Cloudstack agents cannot start without both primary and secondary storage available - backup/failover question

[Jayd603]

Ok so it seems like the agent will not even start without both storage mounts available. I figured I could at least start some kvm VMs if one or both were missing if they were on local host storage. Oh well. So my question is, if storage is hosed for whatever reason, what is the quickest way back up? If I have a mirrored storage server on a different IP, do I simply update the IP address in the cloud database and restart the management server? Any better ways for failover or is there a way to update it without doing it direct in the database?

[DaanHoogland]

@Jayd603 , you can try a hack like that or add the storage as a new storage and initialise it / seedd it with the system vm template. I would prefer the latter way.

[Jayd603]

The other storage server would already have the system VM. Can I add the new/mirrored storage device and mark it as disabled in cloudstack then easily disable the down/broken server and enable the new one? I believe I tried to do this 1-2 years ago and I still had to go into the database to make it work. I will give it another shot though.

[DaanHoogland]

doesn’t sound like this would work, all snapshot- and template-records would still point to the old location. so you would need to update a lot in the DB. We are talking about secondary storage are we?

[Jayd603]

Right. I'm talking about both. Quickest way to handle needing to change to a backup storage server (that already has basically everything mirrored from the old one). Seems like still just updating the IP address in the database is the best or only way to do it. I think it maintains the same uuid/locations if I just change the IP so as long as those same files exist on the backup machine it should work... I was just wondering if there was a better way in cstack I did not know about. Really the UI should probably just add the ability to edit the IP address of the storage server.

[DaanHoogland]

this would be a change to the API. Feel free to enter a feature request.