[RLlib] Don't sync (old API stack) connector states after worker restore. #47021
+42
−15
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
…gorithm.restore_workers) directly after their recovery. Signed-off-by: sven1977 <[email protected]>
…gorithm.restore_workers) directly after their recovery. Signed-off-by: sven1977 <[email protected]>
…rker's state after a worker has been restored and before it is synched with the local worker's state. The re-creation of the connectors on the already instantiated (restored) new worker broke things, b/c the AgentCollector that one of the "agent" connector pieces points to has already undergone some change and is then re-created, which breaks the stored vf-pred value (and other "extra fetches"). Signed-off-by: sven1977 <[email protected]>
Signed-off-by: sven1977 <[email protected]>
sven1977
requested review from
ArturNiederfahrenhorst and
simonsays1980
as code owners
simonsays1980
approved these changes
Aug 16, 2024
| @@ -1073,6 +1073,7 @@ def _training_step_old_api_stack(self): | |||
| if self._aggregator_actor_manager: | |||
| self._aggregator_actor_manager.probe_unhealthy_actors( | |||
| timeout_seconds=self.config.env_runner_health_probe_timeout_s, | |||
| mark_healthy=True, | |||
There was a problem hiding this comment.
Does this only look unformatted?
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Outdated suggestions cannot be applied.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.
This PR changes:
mark_healthymethod args are set to False. We do not want to mark any actors back from unhealthy to healthy, unless a successful "probe" request has been sent in a controlled fashion by the Algorithm.EnvRunnerGroup.probe_unhealthy_actorsnow manually set this arg to True. There are only 2 instances of this call: algorithm.py and in IMPALA (for the aggregation worker set).Algorithm.restore_workers()).TODO: File issue.
TODO: Add a repro script as test case for the issue that originally led us to investigate this issue.
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.