This is an automated cherry-pick of pingcap#5186

Signed-off-by: ti-chi-bot <[email protected]>

cdc/owner/ddl_sink.go

@@ -0,0 +1,271 @@
+// Copyright 2021 PingCAP, Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package owner
+
+import (
+	"context"
+	"sync"
+	"sync/atomic"
+	"time"
+
+	"github.com/pingcap/errors"
+	"github.com/pingcap/failpoint"
+	"github.com/pingcap/log"
+	"github.com/pingcap/tiflow/cdc/model"
+	"github.com/pingcap/tiflow/cdc/sink"
+	cdcContext "github.com/pingcap/tiflow/pkg/context"
+	cerror "github.com/pingcap/tiflow/pkg/errors"
+	"github.com/pingcap/tiflow/pkg/filter"
+	"github.com/pingcap/tiflow/pkg/util"
+	"go.uber.org/zap"
+)
+
+const (
+	defaultErrChSize = 1024
+)
+
+// DDLSink is a wrapper of the `Sink` interface for the owner
+// DDLSink should send `DDLEvent` and `CheckpointTs` to downstream sink,
+// If `SyncPointEnabled`, also send `syncPoint` to downstream.
+type DDLSink interface {
+	// run the DDLSink
+	run(ctx cdcContext.Context, id model.ChangeFeedID, info *model.ChangeFeedInfo)
+	// emitCheckpointTs emits the checkpoint Ts to downstream data source
+	// this function will return after recording the checkpointTs specified in memory immediately
+	// and the recorded checkpointTs will be sent and updated to downstream data source every second
+	emitCheckpointTs(ts uint64, tableNames []model.TableName)
+	// emitDDLEvent emits DDL event and return true if the DDL is executed
+	// the DDL event will be sent to another goroutine and execute to downstream
+	// the caller of this function can call again and again until a true returned
+	emitDDLEvent(ctx cdcContext.Context, ddl *model.DDLEvent) (bool, error)
+	emitSyncPoint(ctx cdcContext.Context, checkpointTs uint64) error
+	// close the sink, cancel running goroutine.
+	close(ctx context.Context) error
+}
+
+type ddlSinkImpl struct {
+	lastSyncPoint  model.Ts
+	syncPointStore sink.SyncpointStore
+
+	// It is used to record the checkpointTs and the names of the table at that time.
+	mu struct {
+		sync.Mutex
+		checkpointTs      model.Ts
+		currentTableNames []model.TableName
+	}
+	ddlFinishedTs model.Ts
+	ddlSentTs     model.Ts
+
+	ddlCh chan *model.DDLEvent
+	errCh chan error
+
+	sink sink.Sink
+	// `sinkInitHandler` can be helpful in unit testing.
+	sinkInitHandler ddlSinkInitHandler
+
+	// cancel would be used to cancel the goroutine start by `run`
+	cancel context.CancelFunc
+	wg     sync.WaitGroup
+}
+
+func newDDLSink() DDLSink {
+	return &ddlSinkImpl{
+		ddlCh:           make(chan *model.DDLEvent, 1),
+		errCh:           make(chan error, defaultErrChSize),
+		sinkInitHandler: ddlSinkInitializer,
+		cancel:          func() {},
+	}
+}
+
+type ddlSinkInitHandler func(ctx cdcContext.Context, a *ddlSinkImpl, id model.ChangeFeedID, info *model.ChangeFeedInfo) error
+
+func ddlSinkInitializer(ctx cdcContext.Context, a *ddlSinkImpl, id model.ChangeFeedID, info *model.ChangeFeedInfo) error {
+	filter, err := filter.NewFilter(info.Config)
+	if err != nil {
+		return errors.Trace(err)
+	}
+
+	stdCtx := util.PutChangefeedIDInCtx(ctx, id)
+	stdCtx = util.PutRoleInCtx(stdCtx, util.RoleOwner)
+	s, err := sink.New(stdCtx, id, info.SinkURI, filter, info.Config, info.Opts, a.errCh)
+	if err != nil {
+		return errors.Trace(err)
+	}
+	a.sink = s
+
+	if !info.SyncPointEnabled {
+		return nil
+	}
+	syncPointStore, err := sink.NewSyncpointStore(stdCtx, id, info.SinkURI)
+	if err != nil {
+		return errors.Trace(err)
+	}
+	a.syncPointStore = syncPointStore
+
+	if err := a.syncPointStore.CreateSynctable(stdCtx); err != nil {
+		return errors.Trace(err)
+	}
+	return nil
+}
+
+func (s *ddlSinkImpl) run(ctx cdcContext.Context, id model.ChangeFeedID, info *model.ChangeFeedInfo) {
+	ctx, cancel := cdcContext.WithCancel(ctx)
+	s.cancel = cancel
+
+	s.wg.Add(1)
+	go func() {
+		defer s.wg.Done()
+
+		start := time.Now()
+		if err := s.sinkInitHandler(ctx, s, id, info); err != nil {
+			log.Warn("ddl sink initialize failed",
+				zap.String("changefeed", ctx.ChangefeedVars().ID),
+				zap.Duration("duration", time.Since(start)))
+			ctx.Throw(err)
+			return
+		}
+		log.Info("ddl sink initialized, start processing...",
+			zap.String("changefeed", ctx.ChangefeedVars().ID),
+			zap.Duration("duration", time.Since(start)))
+
+		// TODO make the tick duration configurable
+		ticker := time.NewTicker(time.Second)
+		defer ticker.Stop()
+		var lastCheckpointTs model.Ts
+		for {
+			select {
+			case <-ctx.Done():
+				return
+			case err := <-s.errCh:
+				ctx.Throw(err)
+				return
+			default:
+			}
+			// `ticker.C` and `ddlCh` may can be triggered at the same time, it
+			// does not matter which one emit first, since TiCDC allow DDL with
+			// CommitTs equal to the last CheckpointTs be emitted later.
+			select {
+			case <-ctx.Done():
+				return
+			case err := <-s.errCh:
+				ctx.Throw(err)
+				return
+			case <-ticker.C:
+				s.mu.Lock()
+				checkpointTs := s.mu.checkpointTs
+				if checkpointTs == 0 || checkpointTs <= lastCheckpointTs {
+					s.mu.Unlock()
+					continue
+				}
+				tables := s.mu.currentTableNames
+				s.mu.Unlock()
+				lastCheckpointTs = checkpointTs
+				if err := s.sink.EmitCheckpointTs(ctx, checkpointTs, tables); err != nil {
+					ctx.Throw(errors.Trace(err))
+					return
+				}
+			case ddl := <-s.ddlCh:
+				log.Info("begin emit ddl event",
+					zap.String("changefeed", ctx.ChangefeedVars().ID),
+					zap.Any("DDL", ddl))
+				err := s.sink.EmitDDLEvent(ctx, ddl)
+				failpoint.Inject("InjectChangefeedDDLError", func() {
+					err = cerror.ErrExecDDLFailed.GenWithStackByArgs()
+				})
+				if err == nil || cerror.ErrDDLEventIgnored.Equal(errors.Cause(err)) {
+					log.Info("Execute DDL succeeded",
+						zap.String("changefeed", ctx.ChangefeedVars().ID),
+						zap.Bool("ignored", err != nil),
+						zap.Any("ddl", ddl))
+					atomic.StoreUint64(&s.ddlFinishedTs, ddl.CommitTs)
+					continue
+				}
+				// If DDL executing failed, and the error can not be ignored,
+				// throw an error and pause the changefeed
+				log.Error("Execute DDL failed",
+					zap.String("changefeed", ctx.ChangefeedVars().ID),
+					zap.Error(err),
+					zap.Any("ddl", ddl))
+				ctx.Throw(errors.Trace(err))
+				return
+			}
+		}
+	}()
+}
+
+func (s *ddlSinkImpl) emitCheckpointTs(ts uint64, tableNames []model.TableName) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.mu.checkpointTs = ts
+	s.mu.currentTableNames = tableNames
+}
+
+func (s *ddlSinkImpl) emitDDLEvent(ctx cdcContext.Context, ddl *model.DDLEvent) (bool, error) {
+	ddlFinishedTs := atomic.LoadUint64(&s.ddlFinishedTs)
+	if ddl.CommitTs <= ddlFinishedTs {
+		// the DDL event is executed successfully, and done is true
+		log.Info("ddl already executed",
+			zap.String("changefeed", ctx.ChangefeedVars().ID),
+			zap.Uint64("ddlFinishedTs", ddlFinishedTs), zap.Any("DDL", ddl))
+		return true, nil
+	}
+	if ddl.CommitTs <= s.ddlSentTs {
+		log.Debug("ddl is not finished yet",
+			zap.String("changefeed", ctx.ChangefeedVars().ID),
+			zap.Uint64("ddlSentTs", s.ddlSentTs), zap.Any("DDL", ddl))
+		// the DDL event is executing and not finished yet, return false
+		return false, nil
+	}
+	select {
+	case <-ctx.Done():
+		return false, errors.Trace(ctx.Err())
+	case s.ddlCh <- ddl:
+		s.ddlSentTs = ddl.CommitTs
+		log.Info("ddl is sent",
+			zap.String("changefeed", ctx.ChangefeedVars().ID),
+			zap.Uint64("ddlSentTs", s.ddlSentTs))
+	default:
+		log.Warn("ddl chan full, send it the next round",
+			zap.String("changefeed", ctx.ChangefeedVars().ID),
+			zap.Uint64("ddlSentTs", s.ddlSentTs),
+			zap.Uint64("ddlFinishedTs", s.ddlFinishedTs), zap.Any("DDL", ddl))
+		// if this hit, we think that ddlCh is full,
+		// just return false and send the ddl in the next round.
+	}
+	return false, nil
+}
+
+func (s *ddlSinkImpl) emitSyncPoint(ctx cdcContext.Context, checkpointTs uint64) error {
+	if checkpointTs == s.lastSyncPoint {
+		return nil
+	}
+	s.lastSyncPoint = checkpointTs
+	// TODO implement async sink syncPoint
+	return s.syncPointStore.SinkSyncpoint(ctx, ctx.ChangefeedVars().ID, checkpointTs)
+}
+
+func (s *ddlSinkImpl) close(ctx context.Context) (err error) {
+	s.cancel()
+	if s.sink != nil {
+		err = s.sink.Close(ctx)
+	}
+	if s.syncPointStore != nil {
+		err = s.syncPointStore.Close()
+	}
+	s.wg.Wait()
+	if err != nil && errors.Cause(err) != context.Canceled {
+		return err
+	}
+	return nil
+}