talosctl recover after VMs restart

[skillpoint-dev]

Here is what I have written down for what to do when I restart the VMs where I have Talos running(I got this info in the Talos Slack channel)...

As of "Talos OS 0.8.0-alpha3(k8s 1.20.0)" there is a "Node Affinity"-Bug in k8s for single control plane node clusters.
After restart the pods wont start correctly.

Remedy

talosctl recover --nodes [TALOS-CONTROL-PLANE-IP]

or

talosctl recover --nodes [TALOS-CONTROL-PLANE-IP] --source etcd

Unfortunately this doesn't work anymore but I still face the problem that most of the applications don't recover from a VM restart.

Is there anything that is now in place instead of "talosctl recover" - I only found the disaster recovery page and this seems the wrong place to look for that.

[andrewrynhard]

Which version of Talos? The recover command worked in the self-hosted Kubernetes format. That currently isn't the way Talos deploys things so that command is a noop on clusters that aren't self-hosted.

[skillpoint-dev]

Last try of a recovery after a VM restart was with 0.11.2 but I don't remember at which version the suggested remedy of the problem above was not working anymore.

[smira]

Was the cluster bootstrapped on Talos <= 0.8. If not, it's not using self-hosted control plane, so talosctl recover won't work.

If the cluster was bootstrapped on Talos <= 0.8, it should be converted to the new control plane format: https://www.talos.dev/docs/v0.11/guides/converting-control-plane/