OOMKilled CrashLoop: right-size the memory limit

The situation

The cache Deployment in kubelings never stays up — pods start, then die with OOMKilled and fall into CrashLoopBackOff. The container needs roughly 50Mi of working memory, but its limits.memory is set to 20Mi, so the kernel kills it the moment it allocates.

Your task

Right-size the memory so cache runs steadily:

Inspect why the pod is being killed.
Raise the memory requests/limits to fit the ~50Mi workload (give headroom).
The Deployment must become Available and stop restarting.

kubectl -n kubelings get pods -l app=cache
kubectl -n kubelings describe pod -l app=cache | grep -A3 -i 'last state\|reason'

Hint

Reason: OOMKilled confirms the memory cap. Raise it:

kubectl -n kubelings set resources deploy/cache \
  --requests=memory=64Mi --limits=memory=128Mi

Solution

Root cause

The cache container needs ~50Mi but limits.memory was 20Mi. When a container exceeds its memory limit the kernel OOM-kills it, producing Last State: Terminated / Reason: OOMKilled and CrashLoopBackOff.

Fix

Raise the request and limit above the working-set size (with headroom):

kubectl -n kubelings set resources deploy/cache \
  --requests=memory=64Mi --limits=memory=128Mi

or kubectl -n kubelings edit deploy cache and set:

resources:
  requests: {memory: 64Mi}
  limits:   {memory: 128Mi}

Verify

kubectl -n kubelings rollout status deploy/cache
kubectl -n kubelings get pods -l app=cache   # Running, RESTARTS stays 0

Prevention

Set memory requests from observed usage and limits with headroom; alert on container_memory_working_set_bytes approaching the limit before pods OOM.

Course progress