Make remote flush deadline a command line param.

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

cmd/prometheus/main.go

@@ -92,6 +92,7 @@ func main() {
 		webTimeout          model.Duration
 		queryTimeout        model.Duration
 		queryConcurrency    int
+		RemoteFlushDeadline model.Duration
 
 		prometheusURL string
 
@@ -160,6 +161,9 @@ func main() {
 	a.Flag("storage.tsdb.no-lockfile", "Do not create lockfile in data directory.").
 		Default("false").BoolVar(&cfg.tsdb.NoLockfile)
 
+	a.Flag("storage.remote.flush-deadline", "How long to wait flushing sample on shutdown or config reload.").
+		Default(1 * time.Minute).PlaceHolder("<duration>").SetValue(&cfg.RemoteFlushDeadline)
+
 	a.Flag("alertmanager.notification-queue-capacity", "The capacity of the queue for pending Alertmanager notifications.").
 		Default("10000").IntVar(&cfg.notifier.QueueCapacity)
 
@@ -222,7 +226,7 @@ func main() {
 
 	var (
 		localStorage  = &tsdb.ReadyStorage{}
-		remoteStorage = remote.NewStorage(log.With(logger, "component", "remote"), localStorage.StartTime)
+		remoteStorage = remote.NewStorage(log.With(logger, "component", "remote"), localStorage.StartTime, cfg.RemoteFlushDeadline)
 		fanoutStorage = storage.NewFanout(logger, localStorage, remoteStorage)
 	)
 

config/config.go

@@ -122,8 +122,6 @@ var (
 		MaxRetries: 10,
 		MinBackoff: 30 * time.Millisecond,
 		MaxBackoff: 100 * time.Millisecond,
-
-		FlushDeadline: 1 * time.Minute,
 	}
 
 	// DefaultRemoteReadConfig is the default remote read configuration.
@@ -648,10 +646,6 @@ type QueueConfig struct {
 	// On recoverable errors, backoff exponentially.
 	MinBackoff time.Duration `yaml:"min_backoff,omitempty"`
 	MaxBackoff time.Duration `yaml:"max_backoff,omitempty"`
-
-	// On shutdown or config reload allow the following duration for flushing
-	// pending samples, otherwise continue without waiting.
-	FlushDeadline time.Duration `yaml:"flush_deadline"`
 }
 
 // RemoteReadConfig is the configuration for reading from remote storage.

storage/remote/queue_manager.go

@@ -140,6 +140,7 @@ type StorageClient interface {
 type QueueManager struct {
 	logger log.Logger
 
+	flushDeadline  time.Duration
 	cfg            config.QueueConfig
 	externalLabels model.LabelSet
 	relabelConfigs []*config.RelabelConfig
@@ -159,12 +160,13 @@ type QueueManager struct {
 }
 
 // NewQueueManager builds a new QueueManager.
-func NewQueueManager(logger log.Logger, cfg config.QueueConfig, externalLabels model.LabelSet, relabelConfigs []*config.RelabelConfig, client StorageClient) *QueueManager {
+func NewQueueManager(logger log.Logger, cfg config.QueueConfig, externalLabels model.LabelSet, relabelConfigs []*config.RelabelConfig, client StorageClient, flushDeadline time.Duration) *QueueManager {
 	if logger == nil {
 		logger = log.NewNopLogger()
 	}
 	t := &QueueManager{
 		logger:         logger,
+		flushDeadline:  flushDeadline,
 		cfg:            cfg,
 		externalLabels: externalLabels,
 		relabelConfigs: relabelConfigs,
@@ -256,7 +258,7 @@ func (t *QueueManager) Stop() {
 
 	t.shardsMtx.Lock()
 	defer t.shardsMtx.Unlock()
-	t.shards.stop(t.cfg.FlushDeadline)
+	t.shards.stop(t.flushDeadline)
 
 	level.Info(t.logger).Log("msg", "Remote storage stopped.")
 }
@@ -361,7 +363,7 @@ func (t *QueueManager) reshard(n int) {
 	t.shards = newShards
 	t.shardsMtx.Unlock()
 
-	oldShards.stop(t.cfg.FlushDeadline)
+	oldShards.stop(t.flushDeadline)
 
 	// We start the newShards after we have stopped (the therefore completely
 	// flushed) the oldShards, to guarantee we only every deliver samples in

storage/remote/queue_manager_test.go

@@ -26,6 +26,8 @@ import (
 	"github.com/prometheus/prometheus/prompb"
 )
 
+const defaultFlushDeadline = 1 * time.Minute
+
 type TestStorageClient struct {
 	receivedSamples map[string][]*prompb.Sample
 	expectedSamples map[string][]*prompb.Sample
@@ -109,7 +111,7 @@ func TestSampleDelivery(t *testing.T) {
 
 	cfg := config.DefaultQueueConfig
 	cfg.MaxShards = 1
-	m := NewQueueManager(nil, cfg, nil, nil, c)
+	m := NewQueueManager(nil, cfg, nil, nil, c, defaultFlushDeadline)
 
 	// These should be received by the client.
 	for _, s := range samples[:len(samples)/2] {
@@ -145,7 +147,7 @@ func TestSampleDeliveryTimeout(t *testing.T) {
 	cfg := config.DefaultQueueConfig
 	cfg.MaxShards = 1
 	cfg.BatchSendDeadline = 100 * time.Millisecond
-	m := NewQueueManager(nil, cfg, nil, nil, c)
+	m := NewQueueManager(nil, cfg, nil, nil, c, defaultFlushDeadline)
 	m.Start()
 	defer m.Stop()
 
@@ -181,7 +183,7 @@ func TestSampleDeliveryOrder(t *testing.T) {
 
 	c := NewTestStorageClient()
 	c.expectSamples(samples)
-	m := NewQueueManager(nil, config.DefaultQueueConfig, nil, nil, c)
+	m := NewQueueManager(nil, config.DefaultQueueConfig, nil, nil, c, defaultFlushDeadline)
 
 	// These should be received by the client.
 	for _, s := range samples {
@@ -259,7 +261,7 @@ func TestSpawnNotMoreThanMaxConcurrentSendsGoroutines(t *testing.T) {
 	cfg := config.DefaultQueueConfig
 	cfg.MaxShards = 1
 	cfg.Capacity = n
-	m := NewQueueManager(nil, cfg, nil, nil, c)
+	m := NewQueueManager(nil, cfg, nil, nil, c, defaultFlushDeadline)
 
 	m.Start()
 

storage/remote/storage.go

@@ -16,6 +16,7 @@ package remote
 import (
 	"context"
 	"sync"
+	"time"
 
 	"github.com/go-kit/kit/log"
 	"github.com/prometheus/common/model"
@@ -39,14 +40,19 @@ type Storage struct {
 	// For reads
 	queryables             []storage.Queryable
 	localStartTimeCallback startTimeCallback
+	flushDeadline          time.Duration
 }
 
 // NewStorage returns a remote.Storage.
-func NewStorage(l log.Logger, stCallback startTimeCallback) *Storage {
+func NewStorage(l log.Logger, stCallback startTimeCallback, flushDeadline time.Duration) *Storage {
 	if l == nil {
 		l = log.NewNopLogger()
 	}
-	return &Storage{logger: l, localStartTimeCallback: stCallback}
+	return &Storage{
+		logger:                 l,
+		localStartTimeCallback: stCallback,
+		flushDeadline:          flushDeadline,
+	}
 }
 
 // ApplyConfig updates the state as the new config requires.
@@ -74,6 +80,7 @@ func (s *Storage) ApplyConfig(conf *config.Config) error {
 			conf.GlobalConfig.ExternalLabels,
 			rwConf.WriteRelabelConfigs,
 			c,
+			s.flushDeadline,
 		))
 	}