Comment 6 for bug 1922578

It looks like the sigterm timeout is actually pretty common. I can reproduce the sigterm timeout on a standard cs:charmed-kubernetes deployment, no hacluster, running Kubernetes 1.20.5, by running `sudo service snap.kube-apiserver.daemon stop`.

In the originally reported crashdump, I see that kubernetes-master/0 also hit the stop-sigterm timeout on kube-apiserver. But it recovered just fine because kube-apiserver was restarted shortly after.

So the weird part here is that pacemaker never restarted kube-apiserver on kubernetes-master/2, even after hours of it being stopped.