Merge pull request etcd-io#12795 from wpedrak/resend-read-index-on-fi…

…rst-commit-in-term

etcdserver: resend ReadIndex request on empty apply request

server/etcdserver/server.go

@@ -273,10 +273,6 @@ type EtcdServer struct {
 	peerRt   http.RoundTripper
 	reqIDGen *idutil.Generator
 
-	// forceVersionC is used to force the version monitor loop
-	// to detect the cluster version immediately.
-	forceVersionC chan struct{}
-
 	// wgMu blocks concurrent waitgroup mutation while server stopping
 	wgMu sync.RWMutex
 	// wg is used to wait for the goroutines that depends on the server state
@@ -291,6 +287,9 @@ type EtcdServer struct {
 	leadTimeMu      sync.RWMutex
 	leadElectedTime time.Time
 
+	firstCommitInTermMu sync.RWMutex
+	firstCommitInTermC  chan struct{}
+
 	*AccessController
 }
 
@@ -517,17 +516,17 @@ func NewServer(cfg config.ServerConfig) (srv *EtcdServer, err error) {
 				storage:     NewStorage(w, ss),
 			},
 		),
-		id:               id,
-		attributes:       membership.Attributes{Name: cfg.Name, ClientURLs: cfg.ClientURLs.StringSlice()},
-		cluster:          cl,
-		stats:            sstats,
-		lstats:           lstats,
-		SyncTicker:       time.NewTicker(500 * time.Millisecond),
-		peerRt:           prt,
-		reqIDGen:         idutil.NewGenerator(uint16(id), time.Now()),
-		forceVersionC:    make(chan struct{}),
-		AccessController: &AccessController{CORS: cfg.CORS, HostWhitelist: cfg.HostWhitelist},
-		consistIndex:     cindex.NewConsistentIndex(be.BatchTx()),
+		id:                 id,
+		attributes:         membership.Attributes{Name: cfg.Name, ClientURLs: cfg.ClientURLs.StringSlice()},
+		cluster:            cl,
+		stats:              sstats,
+		lstats:             lstats,
+		SyncTicker:         time.NewTicker(500 * time.Millisecond),
+		peerRt:             prt,
+		reqIDGen:           idutil.NewGenerator(uint16(id), time.Now()),
+		AccessController:   &AccessController{CORS: cfg.CORS, HostWhitelist: cfg.HostWhitelist},
+		consistIndex:       cindex.NewConsistentIndex(be.BatchTx()),
+		firstCommitInTermC: make(chan struct{}),
 	}
 	serverID.With(prometheus.Labels{"server_id": id.String()}).Set(1)
 
@@ -1770,6 +1769,16 @@ func (s *EtcdServer) LeaderChangedNotify() <-chan struct{} {
 	return s.leaderChanged
 }
 
+// FirstCommitInTermNotify returns channel that will be unlocked on first
+// entry committed in new term, which is necessary for new leader to answer
+// read-only requests (leader is not able to respond any read-only requests
+// as long as linearizable semantic is required)
+func (s *EtcdServer) FirstCommitInTermNotify() <-chan struct{} {
+	s.firstCommitInTermMu.RLock()
+	defer s.firstCommitInTermMu.RUnlock()
+	return s.firstCommitInTermC
+}
+
 // RaftStatusGetter represents etcd server and Raft progress.
 type RaftStatusGetter interface {
 	ID() types.ID
@@ -2068,10 +2077,8 @@ func (s *EtcdServer) applyEntryNormal(e *raftpb.Entry) {
 	// raft state machine may generate noop entry when leader confirmation.
 	// skip it in advance to avoid some potential bug in the future
 	if len(e.Data) == 0 {
-		select {
-		case s.forceVersionC <- struct{}{}:
-		default:
-		}
+		s.notifyAboutFirstCommitInTerm()
+
 		// promote lessor when the local member is leader and finished
 		// applying all entries from the last term.
 		if s.isLeader() {
@@ -2140,6 +2147,15 @@ func (s *EtcdServer) applyEntryNormal(e *raftpb.Entry) {
 	})
 }
 
+func (s *EtcdServer) notifyAboutFirstCommitInTerm() {
+	newNotifier := make(chan struct{})
+	s.firstCommitInTermMu.Lock()
+	notifierToClose := s.firstCommitInTermC
+	s.firstCommitInTermC = newNotifier
+	s.firstCommitInTermMu.Unlock()
+	close(notifierToClose)
+}
+
 // applyConfChange applies a ConfChange to the server. It is only
 // invoked with a ConfChange that has already passed through Raft
 func (s *EtcdServer) applyConfChange(cc raftpb.ConfChange, confState *raftpb.ConfState) (bool, error) {
@@ -2319,7 +2335,7 @@ func (s *EtcdServer) ClusterVersion() *semver.Version {
 func (s *EtcdServer) monitorVersions() {
 	for {
 		select {
-		case <-s.forceVersionC:
+		case <-s.FirstCommitInTermNotify():
 		case <-time.After(monitorVersionInterval):
 		case <-s.stopping:
 			return

server/etcdserver/v3_server.go

@@ -775,6 +775,8 @@ func (s *EtcdServer) requestCurrentIndex(leaderChangedNotifier <-chan struct{},
 	retryTimer := time.NewTimer(readIndexRetryTime)
 	defer retryTimer.Stop()
 
+	firstCommitInTermNotifier := s.FirstCommitInTermNotify()
+
 	for {
 		select {
 		case rs := <-s.r.readStateC:
@@ -800,6 +802,15 @@ func (s *EtcdServer) requestCurrentIndex(leaderChangedNotifier <-chan struct{},
 			readIndexFailed.Inc()
 			// return a retryable error.
 			return 0, ErrLeaderChanged
+		case <-firstCommitInTermNotifier:
+			firstCommitInTermNotifier = s.FirstCommitInTermNotify()
+			lg.Info("first commit in current term: resending ReadIndex request")
+			err := s.sendReadIndex(requestId)
+			if err != nil {
+				return 0, err
+			}
+			retryTimer.Reset(readIndexRetryTime)
+			continue
 		case <-retryTimer.C:
 			lg.Warn(
 				"waiting for ReadIndex response took too long, retrying",

tests/go.mod

@@ -37,6 +37,7 @@ require (
 	go.etcd.io/etcd/server/v3 v3.5.0-alpha.0
 	go.uber.org/zap v1.16.0
 	golang.org/x/crypto v0.0.0-20201002170205-7f63de1d35b0
+	golang.org/x/sync v0.0.0-20201020160332-67f06af15bc9
 	golang.org/x/time v0.0.0-20210220033141-f8bda1e9f3ba
 	google.golang.org/grpc v1.36.1
 	gopkg.in/yaml.v2 v2.3.0

tests/go.sum

@@ -367,6 +367,7 @@ golang.org/x/sync v0.0.0-20181221193216-37e7f081c4d4/go.mod h1:RxMgew5VJxzue5/jJ
 golang.org/x/sync v0.0.0-20190227155943-e225da77a7e6/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20190423024810-112230192c58/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20190911185100-cd5d95a43a6e/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
+golang.org/x/sync v0.0.0-20201020160332-67f06af15bc9 h1:SQFwaSi55rU7vdNs9Yr0Z324VNlrF+0wMqRXT4St8ck=
 golang.org/x/sync v0.0.0-20201020160332-67f06af15bc9/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sys v0.0.0-20180823144017-11551d06cbcc/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20180830151530-49385e6e1522/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=

tests/integration/v3_leadership_test.go

@@ -16,12 +16,14 @@ package integration
 
 import (
 	"context"
+	"fmt"
 	"strings"
 	"testing"
 	"time"
 
 	pb "go.etcd.io/etcd/api/v3/etcdserverpb"
 	"go.etcd.io/etcd/api/v3/v3rpc/rpctypes"
+	"golang.org/x/sync/errgroup"
 )
 
 func TestMoveLeader(t *testing.T)        { testMoveLeader(t, true) }
@@ -36,7 +38,7 @@ func testMoveLeader(t *testing.T, auto bool) {
 	oldLeadIdx := clus.WaitLeader(t)
 	oldLeadID := uint64(clus.Members[oldLeadIdx].s.ID())
 
-	// ensure followers go through leader transition while learship transfer
+	// ensure followers go through leader transition while leadership transfer
 	idc := make(chan uint64)
 	stopc := make(chan struct{})
 	defer close(stopc)
@@ -179,3 +181,81 @@ func TestTransferLeadershipWithLearner(t *testing.T) {
 		t.Error("timed out waiting for leader transition")
 	}
 }
+
+func TestFirstCommitNotification(t *testing.T) {
+	BeforeTest(t)
+	clusterSize := 3
+	cluster := NewClusterV3(t, &ClusterConfig{Size: clusterSize})
+	defer cluster.Terminate(t)
+
+	oldLeaderIdx := cluster.WaitLeader(t)
+	oldLeaderClient := cluster.Client(oldLeaderIdx)
+
+	newLeaderIdx := (oldLeaderIdx + 1) % clusterSize
+	newLeaderId := uint64(cluster.Members[newLeaderIdx].ID())
+
+	notifiers := make(map[int]<-chan struct{}, clusterSize)
+	for i, clusterMember := range cluster.Members {
+		notifiers[i] = clusterMember.s.FirstCommitInTermNotify()
+	}
+
+	_, err := oldLeaderClient.MoveLeader(context.Background(), newLeaderId)
+
+	if err != nil {
+		t.Errorf("got error during leadership transfer: %v", err)
+	}
+
+	leaderAppliedIndex := cluster.Members[newLeaderIdx].s.AppliedIndex()
+
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	group, groupContext := errgroup.WithContext(ctx)
+
+	for i, notifier := range notifiers {
+		member, notifier := cluster.Members[i], notifier
+		group.Go(func() error {
+			return checkFirstCommitNotification(groupContext, member, leaderAppliedIndex, notifier)
+		})
+	}
+
+	err = group.Wait()
+	if err != nil {
+		t.Error(err)
+	}
+}
+
+func checkFirstCommitNotification(
+	ctx context.Context,
+	member *member,
+	leaderAppliedIndex uint64,
+	notifier <-chan struct{},
+) error {
+	// wait until server applies all the changes of leader
+	for member.s.AppliedIndex() < leaderAppliedIndex {
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		default:
+			time.Sleep(100 * time.Millisecond)
+		}
+	}
+
+	select {
+	case msg, ok := <-notifier:
+		if ok {
+			return fmt.Errorf(
+				"member with ID %d got message via notifier, msg: %v",
+				member.ID(),
+				msg,
+			)
+		}
+	default:
+		return fmt.Errorf(
+			"notification was not triggered, member ID: %d",
+			member.ID(),
+		)
+	}
+
+	return nil
+}