Skip to main content
DevFixHub
Kubernetes5 min read2026-03-01

Kubernetes CrashLoopBackOff

Fix Pod CrashLoopBackOff errors caused by application exceptions, failed liveness probes, or OOMKilled events.

Error Code / Stack Trace

NAME READY STATUS RESTARTS AGE api-pod 0/1 CrashLoopBackOff 5 3m

Problem Overview

Kubernetes continuously attempts to restart a pod container that repeatedly crashes after launching.

Why Does This Happen?

  • Application throws an unhandled exception or missing environment variable on startup.
  • Liveness or readiness probe points to an incorrect endpoint or has too short a timeout.
  • Container exceeded memory limit and was terminated with OOMKilled.

Step-by-Step Solution

Step 1: Inspect pod events with kubectl describe

Check the Last State, Exit Code, and Recent Events section.

bash
kubectl describe pod <pod-name>

Step 2: Read logs from the previously crashed container

Use the --previous flag to inspect the log output right before the crash.

bash
kubectl logs <pod-name> --previous

Step 3: Tune liveness probe initialDelaySeconds

Give slower JVM or Node apps adequate warm-up time before probes begin.

yaml
livenessProbe:
  httpGet:
    path: /actuator/health/liveness
    port: 8080
  initialDelaySeconds: 30
  periodSeconds: 10

Common Mistakes to Avoid

  • •Running 'kubectl logs <pod-name>' without '--previous', which only shows logs of the currently restarting empty container.

Prevention & Best Practices

  • Always set realistic resources.requests and resources.limits in pod specifications.

Frequently Asked Questions

What is the backoff exponential timing in CrashLoopBackOff?

Kubernetes delays restarts exponentially: 10s, 20s, 40s, up to a maximum delay of 300 seconds (5 minutes).

Related Developer Solutions & Tools

Recommended Tools
Related Error Fixes
In-Depth Tutorials