collect Windows Event Log messages
+//
+//revive:disable-next-line:var-naming
 package win_eventlog
 
 import (

diff --git a/clients/pkg/promtail/targets/windows/win_eventlog/util_test.go b/clients/pkg/promtail/targets/windows/win_eventlog/util_test.go
@@ -22,8 +22,9 @@
 //go:build windows
 // +build windows
 
-//revive:disable-next-line:var-naming
 // Package win_eventlog Input plugin to collect Windows Event Log messages
+//
+//revive:disable-next-line:var-naming
 package win_eventlog
 
 import (

diff --git a/clients/pkg/promtail/targets/windows/win_eventlog/win_eventlog.go b/clients/pkg/promtail/targets/windows/win_eventlog/win_eventlog.go
@@ -22,8 +22,9 @@
 //go:build windows
 // +build windows
 
-//revive:disable-next-line:var-naming
 // Package win_eventlog Input plugin to collect Windows Event Log messages
+//
+//revive:disable-next-line:var-naming
 package win_eventlog
 
 import (

diff --git a/clients/pkg/promtail/targets/windows/win_eventlog/win_eventlog_notwindows.go b/clients/pkg/promtail/targets/windows/win_eventlog/win_eventlog_notwindows.go
@@ -22,6 +22,7 @@
 //go:build !windows
 // +build !windows
 
-//revive:disable-next-line:var-naming
 // Package win_eventlog Input plugin to collect Windows Event Log messages
+//
+//revive:disable-next-line:var-naming
 package win_eventlog
diff --git a/clients/pkg/promtail/targets/windows/win_eventlog/win_eventlog_test.go b/clients/pkg/promtail/targets/windows/win_eventlog/win_eventlog_test.go
@@ -22,8 +22,9 @@
 //go:build windows
 // +build windows
 
-//revive:disable-next-line:var-naming
 // Package win_eventlog Input plugin to collect Windows Event Log messages
+//
+//revive:disable-next-line:var-naming
 package win_eventlog
 
 import (

diff --git a/clients/pkg/promtail/targets/windows/win_eventlog/zsyscall_windows.go b/clients/pkg/promtail/targets/windows/win_eventlog/zsyscall_windows.go
@@ -22,8 +22,9 @@
 //go:build windows
 // +build windows
 
-//revive:disable-next-line:var-naming
 // Package win_eventlog Input plugin to collect Windows Event Log messages
+//
+//revive:disable-next-line:var-naming
 package win_eventlog
 
 import (

diff --git a/clients/pkg/promtail/wal/watcher_test.go b/clients/pkg/promtail/wal/watcher_test.go
@@ -14,6 +14,7 @@ import (
 	"github.com/stretchr/testify/require"
 
 	"github.com/grafana/loki/clients/pkg/promtail/api"
+
 	"github.com/grafana/loki/pkg/ingester/wal"
 	"github.com/grafana/loki/pkg/logproto"
 	"github.com/grafana/loki/pkg/util"

diff --git a/integration/loki_micro_services_delete_test.go b/integration/loki_micro_services_delete_test.go
@@ -12,6 +12,7 @@ import (
 
 	"github.com/grafana/loki/integration/client"
 	"github.com/grafana/loki/integration/cluster"
+
 	"github.com/grafana/loki/pkg/storage"
 )
 

diff --git a/integration/loki_micro_services_test.go b/integration/loki_micro_services_test.go
@@ -10,6 +10,7 @@ import (
 
 	"github.com/grafana/loki/integration/client"
 	"github.com/grafana/loki/integration/cluster"
+
 	"github.com/grafana/loki/pkg/storage"
 	"github.com/grafana/loki/pkg/util/querylimits"
 )

diff --git a/pkg/distributor/writefailures/manager.go b/pkg/distributor/writefailures/manager.go
@@ -22,10 +22,10 @@ func NewManager(logger log.Logger, cfg Cfg, tenants *runtime.TenantConfigs) *Man
 		logger = log.With(logger, "insight", "true")
 	}
 
-	strat := newStrategy(cfg.LogRate.Val(), float64(cfg.LogRate.Val()))
+	strategy := newStrategy(cfg.LogRate.Val(), float64(cfg.LogRate.Val()))
 
 	return &Manager{
-		limiter:    limiter.NewRateLimiter(strat, time.Minute),
+		limiter:    limiter.NewRateLimiter(strategy, time.Minute),
 		logger:     logger,
 		tenantCfgs: tenants,
 	}

diff --git a/pkg/ingester/flush_test.go b/pkg/ingester/flush_test.go
@@ -351,6 +351,10 @@ func (s *testStore) Stats(ctx context.Context, userID string, from, through mode
 	return &stats.Stats{}, nil
 }
 
+func (s *testStore) LabelVolume(ctx context.Context, userID string, from, through model.Time, limit int32, matchers ...*labels.Matcher) (*logproto.LabelVolumeResponse, error) {
+	return &logproto.LabelVolumeResponse{}, nil
+}
+
 func pushTestSamples(t *testing.T, ing logproto.PusherServer) map[string][]logproto.Stream {
 	userIDs := []string{"1", "2", "3"}
 

diff --git a/pkg/ingester/ingester.go b/pkg/ingester/ingester.go
@@ -11,6 +11,8 @@ import (
 	"sync"
 	"time"
 
+	"github.com/grafana/loki/pkg/storage/stores/index/labelvolume"
+
 	"github.com/go-kit/log/level"
 	"github.com/grafana/dskit/concurrency"
 	"github.com/grafana/dskit/modules"
@@ -171,6 +173,7 @@ type ChunkStore interface {
 	GetChunkRefs(ctx context.Context, userID string, from, through model.Time, matchers ...*labels.Matcher) ([][]chunk.Chunk, []*fetcher.Fetcher, error)
 	GetSchemaConfigs() []config.PeriodConfig
 	Stats(ctx context.Context, userID string, from, through model.Time, matchers ...*labels.Matcher) (*index_stats.Stats, error)
+	LabelVolume(ctx context.Context, userID string, from, through model.Time, limit int32, matchers ...*labels.Matcher) (*logproto.LabelVolumeResponse, error)
 }
 
 // Interface is an interface for the Ingester
@@ -1137,6 +1140,50 @@ func (i *Ingester) GetStats(ctx context.Context, req *logproto.IndexStatsRequest
 	return &merged, nil
 }
 
+func (i *Ingester) GetLabelVolume(ctx context.Context, req *logproto.LabelVolumeRequest) (*logproto.LabelVolumeResponse, error) {
+	user, err := tenant.TenantID(ctx)
+	if err != nil {
+		return nil, err
+	}
+
+	instance, err := i.GetOrCreateInstance(user)
+	if err != nil {
+		return nil, err
+	}
+
+	matchers, err := syntax.ParseMatchers(req.Matchers)
+	if err != nil && req.Matchers != labelvolume.MatchAny {
+		return nil, err
+	}
+
+	type f func() (*logproto.LabelVolumeResponse, error)
+	jobs := []f{
+		f(func() (*logproto.LabelVolumeResponse, error) {
+			return instance.GetLabelVolume(ctx, req)
+		}),
+		f(func() (*logproto.LabelVolumeResponse, error) {
+			return i.store.LabelVolume(ctx, user, req.From, req.Through, req.Limit, matchers...)
+		}),
+	}
+	resps := make([]*logproto.LabelVolumeResponse, len(jobs))
+
+	if err := concurrency.ForEachJob(
+		ctx,
+		len(jobs),
+		2,
+		func(ctx context.Context, idx int) error {
+			res, err := jobs[idx]()
+			resps[idx] = res
+			return err
+		},
+	); err != nil {
+		return nil, err
+	}
+
+	merged := labelvolume.Merge(resps, req.Limit)
+	return merged, nil
+}
+
 // Watch implements grpc_health_v1.HealthCheck.
 func (*Ingester) Watch(*grpc_health_v1.HealthCheckRequest, grpc_health_v1.Health_WatchServer) error {
 	return nil

diff --git a/pkg/ingester/ingester_test.go b/pkg/ingester/ingester_test.go
@@ -461,7 +461,21 @@ func (s *mockStore) GetChunkFetcher(tm model.Time) *fetcher.Fetcher {
 }
 
 func (s *mockStore) Stats(ctx context.Context, userID string, from, through model.Time, matchers ...*labels.Matcher) (*stats.Stats, error) {
-	return &stats.Stats{}, nil
+	return &stats.Stats{
+		Streams: 2,
+		Chunks:  5,
+		Bytes:   25,
+		Entries: 100,
+	}, nil
+}
+
+func (s *mockStore) LabelVolume(ctx context.Context, userID string, from, through model.Time, limit int32, matchers ...*labels.Matcher) (*logproto.LabelVolumeResponse, error) {
+	return &logproto.LabelVolumeResponse{
+		Volumes: []logproto.LabelVolume{
+			{Name: "foo", Value: "bar", Volume: 38},
+		},
+		Limit: limit,
+	}, nil
 }
 
 func (s *mockStore) Stop() {}
@@ -1040,6 +1054,60 @@ func Test_DedupeIngesterParser(t *testing.T) {
 	})
 }
 
+func TestStats(t *testing.T) {
+	ingesterConfig := defaultIngesterTestConfig(t)
+	limits, err := validation.NewOverrides(defaultLimitsTestConfig(), nil)
+	require.NoError(t, err)
+
+	i, err := New(ingesterConfig, client.Config{}, &mockStore{}, limits, runtime.DefaultTenantConfigs(), nil)
+	require.NoError(t, err)
+
+	i.instances["test"] = defaultInstance(t)
+
+	ctx := user.InjectOrgID(context.Background(), "test")
+
+	resp, err := i.GetStats(ctx, &logproto.IndexStatsRequest{
+		From:     0,
+		Through:  11000,
+		Matchers: `{host="agent"}`,
+	})
+	require.NoError(t, err)
+
+	require.Equal(t, &logproto.IndexStatsResponse{
+		Streams: 4,
+		Chunks:  7,
+		Bytes:   185,
+		Entries: 110,
+	}, resp)
+}
+
+func TestLabelVolume(t *testing.T) {
+	ingesterConfig := defaultIngesterTestConfig(t)
+	limits, err := validation.NewOverrides(defaultLimitsTestConfig(), nil)
+	require.NoError(t, err)
+
+	i, err := New(ingesterConfig, client.Config{}, &mockStore{}, limits, runtime.DefaultTenantConfigs(), nil)
+	require.NoError(t, err)
+
+	i.instances["test"] = defaultInstance(t)
+
+	ctx := user.InjectOrgID(context.Background(), "test")
+	volumes, err := i.GetLabelVolume(ctx, &logproto.LabelVolumeRequest{
+		From:     0,
+		Through:  10000,
+		Matchers: "{}",
+		Limit:    4,
+	})
+	require.NoError(t, err)
+
+	require.Equal(t, []logproto.LabelVolume{
+		{Name: "host", Value: "agent", Volume: 160},
+		{Name: "job", Value: "3", Volume: 160},
+		{Name: "log_stream", Value: "dispatcher", Volume: 90},
+		{Name: "log_stream", Value: "worker", Volume: 70},
+	}, volumes.Volumes)
+}
+
 type ingesterClient struct {
 	logproto.PusherClient
 	logproto.QuerierClient