Make testRetriesCantBeExecutedForTooManyRequestMessages more reliable #1781
Conversation
This file contains hidden or 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
Motivation: The 'testRetriesCantBeExecutedForTooManyRequestMessages' test wedges infrequently. After some diagnosis this appears to be due to the intersection of two bugs. The first is that when yielding an element to the sequence, if some consumers exist, of which a subset are waiting for the next element and the remainder are not 'slow', the continuations of the fastest consumers would not be resumed when a new element was added. The other bug is a timing issue: when waiting for the next element a subscriber may be told to suspend. However, the subscriber must drop and then reacquire the lock to store the continuation. The state wasn't re-checked on storing the continuation which opened up a window where the element may have become present between asking for it and storing the continuation. Modifications: - Don't release the lock between attempting to consume an element and suspending to wait for it. - Resume waiting consumers more frequently. - Make a few tests less flaky Result: Test didn't wedge in 10k iterations
gjcairo
approved these changes
Jan 30, 2024
glbrntt
added a commit
to glbrntt/grpc-swift
that referenced
this pull request
Feb 5, 2024
…grpc#1781) Motivation: The 'testRetriesCantBeExecutedForTooManyRequestMessages' test wedges infrequently. After some diagnosis this appears to be due to the intersection of two bugs. The first is that when yielding an element to the sequence, if some consumers exist, of which a subset are waiting for the next element and the remainder are not 'slow', the continuations of the fastest consumers would not be resumed when a new element was added. The other bug is a timing issue: when waiting for the next element a subscriber may be told to suspend. However, the subscriber must drop and then reacquire the lock to store the continuation. The state wasn't re-checked on storing the continuation which opened up a window where the element may have become present between asking for it and storing the continuation. Modifications: - Don't release the lock between attempting to consume an element and suspending to wait for it. - Resume waiting consumers more frequently. - Make a few tests less flaky Result: Test didn't wedge in 10k iterations
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Motivation:
The 'testRetriesCantBeExecutedForTooManyRequestMessages' test wedges infrequently. After some diagnosis this appears to be due to the intersection of two bugs. The first is that when yielding an element to the sequence, if some consumers exist, of which a subset are waiting for the next element and the remainder are not 'slow', the continuations of the fastest consumers would not be resumed when a new element was added.
The other bug is a timing issue: when waiting for the next element a subscriber may be told to suspend. However, the subscriber must drop and then reacquire the lock to store the continuation. The state wasn't re-checked on storing the continuation which opened up a window where the element may have become present between asking for it and storing the continuation.
Modifications:
Result:
Test didn't wedge in 10k iterations