Cronjob Should have limit on Active Jobs about kubernetes

[k8s-ci-robot]

This issue is currently awaiting triage.

If a SIG or subproject determines this is a relevant issue, they will accept it by applying the triage/accepted label and provide further guidance.

The triage/accepted label can be added by org members by writing /triage accepted in a comment.

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-sigs/prow repository.

[neolit123]

/sig apps

[kannon92]

Have you looked into Kueue?

[btwseeu78]

We provide kueue to our users as well but as a service provider we can't force it into projects,also we are using it for some other use cases, as I know queue has something to back-off scheduling.

But my question is is it not worthy to be part of vanila k8s job also as a platform provider it's heafty job to convince project for something since it's a controller at the end. Also queue has some issues as well ,it can block entire cluster(since the entire stuff runs via a webhook )if it's controller is dead .so kueue is good but should not replace job.

[kannon92]

Tbh I don’t really understand what you are asking for. If you want to limit the number of active pods in a job one could set parallelism lower. If a user species a parallism greater than max jobs I don’t really know what you want to happen.

Maybe you define a CRD for a Job or CronJob where you can set maximum parallelism. If you want some kind of quota management and to not overwhelm the cluster I think Kueue would be your best work.

@alculquicondor @mimowo wdyt? Do you see this as possible?

[alculquicondor]

I think the ask is pretty clear: they want to limit the number of active jobs created by the CronJob.

OTOH, an excess of Jobs could only happen if Jobs never terminate in the first place.

One mitigation you can do is to use ResourceQuotas to limit the number of jobs in an entire namespace https://kubernetes.io/docs/concepts/policy/resource-quotas/#object-count-quota

In any case, I would recommend you start a proposal and write a KEP to target v1.32 (enhancements freeze for v1.31 already started).

[btwseeu78]

Though resource quota can be a solution, i am running some mutating webhooks for now as default since most of our uses does it really need jobs in parallel and its happening due and error on their end but the fix will take too long since it involved some communication and changes.

The only problem i see resource quota would namespace wide , and we really want something that can control per job wise+ a quota for upper limit a large upper limit lets say.

@alculquicondor for kep im not sure who should writw it ,will it be me ,i'm not sure i have the required development capability to implement this proposal .