Hi folks, we have a self-hosted flyte deployment (specifically on Oracle cloud), and seems like the tasks pods aren't being cleared when completed and that's preventing the auto-scaler to scale down the k8s nodes. Any idea how to make the being cleared quickly? is this something the propeller has control over?

You might want to try to set the k8s finalizer options. e.g.:

k8s:
    plugins:
      k8s:
        inject-finalizer: true
        delete-resource-on-finalize: true

thanks @ripe-smartphone-56353! do you know if what happens if the task fails? will the pod still be deleted?

yes in my experience it will

did you use oracle storage etc?

@freezing-airport-6809 no particular issues with Oracle storage (we use the S3-compatible API and configured an endpoint). We have other issues but not sure if it's related to Oracle Cloud. I will share details in another post.

Please let us know