ifndef-SleePy commented on code in PR #8401: URL: https://github.com/apache/paimon/pull/8401#discussion_r3526995118
########## paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/coordinator/CommittingWriteOperatorCoordinator.java: ########## @@ -0,0 +1,561 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you 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, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.paimon.flink.sink.coordinator; + +import org.apache.paimon.annotation.VisibleForTesting; +import org.apache.paimon.flink.sink.Committable; +import org.apache.paimon.flink.sink.Committer; +import org.apache.paimon.flink.sink.state.CoordinatorState; +import org.apache.paimon.flink.sink.state.CoordinatorStateSerializer; +import org.apache.paimon.flink.sink.state.MemoryBackendStateStore; +import org.apache.paimon.manifest.ManifestCommittable; + +import org.apache.flink.api.common.typeutils.base.ListSerializer; +import org.apache.flink.core.io.SimpleVersionedSerialization; +import org.apache.flink.runtime.jobgraph.OperatorID; +import org.apache.flink.runtime.operators.coordination.OperatorCoordinator; +import org.apache.flink.runtime.operators.coordination.OperatorEvent; +import org.apache.flink.runtime.operators.coordination.RecreateOnResetOperatorCoordinator; +import org.apache.flink.util.function.ThrowingRunnable; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Collections; +import java.util.List; +import java.util.Map; +import java.util.NavigableMap; +import java.util.TreeMap; +import java.util.concurrent.CompletableFuture; +import java.util.concurrent.ExecutorService; +import java.util.concurrent.Executors; + +import static org.apache.paimon.utils.Preconditions.checkState; + +/** + * {@link OperatorCoordinator} that runs the Paimon committer on the JobManager for the + * unaware-bucket append write path. Writers stream per-checkpoint committables to this coordinator + * over the {@link OperatorEvent} channel; on {@link #notifyCheckpointComplete} the coordinator + * aligns committables across subtasks and commits them from a dedicated single-thread executor, so + * the JM main thread is never blocked by table I/O. + * + * <p>Wrap this class with a {@link RecreateOnResetOperatorCoordinator} (see {@link Provider}). The + * wrapper discards this instance on global failover and creates a new one in its place, which keeps + * the lifecycle inside a single instance simple. See {@link #resetToCheckpoint} and {@link State} + * for the resulting state machine. + */ +public class CommittingWriteOperatorCoordinator implements OperatorCoordinator { + + private static final Logger LOG = + LoggerFactory.getLogger(CommittingWriteOperatorCoordinator.class); + + private static final long NO_WATERMARK = Long.MIN_VALUE; + + private final OperatorCoordinator.Context context; + private final Committer.Factory<Committable, ManifestCommittable> committerFactory; + private final boolean streamingCheckpointEnabled; + private final boolean failoverAfterRecovery; + private final int parallelism; + + private final WriterCommittables[] subtaskCommittables; + private final ListSerializer<Committable> committableSerializer; + private final CoordinatorStateSerializer stateSerializer; + private final ExecutorService commitExecutor; + + // Populated by resetToCheckpoint and consumed by start. Plain fields are sufficient: both + // callbacks run on the same scheduler thread in order. + private long restoredCheckpointId = OperatorCoordinator.NO_CHECKPOINT; + private byte[] restoredCheckpointData; + + private State state; + private Committer<Committable, ManifestCommittable> committer; + private String commitUser; + private MemoryBackendStateStore stateStore; + + public CommittingWriteOperatorCoordinator( + OperatorCoordinator.Context context, + Committer.Factory<Committable, ManifestCommittable> committerFactory, + boolean streamingCheckpointEnabled, + String initialCommitUser, + boolean failoverAfterRecovery) { + this.context = context; + this.committerFactory = committerFactory; + this.streamingCheckpointEnabled = streamingCheckpointEnabled; + this.commitUser = initialCommitUser; + this.failoverAfterRecovery = failoverAfterRecovery; + this.parallelism = context.currentParallelism(); + this.subtaskCommittables = new WriterCommittables[parallelism]; + this.committableSerializer = CommittableEvent.createCommittableListSerializer(); + this.stateSerializer = new CoordinatorStateSerializer(); + this.commitExecutor = + Executors.newSingleThreadExecutor( + new CoordinatorExecutorThreadFactory("WriteCommitCoordinator", context)); + this.state = State.CREATED; + } + + @Override + public void start() throws Exception { + // Invoked at most once. If resetToCheckpoint ran first it already moved state to RESTORING + // and stashed the bytes; otherwise we are in CREATED and there is nothing to restore. + checkState( + state == State.CREATED || state == State.RESTORING, + "Coordinator already started, illegal state %s", + state); + runInEventLoop( + () -> { + if (state == State.RESTORING) { + restoreState(restoredCheckpointId, restoredCheckpointData); + // not needed after deserialization; release the reference + restoredCheckpointData = null; + initializeCommitter(); + // stay in RESTORING until writers re-emit committables and align catches up + } else { + restoreState(OperatorCoordinator.NO_CHECKPOINT, null); + initializeCommitter(); + transitionState(State.RUNNING); + } + }, + "starting"); + } + + @Override + public void close() throws Exception { + transitionState(State.CLOSED); + if (commitExecutor != null) { + commitExecutor.shutdownNow(); + } + if (committer != null) { + committer.close(); + committer = null; + } + } + + @Override + public void checkpointCoordinator(long checkpointId, CompletableFuture<byte[]> result) { + runCheckpointInEventLoop( + () -> { + if (state != State.RUNNING) { + // if checkpoint is executed before finishing restoring, just fail it + result.completeExceptionally( + new IllegalStateException( + "Checkpoint of commit coordinator should be taken in RUNNING state, while current state is " + + state)); + return; + } + committer.snapshotState(); + byte[] checkpointData = + SimpleVersionedSerialization.writeVersionAndSerialize( + stateSerializer, + new CoordinatorState( + commitUser, stateStore.getSerializedStates())); + result.complete(checkpointData); + }, + result, + "taking checkpoint %d", + checkpointId); + } + + @Override + public void handleEventFromOperator(int subtask, int attemptNumber, OperatorEvent event) { + runInEventLoop( + () -> { + if (event instanceof CommittableEvent) { + handleCommittableEvent(subtask, (CommittableEvent) event); + } else { + // TODO: watermark handling + // TODO: end input handling + throw new UnsupportedOperationException("Unsupported event type: " + event); + } + }, + "handling operator event %s from subtask %d (#%d)", + event, + subtask, + attemptNumber); + } + + @Override + public void notifyCheckpointComplete(long checkpointId) { + runInEventLoop( + () -> { + if (state != State.RUNNING) { + throw new IllegalStateException( + "Completing checkpoint should be notified in RUNNING state, while current state is " + + state); + } + // writers always report a committable per (subtask, checkpoint) during + // snapshot, even if empty; missing means the writer is broken + if (!alignCommittables(checkpointId)) { + throw new IllegalStateException("Not all committables reported by writer"); + } + commitUpToCheckpoint( + checkpointId, + pollManifestCommittablesForCheckpoint( + checkpointId, subtaskCommittables, committer, NO_WATERMARK), + committables -> { + try { + committer.commit(committables); + } catch (Exception e) { + throw new RuntimeException(e); + } + }); + }, + "completing checkpoint %d", + checkpointId); + } + + /** + * Called by the framework at most once, before {@link #start()}. May be skipped entirely if the + * job has no checkpoint or savepoint to restore from; in that case the coordinator goes + * straight from {@code CREATED} to {@code RUNNING} in {@code start()}. + * + * <p>When invoked, {@code checkpointId} is the persisted checkpoint id and {@code + * checkpointData} is its bytes; for batch mode, disabled checkpointing, or no completed + * checkpoint, the framework calls it with {@link OperatorCoordinator#NO_CHECKPOINT} and {@code + * null}, which is treated as a no-op. + * + * <p>The wrapping {@link RecreateOnResetOperatorCoordinator} replaces this instance with a + * fresh one before any further reset, so this method is never called on an already-started + * coordinator. + */ + @Override + public void resetToCheckpoint(long checkpointId, byte[] checkpointData) { + checkState( + state == State.CREATED, + "resetToCheckpoint must run before start, but current state is %s", + state); + if (checkpointId == OperatorCoordinator.NO_CHECKPOINT) { + // nothing to restore; start() will initialize an empty committer + return; + } + restoredCheckpointId = checkpointId; + restoredCheckpointData = checkpointData; + transitionState(State.RESTORING); + } + + @Override + public void subtaskReset(int subtask, long checkpointId) { + runInEventLoop( + () -> { + WriterCommittables writerCommittables = subtaskCommittables[subtask]; + if (writerCommittables != null) { + // sanity check subtask state + Map<Long, List<Committable>> committables = + writerCommittables.getCommittablesBeforeCheckpoint( + checkpointId, false); + if (!committables.isEmpty()) { + throw new IllegalStateException( + String.format( + "Writer [%d] contains invalid committables before checkpoint %d", + subtask, checkpointId)); + } + writerCommittables.reset(); + } + }, + "resetting subtask %d to checkpoint %d", + subtask, + checkpointId); + } + + @Override + public void executionAttemptFailed(int subtask, int attemptNumber, Throwable reason) {} + + @Override + public void executionAttemptReady(int subtask, int attemptNumber, SubtaskGateway gateway) {} + + private void handleCommittableEvent(int subtask, CommittableEvent event) throws Exception { + if (state == State.RESTORING) { + updateSubtaskCommittables(subtask, event); + if (alignCommittables(event.getCheckpointId())) { + recover(event.getCheckpointId()); + transitionState(State.RUNNING); + } + } else if (state == State.RUNNING) { + if (event.isRestoring()) { + // a region failover replayed restoring committables while the coordinator itself is + // not restoring; it already holds the committed state, so ignore them + LOG.info( + "Ignore restoring committables from subtask {} of checkpoint {}, coordinator is running.", + subtask, + event.getCheckpointId()); + } else { + updateSubtaskCommittables(subtask, event); + } + } else { + throw new IllegalStateException( + "Illegal state " + state + " while handling committable event " + event); + } + } + + private void updateSubtaskCommittables(int subtask, CommittableEvent event) throws IOException { + WriterCommittables incoming = WriterCommittables.from(event, committableSerializer); + if (subtaskCommittables[subtask] != null) { + subtaskCommittables[subtask].mergeWith(incoming); + } else { + subtaskCommittables[subtask] = incoming; + } + } + + private boolean alignCommittables(long checkpointId) { + for (WriterCommittables committables : subtaskCommittables) { + if (committables == null || committables.getMaxCheckpointId() < checkpointId) { + return false; + } + } + return true; + } + + // replaces CommittableStateManager because committables are not stored in the committer + private void recover(long checkpointId) throws Exception { + if (failoverAfterRecovery) { + // recommit the restored committables and trigger a failover to reinitialize all writers + commitUpToCheckpoint( + checkpointId, + pollManifestCommittablesForCheckpoint( + checkpointId, subtaskCommittables, committer, NO_WATERMARK), + committables -> { + int numCommitted = committer.filterAndCommit(committables, true, true); + if (numCommitted > 0) { + throw new RuntimeException( + "This exception is intentionally thrown after committing the " + + "restored checkpoints. By restarting the job we hope " + + "that writers can start writing based on these new commits."); + } + }); + } else { + // just abandon the restoring committables + for (WriterCommittables subtaskCommit : subtaskCommittables) { + subtaskCommit.clearCommittablesBeforeCheckpoint(checkpointId, true); + } + } + } + + @VisibleForTesting + static NavigableMap<Long, ManifestCommittable> pollManifestCommittablesForCheckpoint( + long checkpointId, + WriterCommittables[] subtaskCommittables, + Committer<Committable, ManifestCommittable> committer, + long watermark) + throws IOException { + NavigableMap<Long, ManifestCommittable> committablesPerCheckpoint = new TreeMap<>(); + for (WriterCommittables committables : subtaskCommittables) { + NavigableMap<Long, List<Committable>> perCheckpoint = + committables.getCommittablesBeforeCheckpoint(checkpointId, true); + for (Map.Entry<Long, List<Committable>> entry : perCheckpoint.entrySet()) { + long currentCheckpointId = entry.getKey(); + List<Committable> currentCommittables = entry.getValue(); + if (currentCommittables.isEmpty()) { + continue; + } + ManifestCommittable manifestCommittable = + committablesPerCheckpoint.get(currentCheckpointId); + if (manifestCommittable == null) { + committablesPerCheckpoint.put( + currentCheckpointId, + committer.combine(currentCheckpointId, watermark, currentCommittables)); + } else { + committer.combine( + currentCheckpointId, + watermark, + manifestCommittable, + currentCommittables); + } + } + committables.clearCommittablesBeforeCheckpoint(checkpointId, true); + } + return committablesPerCheckpoint; + } + + private void commitUpToCheckpoint( + long checkpointId, Map<Long, ManifestCommittable> toCommit, CommitAction commitAction) + throws Exception { + List<ManifestCommittable> committables = new ArrayList<>(toCommit.values()); + if (committables.isEmpty() && committer.forceCreatingSnapshot()) { + committables = + Collections.singletonList( + committer.combine(checkpointId, NO_WATERMARK, Collections.emptyList())); + } + commitAction.accept(committables); + } + + private void restoreState(long checkpointId, byte[] checkpointData) throws IOException { + if (checkpointData == null) { + stateStore = new MemoryBackendStateStore(); + } else { + CoordinatorState coordinatorState = + SimpleVersionedSerialization.readVersionAndDeSerialize( + stateSerializer, checkpointData); + commitUser = coordinatorState.getCommitUser(); + stateStore = new MemoryBackendStateStore(coordinatorState.getCommitterStates()); + } + } + + private void initializeCommitter() { + Committer.Context committerContext = + Committer.createContext( + commitUser, + context.metricGroup(), + streamingCheckpointEnabled, + true, Review Comment: @JingsongLi The pipeline failed due to a connection error with the Maven repo. Could you help re-trigger it? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
