Fix race in AbstractSearchAsyncAction request throttling #116264
Merged
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
We had a race here where the non-blocking pending execution would be starved of executing threads. This happened when all the current holders of permits from the semaphore would release their permit after a producer thread failed to acquire a permit and then enqueued its task. => need to peek the queue again after releasing the permit and try to acquire a new permit if there's work left to be done to avoid this scenario.
original-brownbear
added
>non-issue
v9.0.0
:Search Foundations/Search
elasticsearchmachine
added
the
Team:Search Foundations
|
Pinging @elastic/es-search-foundations (Team:Search Foundations) |
|
Maybe important to note, this fix is a stop-gap measure only to fix tests. A follow-up that is incoming will just simplify this to not be as much of an issue any longer by pre-computing the tasks per node and then running them (for the most part at least, some you can't pre-compute). |
javanna
reviewed
Nov 7, 2024
| } | ||
| if (firstError != null) { | ||
| fail(firstError.getMessage()); | ||
| assertHitCount(prepareSearch("test"), numOfDocs); |
There was a problem hiding this comment.
what the rationale behind the changes in this test? I get confused because I thought the code fix below would be enough to fix this test failure that gets unmuted in the same PR.
There was a problem hiding this comment.
Ah sorry I had this cleanup in my reproducer branch and kept it because the test was super dirty from before with the types removal leftover etc. Let me revert this real quick, I'll open a separate test cleanup :)
| } | ||
|
|
||
| private void executeAndRelease(Consumer<Releasable> task) { | ||
| while (task != null) { | ||
| do { |
There was a problem hiding this comment.
This is not strictly necessary but I figured it'd be nice to make it obvious that we never work on null here when entering the loop :)
|
All cleaned up! Thanks Luca :) |
javanna
approved these changes
Nov 7, 2024
|
Thanks Luca, I'll look into a reproducer UT :) |
kderusso
pushed a commit
to kderusso/elasticsearch
that referenced
this pull request
Nov 7, 2024
) We had a race here where the non-blocking pending execution would be starved of executing threads. This happened when all the current holders of permits from the semaphore would release their permit after a producer thread failed to acquire a permit and then enqueued its task. => need to peek the queue again after releasing the permit and try to acquire a new permit if there's work left to be done to avoid this scenario.
jozala
pushed a commit
that referenced
this pull request
Nov 13, 2024
We had a race here where the non-blocking pending execution would be starved of executing threads. This happened when all the current holders of permits from the semaphore would release their permit after a producer thread failed to acquire a permit and then enqueued its task. => need to peek the queue again after releasing the permit and try to acquire a new permit if there's work left to be done to avoid this scenario.
original-brownbear
added a commit
to original-brownbear/elasticsearch
that referenced
this pull request
Nov 24, 2024
) We had a race here where the non-blocking pending execution would be starved of executing threads. This happened when all the current holders of permits from the semaphore would release their permit after a producer thread failed to acquire a permit and then enqueued its task. => need to peek the queue again after releasing the permit and try to acquire a new permit if there's work left to be done to avoid this scenario.
elasticsearchmachine
pushed a commit
that referenced
this pull request
Nov 24, 2024
…117426) We had a race here where the non-blocking pending execution would be starved of executing threads. This happened when all the current holders of permits from the semaphore would release their permit after a producer thread failed to acquire a permit and then enqueued its task. => need to peek the queue again after releasing the permit and try to acquire a new permit if there's work left to be done to avoid this scenario.
original-brownbear
added a commit
to original-brownbear/elasticsearch
that referenced
this pull request
Nov 24, 2024
This is long fixed by elastic#116264
elasticsearchmachine
pushed a commit
that referenced
this pull request
Nov 24, 2024
original-brownbear
added a commit
to original-brownbear/elasticsearch
that referenced
this pull request
Nov 27, 2024
) We had a race here where the non-blocking pending execution would be starved of executing threads. This happened when all the current holders of permits from the semaphore would release their permit after a producer thread failed to acquire a permit and then enqueued its task. => need to peek the queue again after releasing the permit and try to acquire a new permit if there's work left to be done to avoid this scenario.
alexey-ivanov-es
pushed a commit
to alexey-ivanov-es/elasticsearch
that referenced
this pull request
Nov 28, 2024
) We had a race here where the non-blocking pending execution would be starved of executing threads. This happened when all the current holders of permits from the semaphore would release their permit after a producer thread failed to acquire a permit and then enqueued its task. => need to peek the queue again after releasing the permit and try to acquire a new permit if there's work left to be done to avoid this scenario.
alexey-ivanov-es
pushed a commit
to alexey-ivanov-es/elasticsearch
that referenced
this pull request
Nov 28, 2024
alexey-ivanov-es
pushed a commit
to alexey-ivanov-es/elasticsearch
that referenced
this pull request
Nov 28, 2024
This is long fixed by elastic#116264 Fixes elastic#115728
original-brownbear
restored the
fix-race-abstract-search-async-action
branch
elasticsearchmachine
pushed a commit
that referenced
this pull request
Dec 2, 2024
…117638) We had a race here where the non-blocking pending execution would be starved of executing threads. This happened when all the current holders of permits from the semaphore would release their permit after a producer thread failed to acquire a permit and then enqueued its task. => need to peek the queue again after releasing the permit and try to acquire a new permit if there's work left to be done to avoid this scenario.
This was referenced Dec 5, 2024
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.
We had a race here where the non-blocking pending execution would be starved of executing threads.
This happened when all the current holders of permits from the semaphore would release their permit after a producer thread failed to acquire a permit and then enqueued its task.
=> need to peek the queue again after releasing the permit and try to acquire a new permit if there's work left to be done to avoid this scenario.
Also, modernised the test a little. Ever since there's no more types the loop can be cut in half and run for twice as many iterations + the assertion is just a hit count assertion :)
non-issue as this hasn't been released yet.