fix the potential data loss for clusters with only one member

For a cluster with only one member, the raft always send identical
unstable entries and committed entries to etcdserver, and etcd
responds to the client once it finishes (actually partially) the
applying workflow.

When the client receives the response, it doesn't mean etcd has already
successfully saved the data, including BoltDB and WAL, because:
   1. etcd commits the boltDB transaction periodically instead of on each request;
   2. etcd saves WAL entries in parallel with applying the committed entries.
Accordingly, it may run into a situation of data loss when the etcd crashes
immediately after responding to the client and before the boltDB and WAL
successfully save the data to disk.
Note that this issue can only happen for clusters with only one member.

For clusters with multiple members, it isn't an issue, because etcd will
not commit & apply the data before it being replicated to majority members.
When the client receives the response, it means the data must have been applied.
It further means the data must have been committed.
Note: for clusters with multiple members, the raft will never send identical
unstable entries and committed entries to etcdserver.

Signed-off-by: Benjamin Wang <wachao@vmware.com>

server/etcdserver/raft.go

@@ -68,8 +68,10 @@ func init() {
 type toApply struct {
 	entries  []raftpb.Entry
 	snapshot raftpb.Snapshot
-	// notifyc synchronizes etcd server applies with the raft node
-	notifyc chan struct{}
+	// snapNotifyc synchronizes etcd server applies with the raft node
+	snapNotifyc chan struct{}
+	// walNotifyc synchronizes etcd server applies with the WAL persistence
+	walNotifyc chan struct{}
 }
 
 type raftNode struct {
@@ -200,11 +202,16 @@ func (r *raftNode) start(rh *raftReadyHandler) {
 					}
 				}
 
-				notifyc := make(chan struct{}, 1)
+				snapNotifyc := make(chan struct{}, 1)
+				var walNotifyc chan struct{}
+				if shouldWaitWALSync(rd.Entries, rd.CommittedEntries) {
+					walNotifyc = make(chan struct{}, 1)
+				}
 				ap := toApply{
-					entries:  rd.CommittedEntries,
-					snapshot: rd.Snapshot,
-					notifyc:  notifyc,
+					entries:     rd.CommittedEntries,
+					snapshot:    rd.Snapshot,
+					snapNotifyc: snapNotifyc,
+					walNotifyc:  walNotifyc,
 				}
 
 				updateCommittedIndex(&ap, rh)
@@ -237,6 +244,12 @@ func (r *raftNode) start(rh *raftReadyHandler) {
 				if err := r.storage.Save(rd.HardState, rd.Entries); err != nil {
 					r.lg.Fatal("failed to save Raft hard state and entries", zap.Error(err))
 				}
+
+				if walNotifyc != nil {
+					// etcdserver should wait for this notification before responding to client.
+					walNotifyc <- struct{}{}
+				}
+
 				if !raft.IsEmptyHardState(rd.HardState) {
 					proposalsCommitted.Set(float64(rd.HardState.Commit))
 				}
@@ -252,7 +265,7 @@ func (r *raftNode) start(rh *raftReadyHandler) {
 					}
 
 					// etcdserver now claim the snapshot has been persisted onto the disk
-					notifyc <- struct{}{}
+					snapNotifyc <- struct{}{}
 
 					// gofail: var raftBeforeApplySnap struct{}
 					r.raftStorage.ApplySnapshot(rd.Snapshot)
@@ -272,7 +285,7 @@ func (r *raftNode) start(rh *raftReadyHandler) {
 					msgs := r.processMessages(rd.Messages)
 
 					// now unblocks 'applyAll' that waits on Raft log disk writes before triggering snapshots
-					notifyc <- struct{}{}
+					snapNotifyc <- struct{}{}
 
 					// Candidate or follower needs to wait for all pending configuration
 					// changes to be applied before sending messages.
@@ -293,7 +306,7 @@ func (r *raftNode) start(rh *raftReadyHandler) {
 						// to be in sync with scheduled config-change job
 						// (assume notifyc has cap of 1)
 						select {
-						case notifyc <- struct{}{}:
+						case snapNotifyc <- struct{}{}:
 						case <-r.stopped:
 							return
 						}
@@ -303,7 +316,7 @@ func (r *raftNode) start(rh *raftReadyHandler) {
 					r.transport.Send(msgs)
 				} else {
 					// leader already processed 'MsgSnap' and signaled
-					notifyc <- struct{}{}
+					snapNotifyc <- struct{}{}
 				}
 
 				r.Advance()
@@ -314,6 +327,43 @@ func (r *raftNode) start(rh *raftReadyHandler) {
 	}()
 }
 
+// For a cluster with only one member, the raft always send identical
+// unstable entries and committed entries to etcdserver, and etcd
+// responds to the client once it finishes (actually partially) the
+// applying workflow.
+//
+// When the client receives the response, it doesn't mean etcd has already
+// successfully saved the data, including BoltDB and WAL, because:
+//    1. etcd commits the boltDB transaction periodically instead of on each request;
+//    2. etcd saves WAL entries in parallel with applying the committed entries.
+// Accordingly, it might run into a situation of data loss when the etcd crashes
+// immediately after responding to the client and before the boltDB and WAL
+// successfully save the data to disk.
+// Note that this issue can only happen for clusters with only one member.
+//
+// For clusters with multiple members, it isn't an issue, because etcd will
+// not commit & apply the data before it being replicated to majority members.
+// When the client receives the response, it means the data must have been applied.
+// It further means the data must have been committed.
+// Note: for clusters with multiple members, the raft will never send identical
+// unstable entries and committed entries to etcdserver.
+func shouldWaitWALSync(unstableEntries []raftpb.Entry, committedEntries []raftpb.Entry) bool {
+	if len(unstableEntries) != len(committedEntries) {
+		return false
+	}
+
+	if len(unstableEntries) == 0 {
+		return false
+	}
+
+	// Previously `reflect.DeepEqual` is used to compare the `unstableEntries`
+	// and `committedEntries`, but it's a little expensive, so we just compare
+	// the log index. The applying workflow should wait for the WAL sync if
+	// there is any log index overlap between them.
+	lastCommittedEntry := committedEntries[len(committedEntries)-1]
+	return lastCommittedEntry.Term >= unstableEntries[0].Term && lastCommittedEntry.Index >= unstableEntries[0].Index
+}
+
 func updateCommittedIndex(ap *toApply, rh *raftReadyHandler) {
 	var ci uint64
 	if len(ap.entries) != 0 {

server/etcdserver/raft_test.go

@@ -22,6 +22,7 @@ import (
 	"testing"
 	"time"
 
+	"github.com/stretchr/testify/assert"
 	"go.etcd.io/etcd/client/pkg/v3/types"
 	"go.etcd.io/etcd/pkg/v3/pbutil"
 	"go.etcd.io/etcd/raft/v3"
@@ -190,6 +191,52 @@ func TestStopRaftWhenWaitingForApplyDone(t *testing.T) {
 
 // TestConfigChangeBlocksApply ensures toApply blocks if committed entries contain config-change.
 func TestConfigChangeBlocksApply(t *testing.T) {
+	testEtcdserverAndRaftInteraction(t, testConfigChangeBlockApply)
+}
+
+func TestWALSyncNotBlockApply(t *testing.T) {
+	testEtcdserverAndRaftInteraction(t, testWALSyncNotBlockApply)
+}
+
+func TestWALSyncBlockApply(t *testing.T) {
+	testEtcdserverAndRaftInteraction(t, testWALSyncBlockApply)
+}
+
+func testConfigChangeBlockApply(t *testing.T, srv *EtcdServer, n *readyNode) toApply {
+	n.readyc <- raft.Ready{
+		SoftState:        &raft.SoftState{RaftState: raft.StateFollower},
+		CommittedEntries: []raftpb.Entry{{Type: raftpb.EntryConfChange}},
+	}
+	return <-srv.r.applyc
+}
+
+func testWALSyncNotBlockApply(t *testing.T, srv *EtcdServer, n *readyNode) toApply {
+	n.readyc <- raft.Ready{
+		SoftState:        &raft.SoftState{RaftState: raft.StateFollower},
+		CommittedEntries: []raftpb.Entry{{Type: raftpb.EntryConfChange}},
+	}
+	ap := <-srv.r.applyc
+	if ap.walNotifyc != nil {
+		t.Error("unexpected ap.walNotifyc, expected nil")
+	}
+	return ap
+}
+
+func testWALSyncBlockApply(t *testing.T, srv *EtcdServer, n *readyNode) toApply {
+	n.readyc <- raft.Ready{
+		SoftState:        &raft.SoftState{RaftState: raft.StateFollower},
+		Entries:          []raftpb.Entry{{Type: raftpb.EntryConfChange}},
+		CommittedEntries: []raftpb.Entry{{Type: raftpb.EntryConfChange}},
+	}
+	ap := <-srv.r.applyc
+	if ap.walNotifyc == nil {
+		t.Error("unexpected ap.walNotifyc, expected not nil")
+	}
+	assert.NotEqual(t, nil, ap.walNotifyc)
+	return ap
+}
+
+func testEtcdserverAndRaftInteraction(t *testing.T, testFunc func(*testing.T, *EtcdServer, *readyNode) toApply) {
 	n := newNopReadyNode()
 
 	r := newRaftNode(raftNodeConfig{
@@ -208,11 +255,7 @@ func TestConfigChangeBlocksApply(t *testing.T) {
 	})
 	defer srv.r.Stop()
 
-	n.readyc <- raft.Ready{
-		SoftState:        &raft.SoftState{RaftState: raft.StateFollower},
-		CommittedEntries: []raftpb.Entry{{Type: raftpb.EntryConfChange}},
-	}
-	ap := <-srv.r.applyc
+	ap := testFunc(t, srv, n)
 
 	continueC := make(chan struct{})
 	go func() {
@@ -228,7 +271,7 @@ func TestConfigChangeBlocksApply(t *testing.T) {
 	}
 
 	// finish toApply, unblock raft routine
-	<-ap.notifyc
+	<-ap.snapNotifyc
 
 	select {
 	case <-continueC:
@@ -317,3 +360,62 @@ func TestStopRaftNodeMoreThanOnce(t *testing.T) {
 		}
 	}
 }
+
+func TestShouldWaitWALSync(t *testing.T) {
+	testcases := []struct {
+		name            string
+		unstableEntries []raftpb.Entry
+		commitedEntries []raftpb.Entry
+		expectedResult  bool
+	}{
+		{
+			name:            "both entries are nil",
+			unstableEntries: nil,
+			commitedEntries: nil,
+			expectedResult:  false,
+		},
+		{
+			name:            "both entries are empty slices",
+			unstableEntries: []raftpb.Entry{},
+			commitedEntries: []raftpb.Entry{},
+			expectedResult:  false,
+		},
+		{
+			name:            "one nil and the other empty",
+			unstableEntries: nil,
+			commitedEntries: []raftpb.Entry{},
+			expectedResult:  false,
+		},
+		{
+			name:            "one nil and the other has data",
+			unstableEntries: nil,
+			commitedEntries: []raftpb.Entry{{Term: 4, Index: 10, Type: raftpb.EntryNormal, Data: []byte{0x11, 0x22, 0x33}}},
+			expectedResult:  false,
+		},
+		{
+			name:            "one empty and the other has data",
+			unstableEntries: []raftpb.Entry{},
+			commitedEntries: []raftpb.Entry{{Term: 4, Index: 10, Type: raftpb.EntryNormal, Data: []byte{0x11, 0x22, 0x33}}},
+			expectedResult:  false,
+		},
+		{
+			name:            "has different term and index",
+			unstableEntries: []raftpb.Entry{{Term: 3, Index: 11, Type: raftpb.EntryNormal, Data: []byte{0x11, 0x22, 0x33}}},
+			commitedEntries: []raftpb.Entry{{Term: 4, Index: 10, Type: raftpb.EntryNormal, Data: []byte{0x11, 0x22, 0x33}}},
+			expectedResult:  false,
+		},
+		{
+			name:            "has identical data",
+			unstableEntries: []raftpb.Entry{{Term: 4, Index: 10, Type: raftpb.EntryNormal, Data: []byte{0x11, 0x22, 0x33}}},
+			commitedEntries: []raftpb.Entry{{Term: 4, Index: 10, Type: raftpb.EntryNormal, Data: []byte{0x11, 0x22, 0x33}}},
+			expectedResult:  true,
+		},
+	}
+
+	for _, tc := range testcases {
+		t.Run(tc.name, func(t *testing.T) {
+			shouldWALSync := shouldWaitWALSync(tc.unstableEntries, tc.commitedEntries)
+			assert.Equal(t, tc.expectedResult, shouldWALSync)
+		})
+	}
+}