Skip to content

Conversation

@tillrohrmann
Copy link
Contributor

What is the purpose of the change

Due to changes how the slot futures are completed and due to the fact that the
ResultConjunctFuture does not maintain the order in which the futures were specified,
it could happen that executions were not deployed in topological order. This commit
fixes this problem by changing the ResultConjunctFuture so that it maintains the order
of the specified futures in its result collection.

This also fixes a performance regression when deploying tasks since deploying tasks in non-topological order can lead to additional result partition lookups.

Verifying this change

  • Added ExecutionGraphDeploymentTest#testExecutionGraphIsDeployedInTopologicalOrder and ConjunctFutureTest#testConjunctFutureCompletion.

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, Yarn/Mesos, ZooKeeper: (yes)
  • The S3 file system connector: (no)

Documentation

  • Does this pull request introduce a new feature? (no)
  • If yes, how is the feature documented? (not applicable)

@flinkbot
Copy link
Collaborator

flinkbot commented Mar 27, 2019

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.

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.

Details
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

Copy link
Contributor

@zentol zentol left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

one comment as to whether the ResultConjunctFuture constructor should only accept Lists, otherwise +1.

@flinkbot approve all

}
else {
int counter = 0;
for (CompletableFuture<? extends T> future : resultFutures) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Isn't it common to only accept a List if the order of elements is relevant?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I would not restrict the user on the type of collection he wants to use. The contract is that the result collection will have the same order as the input collection.

}

final Iterator<CompletableFuture<LogicalSlot>> slotIterator = slotFutures.iterator();
final SlotProvider slotProvider = new TestingSlotProvider(slotRequestId -> {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks like a basic use-case that we could move into a reusable FixedSizeTestingSlotProvider extends TestingSlotProvider class with a Collection<CompletableFuture<LogicalSlot>> getSlotFutures() method.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good point. I will add this utility and refactor the test.

Copy link
Contributor

@shuai-xu shuai-xu left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Cool, it is useful for us.

@tillrohrmann
Copy link
Contributor Author

Thanks for the review @zentol and @shuai-xu. Merging this PR.

…shipCall

Fix the race condition between executing EmbeddedLeaderService#GrantLeadershipCall
and a concurrent shutdown of the leader service by making GrantLeadershipCall not
accessing mutable state outside of a lock.

This closes apache#7935.
Due to changes how the slot futures are completed and due to the fact that the
ResultConjunctFuture does not maintain the order in which the futures were specified,
it could happen that executions were not deployed in topological order. This commit
fixes this problem by changing the ResultConjunctFuture so that it maintains the order
of the specified futures in its result collection.

This closes apache#8060.
@zentol zentol self-assigned this Mar 28, 2019
@asfgit asfgit closed this in 5e498dc Mar 29, 2019
@tillrohrmann tillrohrmann deleted the FLINK-12021 branch March 29, 2019 09:43
HuangZhenQiu pushed a commit to HuangZhenQiu/flink that referenced this pull request Apr 22, 2019
Due to changes how the slot futures are completed and due to the fact that the
ResultConjunctFuture does not maintain the order in which the futures were specified,
it could happen that executions were not deployed in topological order. This commit
fixes this problem by changing the ResultConjunctFuture so that it maintains the order
of the specified futures in its result collection.

This closes apache#8060.
sunhaibotb pushed a commit to sunhaibotb/flink that referenced this pull request May 8, 2019
Due to changes how the slot futures are completed and due to the fact that the
ResultConjunctFuture does not maintain the order in which the futures were specified,
it could happen that executions were not deployed in topological order. This commit
fixes this problem by changing the ResultConjunctFuture so that it maintains the order
of the specified futures in its result collection.

This closes apache#8060.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants