Add "wait ring stability" to store-gateway and fix cold start issue

CHANGELOG.md

@@ -40,6 +40,7 @@
   * `memberlist_client_messages_to_broadcast_dropped_total`
 * [ENHANCEMENT] Alertmanager: Added `-alertmanager.max-dispatcher-aggregation-groups` option to control max number of active dispatcher groups in Alertmanager (per tenant, also overrideable). When the limit is reached, Dispatcher produces log message and increases `cortex_alertmanager_dispatcher_aggregation_group_limit_reached_total` metric. #4254
 * [ENHANCEMENT] Alertmanager: Added `-alertmanager.max-alerts-count` and `-alertmanager.max-alerts-size-bytes` to control max number of alerts and total size of alerts that a single user can have in Alertmanager's memory. Adding more alerts will fail with a log message and incrementing `cortex_alertmanager_alerts_insert_limited_total` metric (per-user). These limits can be overrided by using per-tenant overrides. Current values are tracked in `cortex_alertmanager_alerts_limiter_current_alerts` and `cortex_alertmanager_alerts_limiter_current_alerts_size_bytes` metrics. #4253
+* [ENHANCEMENT] Store-gateway: added `-store-gateway.sharding-ring.wait-stability-min-duration` and `-store-gateway.sharding-ring.wait-stability-max-duration` support to store-gateway, to wait for ring stability at startup. #4271
 * [BUGFIX] Purger: fix `Invalid null value in condition for column range` caused by `nil` value in range for WriteBatch query. #4128
 * [BUGFIX] Ingester: fixed infrequent panic caused by a race condition between TSDB mmap-ed head chunks truncation and queries. #4176
 * [BUGFIX] Alertmanager: fix Alertmanager status page if clustering via gossip is disabled or sharding is enabled. #4184
@@ -49,6 +50,7 @@
 * [BUGFIX] Ingester: fix issue where runtime limits erroneously override default limits. #4246
 * [BUGFIX] Ruler: fix startup in single-binary mode when the new `ruler_storage` is used. #4252
 * [BUGFIX] Querier: fix queries failing with "at least 1 healthy replica required, could only find 0" error right after scaling up store-gateways until they're ACTIVE in the ring. #4263
+* [BUGFIX] Store-gateway: when blocks sharding is enabled, do not load all blocks in each store-gateway in case of a cold startup, but load only blocks owned by the store-gateway replica. #4271
 * [BUGFIX] Memberlist: fix to setting the default configuration value for `-memberlist.retransmit-factor` when not provided. This should improve propagation delay of the ring state (including, but not limited to, tombstones). Note that if the configuration is already explicitly given, this fix has no effect. #4269
 
 ## Blocksconvert

docs/blocks-storage/compactor.md

@@ -223,7 +223,7 @@ compactor:
     [wait_stability_min_duration: <duration> | default = 1m]
 
     # Maximum time to wait for ring stability at startup. If the compactor ring
-    # keep changing after this period of time, the compactor will start anyway.
+    # keeps changing after this period of time, the compactor will start anyway.
     # CLI flag: -compactor.ring.wait-stability-max-duration
     [wait_stability_max_duration: <duration> | default = 5m]
 

docs/blocks-storage/store-gateway.md

@@ -81,6 +81,14 @@ The store-gateway replication optionally supports [zone-awareness](../guides/zon
 2. Enable blocks zone-aware replication via the `-store-gateway.sharding-ring.zone-awareness-enabled` CLI flag (or its respective YAML config option). Please be aware this configuration option should be set to store-gateways, queriers and rulers.
 3. Rollout store-gateways, queriers and rulers to apply the new configuration
 
+### Waiting for stable ring at startup
+
+In the event of a cluster cold start or scale up of 2+ store-gateway instances at the same time we may end up in a situation where each new store-gateway instance starts at a slightly different time and thus each one runs the initial blocks sync based on a different state of the ring. For example, in case of a cold start, the first store-gateway joining the ring may load all blocks since the sharding logic runs based on the current state of the ring, which is 1 single store-gateway.
+
+To reduce the likelihood this could happen, the store-gateway waits for a stable ring at startup. A ring is considered stable if no instance is added/removed to the ring for at least `-store-gateway.sharding-ring.wait-stability-min-duration`. If the ring keep getting changed after `-store-gateway.sharding-ring.wait-stability-max-duration`, the store-gateway will stop waiting for a stable ring and will proceed starting up normally.
+
+To disable this waiting logic, you can start the store-gateway with `-store-gateway.sharding-ring.wait-stability-min-duration=0`.
+
 ## Blocks index-header
 
 The [index-header](./binary-index-header.md) is a subset of the block index which the store-gateway downloads from the object storage and keeps on the local disk in order to speed up queries.
@@ -250,6 +258,16 @@ store_gateway:
     # CLI flag: -store-gateway.sharding-ring.zone-awareness-enabled
     [zone_awareness_enabled: <boolean> | default = false]
 
+    # Minimum time to wait for ring stability at startup. 0 to disable.
+    # CLI flag: -store-gateway.sharding-ring.wait-stability-min-duration
+    [wait_stability_min_duration: <duration> | default = 1m]
+
+    # Maximum time to wait for ring stability at startup. If the store-gateway
+    # ring keeps changing after this period of time, the store-gateway will
+    # start anyway.
+    # CLI flag: -store-gateway.sharding-ring.wait-stability-max-duration
+    [wait_stability_max_duration: <duration> | default = 5m]
+
     # Name of network interface to read address from.
     # CLI flag: -store-gateway.sharding-ring.instance-interface-names
     [instance_interface_names: <list of string> | default = [eth0 en0]]

docs/blocks-storage/store-gateway.template

@@ -81,6 +81,14 @@ The store-gateway replication optionally supports [zone-awareness](../guides/zon
 2. Enable blocks zone-aware replication via the `-store-gateway.sharding-ring.zone-awareness-enabled` CLI flag (or its respective YAML config option). Please be aware this configuration option should be set to store-gateways, queriers and rulers.
 3. Rollout store-gateways, queriers and rulers to apply the new configuration
 
+### Waiting for stable ring at startup
+
+In the event of a cluster cold start or scale up of 2+ store-gateway instances at the same time we may end up in a situation where each new store-gateway instance starts at a slightly different time and thus each one runs the initial blocks sync based on a different state of the ring. For example, in case of a cold start, the first store-gateway joining the ring may load all blocks since the sharding logic runs based on the current state of the ring, which is 1 single store-gateway.
+
+To reduce the likelihood this could happen, the store-gateway waits for a stable ring at startup. A ring is considered stable if no instance is added/removed to the ring for at least `-store-gateway.sharding-ring.wait-stability-min-duration`. If the ring keep getting changed after `-store-gateway.sharding-ring.wait-stability-max-duration`, the store-gateway will stop waiting for a stable ring and will proceed starting up normally.
+
+To disable this waiting logic, you can start the store-gateway with `-store-gateway.sharding-ring.wait-stability-min-duration=0`.
+
 ## Blocks index-header
 
 The [index-header](./binary-index-header.md) is a subset of the block index which the store-gateway downloads from the object storage and keeps on the local disk in order to speed up queries.

docs/configuration/config-file-reference.md

@@ -5175,7 +5175,7 @@ sharding_ring:
   [wait_stability_min_duration: <duration> | default = 1m]
 
   # Maximum time to wait for ring stability at startup. If the compactor ring
-  # keep changing after this period of time, the compactor will start anyway.
+  # keeps changing after this period of time, the compactor will start anyway.
   # CLI flag: -compactor.ring.wait-stability-max-duration
   [wait_stability_max_duration: <duration> | default = 5m]
 
@@ -5260,6 +5260,16 @@ sharding_ring:
   # CLI flag: -store-gateway.sharding-ring.zone-awareness-enabled
   [zone_awareness_enabled: <boolean> | default = false]
 
+  # Minimum time to wait for ring stability at startup. 0 to disable.
+  # CLI flag: -store-gateway.sharding-ring.wait-stability-min-duration
+  [wait_stability_min_duration: <duration> | default = 1m]
+
+  # Maximum time to wait for ring stability at startup. If the store-gateway
+  # ring keeps changing after this period of time, the store-gateway will start
+  # anyway.
+  # CLI flag: -store-gateway.sharding-ring.wait-stability-max-duration
+  [wait_stability_max_duration: <duration> | default = 5m]
+
   # Name of network interface to read address from.
   # CLI flag: -store-gateway.sharding-ring.instance-interface-names
   [instance_interface_names: <list of string> | default = [eth0 en0]]

integration/backward_compatibility_test.go

@@ -26,10 +26,10 @@ var (
 		"quay.io/cortexproject/cortex:v1.3.0": preCortex14Flags,
 		"quay.io/cortexproject/cortex:v1.4.0": preCortex16Flags,
 		"quay.io/cortexproject/cortex:v1.5.0": preCortex16Flags,
-		"quay.io/cortexproject/cortex:v1.6.0": nil,
-		"quay.io/cortexproject/cortex:v1.7.0": nil,
-		"quay.io/cortexproject/cortex:v1.8.0": nil,
-		"quay.io/cortexproject/cortex:v1.9.0": nil,
+		"quay.io/cortexproject/cortex:v1.6.0": preCortex110Flags,
+		"quay.io/cortexproject/cortex:v1.7.0": preCortex110Flags,
+		"quay.io/cortexproject/cortex:v1.8.0": preCortex110Flags,
+		"quay.io/cortexproject/cortex:v1.9.0": preCortex110Flags,
 	}
 )
 
@@ -42,13 +42,27 @@ func preCortex14Flags(flags map[string]string) map[string]string {
 		"-store-gateway.sharding-ring.replication-factor": "",
 		// Query-scheduler has been introduced in 1.6.0
 		"-frontend.scheduler-dns-lookup-period": "",
+		// Store-gateway "wait ring stability" has been introduced in 1.10.0
+		"-store-gateway.sharding-ring.wait-stability-min-duration": "",
+		"-store-gateway.sharding-ring.wait-stability-max-duration": "",
 	})
 }
 
 func preCortex16Flags(flags map[string]string) map[string]string {
 	return e2e.MergeFlagsWithoutRemovingEmpty(flags, map[string]string{
 		// Query-scheduler has been introduced in 1.6.0
 		"-frontend.scheduler-dns-lookup-period": "",
+		// Store-gateway "wait ring stability" has been introduced in 1.10.0
+		"-store-gateway.sharding-ring.wait-stability-min-duration": "",
+		"-store-gateway.sharding-ring.wait-stability-max-duration": "",
+	})
+}
+
+func preCortex110Flags(flags map[string]string) map[string]string {
+	return e2e.MergeFlagsWithoutRemovingEmpty(flags, map[string]string{
+		// Store-gateway "wait ring stability" has been introduced in 1.10.0
+		"-store-gateway.sharding-ring.wait-stability-min-duration": "",
+		"-store-gateway.sharding-ring.wait-stability-max-duration": "",
 	})
 }
 

integration/e2ecortex/services.go

@@ -121,6 +121,9 @@ func NewStoreGatewayWithConfigFile(name, consulAddress, configFile string, flags
 			"-store-gateway.sharding-ring.store":              "consul",
 			"-store-gateway.sharding-ring.consul.hostname":    consulAddress,
 			"-store-gateway.sharding-ring.replication-factor": "1",
+			// Startup quickly.
+			"-store-gateway.sharding-ring.wait-stability-min-duration": "0",
+			"-store-gateway.sharding-ring.wait-stability-max-duration": "0",
 		}, flags))...),
 		e2e.NewHTTPReadinessProbe(httpPort, "/ready", 200, 299),
 		httpPort,
@@ -301,6 +304,9 @@ func NewSingleBinary(name string, flags map[string]string, image string, otherPo
 			"-ingester.concurrent-flushes":   "10",
 			"-ingester.max-transfer-retries": "10",
 			"-ingester.num-tokens":           "512",
+			// Startup quickly.
+			"-store-gateway.sharding-ring.wait-stability-min-duration": "0",
+			"-store-gateway.sharding-ring.wait-stability-max-duration": "0",
 		}, flags))...),
 		e2e.NewHTTPReadinessProbe(httpPort, "/ready", 200, 299),
 		httpPort,

integration/getting_started_with_gossiped_ring_test.go

@@ -33,15 +33,17 @@ func TestGettingStartedWithGossipedRing(t *testing.T) {
 	// We don't care for storage part too much here. Both Cortex instances will write new blocks to /tmp, but that's fine.
 	flags := map[string]string{
 		// decrease timeouts to make test faster. should still be fine with two instances only
-		"-ingester.join-after":                       "0s", // join quickly
-		"-ingester.observe-period":                   "5s", // to avoid conflicts in tokens
-		"-blocks-storage.bucket-store.sync-interval": "1s", // sync continuously
-		"-blocks-storage.backend":                    "s3",
-		"-blocks-storage.s3.bucket-name":             bucketName,
-		"-blocks-storage.s3.access-key-id":           e2edb.MinioAccessKey,
-		"-blocks-storage.s3.secret-access-key":       e2edb.MinioSecretKey,
-		"-blocks-storage.s3.endpoint":                fmt.Sprintf("%s-minio-9000:9000", networkName),
-		"-blocks-storage.s3.insecure":                "true",
+		"-ingester.join-after":                                     "0s", // join quickly
+		"-ingester.observe-period":                                 "5s", // to avoid conflicts in tokens
+		"-blocks-storage.bucket-store.sync-interval":               "1s", // sync continuously
+		"-blocks-storage.backend":                                  "s3",
+		"-blocks-storage.s3.bucket-name":                           bucketName,
+		"-blocks-storage.s3.access-key-id":                         e2edb.MinioAccessKey,
+		"-blocks-storage.s3.secret-access-key":                     e2edb.MinioSecretKey,
+		"-blocks-storage.s3.endpoint":                              fmt.Sprintf("%s-minio-9000:9000", networkName),
+		"-blocks-storage.s3.insecure":                              "true",
+		"-store-gateway.sharding-ring.wait-stability-min-duration": "0", // start quickly
+		"-store-gateway.sharding-ring.wait-stability-max-duration": "0", // start quickly
 	}
 
 	// This cortex will fail to join the cluster configured in yaml file. That's fine.

pkg/compactor/compactor.go

@@ -400,7 +400,7 @@ func (c *Compactor) starting(ctx context.Context) error {
 
 		// In the event of a cluster cold start or scale up of 2+ compactor instances at the same
 		// time, we may end up in a situation where each new compactor instance starts at a slightly
-		// different time and thus each one starts with on a different state of the ring. It's better
+		// different time and thus each one starts with a different state of the ring. It's better
 		// to just wait the ring stability for a short time.
 		if c.compactorCfg.ShardingRing.WaitStabilityMinDuration > 0 {
 			minWaiting := c.compactorCfg.ShardingRing.WaitStabilityMinDuration

pkg/compactor/compactor_ring.go

@@ -51,7 +51,7 @@ func (cfg *RingConfig) RegisterFlags(f *flag.FlagSet) {
 
 	// Wait stability flags.
 	f.DurationVar(&cfg.WaitStabilityMinDuration, "compactor.ring.wait-stability-min-duration", time.Minute, "Minimum time to wait for ring stability at startup. 0 to disable.")
-	f.DurationVar(&cfg.WaitStabilityMaxDuration, "compactor.ring.wait-stability-max-duration", 5*time.Minute, "Maximum time to wait for ring stability at startup. If the compactor ring keep changing after this period of time, the compactor will start anyway.")
+	f.DurationVar(&cfg.WaitStabilityMaxDuration, "compactor.ring.wait-stability-max-duration", 5*time.Minute, "Maximum time to wait for ring stability at startup. If the compactor ring keeps changing after this period of time, the compactor will start anyway.")
 
 	// Instance flags
 	cfg.InstanceInterfaceNames = []string{"eth0", "en0"}

pkg/storegateway/bucket_store_inmemory_server.go

@@ -3,8 +3,10 @@ package storegateway
 import (
 	"context"
 
+	"github.com/gogo/protobuf/types"
 	"github.com/pkg/errors"
 	"github.com/prometheus/prometheus/storage"
+	"github.com/thanos-io/thanos/pkg/store/hintspb"
 	"github.com/thanos-io/thanos/pkg/store/storepb"
 )
 
@@ -19,6 +21,7 @@ type bucketStoreSeriesServer struct {
 
 	SeriesSet []*storepb.Series
 	Warnings  storage.Warnings
+	Hints     hintspb.SeriesResponseHints
 }
 
 func newBucketStoreSeriesServer(ctx context.Context) *bucketStoreSeriesServer {
@@ -30,6 +33,13 @@ func (s *bucketStoreSeriesServer) Send(r *storepb.SeriesResponse) error {
 		s.Warnings = append(s.Warnings, errors.New(r.GetWarning()))
 	}
 
+	if rawHints := r.GetHints(); rawHints != nil {
+		// We expect only 1 hints entry so we just keep 1.
+		if err := types.UnmarshalAny(rawHints, &s.Hints); err != nil {
+			return errors.Wrap(err, "failed to unmarshal series hints")
+		}
+	}
+
 	if recvSeries := r.GetSeries(); recvSeries != nil {
 		// Thanos uses a pool for the chunks and may use other pools in the future.
 		// Given we need to retain the reference after the pooled slices are recycled,

pkg/storegateway/bucket_stores_test.go

@@ -589,7 +589,7 @@ func (u *userShardingStrategy) FilterUsers(ctx context.Context, userIDs []string
 	return u.users
 }
 
-func (u *userShardingStrategy) FilterBlocks(ctx context.Context, userID string, metas map[ulid.ULID]*thanos_metadata.Meta, synced *extprom.TxGaugeVec) error {
+func (u *userShardingStrategy) FilterBlocks(ctx context.Context, userID string, metas map[ulid.ULID]*thanos_metadata.Meta, loaded map[ulid.ULID]struct{}, synced *extprom.TxGaugeVec) error {
 	if util.StringsContain(u.users, userID) {
 		return nil
 	}

pkg/storegateway/gateway.go

@@ -231,6 +231,22 @@ func (g *StoreGateway) starting(ctx context.Context) (err error) {
 			return err
 		}
 		level.Info(g.logger).Log("msg", "store-gateway is JOINING in the ring")
+
+		// In the event of a cluster cold start or scale up of 2+ store-gateway instances at the same
+		// time, we may end up in a situation where each new store-gateway instance starts at a slightly
+		// different time and thus each one starts with a different state of the ring. It's better
+		// to just wait the ring stability for a short time.
+		if g.gatewayCfg.ShardingRing.WaitStabilityMinDuration > 0 {
+			minWaiting := g.gatewayCfg.ShardingRing.WaitStabilityMinDuration
+			maxWaiting := g.gatewayCfg.ShardingRing.WaitStabilityMaxDuration
+
+			level.Info(g.logger).Log("msg", "waiting until store-gateway ring topology is stable", "min_waiting", minWaiting.String(), "max_waiting", maxWaiting.String())
+			if err := ring.WaitRingStability(ctx, g.ring, BlocksOwnerSync, minWaiting, maxWaiting); err != nil {
+				level.Warn(g.logger).Log("msg", "store-gateway is ring topology is not stable after the max waiting time, proceeding anyway")
+			} else {
+				level.Info(g.logger).Log("msg", "store-gateway is ring topology is stable")
+			}
+		}
 	}
 
 	// At this point, if sharding is enabled, the instance is registered with some tokens
@@ -271,7 +287,7 @@ func (g *StoreGateway) running(ctx context.Context) error {
 	defer syncTicker.Stop()
 
 	if g.gatewayCfg.ShardingEnabled {
-		ringLastState, _ = g.ring.GetAllHealthy(BlocksSync) // nolint:errcheck
+		ringLastState, _ = g.ring.GetAllHealthy(BlocksOwnerSync) // nolint:errcheck
 		ringTicker := time.NewTicker(util.DurationWithJitter(g.gatewayCfg.ShardingRing.RingCheckPeriod, 0.2))
 		defer ringTicker.Stop()
 		ringTickerChan = ringTicker.C
@@ -284,7 +300,7 @@ func (g *StoreGateway) running(ctx context.Context) error {
 		case <-ringTickerChan:
 			// We ignore the error because in case of error it will return an empty
 			// replication set which we use to compare with the previous state.
-			currRingState, _ := g.ring.GetAllHealthy(BlocksSync) // nolint:errcheck
+			currRingState, _ := g.ring.GetAllHealthy(BlocksOwnerSync) // nolint:errcheck
 
 			if ring.HasReplicationSetChanged(ringLastState, currRingState) {
 				ringLastState = currRingState