Merge pull request #10797 from jingyih/lease_checkpoint_enabled_by_experimental_flag

*: enable lease checkpoint via experimental flag

CHANGELOG-3.4.md

@@ -53,6 +53,7 @@ See [code changes](https://github.com/etcd-io/etcd/compare/v3.3.0...v3.4.0) and
 - Improve [index compaction blocking](https://github.com/etcd-io/etcd/pull/9511) by using a copy on write clone to avoid holding the lock for the traversal of the entire index.
 - Update [JWT methods](https://github.com/etcd-io/etcd/pull/9883) to allow for use of any supported signature method/algorithm.
 - Add [Lease checkpointing](https://github.com/etcd-io/etcd/pull/9924) to persist remaining TTLs to the consensus log periodically so that long lived leases progress toward expiry in the presence of leader elections and server restarts.
+  - Enabled by experimental flag "--experimental-enable-lease-checkpoint".
 - Add [gRPC interceptor for debugging logs](https://github.com/etcd-io/etcd/pull/9990); enable `etcd --debug` flag to see per-request debug information.
 - Add [consistency check in snapshot status](https://github.com/etcd-io/etcd/pull/10109). If consistency check on snapshot file fails, `snapshot status` returns `"snapshot file integrity check failed..."` error.
 - Add [`Verify` function to perform corruption check on WAL contents](https://github.com/etcd-io/etcd/pull/10603).

embed/config.go

@@ -278,6 +278,8 @@ type Config struct {
 	ExperimentalEnableV2V3          string        `json:"experimental-enable-v2v3"`
 	// ExperimentalBackendFreelistType specifies the type of freelist that boltdb backend uses (array and map are supported types).
 	ExperimentalBackendFreelistType string `json:"experimental-backend-bbolt-freelist-type"`
+	// ExperimentalEnableLeaseCheckpoint enables primary lessor to persist lease remainingTTL to prevent indefinite auto-renewal of long lived leases.
+	ExperimentalEnableLeaseCheckpoint bool `json:"experimental-enable-lease-checkpoint"`
 
 	// ForceNewCluster starts a new cluster even if previously started; unsafe.
 	ForceNewCluster bool `json:"force-new-cluster"`

embed/etcd.go

@@ -204,6 +204,7 @@ func StartEtcd(inCfg *Config) (e *Etcd, err error) {
 		Debug:                      cfg.Debug,
 		ForceNewCluster:            cfg.ForceNewCluster,
 		EnableGRPCGateway:          cfg.EnableGRPCGateway,
+		EnableLeaseCheckpoint:      cfg.ExperimentalEnableLeaseCheckpoint,
 	}
 	print(e.cfg.logger, *cfg, srvcfg, memberInitialized)
 	if e.Server, err = etcdserver.NewServer(srvcfg); err != nil {

etcdmain/config.go

@@ -249,6 +249,7 @@ func newConfig() *config {
 	fs.DurationVar(&cfg.ec.ExperimentalCorruptCheckTime, "experimental-corrupt-check-time", cfg.ec.ExperimentalCorruptCheckTime, "Duration of time between cluster corruption check passes.")
 	fs.StringVar(&cfg.ec.ExperimentalEnableV2V3, "experimental-enable-v2v3", cfg.ec.ExperimentalEnableV2V3, "v3 prefix for serving emulated v2 state.")
 	fs.StringVar(&cfg.ec.ExperimentalBackendFreelistType, "experimental-backend-bbolt-freelist-type", cfg.ec.ExperimentalBackendFreelistType, "ExperimentalBackendFreelistType specifies the type of freelist that boltdb backend uses(array and map are supported types)")
+	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.")
 
 	// unsafe
 	fs.BoolVar(&cfg.ec.ForceNewCluster, "force-new-cluster", false, "Force to create a new one member cluster.")

etcdmain/help.go

@@ -202,6 +202,8 @@ Experimental feature:
     Serve v2 requests through the v3 backend under a given prefix.
   --experimental-backend-bbolt-freelist-type
     ExperimentalBackendFreelistType specifies the type of freelist that boltdb backend uses(array and map are supported types).
+  --experimental-enable-lease-checkpoint
+    ExperimentalEnableLeaseCheckpoint enables primary lessor to persist lease remainingTTL to prevent indefinite auto-renewal of long lived leases.
 
 Unsafe feature:
   --force-new-cluster 'false'

etcdserver/config.go

@@ -150,6 +150,8 @@ type ServerConfig struct {
 
 	ForceNewCluster bool
 
+	// EnableLeaseCheckpoint enables primary lessor to persist lease remainingTTL to prevent indefinite auto-renewal of long lived leases.
+	EnableLeaseCheckpoint bool
 	// LeaseCheckpointInterval time.Duration is the wait duration between lease checkpoints.
 	LeaseCheckpointInterval time.Duration
 

etcdserver/server.go

@@ -595,9 +595,12 @@ func NewServer(cfg ServerConfig) (srv *EtcdServer, err error) {
 		return nil, err
 	}
 
-	srv.lessor.SetCheckpointer(func(ctx context.Context, cp *pb.LeaseCheckpointRequest) {
-		srv.raftRequestOnce(ctx, pb.InternalRaftRequest{LeaseCheckpoint: cp})
-	})
+	if srv.Cfg.EnableLeaseCheckpoint {
+		// setting checkpointer enables lease checkpoint feature.
+		srv.lessor.SetCheckpointer(func(ctx context.Context, cp *pb.LeaseCheckpointRequest) {
+			srv.raftRequestOnce(ctx, pb.InternalRaftRequest{LeaseCheckpoint: cp})
+		})
+	}
 
 	// TODO: move transport initialization near the definition of remote
 	tr := &rafthttp.Transport{

integration/cluster.go

@@ -150,6 +150,7 @@ type ClusterConfig struct {
 	// UseIP is true to use only IP for gRPC requests.
 	UseIP bool
 
+	EnableLeaseCheckpoint   bool
 	LeaseCheckpointInterval time.Duration
 }
 
@@ -293,6 +294,7 @@ func (c *cluster) mustNewMember(t testing.TB) *member {
 			clientMaxCallSendMsgSize: c.cfg.ClientMaxCallSendMsgSize,
 			clientMaxCallRecvMsgSize: c.cfg.ClientMaxCallRecvMsgSize,
 			useIP:                    c.cfg.UseIP,
+			enableLeaseCheckpoint:    c.cfg.EnableLeaseCheckpoint,
 			leaseCheckpointInterval:  c.cfg.LeaseCheckpointInterval,
 		})
 	m.DiscoveryURL = c.cfg.DiscoveryURL
@@ -581,6 +583,7 @@ type memberConfig struct {
 	clientMaxCallSendMsgSize int
 	clientMaxCallRecvMsgSize int
 	useIP                    bool
+	enableLeaseCheckpoint    bool
 	leaseCheckpointInterval  time.Duration
 }
 
@@ -672,6 +675,7 @@ func mustNewMember(t testing.TB, mcfg memberConfig) *member {
 	m.clientMaxCallSendMsgSize = mcfg.clientMaxCallSendMsgSize
 	m.clientMaxCallRecvMsgSize = mcfg.clientMaxCallRecvMsgSize
 	m.useIP = mcfg.useIP
+	m.EnableLeaseCheckpoint = mcfg.enableLeaseCheckpoint
 	m.LeaseCheckpointInterval = mcfg.leaseCheckpointInterval
 
 	m.InitialCorruptCheck = true

integration/v3_lease_test.go

@@ -230,7 +230,11 @@ func TestV3LeaseCheckpoint(t *testing.T) {
 	var ttl int64 = 300
 	leaseInterval := 2 * time.Second
 	defer testutil.AfterTest(t)
-	clus := NewClusterV3(t, &ClusterConfig{Size: 3, LeaseCheckpointInterval: leaseInterval})
+	clus := NewClusterV3(t, &ClusterConfig{
+		Size:                    3,
+		EnableLeaseCheckpoint:   true,
+		LeaseCheckpointInterval: leaseInterval,
+	})
 	defer clus.Terminate(t)
 
 	// create lease