Juputerhub on K8s user restart policy

JupyterHub
, , ,
1

Hi all, I am following the “Zero to JupyterHub with Kubernetes” guide. I’m not clear though how any crashes of user pods are handled. Are there any additional configurations that need to be entered? I have tried deleting the pod and it disappears, it is not automatically recreated.
In documentation I only found this hint “ 6. A container that runs out of memory is killed by a Linux OOMKiller (Out Of Memory Killer). When this happens, you should see a trace of it using kubectl describe pod --namespace and kubectl logs --previous --namespace .

When a container’s process is killed, the container restarts if the container’s restartPolicy allows it, otherwise the pod is evicted.”
source: Optimizations — Zero to JupyterHub with Kubernetes documentation (in Balancing “guaranteed” vs “maximum” memory and CPU)

So how can this configuration be modified?
Thanks in advance :slight_smile:

2

Restart policies aren’t supported for singleuser pods. Although technically it could be added it’s confusing for users since their pod would restart for no apparent reason, losing most of their state. If the user repeated what they were doing before the pod would most likely crash again.

There are some JupyterLab extensions to help with monitoring memory usage, for example