[Umbrella] ☂️ Requirements for release 0.1.0

[alculquicondor]

Deadline: May 16th Kubecon EU

Issues that we need to complete to consider kueue ready for a first release:

• Match workload affinities with flavors Match workload affinity with capacity labels #3
• Single heap per Capacity Introduce a single heap for per ClusterQueue #87
• Consistent flavors in a cohort Flavors with matching names should have identical labels/taints #59
• Queue status Add info to Queue status #5
• Capacity status Add info to ClusterQueue status #7
• Event for unschedulable workloads Create an event and update workload status when failing to schedule a workload #91
• Capacity namespace selector Add NamespaceSelector to ClusterQueue #4
• Efficient requeuing Efficient re-queueing of unschedulable workloads #8
• User guide Add user guide #64
• Publish image Publish kueue in GCR #52

Nice to have:

• Add borrowing weight Replace borrowing ceiling with weight #62
• E2E test Add E2E test setup and first test #61
• Use kueue.sigs.k8s.io API group Graduate API to beta #23
• Support for one custom job Support kubeflow's MPIJob #65

[alculquicondor]

/kind feature

[alculquicondor]

cc @ahg-g @ArangoGutierrez @denkensk thoughts? Did I miss something?

[ArangoGutierrez]

Ohhhh that smell of a new release in the air.... uh lala

[denkensk]

User guide is also needed.

[ArangoGutierrez]

something we can quickly implement as well could be github-pages for small documentation? (I can volunteer have done that for other sigs)

[denkensk]

#14 Is there any progress on this？ @ArangoGutierrez

[ArangoGutierrez]

we agree to wait for klog/v3 no?

But this is changing soon https://github.com/kubernetes/enhancements/tree/master/keps/sig-instrumentation/3077-contextual-logging

In the meantime, if we stick with logr for now, I think it will be easier to switch to klog (v3?) once it supports contextual logging.

[ArangoGutierrez]

I would add #52 to that checklist

[ahg-g]

I don't think we should block the first release on #62, ceiling is workable for a 0.0.1 version .

[ArangoGutierrez]

can we set KubeCon EU as dead line for this? would be a nice thing to achieve

[ahg-g]

Yeah, I submitted a proposal to present it at KubeCon the general event, but will also be discussing it in the colo event.

[denkensk]

Great

[alculquicondor]

I think we should be able to have something sooner. But kubecon EU is a nice hard deadline :) Added to the description.

I don't think we should block the first release on #62, ceiling is workable for a 0.0.1 version .

I think it is indeed usable, but migrating to weights from the first version will avoid one breaking change. Sure... they are alpha APIs, but if we can avoid it, it would be great. Moved to nice-to-have.

[ArangoGutierrez]

@alculquicondor can you pin this? so is easier to find

[ahg-g]

I think we need to add #4 to the list, I will work on that.

[ahg-g]

also #91

[ArangoGutierrez]

check mark Publish image https://github.com/kubernetes-sigs/kueue/issues/52

[ArangoGutierrez]

Check Single heap per Capacity https://github.com/kubernetes-sigs/kueue/issues/87 ✅

[ArangoGutierrez]

we have an image!!!
gcr.io/k8s-staging-kueue/kueue:devel

[eduardo@fedora-workstation kueue]$ skopeo list-tags docker://gcr.io/k8s-staging-kueue/kueue
{
    "Repository": "gcr.io/k8s-staging-kueue/kueue",
    "Tags": [
        "devel"
    ]
}

Prow job: https://prow.k8s.io/view/gs/kubernetes-jenkins/logs/post-kueue-push-images/1502394011774619648

[ArangoGutierrez]

CHeckMark -> Queue status https://github.com/kubernetes-sigs/kueue/issues/5

[ArangoGutierrez]

Checkmark Event for unschedulable workloads #91

[ArangoGutierrez]

checkmark User guide https://github.com/kubernetes-sigs/kueue/issues/64

[alculquicondor]

@ahg-g @denkensk @ArangoGutierrez PTAL at the draft for release notes in the issue description.

[ArangoGutierrez]

can we use https://github.com/kubernetes-sigs/kueue/blob/main/.github/ISSUE_TEMPLATE/NEW_RELEASE.md ?

[alculquicondor]

Opened #197

[ahg-g]

A couple of other renames we should do:

1. cq.requestableResources -> resources
2. cq.flavor.resourceFlavor -> cq.flavor.name

[ahg-g]

Before the first release we need to ignore jobs with queue name not set; this is important so users are not surprised that their existing running jobs get suspended when they first try Kueue.

We will add a flag to ComponentConfig to disable this behavior, and a namespace selector to select which namespaces the job-controller will take over. We can do that in the next release.

[alculquicondor]

sgtm, can you open an issue and/or PR?

[ahg-g]

I will address the two comments above once the multiple api groups pr merges

[alculquicondor]

A couple of other renames we should do:

1. cq.requestableResources -> resources
2. cq.flavor.resourceFlavor -> cq.flavor.name

+1 to both

[ahg-g]

#205 disables reconciling jobs with no queue name set.

[ahg-g]

and #206 for renaming requestableResources in the api.

[ahg-g]

and #207 for component config

[alculquicondor]

All items are complete. We probably leave the nice-to-have issues for the next release.

/close

[k8s-ci-robot]

@alculquicondor: Closing this issue.

In response to this:

All items are complete. We probably leave the nice-to-have issues for the next release.

/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.