Skip to content

[SPARK-20282][SS][Tests]Write the commit log first to fix a race contion in tests #17594

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from
Closed

[SPARK-20282][SS][Tests]Write the commit log first to fix a race contion in tests #17594

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/StreamExecution.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -304,8 +304,8 @@ class StreamExecution(
finishTrigger(dataAvailable)
if (dataAvailable) {
// Update committed offsets.
committedOffsets ++= availableOffsets
batchCommitLog.add(currentBatchId)
Copy link
Contributor

@marmbrus marmbrus Apr 10, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is existing, but do we not write to the commit log if there is no new data in the next batch?

Copy link
Member Author

@zsxwing zsxwing Apr 10, 2017 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it's because we only update currentBatchId when data is available. cc @tdas

Copy link
Contributor

@tdas tdas Apr 10, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Technically, when there is a batch with data, it finishes executing and then increments the batch id for the next batch. But if next batch has no data, then the batchid not further increment, and gets eventually used in a future batch with data.

So i think it is correct to update the commit log as soon this batch is done, and before the batch id is incremented for next batch. This change maintains that invariant as far as i think.

committedOffsets ++= availableOffsets
logDebug(s"batch ${currentBatchId} committed")
// We'll increase currentBatchId after we complete processing current batch's data
currentBatchId += 1
Expand Down