How to determine when a reboot is required?

[jkossis]

Other Linux distributions use the /var/run/reboot-required convention in order to signal that a reboot is required, useful for applications like unattended-upgrades or kured.

Does Talos Linux have a similar convention?

[smira]

No, Talos only requires reboot hard for upgrades of Talos, and as Talos is minimal and immutable, you get the whole base OS updated.

Some component updates (e.g. kubelet upgrades) don't require a reboot at all.

[SISheogorath]

I'm joining in here, after I adjusted my system-upgrade-controller config assuming talosctl upgrade --stage would, I don't know, "stage an upgrade" instead of triggering a reboot, which turned out to be incredibly wrong. (You can see all details here)

Intention was to stage all upgrades without draining or rebooting the nodes and let kured, which is much better at handling reboots that system-upgrade-controller, take care of it.

Either way, I think it would be really useful to get this functionality, since it would allow so much nicer self-contained upgrade mechanisms without running talosctl upgrade either outside the cluster or with some kind of fancy job definition that avoids scheduling the pod on the node that is supposed to be rebooted.