Consider draining and rebooting the node when deploying the Operator

[fidencio]

Sasha mentioned that restarting basic services in the node may lead to Kubernetes failures that would be avoided by draining and rebooting the node.

We should consider doing that as part of the deployment.

[bpradipt]

@fidencio agree that using draining and node reboot is safer. We will need to improve the pre-install and post-uninstall flow to incorporate this.

[bpradipt]

We might use something like this - https://github.com/weaveworks/kured

[fidencio]

@fidencio agree that using draining and node reboot is safer. We will need to improve the pre-install and post-uninstall flow to incorporate this.

Not only pre-install / post-uninstall flow, but also the normal flow where we add / remove a new runtimeclass to the containerd configuration.

[ariel-adam]

@fidencio is this issue still relevant or can be closed?
If it's still relevant to what release do you think we should map it to (mid-November, end-December, mid-February etc...)?

[fidencio]

This is still relevant and should be addressed in the short-term future. I'm labelling it to be part of the 0.3.0, but we'll re-asses when that time comes.

[ariel-adam]

Cool, thanks