[RLlib] Enhance node-failure tolerance. #50007
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
Signed-off-by: sven1977 <[email protected]>
sven1977
added
rllib
sven1977
commented
Jan 22, 2025
| def fetch_ready_async_reqs( | ||
| self, | ||
| *, | ||
| timeout_seconds: Optional[float] = 0.0, | ||
| return_obj_refs: bool = False, | ||
| mark_healthy: bool = True, | ||
| mark_healthy: bool = False, |
There was a problem hiding this comment.
Most important change here:
It's possible that an async result is still in the object store as the actor has crashed and been restarted. Fetching a valid result from such a restarted actor should NOT by default mark it as healthy. Only dedicated health checks, such as EnvRunnerGroup.probe_unhealthy_workers should set mark_healthy=True.
Signed-off-by: sven1977 <[email protected]>
…nce_spot_failure_support
simonsays1980
approved these changes
Jan 22, 2025
| # In this case, try each ref individually and collect only valid results. | ||
| try: | ||
| episodes = tree.flatten(ray.get(episode_refs)) | ||
| except ray.exceptions.OwnerDiedError: |
There was a problem hiding this comment.
Very nice! Didn't know about this error type
Signed-off-by: sven1977 <[email protected]>
…to enhance_spot_failure_support
win5923
pushed a commit
to win5923/ray
that referenced
this pull request
Jan 23, 2025
simonsays1980
pushed a commit
to simonsays1980/ray
that referenced
this pull request
Jan 23, 2025
anson627
pushed a commit
to anson627/ray
that referenced
this pull request
Jan 31, 2025
Signed-off-by: Anson Qian <[email protected]>
anson627
pushed a commit
to anson627/ray
that referenced
this pull request
Jan 31, 2025
Signed-off-by: Anson Qian <[email protected]>
srinathk10
pushed a commit
that referenced
this pull request
Feb 2, 2025
anyadontfly
pushed a commit
to anyadontfly/ray
that referenced
this pull request
Feb 13, 2025
Signed-off-by: Puyuan Yao <[email protected]>
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.
Enhance node-failure tolerance.
...worker...to..env_runner..for clarification.EnvRunnerGroup.fetch_ready_async_reqs(mark_healthy=)fromTruetoFalse, which is crucial for EnvRunner restoration detection (and the respective callback) insideAlgorithm.Learnertolerant againstray.exceptions.OwnerNotFounderrors in case an entire node goes down and the object reference to an episode list passed to Learner (or AggregatorActor) cannot beray.get()'d.Why are these changes needed?
Related issue number
Checks
git commit -s) in this PR.scripts/format.shto lint the changes in this PR.method in Tune, I've added it in
doc/source/tune/api/under thecorresponding
.rstfile.