Prevent per-node scheduling fallback during batch scheduling retry #719
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Outdated suggestions cannot be applied.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.
Context
Currently, when a new job gets submitted, all the workers get scheduled in batch to have an all-or-nothing manner. However, the job actor heartbeat check will also try to re-schedule a worker if it's "stuck" in the allocation phase for too long (based on the worker-heartbeat-timeout config). Thus the batch scheduling gets invalidated after some timeout (which could be problematic for larger jobs when we need more time to get the requested resource allocated from the cluster auto scaler).
Behavior changes here:
Checklist
./gradlew build
compiles code correctly./gradlew test
passes all tests