Skip to content
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.

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

[GLUTEN-7524][VL][UNIFFLE] Reset rss.row.based configuration of uniffle #7525

Merged
merged 3 commits into from
Oct 15, 2024

Conversation

wForget
Copy link
Member

@wForget wForget commented Oct 14, 2024

What changes were proposed in this pull request?

closes #7524

How was this patch tested?

@github-actions github-actions bot added the RSS label Oct 14, 2024
Copy link

#7524

@@ -57,6 +57,8 @@ public <K, V> ShuffleWriter<K, V> getWriter(
} else {
writeMetrics = context.taskMetrics().shuffleWriteMetrics();
}
sparkConf.set(
RssSparkConfig.SPARK_RSS_CONFIG_PREFIX + RssSparkConfig.RSS_ROW_BASED.key(), "false");
Copy link
Contributor

Choose a reason for hiding this comment

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

It seems we have already set rss.row.based to false while initlizing UniflleShuffleManager, do we still need this line?

Copy link
Member Author

Choose a reason for hiding this comment

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

Is it possible to create a fallbacked shuffle writer and then create a non-fallbacked shuffle writer?

Copy link
Member Author

Choose a reason for hiding this comment

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

For example, there are two stages running on the same executor, one shuffle without fallback and one with fallback

Copy link
Contributor

Choose a reason for hiding this comment

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

Good Catch. Just as @summaryzb said, this may cause some problem when there are two stages, with different rss.row.based, trying to create writer at the same time.

@@ -57,6 +57,7 @@ public <K, V> ShuffleWriter<K, V> getWriter(
} else {
writeMetrics = context.taskMetrics().shuffleWriteMetrics();
}
sparkConf.set(RssSparkConfig.SPARK_RSS_CONFIG_PREFIX + RssSparkConfig.RSS_ROW_BASED.key(), "false");
Copy link
Contributor

Choose a reason for hiding this comment

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

This conf change may effect other columnar write task

Copy link
Member Author

@wForget wForget Oct 14, 2024

Choose a reason for hiding this comment

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

Currently it only affects shuffle writes, I have reset it when creating shuffle writer

Copy link
Member Author

Choose a reason for hiding this comment

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

Oh, I see, how about we use a sparkconf.clone ?

Copy link
Contributor

Choose a reason for hiding this comment

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

Sounds good to me. @summaryzb any better ideas?

Copy link
Contributor

Choose a reason for hiding this comment

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

Sounds good

Copy link
Contributor

@jackylee-ch jackylee-ch left a comment

Choose a reason for hiding this comment

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

LGTM, thanks for your fix.

@jackylee-ch jackylee-ch merged commit b423de3 into apache:main Oct 15, 2024
43 checks passed
weiting-chen pushed a commit to weiting-chen/gluten that referenced this pull request Nov 19, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[VL][UNIFFLE] NPE occurs after uniffle shuffle writer fallback to row based
3 participants