Fix a deadlock when queued tasks are resubmitted quickly in succession #7348
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
fjetter
requested review from
crusaderky and
hendrikmakait
and removed request for
crusaderky
Unit Test ResultsSee test report for an extended history of previous test failures. This is useful for diagnosing flaky tests. 15 files ± 0 15 suites ±0 6h 29m 15s ⏱️ + 14m 34s For more details on these failures, see this check. Results for commit 9026d47. ± Comparison against base commit cff33d5. ♻️ This comment has been updated with latest results. |
hendrikmakait
approved these changes
Nov 25, 2022
There was a problem hiding this comment.
LGTM, great job on the description of what the test is doing. Thanks, @fjetter!
hendrikmakait
requested changes
Nov 25, 2022
ahhh.. thanks, that makes sense. I forgot about the additional CI config and was already concerned |
hendrikmakait
approved these changes
Nov 25, 2022
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Closes #7200
There is a race condition if
task-finishesandfree-keysare submitted concurrently.Queued tasks could end up being transitioned to memory which is wrong because shortly after this the worker will have forgotten the data already.
Sending another free-keys in this situation is not absolutely necessary but safe since the scheduler guarantees ordering of messages to a worker, i.e. if the task is in queued there is no other worker supposed to have or compute this task until it is transitioned out of this state. The free-keys is just there for good measure and will be handled on worker side gracefully.
I added a test for processing as well to be on the safe side. This isn't asserted but the worker just computes the task twice, as it should from our "release and compute task" semantics.
The test is a bit involved due to how we define rootishness but I added hopefully sufficient commentary.