[Data][LLM] Remove DataContext overrides in Ray Data LLM Processor#60142
Merged
kouroshHakha merged 4 commits intoray-project:masterfrom Jan 14, 2026
Merged
Conversation
Issue ray-project#53124 (GPU resource leakage) is now fixed. Remove the workaround that set wait_for_min_actors_s=600. This restores default Ray Data behavior where processing starts as soon as any actor is ready, rather than blocking for min_actors. Signed-off-by: Nikhil Ghosh <nikhil@anyscale.com>
Verify that: - Processor does not override wait_for_min_actors_s (default preserved) - User-set wait_for_min_actors_s values are preserved - Concurrency config correctly maps to ActorPoolStrategy min/max size Signed-off-by: Nikhil Ghosh <nikhil@anyscale.com>
Resource cleanup issues are now fixed (Issue ray-project#53169) including actor __del__ invocation Signed-off-by: Nikhil Ghosh <nikhil@anyscale.com>
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Code Review
This pull request correctly removes two workarounds in the Ray Data LLM Processor that are no longer necessary due to upstream fixes. The removal of the hardcoded wait_for_min_actors_s and _enable_actor_pool_on_exit_hook overrides simplifies the code and restores the default, non-blocking behavior. The newly added tests are comprehensive, effectively verifying that the processor no longer overrides these DataContext settings and that the concurrency configuration is correctly passed through to the ActorPoolStrategy. The changes are clean and well-tested.
Signed-off-by: Nikhil Ghosh <nikhil@anyscale.com>
nrghosh
added
data
kouroshHakha
approved these changes
Jan 14, 2026
kouroshHakha
changed the title
jeffery4011
pushed a commit
to jeffery4011/ray
that referenced
this pull request
Jan 20, 2026
…ay-project#60142) Signed-off-by: Nikhil Ghosh <nikhil@anyscale.com> Signed-off-by: jeffery4011 <jefferyshen1015@gmail.com>
ryanaoleary
pushed a commit
to ryanaoleary/ray
that referenced
this pull request
Feb 3, 2026
…ay-project#60142) Signed-off-by: Nikhil Ghosh <nikhil@anyscale.com>
peterxcli
pushed a commit
to peterxcli/ray
that referenced
this pull request
Feb 25, 2026
…ay-project#60142) Signed-off-by: Nikhil Ghosh <nikhil@anyscale.com> Signed-off-by: peterxcli <peterxcli@gmail.com>
peterxcli
pushed a commit
to peterxcli/ray
that referenced
this pull request
Feb 25, 2026
…ay-project#60142) Signed-off-by: Nikhil Ghosh <nikhil@anyscale.com> Signed-off-by: peterxcli <peterxcli@gmail.com>
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
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.
Summary
Remove two DataContext workarounds in Ray Data LLM that are no longer needed:
wait_for_min_actors_s = 600override (issue [Data] GPU resource leakage after ray.data.llm pipeline is terminated #53124 closed)_enable_actor_pool_on_exit_hook = Trueoverride (issue [Core] Make sure Actor's__del__method invoked on Actor's destruction #53169 closed)Behavior Change
Previously, Ray Data LLM hardcoded
wait_for_min_actors_s = 600, which caused blocking behavior:concurrency=N: blocked until all N actors were readyconcurrency=(1, N): blocked until 1 actor was readyAfter this change,
wait_for_min_actors_sstays at default (-1), so:wait_for_min_actors_smanually if they want blocking/timeout behaviorRelated
concurrency=(N,N)starts processing with 1 actor instead of waiting for N #60124Reproduction / Proof
See: https://gist.github.com/nrghosh/68d63040e92b82987c67e4dee6c8f40f
Test Plan
wait_for_min_actors_s