Describe what you need us to do:
Fix the Fedora InfraCloud to be able to spawn an x86_64 machine.
When do you need this?
If we cannot complete your request, what is the impact?
Nobody will be able to effectively spawn new intel VMs in the cloud, which affects its general usability. Apps in the cloud that depends on dynamic machine spawning are effectively disabled/unusable by this bug.
I tried to spawn a generic Fedora-Cloud-Base-28_Beta-1.3.x86_64 and it won't finish in 8 minutes and even more (it won't probably finish at all). Normally such image is spawned in matter of several seconds. I tried to spawn a copr-builder image. It failed after 15 minutes with:
Error: no valid host was found. Exceeded max scheduling attempts.
ok. I think I have tracked down and fixed this now.
It has to do with updates we did last week. After updates, things started defaulting to a cpu type that the newer compute nodes didn't understand. So, instances would go there and die/not start/get in a weird state. Soon that happened enough to fill up the capacity.
I fixed that and tried to kill all the stuck instances, but some are still in an odd state and likely will require a reboot of the cloud to fix. For now I just upped the quota on copr so it can operate with all those dead instances in it's quota.
Metadata Update from @kevin:
- Issue close_status updated to: Fixed
- Issue priority set to: None (was: Needs Review)
- Issue status updated to: Closed (was: Open)
to comment on this ticket.