Sponsored: backing up and recovering machine configuration

[andrewrynhard]

Description

Another nice thing to have is the ability to reapply the last known good machine configuration to get back the node to a working state.
Save the machine-config.yml when applying a new one, and provide a way to use the backup if the boot went wrong.

[andrewrynhard]

Ideas:

• API for rolling back to a last known configuration
• Storing the machine configuration with each boot option (A, or B)

Try mode may handle some of this. Other parts might have to be manual?

See #4591, #4628

[sbskas]

Try mode is not enough for testing all the modification of a machine config.
Some need a restart of the node which is prevented by the volatile nature of the try mode.
Something simple and robust (manual) would be enough.

[PrivatePuffin]

Try mode is not enough for testing all the modification of a machine config. Some need a restart of the node which is prevented by the volatile nature of the try mode. Something simple and robust (manual) would be enough.

Bit of a necro, but agreed...
"Try" is still too much "hit or miss" to rely on...
At least when it comes to node-bricking.

Talos has 1 single actual weakness and thats the fact that a bad apply can cause a node to go down without a way to get it back-up without manual intervention on-site.

It would be nice having a KVM key combo that can be hit to force a revert to previous known-good config.

[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.

[DmitriyMV]

Didn't we do something with this?