排查常见 K8s 问题

Kubernetes is powerful, but things can go wrong! Understanding how to diagnose issues is a vital skill for any developer or operator.

This lesson will equip you with essential tools and techniques to identify and resolve common problems in your Kubernetes clusters.

The kubectl command-line tool is your primary interface for interacting with a Kubernetes cluster. It's indispensable for troubleshooting.

• Inspect: View the status and details of your resources.
• Monitor: Check logs and events for clues.
• Interact: Run commands inside your containers for deeper debugging.

kubectl describe provides a wealth of information about a specific resource, such as a Pod, Deployment, or Service.

It shows current status, configurations, resource limits, and crucially, recent events related to that resource. Look for warnings or errors in the "Events" section.

Application logs are often the first place to look when a container isn't behaving as expected. Use kubectl logs to see what your application is outputting.

Kubernetes generates events for significant occurrences, like Pod scheduling, image pulling, or volume mounting. These events are crucial for understanding the cluster's actions.

Use kubectl get events to see a timeline of what's happening. Filter by namespace or resource for relevance.

Sometimes you need to go inside a running container to debug. kubectl exec lets you run commands directly within a container, like checking file paths or network connectivity.

These are two very common Pod status issues:

• Pending: Often due to insufficient resources (CPU/memory), node affinity issues, or a missing Persistent Volume Claim. Check kubectl describe pod events.
• CrashLoopBackOff: The container starts, crashes, and Kubernetes tries to restart it repeatedly. This usually means an application error. Check kubectl logs.

Is your application unreachable? Network issues are common. Check these:

• Service Endpoints: Use kubectl describe service your-service to ensure it has active Pod endpoints.
• Ingress Rules: For external access, verify your Ingress resource rules and the Ingress controller's status.
• Network Policies: Ensure no Network Policy is blocking traffic unintentionally.

If your Pods are consuming too much CPU or memory, Kubernetes might terminate them. For memory, this is often seen as OOMKilled (Out Of Memory Killed).

Check your Pod's resource requests and limits in its YAML definition. Use kubectl describe pod to see if a container was OOMKilled.

You've deployed a new Pod, but it's stuck in Pending status. What are the most likely first steps you should take to diagnose the problem?

You've learned to approach Kubernetes troubleshooting systematically. Remember your key tools:

• kubectl describe for detailed resource info and events.
• kubectl logs for application output.
• kubectl get events for cluster-wide occurrences.
• kubectl exec for in-container debugging.

Practice these commands to quickly pinpoint and resolve issues in your Kubernetes deployments!