Merge pull request #12216 from jingyih/experimental_flag_for_watch_notify_interval

*: add experimental flag for watch notify interval

CHANGELOG-3.5.md

@@ -136,6 +136,7 @@ Note that any `etcd_debugging_*` metrics are experimental and subject to change.
 - Log [expensive request info in UnaryInterceptor](https://github.com/etcd-io/etcd/pull/12086).
 - [Fix invalid Go type in etcdserverpb](https://github.com/etcd-io/etcd/pull/12000).
 - [Improve healthcheck by using v3 range request and its corresponding timeout](https://github.com/etcd-io/etcd/pull/12195).
+- Add [`etcd --experimental-watch-progress-notify-interval`](https://github.com/etcd-io/etcd/pull/12216) flag to make watch progress notify interval configurable.
 
 ### Package `runtime`
 

embed/config.go

@@ -283,6 +283,7 @@ type Config struct {
 	// ExperimentalEnableLeaseCheckpoint enables primary lessor to persist lease remainingTTL to prevent indefinite auto-renewal of long lived leases.
 	ExperimentalEnableLeaseCheckpoint       bool          `json:"experimental-enable-lease-checkpoint"`
 	ExperimentalCompactionBatchLimit        int           `json:"experimental-compaction-batch-limit"`
+	ExperimentalWatchProgressNotifyInterval time.Duration `json:"experimental-watch-progress-notify-interval"`
 
 	// ForceNewCluster starts a new cluster even if previously started; unsafe.
 	ForceNewCluster bool `json:"force-new-cluster"`

embed/etcd.go

@@ -200,6 +200,7 @@ func StartEtcd(inCfg *Config) (e *Etcd, err error) {
 		UnsafeNoFsync:               cfg.UnsafeNoFsync,
 		EnableLeaseCheckpoint:       cfg.ExperimentalEnableLeaseCheckpoint,
 		CompactionBatchLimit:        cfg.ExperimentalCompactionBatchLimit,
+		WatchProgressNotifyInterval: cfg.ExperimentalWatchProgressNotifyInterval,
 	}
 	print(e.cfg.logger, *cfg, srvcfg, memberInitialized)
 	if e.Server, err = etcdserver.NewServer(srvcfg); err != nil {

etcdmain/config.go

@@ -251,6 +251,7 @@ func newConfig() *config {
 	fs.StringVar(&cfg.ec.ExperimentalEnableV2V3, "experimental-enable-v2v3", cfg.ec.ExperimentalEnableV2V3, "v3 prefix for serving emulated v2 state.")
 	fs.BoolVar(&cfg.ec.ExperimentalEnableLeaseCheckpoint, "experimental-enable-lease-checkpoint", false, "Enable to persist lease remaining TTL to prevent indefinite auto-renewal of long lived leases.")
 	fs.IntVar(&cfg.ec.ExperimentalCompactionBatchLimit, "experimental-compaction-batch-limit", cfg.ec.ExperimentalCompactionBatchLimit, "Sets the maximum revisions deleted in each compaction batch.")
+	fs.DurationVar(&cfg.ec.ExperimentalWatchProgressNotifyInterval, "experimental-watch-progress-notify-interval", cfg.ec.ExperimentalWatchProgressNotifyInterval, "Duration of periodic watch progress notifications.")
 
 	// unsafe
 	fs.BoolVar(&cfg.ec.UnsafeNoFsync, "unsafe-no-fsync", false, "Disables fsync, unsafe, will cause data loss.")

etcdmain/help.go

@@ -210,6 +210,8 @@ Experimental feature:
     ExperimentalCompactionBatchLimit sets the maximum revisions deleted in each compaction batch.
   --experimental-peer-skip-client-san-verification 'false'
     Skip verification of SAN field in client certificate for peer connections.
+  --experimental-watch-progress-notify-interval '10m'
+    Duration of periodical watch progress notification.
 
 Unsafe feature:
   --force-new-cluster 'false'

etcdserver/api/v3rpc/watch.go

@@ -31,6 +31,8 @@ import (
 	"go.uber.org/zap"
 )
 
+const minWatchProgressInterval = 100 * time.Millisecond
+
 type watchServer struct {
 	lg *zap.Logger
 
@@ -61,6 +63,16 @@ func NewWatchServer(s *etcdserver.EtcdServer) pb.WatchServer {
 	if srv.lg == nil {
 		srv.lg = zap.NewNop()
 	}
+	if s.Cfg.WatchProgressNotifyInterval > 0 {
+		if s.Cfg.WatchProgressNotifyInterval < minWatchProgressInterval {
+			srv.lg.Warn(
+				"adjusting watch progress notify interval to minimum period",
+				zap.Duration("min-watch-progress-notify-interval", minWatchProgressInterval),
+			)
+			s.Cfg.WatchProgressNotifyInterval = minWatchProgressInterval
+		}
+		SetProgressReportInterval(s.Cfg.WatchProgressNotifyInterval)
+	}
 	return srv
 }
 

etcdserver/config.go

@@ -157,6 +157,8 @@ type ServerConfig struct {
 
 	EnableGRPCGateway bool
 
+	WatchProgressNotifyInterval time.Duration
+
 	// UnsafeNoFsync disables all uses of fsync.
 	// Setting this is unsafe and will cause data loss.
 	UnsafeNoFsync bool `json:"unsafe-no-fsync"`