[release-19.0] VReplication: Take replication lag into account in VSt…

…reamManager healthcheck result processing (#15761) (#15774) Signed-off-by: Matt Lord <[email protected]> Co-authored-by: vitess-bot[bot] <108069721+vitess-bot[bot]@users.noreply.github.com> Co-authored-by: Matt Lord <[email protected]>
vitessio · Apr 22, 2024 · cc64915 · cc64915
1 parent 4ec2bcf
commit cc64915
Show file tree

Hide file tree

Showing 3 changed files with 155 additions and 10 deletions.
diff --git a/go/vt/vtgate/vstream_manager.go b/go/vt/vtgate/vstream_manager.go
@@ -31,6 +31,7 @@ import (
 	"vitess.io/vitess/go/vt/servenv"
 	"vitess.io/vitess/go/vt/srvtopo"
 	"vitess.io/vitess/go/vt/topo"
+	"vitess.io/vitess/go/vt/topo/topoproto"
 	"vitess.io/vitess/go/vt/vterrors"
 
 	binlogdatapb "vitess.io/vitess/go/vt/proto/binlogdata"
@@ -531,18 +532,23 @@ func (vs *vstream) streamFromTablet(ctx context.Context, sgtid *binlogdatapb.Sha
 		go func() {
 			_ = tabletConn.StreamHealth(ctx, func(shr *querypb.StreamHealthResponse) error {
 				var err error
-				if ctx.Err() != nil {
+				switch {
+				case ctx.Err() != nil:
 					err = fmt.Errorf("context has ended")
-				} else if shr == nil || shr.RealtimeStats == nil || shr.Target == nil {
-					err = fmt.Errorf("health check failed")
-				} else if vs.tabletType != shr.Target.TabletType {
-					err = fmt.Errorf("tablet type has changed from %s to %s, restarting vstream",
-						vs.tabletType, shr.Target.TabletType)
-				} else if shr.RealtimeStats.HealthError != "" {
+				case shr == nil || shr.RealtimeStats == nil || shr.Target == nil:
+					err = fmt.Errorf("health check failed on %s", topoproto.TabletAliasString(tablet.Alias))
+				case vs.tabletType != shr.Target.TabletType:
+					err = fmt.Errorf("tablet %s type has changed from %s to %s, restarting vstream",
+						topoproto.TabletAliasString(tablet.Alias), vs.tabletType, shr.Target.TabletType)
+				case shr.RealtimeStats.HealthError != "":
 					err = fmt.Errorf("tablet %s is no longer healthy: %s, restarting vstream",
-						tablet.Alias, shr.RealtimeStats.HealthError)
+						topoproto.TabletAliasString(tablet.Alias), shr.RealtimeStats.HealthError)
+				case shr.RealtimeStats.ReplicationLagSeconds > uint32(discovery.GetLowReplicationLag().Seconds()):
+					err = fmt.Errorf("tablet %s has a replication lag of %d seconds which is beyond the value provided in --discovery_low_replication_lag of %s so the tablet is no longer considered healthy, restarting vstream",
+						topoproto.TabletAliasString(tablet.Alias), shr.RealtimeStats.ReplicationLagSeconds, discovery.GetLowReplicationLag())
 				}
 				if err != nil {
+					log.Warningf("Tablet state changed: %s, attempting to restart", err)
 					errCh <- err
 					return err
 				}
@@ -573,7 +579,6 @@ func (vs *vstream) streamFromTablet(ctx context.Context, sgtid *binlogdatapb.Sha
 			case <-ctx.Done():
 				return ctx.Err()
 			case streamErr := <-errCh:
-				log.Warningf("Tablet state changed: %s, attempting to restart", streamErr)
 				return vterrors.New(vtrpcpb.Code_UNAVAILABLE, streamErr.Error())
 			case <-journalDone:
 				// Unreachable.

diff --git a/go/vt/vtgate/vstream_manager_test.go b/go/vt/vtgate/vstream_manager_test.go
@@ -32,12 +32,16 @@ import (
 
 	"vitess.io/vitess/go/stats"
 	"vitess.io/vitess/go/vt/discovery"
+	"vitess.io/vitess/go/vt/log"
+	"vitess.io/vitess/go/vt/logutil"
 	"vitess.io/vitess/go/vt/srvtopo"
 	"vitess.io/vitess/go/vt/topo"
+	"vitess.io/vitess/go/vt/topo/topoproto"
 	"vitess.io/vitess/go/vt/vterrors"
 	"vitess.io/vitess/go/vt/vttablet/sandboxconn"
 
 	binlogdatapb "vitess.io/vitess/go/vt/proto/binlogdata"
+	querypb "vitess.io/vitess/go/vt/proto/query"
 	topodatapb "vitess.io/vitess/go/vt/proto/topodata"
 	vtgatepb "vitess.io/vitess/go/vt/proto/vtgate"
 	vtrpcpb "vitess.io/vitess/go/vt/proto/vtrpc"
@@ -1279,6 +1283,127 @@ func TestVStreamIdleHeartbeat(t *testing.T) {
 	}
 }
 
+// TestVStreamManagerHealthCheckResponseHandling tests the handling of healthcheck responses by
+// the vstream manager to confirm that we are correctly restarting the vstream when we should.
+func TestVStreamManagerHealthCheckResponseHandling(t *testing.T) {
+	ctx := utils.LeakCheckContext(t)
+
+	// Capture the vstream warning log. Otherwise we need to re-implement the vstream error
+	// handling in SandboxConn's implementation and then we're not actually testing the
+	// production code.
+	logger := logutil.NewMemoryLogger()
+	log.Warningf = logger.Warningf
+
+	cell := "aa"
+	ks := "TestVStream"
+	shard := "0"
+	tabletType := topodatapb.TabletType_REPLICA
+	_ = createSandbox(ks)
+	hc := discovery.NewFakeHealthCheck(nil)
+	st := getSandboxTopo(ctx, cell, ks, []string{shard})
+	vsm := newTestVStreamManager(ctx, hc, st, cell)
+	vgtid := &binlogdatapb.VGtid{
+		ShardGtids: []*binlogdatapb.ShardGtid{{
+			Keyspace: ks,
+			Shard:    shard,
+		}},
+	}
+	source := hc.AddTestTablet(cell, "1.1.1.1", 1001, ks, shard, tabletType, true, 0, nil)
+	tabletAlias := topoproto.TabletAliasString(source.Tablet().Alias)
+	addTabletToSandboxTopo(t, ctx, st, ks, shard, source.Tablet())
+	target := &querypb.Target{
+		Cell:       cell,
+		Keyspace:   ks,
+		Shard:      shard,
+		TabletType: tabletType,
+	}
+	highLag := uint32(discovery.GetLowReplicationLag().Seconds()) + 1
+
+	type testcase struct {
+		name    string
+		hcRes   *querypb.StreamHealthResponse
+		wantErr string
+	}
+	testcases := []testcase{
+		{
+			name: "all healthy", // Will hit the context timeout
+		},
+		{
+			name: "failure",
+			hcRes: &querypb.StreamHealthResponse{
+				TabletAlias: source.Tablet().Alias,
+				Target:      nil, // This is seen as a healthcheck stream failure
+			},
+			wantErr: fmt.Sprintf("health check failed on %s", tabletAlias),
+		},
+		{
+			name: "tablet type changed",
+			hcRes: &querypb.StreamHealthResponse{
+				TabletAlias: source.Tablet().Alias,
+				Target: &querypb.Target{
+					Cell:       cell,
+					Keyspace:   ks,
+					Shard:      shard,
+					TabletType: topodatapb.TabletType_PRIMARY,
+				},
+				PrimaryTermStartTimestamp: time.Now().Unix(),
+				RealtimeStats:             &querypb.RealtimeStats{},
+			},
+			wantErr: fmt.Sprintf("tablet %s type has changed from %s to %s",
+				tabletAlias, tabletType, topodatapb.TabletType_PRIMARY.String()),
+		},
+		{
+			name: "unhealthy",
+			hcRes: &querypb.StreamHealthResponse{
+				TabletAlias: source.Tablet().Alias,
+				Target:      target,
+				RealtimeStats: &querypb.RealtimeStats{
+					HealthError: "unhealthy",
+				},
+			},
+			wantErr: fmt.Sprintf("tablet %s is no longer healthy", tabletAlias),
+		},
+		{
+			name: "replication lag too high",
+			hcRes: &querypb.StreamHealthResponse{
+				TabletAlias: source.Tablet().Alias,
+				Target:      target,
+				RealtimeStats: &querypb.RealtimeStats{
+					ReplicationLagSeconds: highLag,
+				},
+			},
+			wantErr: fmt.Sprintf("%s has a replication lag of %d seconds which is beyond the value provided",
+				tabletAlias, highLag),
+		},
+	}
+
+	for _, tc := range testcases {
+		t.Run(tc.name, func(t *testing.T) {
+			done := make(chan struct{})
+			go func() {
+				sctx, cancel := context.WithTimeout(ctx, 5*time.Second)
+				defer cancel()
+				defer close(done)
+				// SandboxConn's VStream implementation always waits for the context to timeout.
+				err := vsm.VStream(sctx, tabletType, vgtid, nil, nil, func(events []*binlogdatapb.VEvent) error {
+					require.Fail(t, "unexpected event", "Received unexpected events: %v", events)
+					return nil
+				})
+				if tc.wantErr != "" { // Otherwise we simply expect the context to timeout
+					if !strings.Contains(logger.String(), tc.wantErr) {
+						require.Fail(t, "unexpected vstream error", "vstream ended with error: %v, which did not contain: %s", err, tc.wantErr)
+					}
+				}
+			}()
+			if tc.wantErr != "" {
+				source.SetStreamHealthResponse(tc.hcRes)
+			}
+			<-done
+			logger.Clear()
+		})
+	}
+}
+
 func newTestVStreamManager(ctx context.Context, hc discovery.HealthCheck, serv srvtopo.Server, cell string) *vstreamManager {
 	gw := NewTabletGateway(ctx, hc, serv, cell)
 	srvResolver := srvtopo.NewResolver(serv, gw, cell)

diff --git a/go/vt/vttablet/sandboxconn/sandboxconn.go b/go/vt/vttablet/sandboxconn/sandboxconn.go
@@ -130,6 +130,8 @@ type SandboxConn struct {
 	getSchemaResult []map[string]string
 
 	parser *sqlparser.Parser
+
+	streamHealthResponse *querypb.StreamHealthResponse
 }
 
 var _ queryservice.QueryService = (*SandboxConn)(nil) // compile-time interface check
@@ -456,8 +458,21 @@ func (sbc *SandboxConn) MessageAck(ctx context.Context, target *querypb.Target,
 // SandboxSQRowCount is the default number of fake splits returned.
 var SandboxSQRowCount = int64(10)
 
-// StreamHealth always mocks a "healthy" result.
+// SetStreamHealthResponse sets the StreamHealthResponse to be returned in StreamHealth.
+func (sbc *SandboxConn) SetStreamHealthResponse(res *querypb.StreamHealthResponse) {
+	sbc.mapMu.Lock()
+	defer sbc.mapMu.Unlock()
+	sbc.streamHealthResponse = res
+}
+
+// StreamHealth always mocks a "healthy" result by default. If you want to override this behavior you
+// can call SetStreamHealthResponse.
 func (sbc *SandboxConn) StreamHealth(ctx context.Context, callback func(*querypb.StreamHealthResponse) error) error {
+	sbc.mapMu.Lock()
+	defer sbc.mapMu.Unlock()
+	if sbc.streamHealthResponse != nil {
+		return callback(sbc.streamHealthResponse)
+	}
 	return nil
 }