Add ingestion rate global limit support (#1486)

* Added ingestion rate global limit support Signed-off-by: Marco Pracucci <[email protected]> * Fixed code comments Signed-off-by: Marco Pracucci <[email protected]> * Updated config doc Signed-off-by: Marco Pracucci <[email protected]> * Updated changelog Signed-off-by: Marco Pracucci <[email protected]> * Updated doc Signed-off-by: Marco Pracucci <[email protected]>
grafana · Jan 9, 2020 · ce407d3 · ce407d3
1 parent 33f70b7
commit ce407d3
Show file tree

Hide file tree

Showing 63 changed files with 11,401 additions and 107 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,6 +1,12 @@
 ## master / unreleased
 
+### Notable Changes
+* [1486](https://github.com/grafana/loki/pull/1486) **pracucci**: Deprecated `-distributor.limiter-reload-period` flag / distributor's `limiter_reload_period` config option.
+
+### Features
+
 * [FEATURE] promtail positions file corruptions can be ignored with the `positions.ignore-invalid-yaml` flag. In the case the positions yaml is corrupted an empty positions config will be used and should later overwrite the malformed yaml.
+* [1486](https://github.com/grafana/loki/pull/1486) **pracucci**: Added `global` ingestion rate limiter strategy support.
 
 # 1.2.0 (2019-12-09)
 

diff --git a/docs/configuration/README.md b/docs/configuration/README.md
@@ -144,8 +144,9 @@ The `server_config` block configures Promtail's behavior as an HTTP server:
 The `distributor_config` block configures the Loki Distributor.
 
 ```yaml
-# Period at which to reload user ingestion limits.
-[limiter_reload_period: <duration> | default = 5m]
+# Configures the distributors ring, used when the "global" ingestion rate
+# strategy is enabled.
+[ring: <ring_config>]
 ```
 
 ## querier_config
@@ -716,12 +717,27 @@ The `limits_config` block configures global and per-tenant limits for ingesting
 logs in Loki.
 
 ```yaml
+# Whether the ingestion rate limit should be applied individually to each
+# distributor instance (local), or evenly shared across the cluster (global).
+# The ingestion rate strategy cannot be overridden on a per-tenant basis.
+#
+# - local: enforces the limit on a per distributor basis. The actual effective
+#   rate limit will be N times higher, where N is the number of distributor
+#   replicas.
+# - global: enforces the limit globally, configuring a per-distributor local
+#   rate limiter as "ingestion_rate / N", where N is the number of distributor
+#   replicas (it's automatically adjusted if the number of replicas change).
+#   The global strategy requires the distributors to form their own ring, which
+#   is used to keep track of the current number of healthy distributor replicas.
+[ingestion_rate_strategy: <string> | default = "local"]
+
 # Per-user ingestion rate limit in sample size per second. Units in MB.
 [ingestion_rate_mb: <float> | default = 4]
 
-# Per-user allowed ingestion burst size (in sample size). Units in MB. Warning,
-# very high limits will be reset every limiter_reload_period defined in
-# distributor_config.
+# Per-user allowed ingestion burst size (in sample size). Units in MB.
+# The burst size refers to the per-distributor local rate limiter even in the
+# case of the "global" strategy, and should be set at least to the maximum logs
+# size expected in a single push request.
 [ingestion_burst_size_mb: <int> | default = 6]
 
 # Maximum length of a label name.

diff --git a/go.mod b/go.mod
@@ -53,7 +53,6 @@ require (
 	golang.org/x/net v0.0.0-20190923162816-aa69164e4478
 	golang.org/x/sync v0.0.0-20190911185100-cd5d95a43a6e // indirect
 	golang.org/x/sys v0.0.0-20191218084908-4a24b4065292 // indirect
-	golang.org/x/time v0.0.0-20191024005414-555d28b269f0
 	golang.org/x/tools v0.0.0-20190925134113-a044388aa56f // indirect
 	google.golang.org/appengine v1.6.3 // indirect
 	google.golang.org/genproto v0.0.0-20190916214212-f660b8655731 // indirect

diff --git a/go.sum b/go.sum
@@ -197,6 +197,7 @@ github.com/fatih/color v1.7.0/go.mod h1:Zm6kSWBoL9eyXnKyktHP6abPY2pDugNf5Kwzbycv
 github.com/fatih/structtag v1.0.0/go.mod h1:IKitwq45uXL/yqi5mYghiD3w9H6eTOvI9vnk8tXMphA=
 github.com/fluent/fluent-bit-go v0.0.0-20190925192703-ea13c021720c h1:QwbffUs/+ptC4kTFPEN9Ej2latTq3bZJ5HO/OwPXYMs=
 github.com/fluent/fluent-bit-go v0.0.0-20190925192703-ea13c021720c/go.mod h1:WQX+afhrekY9rGK+WT4xvKSlzmia9gDoLYu4GGYGASQ=
+github.com/fluent/fluent-logger-golang v1.2.1 h1:CMA+mw2zMiOGEOarZtaqM3GBWT1IVLNncNi0nKELtmU=
 github.com/fluent/fluent-logger-golang v1.2.1/go.mod h1:2/HCT/jTy78yGyeNGQLGQsjF3zzzAuy6Xlk6FCMV5eU=
 github.com/fortytw2/leaktest v1.3.0/go.mod h1:jDsjWgpAGjm2CA7WthBh/CdZYEPF31XHquHwclZch5g=
 github.com/frankban/quicktest v1.7.2 h1:2QxQoC1TS09S7fhCPsrvqYdvP1H5M1P1ih5ABm3BTYk=
@@ -563,6 +564,7 @@ github.com/pascaldekloe/goe v0.1.0 h1:cBOtyMzM9HTpWjXfbbunk26uA6nG3a8n06Wieeh0Mw
 github.com/pascaldekloe/goe v0.1.0/go.mod h1:lzWF7FIEvWOWxwDKqyGYQf6ZUaNfKdP144TG7ZOy1lc=
 github.com/pborman/uuid v1.2.0/go.mod h1:X/NO0urCmaxf9VXbdlT7C2Yzkj2IKimNn4k+gtPdI/k=
 github.com/peterbourgon/diskv v2.0.1+incompatible/go.mod h1:uqqh8zWWbv1HBMNONnaR/tNboyR3/BZd58JJSHlUSCU=
+github.com/philhofer/fwd v0.0.0-20160129035939-98c11a7a6ec8 h1:jkUFVqrKRttbdDqkTrvOmHxfqIsJK0Oe2WGi1ACAE+M=
 github.com/philhofer/fwd v0.0.0-20160129035939-98c11a7a6ec8/go.mod h1:gk3iGcWd9+svBvR0sR+KPcfE+RNWozjowpeBVG3ZVNU=
 github.com/pierrec/lz4 v2.0.5+incompatible/go.mod h1:pdkljMzZIN41W+lC3N2tnIh5sFi+IEE17M5jbnwPHcY=
 github.com/pierrec/lz4 v2.3.1-0.20191115212037-9085dacd1e1e+incompatible h1:5isCJDRADbeSlWx6KVXAYwrcihyCGVXr7GNCdLEVDr8=
@@ -666,6 +668,7 @@ github.com/stretchr/testify v1.4.0/go.mod h1:j7eGeouHqKxXV5pUuKE4zz7dFj8WfuZ+81P
 github.com/thanos-io/thanos v0.8.1/go.mod h1:qQDi/6tgypn96+VzSumlxfJIgFX2y3ablfhHHLZ05cg=
 github.com/tidwall/pretty v0.0.0-20180105212114-65a9db5fad51/go.mod h1:XNkn88O1ChpSDQmQeStsy+sBenx6DDtFZJxhVysOjyk=
 github.com/tidwall/pretty v1.0.0/go.mod h1:XNkn88O1ChpSDQmQeStsy+sBenx6DDtFZJxhVysOjyk=
+github.com/tinylib/msgp v0.0.0-20161221055906-38a6f61a768d h1:Ninez2SUm08xpmnw7kVxCeOc3DahF6IuMuRMCdM4wTQ=
 github.com/tinylib/msgp v0.0.0-20161221055906-38a6f61a768d/go.mod h1:+d+yLhGm8mzTaHzB+wgMYrodPfmZrzkirds8fDWklFE=
 github.com/tmc/grpc-websocket-proxy v0.0.0-20170815181823-89b8d40f7ca8/go.mod h1:ncp9v5uamzpCO7NfCPTXjqaC+bZgJeR0sMTm6dMHP7U=
 github.com/tmc/grpc-websocket-proxy v0.0.0-20190109142713-0ad062ec5ee5 h1:LnC5Kc/wtumK+WB441p7ynQJzVuNRJiqddSIE3IlSEQ=
@@ -687,6 +690,7 @@ github.com/ugorji/go v1.1.7/go.mod h1:kZn38zHttfInRq0xu/PH0az30d+z6vm202qpg1oXVM
 github.com/ugorji/go/codec v1.1.7 h1:2SvQaVZ1ouYrrKKwoSk2pzd4A9evlKJb9oTL+OaLUSs=
 github.com/ugorji/go/codec v1.1.7/go.mod h1:Ax+UKWsSmolVDwsd+7N3ZtXu+yMGCf907BLYF3GoBXY=
 github.com/urfave/cli v1.20.0/go.mod h1:70zkFmudgCuE/ngEzBv17Jvp/497gISqfk5gWijbERA=
+github.com/weaveworks/billing-client v0.0.0-20171006123215-be0d55e547b1 h1:qi+YkNiB7T3Ikw1DoDIFhdAPbDU7fUPDsKrUoZdupnQ=
 github.com/weaveworks/billing-client v0.0.0-20171006123215-be0d55e547b1/go.mod h1:7gGdEUJaCrSlWi/mjd68CZv0sfqektYPDcro9cE+M9k=
 github.com/weaveworks/common v0.0.0-20190822150010-afb9996716e4 h1:O8BmyjqQoByXjAj6XaTfcxxqSIK6DYLmOSiYQPL9yJg=
 github.com/weaveworks/common v0.0.0-20190822150010-afb9996716e4/go.mod h1:pSm+0KR57BG3pvGoJWFXJSAC7+sEPewcvdt5StevL3A=

diff --git a/pkg/distributor/distributor.go b/pkg/distributor/distributor.go
@@ -4,13 +4,14 @@ import (
 	"context"
 	"flag"
 	"net/http"
-	"sync"
 	"sync/atomic"
 	"time"
 
+	cortex_distributor "github.com/cortexproject/cortex/pkg/distributor"
 	cortex_client "github.com/cortexproject/cortex/pkg/ingester/client"
 	"github.com/cortexproject/cortex/pkg/ring"
 	cortex_util "github.com/cortexproject/cortex/pkg/util"
+	"github.com/cortexproject/cortex/pkg/util/limiter"
 	cortex_validation "github.com/cortexproject/cortex/pkg/util/validation"
 
 	"github.com/go-kit/kit/log/level"
@@ -19,7 +20,6 @@ import (
 	"github.com/prometheus/client_golang/prometheus/promauto"
 	"github.com/weaveworks/common/httpgrpc"
 	"github.com/weaveworks/common/user"
-	"golang.org/x/time/rate"
 	"google.golang.org/grpc/health/grpc_health_v1"
 
 	"github.com/grafana/loki/pkg/ingester/client"
@@ -30,7 +30,6 @@ import (
 
 const (
 	metricName = "logs"
-	bytesInMB  = 1048576
 )
 
 var readinessProbeSuccess = []byte("Ready")
@@ -60,77 +59,78 @@ var (
 
 // Config for a Distributor.
 type Config struct {
-	// For testing.
-	factory func(addr string) (grpc_health_v1.HealthClient, error)
+	// Distributors ring
+	DistributorRing cortex_distributor.RingConfig `yaml:"ring,omitempty"`
 
-	LimiterReloadPeriod time.Duration `yaml:"limiter_reload_period"`
+	// For testing.
+	factory func(addr string) (grpc_health_v1.HealthClient, error) `yaml:"-"`
 }
 
 // RegisterFlags registers the flags.
 func (cfg *Config) RegisterFlags(f *flag.FlagSet) {
-	f.DurationVar(&cfg.LimiterReloadPeriod, "distributor.limiter-reload-period", 5*time.Minute, "Period at which to reload user ingestion limits.")
+	cfg.DistributorRing.RegisterFlags(f)
 }
 
 // Distributor coordinates replicates and distribution of log streams.
 type Distributor struct {
-	cfg       Config
-	clientCfg client.Config
-	ring      ring.ReadRing
-	overrides *validation.Overrides
-	pool      *cortex_client.Pool
-
-	ingestLimitersMtx sync.RWMutex
-	ingestLimiters    map[string]*rate.Limiter
-	quit              chan struct{}
+	cfg           Config
+	clientCfg     client.Config
+	ingestersRing ring.ReadRing
+	overrides     *validation.Overrides
+	pool          *cortex_client.Pool
+
+	// The global rate limiter requires a distributors ring to count
+	// the number of healthy instances.
+	distributorsRing *ring.Lifecycler
+
+	// Per-user rate limiter.
+	ingestionRateLimiter *limiter.RateLimiter
 }
 
 // New a distributor creates.
-func New(cfg Config, clientCfg client.Config, ring ring.ReadRing, overrides *validation.Overrides) (*Distributor, error) {
+func New(cfg Config, clientCfg client.Config, ingestersRing ring.ReadRing, overrides *validation.Overrides) (*Distributor, error) {
 	factory := cfg.factory
 	if factory == nil {
 		factory = func(addr string) (grpc_health_v1.HealthClient, error) {
 			return client.New(clientCfg, addr)
 		}
 	}
 
-	d := Distributor{
-		cfg:            cfg,
-		clientCfg:      clientCfg,
-		ring:           ring,
-		overrides:      overrides,
-		pool:           cortex_client.NewPool(clientCfg.PoolConfig, ring, factory, cortex_util.Logger),
-		ingestLimiters: map[string]*rate.Limiter{},
-		quit:           make(chan struct{}),
-	}
+	// Create the configured ingestion rate limit strategy (local or global).
+	var ingestionRateStrategy limiter.RateLimiterStrategy
+	var distributorsRing *ring.Lifecycler
 
-	go d.loop()
+	if overrides.IngestionRateStrategy() == validation.GlobalIngestionRateStrategy {
+		var err error
+		distributorsRing, err = ring.NewLifecycler(cfg.DistributorRing.ToLifecyclerConfig(), nil, "distributor", ring.DistributorRingKey)
+		if err != nil {
+			return nil, err
+		}
 
-	return &d, nil
-}
+		distributorsRing.Start()
 
-func (d *Distributor) loop() {
-	if d.cfg.LimiterReloadPeriod == 0 {
-		return
+		ingestionRateStrategy = newGlobalIngestionRateStrategy(overrides, distributorsRing)
+	} else {
+		ingestionRateStrategy = newLocalIngestionRateStrategy(overrides)
 	}
 
-	ticker := time.NewTicker(d.cfg.LimiterReloadPeriod)
-	defer ticker.Stop()
-
-	for {
-		select {
-		case <-ticker.C:
-			d.ingestLimitersMtx.Lock()
-			d.ingestLimiters = make(map[string]*rate.Limiter, len(d.ingestLimiters))
-			d.ingestLimitersMtx.Unlock()
-
-		case <-d.quit:
-			return
-		}
+	d := Distributor{
+		cfg:                  cfg,
+		clientCfg:            clientCfg,
+		ingestersRing:        ingestersRing,
+		distributorsRing:     distributorsRing,
+		overrides:            overrides,
+		pool:                 cortex_client.NewPool(clientCfg.PoolConfig, ingestersRing, factory, cortex_util.Logger),
+		ingestionRateLimiter: limiter.NewRateLimiter(ingestionRateStrategy, 10*time.Second),
 	}
+
+	return &d, nil
 }
 
 func (d *Distributor) Stop() {
-	close(d.quit)
+	if d.distributorsRing != nil {
+		d.distributorsRing.Shutdown()
+	}
 }
 
 // TODO taken from Cortex, see if we can refactor out an usable interface.
@@ -153,7 +153,7 @@ type pushTracker struct {
 // ReadinessHandler is used to indicate to k8s when the distributor is ready.
 // Returns 200 when the distributor is ready, 500 otherwise.
 func (d *Distributor) ReadinessHandler(w http.ResponseWriter, r *http.Request) {
-	_, err := d.ring.GetAll()
+	_, err := d.ingestersRing.GetAll()
 	if err != nil {
 		http.Error(w, "Not ready: "+err.Error(), http.StatusInternalServerError)
 		return
@@ -226,13 +226,13 @@ func (d *Distributor) Push(ctx context.Context, req *logproto.PushRequest) (*log
 		return &logproto.PushResponse{}, validationErr
 	}
 
-	limiter := d.getOrCreateIngestLimiter(userID)
-	if !limiter.AllowN(time.Now(), validatedSamplesSize) {
+	now := time.Now()
+	if !d.ingestionRateLimiter.AllowN(now, userID, validatedSamplesSize) {
 		// Return a 4xx here to have the client discard the data and not retry. If a client
 		// is sending too much data consistently we will unlikely ever catch up otherwise.
 		validation.DiscardedSamples.WithLabelValues(validation.RateLimited, userID).Add(float64(validatedSamplesCount))
 		validation.DiscardedBytes.WithLabelValues(validation.RateLimited, userID).Add(float64(validatedSamplesSize))
-		return nil, httpgrpc.Errorf(http.StatusTooManyRequests, "ingestion rate limit (%d) exceeded while adding %d lines", int(limiter.Limit()), validatedSamplesCount)
+		return nil, httpgrpc.Errorf(http.StatusTooManyRequests, "ingestion rate limit (%d bytes) exceeded while adding %d lines for a total size of %d bytes", int(d.ingestionRateLimiter.Limit(now, userID)), validatedSamplesCount, validatedSamplesSize)
 	}
 
 	const maxExpectedReplicationSet = 5 // typical replication factor 3 plus one for inactive plus one for luck
@@ -241,7 +241,7 @@ func (d *Distributor) Push(ctx context.Context, req *logproto.PushRequest) (*log
 	samplesByIngester := map[string][]*streamTracker{}
 	ingesterDescs := map[string]ring.IngesterDesc{}
 	for i, key := range keys {
-		replicationSet, err := d.ring.Get(key, ring.Write, descs[:0])
+		replicationSet, err := d.ingestersRing.Get(key, ring.Write, descs[:0])
 		if err != nil {
 			return nil, err
 		}
@@ -349,21 +349,3 @@ func (d *Distributor) sendSamplesErr(ctx context.Context, ingester ring.Ingester
 func (*Distributor) Check(_ context.Context, _ *grpc_health_v1.HealthCheckRequest) (*grpc_health_v1.HealthCheckResponse, error) {
 	return &grpc_health_v1.HealthCheckResponse{Status: grpc_health_v1.HealthCheckResponse_SERVING}, nil
 }
-
-func (d *Distributor) getOrCreateIngestLimiter(userID string) *rate.Limiter {
-	d.ingestLimitersMtx.RLock()
-	limiter, ok := d.ingestLimiters[userID]
-	d.ingestLimitersMtx.RUnlock()
-
-	if ok {
-		return limiter
-	}
-
-	limiter = rate.NewLimiter(rate.Limit(int64(d.overrides.IngestionRate(userID)*bytesInMB)), int(d.overrides.IngestionBurstSize(userID)*bytesInMB))
-
-	d.ingestLimitersMtx.Lock()
-	d.ingestLimiters[userID] = limiter
-	d.ingestLimitersMtx.Unlock()
-
-	return limiter
-}