Removed Terminated pods from Kubernetes

So in recent versions of Kubernetes, when pods are killed from a drain event, they go into a state of Failed with the reason being Terminated. They hang out there for a time, but will eventually get GC’d. This can however mess with some stats and it looks kinda scary if you aren’t prepared to see it.

I prefer keeping a clean slate so I run a cron that cleans up these pods using this lovely command:

kubectl get pods --all-namespaces -o go-template='{{range .items}}{{if eq .status.phase "Failed"}}{{if or (eq .status.reason "Terminated") (eq .status.reason "NodeShutdown")}}{{.metadata.namespace}} {{.metadata.name}}{{printf "\n"}}{{end}}{{end}}{{end}}' | xargs -n2 kubectl delete pod -n

Due to this running kubectl it obviously needs some RBAC, and a container that has kubectl available. Here is the full set of manifests, hopefully it helps someone!

apiVersion: rbac.authorization.k8s.io/v1
metadata:
  namespace: kube-system
  name: shutdown-cleaner
rules:
  - verbs:
      - delete
      - get
      - list
    resources:
      - pods
    apiGroups:
      - ""
---
kind: ClusterRoleBinding
apiVersion: rbac.authorization.k8s.io/v1
metadata:
  namespace: kube-system
  name: shutdown-cleaner
subjects:
  - kind: ServiceAccount
    name: shutdown-cleaner
    namespace: kube-system
roleRef:
  apiGroup: ""
  kind: ClusterRole
  name: shutdown-cleaner
---
kind: ServiceAccount
apiVersion: v1
metadata:
  namespace: kube-system
  name: shutdown-cleaner
---
kind: CronJob
apiVersion: batch/v1beta1
metadata:
  name: shutdown-cleaner
  namespace: kube-system
spec:
  successfulJobsHistoryLimit: 3
  failedJobsHistoryLimit: 1
  suspend: false
  concurrencyPolicy: Forbid
  jobTemplate:
    spec:
      parallelism: 1
      completions: 1
      backoffLimit: 3
      activeDeadlineSeconds: 60
      template:
        spec:
          serviceAccountName: shutdown-cleaner
          containers:
            - name: shutdown-cleaner
              image: bitnami/kubectl:1.21.5
              command:
                - /bin/sh
                - -c
                - kubectl get pods --all-namespaces -o go-template='{{range .items}}{{if eq .status.phase "Failed"}}{{if or (eq .status.reason "Terminated") (eq .status.reason "NodeShutdown")}}{{.metadata.namespace}} {{.metadata.name}}{{printf "\n"}}{{end}}{{end}}{{end}}' | xargs -n2 kubectl delete pod -n
          restartPolicy: Never
  schedule: "0 * * * *"