rkhachatryan commented on a change in pull request #11515: [FLINK-16744][task]
implement channel state persistence for unaligned checkpoints
URL: https://github.com/apache/flink/pull/11515#discussion_r405463652
##########
File path:
flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/tasks/SubtaskCheckpointCoordinatorImpl.java
##########
@@ -130,95 +130,108 @@ public void checkpointState(
// Step (1): Prepare the checkpoint, allow operators to do some
pre-barrier work.
// The pre-barrier work should be nothing or minimal
in the common case.
- operatorChain.prepareSnapshotPreBarrier(checkpointId);
+
operatorChain.prepareSnapshotPreBarrier(metadata.getCheckpointId());
// Step (2): Send the checkpoint barrier downstream
- operatorChain.broadcastCheckpointBarrier(
- checkpointId,
- checkpointMetaData.getTimestamp(),
- checkpointOptions);
+
operatorChain.broadcastCheckpointBarrier(metadata.getCheckpointId(),
metadata.getTimestamp(), options);
- // Step (3): Take the state snapshot. This should be largely
asynchronous, to not
- // impact progress of the streaming topology
+ // Step (3): Take the state snapshot. This should be largely
asynchronous, to not impact progress of the streaming topology
- long startSyncPartNano = System.nanoTime();
-
- HashMap<OperatorID, OperatorSnapshotFutures>
operatorSnapshotsInProgress = new
HashMap<>(operatorChain.getNumberOfOperators());
- ChannelStateWriteResult channelStateWriteResult =
- checkpointOptions.getCheckpointType() == CHECKPOINT ?
channelStateWriter.getWriteResult(checkpointMetaData.getCheckpointId()) :
- ChannelStateWriteResult.EMPTY;
+ Map<OperatorID, OperatorSnapshotFutures> snapshotFutures = new
HashMap<>(operatorChain.getNumberOfOperators());
try {
- for (StreamOperatorWrapper<?, ?> operatorWrapper :
operatorChain.getAllOperators(true)) {
- operatorSnapshotsInProgress.put(
-
operatorWrapper.getStreamOperator().getOperatorID(),
- buildOperatorSnapshotFutures(
- checkpointMetaData,
- checkpointOptions,
- operatorChain,
-
operatorWrapper.getStreamOperator(),
- isCanceled,
- channelStateWriteResult)
- );
- }
-
checkpointStorage.clearCacheFor(checkpointMetaData.getCheckpointId());
+ takeSnapshotSync(snapshotFutures, metadata, metrics,
options, operatorChain, isCanceled);
+ finishAndReportAsync(snapshotFutures, metadata,
metrics);
+ } catch (Exception ex) {
+ cleanup(snapshotFutures, metadata, metrics, options,
ex);
+ }
+ }
- if (LOG.isDebugEnabled()) {
- LOG.debug("Finished synchronous checkpoints for
checkpoint {} on task {}",
- checkpointMetaData.getCheckpointId(),
taskName);
- }
+ private void cleanup(
+ Map<OperatorID, OperatorSnapshotFutures>
operatorSnapshotsInProgress,
+ CheckpointMetaData metadata,
+ CheckpointMetrics metrics, CheckpointOptions options,
+ Exception ex) throws Exception {
- long startAsyncPartNano = System.nanoTime();
-
-
checkpointMetrics.setSyncDurationMillis((startAsyncPartNano -
startSyncPartNano) / 1_000_000);
-
- // we are transferring ownership over
snapshotInProgressList for cleanup to the thread, active on submit
- executorService.execute(new AsyncCheckpointRunnable(
- operatorSnapshotsInProgress,
- checkpointMetaData,
- checkpointMetrics,
- startAsyncPartNano,
- taskName,
- closeableRegistry,
- env,
- asyncExceptionHandler));
-
- if (LOG.isDebugEnabled()) {
- LOG.debug(
- "{} - finished synchronous part of
checkpoint {}. Alignment duration: {} ms, snapshot duration {} ms",
- taskName,
checkpointMetaData.getCheckpointId(),
-
checkpointMetrics.getAlignmentDurationNanos() / 1_000_000,
-
checkpointMetrics.getSyncDurationMillis());
- }
- } catch (Exception ex) {
- // Cleanup to release resources
- for (OperatorSnapshotFutures operatorSnapshotResult :
operatorSnapshotsInProgress.values()) {
- if (null != operatorSnapshotResult) {
- try {
- operatorSnapshotResult.cancel();
- } catch (Exception e) {
- LOG.warn("Could not properly
cancel an operator snapshot result.", e);
- }
+ for (OperatorSnapshotFutures operatorSnapshotResult :
operatorSnapshotsInProgress.values()) {
+ if (operatorSnapshotResult != null) {
+ try {
+ operatorSnapshotResult.cancel();
+ } catch (Exception e) {
+ LOG.warn("Could not properly cancel an
operator snapshot result.", e);
}
}
+ }
- if (LOG.isDebugEnabled()) {
- LOG.debug(
- "{} - did NOT finish synchronous part
of checkpoint {}. Alignment duration: {} ms, snapshot duration {} ms",
- taskName,
checkpointMetaData.getCheckpointId(),
-
checkpointMetrics.getAlignmentDurationNanos() / 1_000_000,
-
checkpointMetrics.getSyncDurationMillis());
- }
+ if (LOG.isDebugEnabled()) {
+ LOG.debug(
+ "{} - did NOT finish synchronous part of
checkpoint {}. Alignment duration: {} ms, snapshot duration {} ms",
+ taskName, metadata.getCheckpointId(),
+ metrics.getAlignmentDurationNanos() / 1_000_000,
+ metrics.getSyncDurationMillis());
+ }
- if
(checkpointOptions.getCheckpointType().isSynchronous()) {
- // in the case of a synchronous checkpoint, we
always rethrow the exception,
- // so that the task fails.
- // this is because the intention is always to
stop the job after this checkpointing
- // operation, and without the failure, the task
would go back to normal execution.
- throw ex;
- } else {
-
env.declineCheckpoint(checkpointMetaData.getCheckpointId(), ex);
- }
+ if (options.getCheckpointType().isSynchronous()) {
+ // in the case of a synchronous checkpoint, we always
rethrow the exception,
+ // so that the task fails.
+ // this is because the intention is always to stop the
job after this checkpointing
+ // operation, and without the failure, the task would
go back to normal execution.
+ throw ex;
+ } else {
+ env.declineCheckpoint(metadata.getCheckpointId(), ex);
+ }
+ }
+
+ private void finishAndReportAsync(Map<OperatorID,
OperatorSnapshotFutures> snapshotFutures, CheckpointMetaData metadata,
CheckpointMetrics metrics) {
+ // we are transferring ownership over snapshotInProgressList
for cleanup to the thread, active on submit
+ executorService.execute(new AsyncCheckpointRunnable(
+ snapshotFutures,
+ metadata,
+ metrics,
+ System.nanoTime(),
+ taskName,
+ closeableRegistry,
+ env,
+ asyncExceptionHandler));
+ }
+
+ private void takeSnapshotSync(
+ Map<OperatorID, OperatorSnapshotFutures>
operatorSnapshotsInProgress, CheckpointMetaData checkpointMetaData,
+ CheckpointMetrics checkpointMetrics, CheckpointOptions
checkpointOptions,
+ OperatorChain<?, ?> operatorChain,
+ Supplier<Boolean> isCanceled) throws Exception {
+
+ long checkpointId = checkpointMetaData.getCheckpointId();
+ long started = System.nanoTime();
+
+ ChannelStateWriteResult channelStateWriteResult =
checkpointOptions.getCheckpointType() == CHECKPOINT ?
+
channelStateWriter.getWriteResult(checkpointId) :
+
ChannelStateWriteResult.EMPTY;
+
+ CheckpointStreamFactory storage =
checkpointStorage.resolveCheckpointStorageLocation(checkpointId,
checkpointOptions.getTargetLocation());
+
+ for (StreamOperatorWrapper<?, ?> operatorWrapper :
operatorChain.getAllOperators(true)) {
+ operatorSnapshotsInProgress.put(
+
operatorWrapper.getStreamOperator().getOperatorID(),
+ buildOperatorSnapshotFutures(
+ checkpointMetaData,
+ checkpointOptions,
+ operatorChain,
+ operatorWrapper.getStreamOperator(),
+ isCanceled,
+ channelStateWriteResult,
+ storage));
}
+
+ checkpointStorage.clearCacheFor(checkpointId);
Review comment:
Yes, will fix this.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
