Homura222
Homura222
> I think it makes sense to `virt-handler` to be aware that `kubelet` has being restarted and avoid disruptions if possible. Not sure for instance, when kubelet's restart, what's the...
> Also, @Homura222 do you see a case not covered by Fabian's [#11662 (comment)](https://github.com/kubevirt/kubevirt/issues/11662#issuecomment-2042447470)? That is, `kubelet` restart in which we should keep VM running. I also think `kubelet` restart...
> > I think it makes sense to `virt-handler` to be aware that `kubelet` has being restarted and avoid disruptions if possible. Not sure for instance, when kubelet's restart, what's...
> that do not match the node label. > Ok, after checking in on [kubernetes/kubernetes#124586](https://github.com/kubernetes/kubernetes/issues/124586), I think our path forward would be that the heartbeat label is requested through a...
> > @rmohr Do you mean to move `kubevirt.io/schedulable: "true"` from pod `nodeSelector` to pod `affinity`? But after kubelet restarts, kubelet will to kill the pod that `affinity` or `nodeSelector`...
> > In the reported case,kubelet also kill running pod when kubelet is restarted. [kubernetes/kubernetes#123980](https://github.com/kubernetes/kubernetes/issues/123980) > > This is definitely happening with `nodeSelector`. Not sure though about affinity, considering [kubernetes/kubernetes#124586...
> > > In the reported case,kubelet also kill running pod when kubelet is restarted. [kubernetes/kubernetes#123980](https://github.com/kubernetes/kubernetes/issues/123980) > > > > > > This is definitely happening with `nodeSelector`. Not sure...
Could you add another e2e test for pod.Spec.NodeSelector, please?
if pod use `preferredDuringSchedulingIgnoredDuringExecution` nodeAffinity , pod will not be killed in this case. Is this expected?