Auto delete failed pods in Kubernetes

auto-delete-failed-pods-cronjob.yaml

# Auto delete pods with status.phase=Failed via CronJob
---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: sa-auto-delete-failed-pods
  namespace: kube-system

---
kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1
metadata:
  name: auto-delete-failed-pods
rules:
- apiGroups: [""] # "" indicates the core API group
  resources: ["pods"]
  verbs: ["get", "watch", "list", "delete"]

---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: auto-delete-failed-pods
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: auto-delete-failed-pods
subjects:
- kind: ServiceAccount
  name: sa-auto-delete-failed-pods
  namespace: kube-system

---
apiVersion: batch/v1
kind: CronJob
metadata:
  name: auto-delete-failed-pods-cronjob
  namespace: kube-system
spec:
  schedule: "*/30 * * * *"
  failedJobsHistoryLimit: 1
  successfulJobsHistoryLimit: 1
  jobTemplate:
    spec:
      template:
        spec:
          serviceAccountName: sa-auto-delete-failed-pods
          containers:
          - name: kubectl-runner
            image: bitnami/kubectl:latest
            command: ["sh", "-c", "kubectl get pods --all-namespaces --field-selector 'status.phase==Failed' -o json | kubectl delete -f -"]
          restartPolicy: Never