search

kubeflow / mpi-operator

Kubernetes Operator for MPI-based applications (distributed training, HPC, etc.)
https://www.kubeflow.org/docs/components/training/mpi/
Apache License 2.0
430 stars 216 forks source link

Pass schedulingPolicy.PriorityClass to Launcher and Workers #536

Closed tenzen-y closed 1 year ago

tenzen-y commented 1 year ago

/kind feature

Since #520, the mpi-operator respects schedulingPolicy when creating PodGroup. However, currently, the mpi-operator doesn't pass schedulingPolicy.PriorityClass to Launcher and Workers even if they don't have priorityClass.

So I would like to pass schedulingPolicy.PriorityClass to ReplicaSpec if they don't have priorityClass.

@alculquicondor @terrytangyuan WDYT?

alculquicondor commented 1 year ago

Do we do the same in other training-operator controllers?

I've heard from some people that they prefer to have a job-level priority that it's independent of the pods priority.

tenzen-y commented 1 year ago

Do we do the same in other training-operator controllers?

No, the training-operator doesn't have the same logic.

I've heard from some people that they prefer to have a job-level priority that it's independent of the pods priority.

It makes sense. It might be better to keep the current logic.

tenzen-y commented 1 year ago

/close

google-oss-prow[bot] commented 1 year ago

@tenzen-y: Closing this issue.

In response to [this](https://github.com/kubeflow/mpi-operator/issues/536#issuecomment-1465858218): >/close > Instructions for interacting with me using PR comments are available [here](https://git.k8s.io/community/contributors/guide/pull-requests.md). If you have questions or suggestions related to my behavior, please file an issue against the [kubernetes/test-infra](https://github.com/kubernetes/test-infra/issues/new?title=Prow%20issue:) repository.