Lesson 4.3

Resource limits and troubleshooting

Requests drive scheduling, limits drive eviction. CrashLoopBackOff and OOMKilled are usually a limits conversation, not a code one.

5mAdvanced3.7k students

Overview

What this lesson covers

  • Set requests and limits that reflect real usage
  • Diagnose CrashLoopBackOff and OOMKilled pods
  • Debug a pod with logs, describe, and exec

This lesson sits in Operating a Cluster, part of DevOps with Docker and Kubernetes. It assumes what came before it and leads directly into the next lesson in the module.

In this lesson you will:

  • Set requests and limits that reflect real usage
  • Diagnose CrashLoopBackOff and OOMKilled pods
  • Debug a pod with logs, describe, and exec
Pro Tip

Requests decide where a pod is scheduled; limits decide when it gets killed. They are not the same number.

Resources

Kubernetes documentation

The official reference for the tools used in this lesson.