Skip to content

Conversation

@ahshahid
Copy link

What changes were proposed in this pull request?

Why are the changes needed?

Does this PR introduce any user-facing change?

How was this patch tested?

Was this patch authored or co-authored using generative AI tooling?

ashahid added 30 commits February 20, 2025 14:19
…in retrying all partitions in case of indeterministic shuffle keys
…on which will fail if all partitions are not retried for indeterministic shuffle stage
…minism via curried constructor of AttributeReference, as per feedback. Renamed the boolean
…minism via curried constructor of AttributeReference, as per feedback. Renamed the boolean
… relying on concurrent processing of DagEvent, instead relying on concurrent addition of event to the event loop queue
… always be in a single thread of EventLoop and that when ResubmitFailedStages event is submitted, though it is done by MessageScheduler ( a separate threadpool, but it is still doing post to the event loop's queue instead of direct onRecieve invocation, the read/write locks in stage are not needed
@ahshahid ahshahid changed the title Spark 51272 51016 combined [WIP]: Spark 51272 51016 combined: For testing of HA Test Mar 13, 2025
ashahid added 23 commits March 13, 2025 11:15
…( stage submitted and stage completed to kill the executors)
…tions is not equal to total partitions and resultstage is inDeterminate. Though with the current PR, this is not expected, unless there still remains a window of race.
…dencies of ResultStage on two Shuffle Stages, instead of the usual transitive dependency of shuffle stage2 on shuffle stage1
@ahshahid ahshahid closed this by deleting the head repository Jun 4, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant