[CELEBORN-1904] Cancel the stage running tasks on stage rerun #3144
+169
−12
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
turboFei
changed the title
turboFei
commented
Mar 11, 2025
| if (shuffleFetchFailureTaskIds != null) { | ||
| shuffleFetchFailureTaskIds.asScala.headOption.foreach { case taskId => | ||
| val taskSetManager = SparkUtils.getTaskSetManager(taskId) | ||
| if (taskSetManager != null && taskSetManager.runningTasks > 0) { |
There was a problem hiding this comment.
even we do not know whether the shuffleFetchFailureTaskIds related task did trigger the FetchFailedException eventually(depends on whether another task attempt is running or has been finished, see #2921).
But it should be safe to cancel the running tasks after checking the taskSetManager.runningTasks > 0.
turboFei
changed the title
turboFei
changed the title
|
cc @SteNicholas |
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Outdated suggestions cannot be applied.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.
What changes were proposed in this pull request?
On
SparkListenerStageCompletedevent, check whether the shuffle fetch failure reported in the stage, if that, cancel the running tasks due celeborn client will rerun the whole stage.Why are the changes needed?
If the task failed due to FetchFailed, dag scheduler would
markStageAsFinished.https://github.com/apache/spark/blob/3a872b7ca11faa128a2667de55f6dca13807057a/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala#L2022-L2060
But it will not cancel the running tasks in the stage.
For example, in below stage, a task failed due to fetch failure, and the stage duration is 39s.
However, it does not cancel the running tasks, the launched 2496 tasks keep running and the maximum task duration is 31 minutes.
It wastes a lot of compute resource.
For celeborn shuffle fetch failure, it will rerun the whole stage, so it is fine to cancel all the running tasks.
Does this PR introduce any user-facing change?
No.
How was this patch tested?
UT & Cluster testing.
The stage terminated quickly.
