PEBCAK! Don’t know if I should request this in Juju or MAAS
So I was cleaning up some machines not controlled by juju, and ended up releasing a machine that was part of a model. (Luckily I didn’t erase the drive, so I was just able to turn it back on)
There’s now a disconnect between what MAAS thinks is not deployed and what Juju knows is deployed.
Anyone know how I can get a machine back to a deployed state?
There will be many situations where a set of machines are provisioned together for a purpose. A Juju model is one example, but I can think of others. I think we should allow the specification of a cohort id, similar to the way machines can have a snap cohort id. Then, when critical operations take place, we require either the cohort-id or a manual override. In the web interface, if you tried to deprovision a machine in a cohort, it would ask for the cohort id. The cohort id could come with some metadata, which gets presented back to you as a reminder. This would prevent a lot of oopsies.
That is wicked cool. Should I post to Juju as a feature request?
Previously I used to lock the machines that were part of my Openstack model, but I redeployed about a month ago, and then never relocked them, and thus stupidity ensued.
Still haven’t had the time to play with maasdb as mentioned.
I eventually managed to try this. I managed to mark the machine as deployed again, it’s better than what it was but it’s not the same , it’s turning into a pet which got bitten by a zombie
The IP addresses are missing, the deployed containers are no longer visible, and I’m too dumb to query the other tables in the maasdb.
I know that I should shoot it and get a new one, just not sure I want to deal with that right now because it is working