[core][compiled graphs] Wait for monitor thread to join before shutting down #48808
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
This reverts commit bbc748a.
stephanie-wang
changed the title
rkooo567
approved these changes
Nov 20, 2024
kevin85421
reviewed
Nov 20, 2024
| monitor.teardown(kill_actors=kill_actors) | ||
| monitor.join(timeout=ctx.teardown_timeout) |
There was a problem hiding this comment.
Based on the PR description,
To prevent this from happening, this fix joins the monitor thread before destructing the CoreWorker.
I don't understand when the CoreWorker destruction happens and why calling monitor.join in __del__ ensures that the monitor thread joins before the CoreWorker is destructed.
I guess the CoreWorker will be terminated when ray.shutdown() is called.
There was a problem hiding this comment.
Yes, CoreWorker is destructed in ray.shutdown(), and the monitor thread could still be running at that point.
jecsand838
pushed a commit
to jecsand838/ray
that referenced
this pull request
Dec 4, 2024
…ng down (ray-project#48808) Each compiled graph starts a monitor thread to tear down the DAG upon detecting an error in one of the workers' task loops. Currently, during driver shutdown, this thread can live past the lifetime of the C++ CoreWorker. This causes a silent process exit when the thread later tries to call on the CoreWorker but it has already been destructed. To prevent this from happening, this fix joins the monitor thread *before* destructing the CoreWorker. ## Related issue number Closes ray-project#48288. --------- Signed-off-by: Stephanie Wang <smwang@cs.washington.edu> Signed-off-by: Connor Sanders <connor@elastiflow.com>
dentiny
pushed a commit
to dentiny/ray
that referenced
this pull request
Dec 7, 2024
…ng down (ray-project#48808) Each compiled graph starts a monitor thread to tear down the DAG upon detecting an error in one of the workers' task loops. Currently, during driver shutdown, this thread can live past the lifetime of the C++ CoreWorker. This causes a silent process exit when the thread later tries to call on the CoreWorker but it has already been destructed. To prevent this from happening, this fix joins the monitor thread *before* destructing the CoreWorker. ## Related issue number Closes ray-project#48288. --------- Signed-off-by: Stephanie Wang <smwang@cs.washington.edu> Signed-off-by: hjiang <dentinyhao@gmail.com>
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.
Why are these changes needed?
Each compiled graph starts a monitor thread to tear down the DAG upon detecting an error in one of the workers' task loops. Currently, during driver shutdown, this thread can live past the lifetime of the C++ CoreWorker. This causes a silent process exit when the thread later tries to call on the CoreWorker but it has already been destructed. To prevent this from happening, this fix joins the monitor thread before destructing the CoreWorker.
Related issue number
Closes #48288.