Prompt if reboot is required when apply affect may need one.

[Jackbennett]

Feature Request

Updated machine registry config requires a reboot, but the cli does not warn you;

> talosctl apply -f controlplane.yaml
Applied configuration without a reboot

Kind of implies to me there's a check if a reboot might be required on apply.

Description

Within the manual;

.machine.registries (CRI containerd plugin will not pick up the registry authentication settings without a reboot)
https://www.talos.dev/v1.2/talos-guides/configuration/editing-machine-configuration/

So upon applying that change the CLI could know it won't take effect until reboot. I think not automatically rolling out reboots is the correct behaviour.

[smira]

It's complicated, as images are pulled by Talos and CRI.

Talos applies any changes on the fly.

CRI applies some changes on the fly, and some not. CRI should be fixed eventually to support reloading everything on the fly, but it's not something we have control over.

[github-actions]

This issue is stale because it has been open 180 days with no activity. Remove stale label or comment or this will be closed in 7 days.

[Jackbennett]

It's not stale but unresolved.
Making the cri dynamic works sure be cool but I think this issue is more about the manual knows a key needs a reboot so the cli could let me know ones needed. Or if not at apply time some reconcile earning, like you get if I use mismatched talosctl versions

[rothgar]

Is this request asking if we can have the CLI print something like configuration applied: reboot needed [will not happen automatically] when the configuration change requires a reboot?

[Jackbennett]

Yeh that's accurate, if there's a modified config that it's known requires a reboot to take effect.

Possibly it might be better worded like "Configuration applied, changes affected on reboot: [.machine.registries, etc...]" with a view to seeing that too with talosctl node status

If this is a whole thing then maybe close the issue. But seeing as the docs know this, maybe the cli types can react to it quite simply?

I've not really had this next example with Talos but before I've seen config changes made weeks/months ago that broke everything, but that only applied on reboot, so once a reboot finally happened, the thing to fix wasn't recent at all.

I think this is important as your config yaml evolves from patches to a git tracked file instead of regenerated from scratch each the time, new defaults can get missed. Maybe I'm always supposed to start patching machine config from newly generated configs.

[smira]

The "defaults" don't change over time, as long as you properly generated "base layer" of the machine configuration:

talosctl gen config --talos-version=v1.8 --with-secrets=secrets.yaml

The --talos-version=v1.8 is the compatibility guarantee, it should be the version of Talos the cluster was created with, and it should never change.

This way even if you update talosctl, the config will stay same.