Add phase/conditions into PodGroup.Status

[k82cn]

Is this a BUG REPORT or FEATURE REQUEST?:

/kind feature

Description:

Currently, kube-batch only raised Unschedulable event for its status; it's better to include phase and conditions in PodGroup.Status to give more detail.

Tasks

• Design Doc (The design doc of PodGroup Phase in Status. #533)
• Add phase/condition into PodGroup.Status CRD (Added Phase and Conditions to PodGroup.Status struct #525)
• Updated phase/condition by kube-batch (Update PodGroup status. #560)

[Zyqsempai]

@k82cn I would like to work on this one, but can you please provide more details.
As far as I can see, currently we have that type of status for PodGroup:
https://github.com/kubernetes-sigs/kube-batch/blob/master/pkg/apis/scheduling/v1alpha1/types.go#L50
You want to extend PodGroupStatus struct and add few fields for additional info?

[k82cn]

You want to extend PodGroupStatus struct and add few fields for additional info?

Yes, we need to add PodGroupPhase and Conditions at least; so the operator/controller knows where it is in its lifecycle. We can refer to PodStatus :)

[Zyqsempai]

Got it, I will take care of it!

[Zyqsempai]

@k82cn Hey, so I opened PR, but I still have some questions, I am not sure how and where we will fill up those new fields, also, I created them based on same Pod.Status fields, not sure that every value is acceptable for PodGroup.

[k82cn]

I am not sure how and where we will fill up those new fields

kube batch will update after each scheduling cycle.

[k82cn]

@Zyqsempai , just go through your PR; I think I need to give a design doc for this, including the feature interaction with cluster autoscaler; so you can create related PR if you're interesting in. I'll create the design doc asap :)

[Zyqsempai]

@k82cn , Great, will be happy to work on it, but what about original PR, should we merge it, or wait for related PR based on your design doc?

[k82cn]

but what about original PR, should we merge it,

For your PR, I think that's a good start; I'll also add some comments according to the design doc. So I'd suggest to update your PR when design doc is ready. I'll try to get doc ready this week and ask you and @Jeffwan to review

[Jeffwan]

@k82cn I think for short term, I will only update podCondition for those pods that fails in predicate action. If that case, cluster autoscaler will scale up right count of new nodes. It would be great to see your doc and work on long term support.

[k82cn]

@Zyqsempai , would you help to update your PRs according to https://github.com/kubernetes-sigs/kube-batch/blob/master/doc/design/podgroup-status.md :)

[k82cn]

I think for short term, I will only update podCondition for those pods that fails in predicate action. If that case, cluster autoscaler will scale up right count of new nodes.

I'm ok with that; we can cover regular case firstly. For gang-scheduling/coscheduling case, we'll discuss it later :)

[Zyqsempai]

@k82cn Done;)

[k82cn]

@Zyqsempai , are you going to work on the other items, e.g. updating status accordingly? If not, someone from our team may help on it :)

[Zyqsempai]

@k82cn Will be glad to, can you point me a little bit? I can't find the proper place where status should be updated.

[k82cn]

/close as PodGroup status is done