So another interesting question. I am following the deployment (more or less) found here for AWS EKS: https://github.com/unionai-oss/deploy-flyte/tree/main/environments/aws/flyte-core I ran a workflow, but I noticed my pods are not actually being evicted ever. I am a bit confused, because I see there is some finalizer k8 piece in the YML file, but I am not sure that it is being utilized properly. I noticed the

<http://cluster-autoscaler.kubernetes.io/safe-to-evict|cluster-autoscaler.kubernetes.io/safe-to-evict>: false

annotation is enabled, which makes sense obviously when the task is running, but not when it is Completed, it should be evicted, right? Is there something core/key here I am missing to make sure I don't have 100s of pods lying around after a workflow is done?

I think once the pod is completed, flyte leaves it up to the k8s garbage collector to clean it up - if it's completed it doesn't take up any resources so it's okay to have a bunch of them

The problem I am having, is that I am creating PVCs, which are assigned to the pods, and do not free up currently because it is still bound to a "completed pod"

I see yeah I'm not too familiar with PVCs but I would try out that config

This option worked btw 🙂 Thank you!