Fix datasets leaking worker processes due to closure capture of stats actor handle#22156
Merged
ericl merged 2 commits intoray-project:masterfrom Feb 7, 2022
Merged
Fix datasets leaking worker processes due to closure capture of stats actor handle#22156ericl merged 2 commits intoray-project:masterfrom
ericl merged 2 commits intoray-project:masterfrom
Conversation
2 tasks
clarkzinzow
approved these changes
Feb 7, 2022
| assert ray.get(actor.method.remote()) == 1 | ||
|
|
||
| # Tests datasets doesn't leak workers. | ||
| ray.data.range(100).map(lambda x: x).take() |
Contributor
There was a problem hiding this comment.
Is it possible to capture this in a unit test, or is it not worth it?
Contributor
Author
There was a problem hiding this comment.
It's a little tricky, so I punted on it for now.
scv119
approved these changes
Feb 7, 2022
rkooo567
added a commit
that referenced
this pull request
Feb 9, 2022
After this PR (#22156), for some reasons the driver script has some string that cannot be encoded with ascii. It seems like using utf-8 solves the problem.
simonsays1980
pushed a commit
to simonsays1980/ray
that referenced
this pull request
Feb 27, 2022
simonsays1980
pushed a commit
to simonsays1980/ray
that referenced
this pull request
Feb 27, 2022
After this PR (ray-project#22156), for some reasons the driver script has some string that cannot be encoded with ascii. It seems like using utf-8 solves the problem.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why are these changes needed?
It looks like an actor handle was closure captured in one of the remote tasks. This prevents the workers of those tasks from exiting once the job finishes, since the reference counter is seeing live references.
This PR changes Datasets to avoid this code path. Separately @stephanie-wang I think ReferenceCounter::DrainAndShutdown() may need to be changed to ignore references that don't have any remote borrowers (in this case, an entry in the function table holding an actor reference doesn't seem like a reason to hold up shutdown).
Related issue number
Closes #22154