GKE CrashLoopBackOff: read the exit code, then the liveness probe
CrashLoopBackOff is not a cause. Check the exit code under Last State: 0 points to the entry point, non-zero to a crash, and OOMKilled to memory.
What you see
State: Waiting,Reason: CrashLoopBackOffLast State: Terminatedwith an exit code
What it means
Google's guide stresses that CrashLoopBackOff does not name a specific error; it means a container is repeatedly crashing after restarts. Restarts happen when a container exits for any reason, even after a successful exit, so a clean exit can loop too.
The exit code under Last State is the first thing to read. Exit code 0 is unexpected for a long-running service and points to the entry point or the app's design. A non-zero code means the app crashed, so look at configuration, dependencies or bugs.
How to confirm the cause
- 1Read the state, last state and exit code
kubectl describe pod POD_NAME
Last Stateshows the previous container's termination. - 2Read the previous container's logs
kubectl logs POD_NAME --previous
--previousretrieves logs from the prior, terminated container. - 3Check whether liveness probes caused the kill
# Logs Explorer filter: jsonPayload.MESSAGE:"failed liveness probe, will be restarted"
If these appear, the probe is killing a container that may be fine.
Causes and fixes
Resource exhaustion
- HOW TO TELL
Reason: OOMKilledunderLast State.- ROOT CAUSE
- The container used more memory than its limit and the kernel killed it.
- FIX
- Raise the memory limit or fix the leak. See the GKE OOM runbook.
Liveness probe failures
- HOW TO TELL
- Probe failure messages precede each restart.
- ROOT CAUSE
- The probe fails because it is misconfigured, too strict, or the app is slow to start or under heavy CPU or disk I/O load.
- FIX
- Review the probe's command, port, path and
timeoutSeconds, and loosen it or add a startup probe.
Exit code 0
- HOW TO TELL
- The container exits cleanly and restarts.
- ROOT CAUSE
- The process finished its work and exited, or the entry point starts something that returns immediately.
- FIX
- Make the container run a long-lived process, or run it as a Job.
Application misconfiguration or crash
- HOW TO TELL
- A non-zero exit code and an error in the previous logs.
- ROOT CAUSE
- A missing configuration value, a failed dependency or a bug.
- FIX
- Fix what the log shows.
How to stop it coming back
- Test the container with the same environment before deploying.
- Alert on restart counts.
- Keep probes simple and tolerant.
Keep going
- Runbook: GKE OOM events: container-level vs node-level kills
- Runbook: CrashLoopBackOff: the container keeps crashing
- Runbook: Liveness probe failed: the container keeps being restarted
- Cheat sheet: kubectl
- Lesson: the kubernetes topic in Learn
Sources
Google Cloud documentation (docs.cloud.google.com), read 18 Nov 2026.