Skip to content

Conversation

@gedeh
Copy link

@gedeh gedeh commented May 22, 2021

What changes were proposed in this pull request?

Backporting changes from #26331 to Spark 2.4, in a similar fashion, executor's bindAddress is passed as an input parameter for RPCEnv.create.

This PR also ensure YarnAllocator use the config when allocating new executor and use default hostname when its not configured.

Why are the changes needed?

Recently we've came across this issue with Spark running on Yarn in Istio enabled Kubernetes cluster. As I understand it, Spark 2.4 is a stepping stone for teams moving away from Scala 2.11 to Scala 2.12 and then moving to Spark 3.

Does this PR introduce any user-facing change?

Yes, new config is added and relevant doc is updated.

How was this patch tested?

I will run this on Kubenetes and Istio.

@AmplabJenkins
Copy link

Can one of the admins verify this patch?

Copy link
Member

@srowen srowen left a comment

Choose a reason for hiding this comment

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

This could be OK, but I'm not sure there will be any more 2.4.x releases after 2.4.8 anyway

@gedeh gedeh force-pushed the executor-bind-address branch from 24b4511 to 99c360c Compare May 22, 2021 16:01
@gedeh gedeh changed the title [WIP] [SPARK-29670][core] Make executor bind address configurable for Spark 2.4 [SPARK-29670][CORE][2.4] Make executor bind address configurable for Spark 2.4 May 22, 2021
@gedeh gedeh marked this pull request as ready for review May 22, 2021 16:31
@gedeh gedeh requested a review from srowen May 22, 2021 16:33
@gedeh
Copy link
Author

gedeh commented May 22, 2021

This could be OK, but I'm not sure there will be any more 2.4.x releases after 2.4.8 anyway

Ow, I didn't aware of this. That's explains 2.4.9-SNAPSHOT changed to 2.4.8 in pom.xml

Copy link
Member

@dongjoon-hyun dongjoon-hyun left a comment

Choose a reason for hiding this comment

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

Thank you for making a PR, @gedeh .

However, this PR is invalid in a few ways.

  1. SPARK-29670 is closed as Duplicated. You cannot ask a backport of unmerged JIRA.
  2. The actual merged JIRA for feature is SPARK-24203 which was an Improvement at Apache Spark 3.0.0. Apache Spark community have a general policy which allows bug-fix backporting only. So, we cannot backport a new feature because we need to protect old branches.
  3. Lastly, 2.4.8 was the EOL release already. Here is some explanation for that.

The last minor release within a major a release will typically be maintained for longer as an “LTS” release. For example, 2.4.0 was released in November 2nd 2018 and had been maintained for 31 months until 2.4.8 was released on May 2021. 2.4.8 is the last release and no more 2.4.x releases should be expected even for bug fixes.

Given that the above reasons, could you close this PR, @gedeh ?

@gedeh
Copy link
Author

gedeh commented May 23, 2021

Thank you @dongjoon-hyun for clarification. I wasn't aware of EOL of Spark 2.4 closing this PR. Thanks!

@gedeh gedeh closed this May 23, 2021
@dongjoon-hyun
Copy link
Member

Thank you so much, @gedeh !

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.

4 participants