☂️ Requirements for v0.5

[alculquicondor]

Accepted

Beta Give feedback

1. Synchronize batch/job.completions with parallelism in case of partial admission #970
   kind/feature +1
   
2. Extend the unit-test coverage for jobframework, #922
   kind/cleanup +1
   
3. Add metrics for ClusterQueue usage #972
   kind/feature +1
   
4. Workload priority #973
   kind/feature +1
   
5. Support kubeflow operator #297
   kind/feature lifecycle/frozen +2
   
6. A summary of Workloads order in queues #168
   kind/feature kind/ux lifecycle/frozen priority/backlog +4
   
7. Managing raw pods #976
   kind/feature +1
   
8. Preemption based on resource flavor order #582
   kind/feature +1
   
9. Introduce a plugin mechanism for arbitrary checks during workload admission. #993
   kind/feature +1
   

[alculquicondor]

This is very much work-in-progress.

Ideally, we should aim for a release in mid September.

I also acknowledge that there are a few feature requests and open KEPs for more policies. I want to discuss with folks which of those policies are most important and can be implemented in a maintainable way. Alternatively, we could look into interfaces that would allow folks to implement their own policies.

cc @tenzen-y @Gekko0114 @KunWuLuan @vsoch @ahg-g for feedback

[tenzen-y]

@alculquicondor Thanks for creating this issue. I would like to add #297.

I also acknowledge that there are a few feature requests an open KEPs for more policies. I want to discuss with folks which of those policies are most important and can be implemented in a maintainable way. Alternatively, we could look into interfaces that would allow folks to implement their own policies.

I agree. It would be nice to define interfaces for queueing and expose extensible points like the kube-scheduler scheduling framework.

[vsoch]

I was planning on starting the plugin soon, although I can't give a proper ETA until I dig in a bit. I will try to post an update on that soon.

[ahg-g]

A couple of features to consider:

1. Managing raw pods  #976
2. Partial preemption of workloads #975

[alculquicondor]

We probably should start with #976.

I don't think we can fit that and partial preemption in one release.

[ahg-g]

I agree that #976 is the higher priority

[KunWuLuan]

#963 can be a candidate I think
#849 I will work on this, try to complete before August

[alculquicondor]

I'm currently hesitant about #963 unless we answer the question about arbitrary hierarchies.

I think we can add #849 as a non-blocking feature for the release.

What do others think?

[alculquicondor]

See updated list. I'm leaving out #963 for now, as we already have a lot in our plates.
I think we can consider it for release 0.6

[tenzen-y]

LGTM

[tenzen-y]

@alculquicondor What are the remaining tasks? I guess it's only #297?

[alculquicondor]

I'm hoping we can also get #1136, as proof that two-stage admission works. But there are some delays in the cluster-autoscaler side.

Which APIs are we missing in #297?

[tenzen-y]

I'm hoping we can also get #1136, as proof that two-stage admission works. But there are some delays in the cluster-autoscaler side.

I see. I'm ok with including #1136 in this release. However, it might be better to cut a new release by KubeCon.

Which APIs are we missing in #297?

All APIs had been done.

[alculquicondor]

I agree, kubecon is an important deadline. Let's take it out if it's not done by the end of this week.

[tenzen-y]

I agree, kubecon is an important deadline. Let's take it out if it's not done by the end of this week.

SGTM

[tenzen-y]

Are these required PRs?

• ProvisioningRequest admission check controller #1154
• Validating admissionchecks #1220
• [webhook] Add admission check webhook #1232
• Two phase admission documentation #1181
• update document for FlavorFungibility #1226

[alculquicondor]

The last 2 (documentation) could be merge afterwards.
But yes, the list is accurate.

[alculquicondor]

I think they can all merge by Monday.

[tenzen-y]

SGTM

[kerthcet]

cc @B1F030

[alculquicondor]

🚀

/close

[k8s-ci-robot]

@alculquicondor: Closing this issue.

In response to this:

🚀

/close

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.