1996fanrui commented on code in PR #21131:
URL: https://github.com/apache/flink/pull/21131#discussion_r1009230664
##########
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/channel/ChannelStateWriteRequestDispatcherImpl.java:
##########
@@ -73,11 +79,27 @@ public void dispatch(ChannelStateWriteRequest request)
throws Exception {
private void dispatchInternal(ChannelStateWriteRequest request) throws
Exception {
if (request instanceof CheckpointStartRequest) {
checkState(
- !writers.containsKey(request.getCheckpointId()),
- "writer not found for request " + request);
- writers.put(request.getCheckpointId(),
buildWriter((CheckpointStartRequest) request));
+ request.getCheckpointId() > ongoingCheckpointId,
+ String.format(
+ "Checkpoint must be incremented,
ongoingCheckpointId is %s, but the request is %s.",
+ ongoingCheckpointId, request));
+ if (writer != null && !writer.isDone()) {
+ writer.fail(
+ new IllegalStateException(
+ String.format(
+ "Task[name=%s, subtaskIndex=%s] has
uncompleted channelState writer of checkpointId=%s, "
+ + "but it received a new
checkpoint start request of checkpointId=%s, it maybe "
+ + "a bug due to currently not
supported concurrent unaligned checkpoint.",
+ taskName,
+ subtaskIndex,
+ ongoingCheckpointId,
+ request.getCheckpointId())));
+ }
Review Comment:
After step2 `abort checkpoint 42`, `writer.isDone()` will be true. So here
won't throw exception during step3.
BTW, `writer` cannot be set to `null` during abort checkpoint. There may be
some `writeInput/writeOutput` after abort. If set to `null`,
`req.onWriterMissing();` will throw exception.
`writer` just be set to `null` when the checkpoint is completed, or
`ChannelStateWriteRequestDispatcherImpl` failed.
--
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]
