Allow failed instances to be stopped

[david-crespo]

Also rebooted maybe?

https://github.com/oxidecomputer/omicron/pull/6652

[hawkw]

Also rebooted maybe?

I believe Failed instances can be restarted by PUT /v1/instances/{id}/start, but not rebooted by PUT /v1/instances/{id}/reboot. The "reboot" API doesn't start a stopped/failed instance on a new VMM, but instead instructs the existing VMM to reboot the guest (as though the guest was power-cycled, essentially). That doesn't work when there is no VMM --- the instance will have to be started first.

[charliepark]

2481 looks to be a duplicate of this, though I don't know that I have enough context on the rebooting conversation above to know if it's an exhaustive duplicate or only a partial duplicate.

[hawkw]

@charliepark I think #2481 is definitely a duplicate of this. Re: rebooting, we should not allow rebooting a Failed instance, since that will just result in Nexus returning an error anyway. I'd probably keep this issue over #2481 since it has that additional context.

[hawkw]

Oh, if #2482 closes #2481, it should definitely also close this --- the rebooting thing is just kinda a red herring.

[charliepark]

Solid; will update #2482 and close #2481.

[askfongjojo]

Sorry about the dup. I should have checked before filing it. 😓

[charliepark]

No worries! It was actually seeing your issue that prompted my working on the fix, so it all worked out. 🙂