[FLINK-20331][checkpointing][task] Don't fail the task if unaligned checkpoint was subsumed #14218

rkhachatryan · 2020-11-25T12:40:01Z

What is the purpose of the change

A check was added recently that the RemoteInputChannel.lastBarrierSequenceNumber was not overwritten by a newer barrier.
The check itself is valid, however the task shouldn't fail in that case.

Relying on the assumption of at most one active unaligned checkpoint at a time, the older checkpoint should simply be declined.
Cleanup is needed for (all) channels, barrier controller, handler and checkpoint writer. Therefore, abort in SingleCheckpointBarrierHandler should be used.

Verifying this change

This change is already covered by existing test: UnalignedCheckpointITCase.parallel pipeline with mixed channels, p = 20

Does this pull request potentially affect one of the following parts:

Dependencies (does it add or upgrade a dependency): no
The public API, i.e., is any changed class annotated with @Public(Evolving): no
The serializers: no
The runtime per-record code paths (performance sensitive): no
Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Kubernetes/Yarn/Mesos, ZooKeeper: no
The S3 file system connector: no

Documentation

Does this pull request introduce a new feature? no
If yes, how is the feature documented? no

flinkbot · 2020-11-25T12:42:11Z

Thanks a lot for your contribution to the Apache Flink project. I'm the @flinkbot. I help the community
to review your pull request. We will use this comment to track the progress of the review.

Automated Checks

Last check on commit 1109ad2 (Wed Nov 25 12:42:11 UTC 2020)

Warnings:

No documentation files were touched! Remember to keep the Flink docs up to date!

_{Mention the bot in a comment to re-run the automated checks.}

Review Progress

❓ 1. The [description] looks good.
❓ 2. There is [consensus] that the contribution should go into to Flink.
❓ 3. Needs [attention] from.
❓ 4. The change fits into the overall [architecture].
❓ 5. Overall code [quality] is good.

Please see the Pull Request Review Guide for a full explanation of the review process.

The Bot is tracking the review progress through labels. Labels are applied according to the order of the review items. For consensus, approval by a Flink committer of PMC member is required

Bot commands

The @flinkbot bot supports the following commands:

@flinkbot approve description to approve one or more aspects (aspects: description, consensus, architecture and quality)
@flinkbot approve all to approve all aspects
@flinkbot approve-until architecture to approve everything until architecture
@flinkbot attention @username1 [@username2 ..] to require somebody's attention
@flinkbot disapprove architecture to remove an approval you gave earlier

rkhachatryan · 2020-11-25T12:43:51Z

...src/main/java/org/apache/flink/runtime/io/network/partition/consumer/RemoteInputChannel.java

 		assert Thread.holdsLock(receivedBuffers);

+		if (checkpointId < lastBarrierId) {
+			throw new CheckpointException(


I was in doubt whether this exception type is appropriate here or whether an Optional should be returned.
But at the upper level, it ultimately should be used, and the reason is only known here (so it would be harder to analyze the exception/empty buffers later).

I think this is exactly the case for which CheckpointException was designed.

rkhachatryan · 2020-11-25T12:45:16Z

...java/src/main/java/org/apache/flink/streaming/runtime/io/SingleCheckpointBarrierHandler.java

@@ -158,14 +163,16 @@ public void processBarrierAnnouncement(
 	@Override
 	public void processCancellationBarrier(CancelCheckpointMarker cancelBarrier) throws IOException {
 		final long cancelledId = cancelBarrier.getCheckpointId();
-		if (currentCheckpointId > cancelledId || (currentCheckpointId == cancelledId && numBarriersReceived == 0)) {
-			return;
+		if (cancelledId >= currentCheckpointId && (cancelledId > currentCheckpointId || numBarriersReceived > 0)) {


Please double-check this @AHeise :)

It's correct but it might be easier to do

cancelledId > currentCheckpointId || (currentCheckpointId == cancelledId && numBarriersReceived > 0)

So either it's a future checkpoint, or it's the current checkpoint and not yet canceled.

rkhachatryan · 2020-11-25T12:47:27Z

...java/src/main/java/org/apache/flink/streaming/runtime/io/SingleCheckpointBarrierHandler.java

 		// by setting the currentCheckpointId to this checkpoint while keeping the numBarriers
 		// at zero means that no checkpoint barrier can start a new alignment
-		currentCheckpointId = cancelledId;
+		currentCheckpointId = Math.max(cancelledId, currentCheckpointId);


This is not necessary currently.
But because I extracted this method the assignment became more error-prone, so I added this check.

AHeise

Thank you very much. The fix looks very good.
I made a suggestion on the more complicated condition.

AHeise · 2020-11-25T13:01:58Z

...java/src/main/java/org/apache/flink/streaming/runtime/io/SingleCheckpointBarrierHandler.java

@@ -158,14 +163,16 @@ public void processBarrierAnnouncement(
 	@Override
 	public void processCancellationBarrier(CancelCheckpointMarker cancelBarrier) throws IOException {
 		final long cancelledId = cancelBarrier.getCheckpointId();
-		if (currentCheckpointId > cancelledId || (currentCheckpointId == cancelledId && numBarriersReceived == 0)) {
-			return;
+		if (cancelledId >= currentCheckpointId && (cancelledId > currentCheckpointId || numBarriersReceived > 0)) {


It's correct but it might be easier to do

cancelledId > currentCheckpointId || (currentCheckpointId == cancelledId && numBarriersReceived > 0)

So either it's a future checkpoint, or it's the current checkpoint and not yet canceled.

flinkbot · 2020-11-25T13:04:20Z

CI report:

cfb594a Azure: SUCCESS

Bot commands

The @flinkbot bot supports the following commands:

@flinkbot run travis re-run the last Travis build
@flinkbot run azure re-run the last Azure build

…heckpoint was subsumed

rkhachatryan changed the title ~~[FLINK-20331][checkpointing][task] Don't fail the task if unaligned c…~~ [FLINK-20331][checkpointing][task] Don't fail the task if unaligned checkpoint was subsumed Nov 25, 2020

rmetzger added the review=description? label Nov 25, 2020

rkhachatryan commented Nov 25, 2020

View reviewed changes

AHeise approved these changes Nov 25, 2020

View reviewed changes

rkhachatryan force-pushed the f20331 branch from 1109ad2 to 34bf5e7 Compare November 25, 2020 13:23

rmetzger added the component=Runtime/Checkpointing label Nov 25, 2020

[FLINK-20331][checkpointing][task] Don't fail the task if unaligned c…

cfb594a

…heckpoint was subsumed

rkhachatryan force-pushed the f20331 branch from 34bf5e7 to cfb594a Compare November 25, 2020 18:34

AHeise merged commit 50af0b1 into apache:master Nov 26, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[FLINK-20331][checkpointing][task] Don't fail the task if unaligned checkpoint was subsumed #14218

[FLINK-20331][checkpointing][task] Don't fail the task if unaligned checkpoint was subsumed #14218

rkhachatryan commented Nov 25, 2020

flinkbot commented Nov 25, 2020

rkhachatryan Nov 25, 2020

AHeise Nov 25, 2020

rkhachatryan Nov 25, 2020

AHeise Nov 25, 2020 •

edited

rkhachatryan Nov 25, 2020

rkhachatryan Nov 25, 2020

AHeise left a comment

AHeise Nov 25, 2020 •

edited

flinkbot commented Nov 25, 2020 •

edited

[FLINK-20331][checkpointing][task] Don't fail the task if unaligned checkpoint was subsumed #14218

[FLINK-20331][checkpointing][task] Don't fail the task if unaligned checkpoint was subsumed #14218

Conversation

rkhachatryan commented Nov 25, 2020

What is the purpose of the change

Verifying this change

Does this pull request potentially affect one of the following parts:

Documentation

flinkbot commented Nov 25, 2020

Automated Checks

Review Progress

rkhachatryan Nov 25, 2020

Choose a reason for hiding this comment

AHeise Nov 25, 2020

Choose a reason for hiding this comment

rkhachatryan Nov 25, 2020

Choose a reason for hiding this comment

AHeise Nov 25, 2020 • edited

Choose a reason for hiding this comment

rkhachatryan Nov 25, 2020

Choose a reason for hiding this comment

rkhachatryan Nov 25, 2020

Choose a reason for hiding this comment

AHeise left a comment

Choose a reason for hiding this comment

AHeise Nov 25, 2020 • edited

Choose a reason for hiding this comment

flinkbot commented Nov 25, 2020 • edited

CI report:

AHeise Nov 25, 2020 •

edited

AHeise Nov 25, 2020 •

edited

flinkbot commented Nov 25, 2020 •

edited