Kubernetes集群中Docker容器启动后立即暂停退出问题求助

It sounds like your user-service pods are hitting a CrashLoopBackOff state—this is a common issue when containers fail to start successfully. Let’s walk through the steps to diagnose and fix this:

1. Gather Detailed Pod Information #

First, identify the problematic pod and collect critical logs and status details:

• List all pods to find the name of your user-service pod:

  kubectl get pods
  

• Check logs from the latest failed container run (replace <pod-name> with the actual name from the previous command):

  kubectl logs <pod-name>
  

• If the pod has restarted multiple times, check logs from the previous failed instance to get more context:

  kubectl logs <pod-name> -p
  

• Get a full breakdown of the pod’s status, events, and configuration—this will reveal key details like image pull errors or container exit codes:

  kubectl describe pod <pod-name>
  

2. Most Likely Root Cause: Image Availability on the Worker Node #

Since you built the user-service image locally on your machine, your worker node (vm-hps11) doesn’t have access to this image by default. Kubernetes tries to pull images from a public registry (like Docker Hub) when it can’t find them locally, which fails if your image isn’t pushed there.

Fix Options: #

Option A: Push the Image to a Container Registry

1. Tag your local image with a registry path (replace your-username with your Docker Hub username):

   docker tag user-service your-username/user-service:latest
   

2. Push the image to Docker Hub:

   docker push your-username/user-service:latest
   

3. Update your deployment to use the full image path:

   kubectl set image deployment/user-service user-service=your-username/user-service:latest
   

Option B: Copy the Image Directly to the Worker Node

If you don’t want to use a public registry, transfer the image to your worker node manually:

1. Save the local image to a tar file:

   docker save user-service -o user-service.tar
   

2. Copy the tar file to your worker node (replace user and vm-hps11 with your worker’s credentials and hostname):

   scp user-service.tar user@vm-hps11:/tmp/
   

3. SSH into the worker node and load the image:

   ssh user@vm-hps11
   docker load -i /tmp/user-service.tar
   

4. Restart your deployment to trigger pod creation with the local image:

   kubectl rollout restart deployment user-service
   

3. Check for Container Startup Errors #

If the image is available on the worker node but the container still crashes, look for application-specific errors in the pod logs. Common issues here include:

• Missing required environment variables
• Incorrect entrypoint/command in the Docker image
• Uninstalled dependencies inside the container
• Port conflicts or misconfigured network settings

Once you’ve addressed the root cause, verify the pods are running successfully with:

kubectl get pods

内容的提问来源于stack exchange，提问作者vaibhav