Skip to content

Conversation

@Yikun
Copy link
Member

@Yikun Yikun commented Mar 21, 2022

What changes were proposed in this pull request?

Add volcano module into release-build.sh

Why are the changes needed?

Add volcano module into release-build.sh

Does this PR introduce any user-facing change?

No

How was this patch tested?

ci passed

@Yikun
Copy link
Member Author

Yikun commented Mar 21, 2022

We added volcano module before, and will introduced spark k8s with volcano support in 3.3.

@dongjoon-hyun @HyukjinKwon @holdenk I'm not very familar with release process, but I think we need this PR to make sure volcano module can be published in final release bin/tar?

@Yikun Yikun changed the title [WIP][K8S] Add volcano module to release-build.sh [SPARK-38621][K8S][BUILD][INFRA] Add volcano module to release-build.sh Mar 22, 2022
@Yikun Yikun marked this pull request as ready for review March 22, 2022 03:16
@Yikun
Copy link
Member Author

Yikun commented Mar 23, 2022

Any comments?

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.

For Apache Spark 3.3, -1 from my side. And, even for master branch, since some release managers still uses master branch release script, I believe we need to hold on this until Apache Spark 3.3.0 is released.

If you don't mind, let's talk later for Apache Spark 3.4 after 3.3.0 release.

@dongjoon-hyun
Copy link
Member

dongjoon-hyun commented Mar 23, 2022

To reviewers. It was intentional not to expose this module into the Apache Spark release artifacts.

While I've been helping this SPIP, there were many issues in Volcano library and volcano module until the last day of branch cut. I'm also not sure about the stability yet. This module and configurations are re-designed to isolate each custom scheduler from the others. This is still in an experiment stage and I believe we need more time in this area. AFAIK, we couldn't add Yunikorn due to the lack of community bandwidth. Hence, we should not include this into binary artifacts in Apache Spark 3.3. After stabilizing more, we can add it at Apache Spark 3.4 officially.

@Yikun
Copy link
Member Author

Yikun commented Mar 24, 2022

@dongjoon-hyun

there were many issues in Volcano library and volcano module until the last day of branch cut.

To be clear, most of these problems because volcano branch selection, recently, we are also continuously verifying the work of volcano and spark, it's stable now.

This is still in an experiment stage and I believe we need more time in this area.

Yes, for sure, we specifically say this in the documentation. For users who want use this experimental feature, what's the best practice? Sounds like we need to mention build release with -Pvolcano in #35870 ?

AFAIK, we couldn't add Yunikorn due to the lack of community bandwidth.

In fact, I have been getting in touch with @yangwwei and working together to make yunikorn integrated better with spark. I will also keeping help yunikorn and spark for further integration in future to make community bandwidth expand a little bit.

@dongjoon-hyun
Copy link
Member

dongjoon-hyun commented Mar 24, 2022

No, there was no available release version, wasn't it. Apache Spark 3.3.0 consumes a set of specific released versions, not branch, @Yikun

To be clear, these problems were most issues because volcano branch selection.

For Yunikorn, thanks. I believe they can follow our latest architecture easily if they start.

@dongjoon-hyun
Copy link
Member

dongjoon-hyun commented Mar 24, 2022

I know that Volcano also has a long dev history. That's the reason why I was surprised during supporting you. We don't know what we are going to hit more in the future until we try. I'm expecting that Apache Spark 3.3.0 will give us more broader and world-wide feedback and help us. What we need is a safe step to introduce a new feature.

we are also continuously verifying the work of volcano and spark, it's stable now.

@Yikun
Copy link
Member Author

Yikun commented Mar 24, 2022

I'm expecting that Apache Spark 3.3.0 will give us more broader and world-wide feedback and help us.

@dongjoon-hyun Totally, agree. then I need to added the some doc like build with -Pvolcano in #35870 .

To be a double confirmed, this doc PR can be backport to 3.3 after getting merged, right? To make sure users can know how to use this experimental features in 3.3.

@dongjoon-hyun
Copy link
Member

Yep, that's right. And, it's up to the author and the committer of that PR. You can ping her.

@dongjoon-hyun
Copy link
Member

BTW, your previous doc PR is also not in branch-3.3, isn't it?

@Yikun
Copy link
Member Author

Yikun commented Mar 24, 2022

BTW, your previous doc PR is also not in branch-3.3, isn't it?

Yep, looks like it is: the #35869 is only landing on master, could you help it? Or I can submit a PR to branch-3.3 (backport to 3.3: #35955).

@github-actions
Copy link

github-actions bot commented Jul 3, 2022

We're closing this PR because it hasn't been updated in a while. This isn't a judgement on the merit of the PR in any way. It's just a way of keeping the PR queue manageable.
If you'd like to revive this PR, please reopen it and ask a committer to remove the Stale tag!

@github-actions github-actions bot added the Stale label Jul 3, 2022
@github-actions github-actions bot closed this Jul 4, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants