如何进入Kubernetes初始化容器（Init Container）执行命令？

Hey there! Let's break down how to get into an init container and run commands, tailored to your kubectl v1.9.6 environment.

First, a quick heads-up: Init containers are built to run once and exit before your main container starts. So you can only access an init container if it's currently running (e.g., stuck waiting on a dependency or in a loop) or if Kubernetes hasn't yet garbage-collected the stopped container (this is temporary and not guaranteed).

Step 1: Locate the Init Container's Name and Status #

First, identify the init container tied to your pod and check its state. Run this command:

kubectl describe pod <your-pod-name>

Look for the Init Containers section in the output. You'll find the container's name (e.g., init-db-setup) and its current status (like Running or Completed).

Step 2: Access the Running Init Container #

If the init container shows a Running status, use kubectl exec with the --container flag (supported in kubectl v1.9+) to launch an interactive shell:

kubectl exec -it <your-pod-name> --container <init-container-name> -- /bin/sh

• Swap /bin/sh with /bin/bash if your init container's image includes bash.
• The -it flags enable interactive mode with a TTY, so you can use the shell just like you would on a local machine.

What if the Init Container Already Completed? #

If the init container has exited (status Completed), Kubernetes has probably cleaned up the container, so you can't access it directly. Instead, try these alternatives:

• View its logs to debug what happened:

  kubectl logs <your-pod-name> --container <init-container-name>
  

• If you need to test commands in the init container, modify your pod's manifest to add a sleep command at the end of the init container's command (e.g., command: ["sh", "-c", "your-existing-command && sleep 3600"]). This keeps the container running long enough for you to exec into it.

Key Notes for Your kubectl v1.9.6 Setup #

• Your client and server versions are matched (v1.9.6), so you won't hit compatibility issues with the --container flag.
• If your init container uses a minimal image without a shell (like distroless images), you won't be able to exec into it—stick to viewing logs or adjusting the image to include a shell for debugging purposes.

内容的提问来源于stack exchange，提问作者Eric Francis