--- slug: "kubernetes-restart-pod-delay" title: "Easily Rolling Restart Kubernetes Pod (Pod Deletion Command)" description: "When `kubectl rollout restart` makes Pods take too long to restart, check `terminationGracePeriodSeconds` and any preStop hooks — both are common bottlenecks." url: "https://www.ytyng.com/en/blog/kubernetes-restart-pod-delay" publish_date: "2021-12-11T13:28:40Z" created: "2021-12-11T13:28:40Z" updated: "2026-05-11T13:21:29.555Z" categories: ["kubernetes"] keywords: "" featured_image_url: "https://media.ytyng.com/resize/20230812/933e18bc4f3b4d04836bd65b1eb16abd.png.webp?width=768" has_video: false has_music: false video_urls: [] music_urls: [] lang: "en" --- # Easily Rolling Restart Kubernetes Pod (Pod Deletion Command)

In Kubernetes, within a Pod's Deployment, the container configuration looks like this:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: myapp-deployment
  namespace: mynamespace
spec:
  replicas: 2
  selector:
    matchLabels:
      app: myapp
  template:
    metadata:
      labels:
        app: myapp
    spec:
      containers:
        - name: myapp
          image: myrepo/myapp:latest
          imagePullPolicy: Always

If you are operating with the container tag set to latest and imagePullPolicy: Always.

In this case, to reflect the latest code, you would use kubectl delete pod, but by deleting the pods at staggered intervals, you can avoid downtime.

While kubectl delete pod itself waits for the deletion and restart of the pod, if there is downtime until the application starts up after the pod restarts, this can be avoided.

A common approach is using this script:

restart-pod.sh

#!/usr/bin/env zsh

if [ -f "${HOME}/.kube/config-xxxx" ]; then
 export KUBECONFIG=${HOME}/.kube/config-xxxx
fi

pods=( $(kubectl get pod -n mynamespace | egrep -o "myapp-deployment-[a-zA-Z0-9-]+") )

_sleep=
for pod in "${pods[@]}" ; do
 ${_sleep}
 kubectl -n mynamespace delete pod ${pod}
 _sleep="sleep 10s"
done