What are the defaults for the Kubernetes CrashLoopBackOff?
Say, I have a pod:
kubectl run mynginx --image nginx -- echo helloAnd I inspect its status:
kubectl get pods -w
NAME READY STATUS RESTARTS AGE
mynginx 0/1 Pending 0 0s
mynginx 0/1 Pending 0 0s
mynginx 0/1 ContainerCreating 0 0s
mynginx 0/1 Completed 0 2s
mynginx 0/1 Completed 1 4s
mynginx 0/1 CrashLoopBackOff 1 5s
mynginx 0/1 Completed 2 20s
mynginx 0/1 CrashLoopBackOff 2 33s
mynginx 0/1 Completed 3 47s
mynginx 0/1 CrashLoopBackOff 3 59s
mynginx 0/1 Completed 4 97s
mynginx 0/1 CrashLoopBackOff 4 109sThis is "expected". Kubernetes starts a pod, it quits "too fast", Kubernetes schedules it again and then Kubernetes sets the state to CrashLoopBackOff.
Now, if i start a pod slightly differently:
kubectl run mynginx3 --image nginx -- /bin/bash -c "sleep 10; echo hello"I get the following
kubectl get pods -w
NAME READY STATUS RESTARTS AGE
mynginx3 0/1 Pending 0 0s
mynginx3 0/1 Pending 0 0s
mynginx3 0/1 ContainerCreating 0 0s
mynginx3 1/1 Running 0 2s
mynginx3 0/1 Completed 0 12s
mynginx3 1/1 Running 1 14s
mynginx3 0/1 Completed 1 24s
mynginx3 0/1 CrashLoopBackOff 1 36s
mynginx3 1/1 Running 2 38s
mynginx3 0/1 Completed 2 48s
mynginx3 0/1 CrashLoopBackOff 2 62s
mynginx3 1/1 Running 3 75s
mynginx3 0/1 Completed 3 85s
mynginx3 0/1 CrashLoopBackOff 3 96s
mynginx3 1/1 Running 4 2m14s
mynginx3 0/1 Completed 4 2m24s
mynginx3 0/1 CrashLoopBackOff 4 2m38sThis is also expected.
But say I set sleep for 24 hours, would I still get the same CrashLoopBackOff after two pod exits initially and then after each next pod exit?
Based on these docs:
The restartPolicy applies to all containers in the Pod. restartPolicy only refers to restarts of the containers by the kubelet on the same node. After containers in a Pod exit, the kubelet restarts them with an exponential back-off delay (10s, 20s, 40s, …), that is capped at five minutes. Once a container has executed for 10 minutes without any problems, the kubelet resets the restart backoff timer for that container.
I think that means that anything that executes for longer than 10 minutes before exiting will not trigger a CrashLoopBackOff status.