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

[BUG] ClusterLoadFallbackPolicy is not strictness when a shuffle with big partitions to register #30

Merged
merged 16 commits into from
Jan 26, 2022

Conversation

TonyDoen
Copy link
Contributor

@TonyDoen TonyDoen commented Jan 4, 2022

[BUG] ClusterLoadFallbackPolicy is not strictness when a shuffle with big partitions to register

What changes were proposed in this pull request?

If a shuffle with 5w partitions to register, and cluster has 3w but the slots usage is not in high load status, 5w allocation slots will failed.

Why are the changes needed?

This change will avoid this circumstance that a shuffle that it will cost too much slots over clusters' continues process of RSS handling

What are the items that need reviewer attention?

Related issues.

#17

Related pull requests.

How was this patch tested?

/cc @FMX

/assign @wangshengjie

@CLAassistant
Copy link

CLAassistant commented Jan 4, 2022

CLA assistant check
All committers have signed the CLA.

@@ -31,7 +31,8 @@ class RssShuffleFallbackPolicyRunner(sparkConf: SparkConf) extends Logging {
def applyAllFallbackPolicy(dependency: ShuffleDependency[_, _, _],
Copy link
Contributor

Choose a reason for hiding this comment

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

dependency.partitioner.numPartitions seems enough, could you please help to optimize this?Thanks

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Yes, I have retrieved the param [dependency] from this function

private def handleGetClusterLoadStatus(context: RpcCallContext): Unit = {
val (_, _, _, result) = getClusterLoad
private def handleGetClusterLoadStatus(context: RpcCallContext, numPartitions: Int): Unit = {
val (_, _, _, result) = getClusterLoad(numPartitions)
Copy link
Contributor

Choose a reason for hiding this comment

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

Why we need so much return values, please remove useless return

Copy link
Contributor Author

@TonyDoen TonyDoen Jan 4, 2022

Choose a reason for hiding this comment

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

The [getClusterLoad] is common function, we may need the top three return-values in the other place. I can separate the value through creating new function.

@FMX
Copy link
Contributor

FMX commented Jan 6, 2022

LGTM

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[BUG] ClusterLoadFallbackPolicy is not strictness when a shuffle with big partitions to register
5 participants