Apparently, when you create a new instance in OpenStack and try to boot it for the first time, it hangs during the boot and waits for entropy. The workaround that was proven to be working is going to dashboard, selecting the instance, going to console and hitting some random keys (ideally with your forehead). That will magically resume the boot.
There is another possibility though. We can specify, that we trust the random generator from the host and fix the issue this way. It needs to be set on an image
openstack image set --property hw_rng_model=virtio <THE_SNAPSHOT_UUID>
Also, to get it working, it needs to be enabled for instance flavor.
Would it be possible for someone, to fix the Fedora images that are available in OpenStack?
It's a nice idea, but we tried to do this after the upstream kernel change landed and our openstack proved to be too old to have the needed properties.
I think you can set the property on the image, but the flavor database simply doesn't have the needed property to allow it. ;(
You're welcome to try and get it working (while being very carefull not to break anything), but I fear it's not gonna work.
Metadata Update from @kevin: - Issue assigned to kevin - Issue close_status updated to: Will Not/Can Not fix - Issue priority set to: Waiting on Assignee (was: Needs Review) - Issue status updated to: Closed (was: Open)
This is probably not truth, because we have it enabled at least for images for copr builders, and on the flavor ms2.builder (used for builders). I think @puiterwijk helped me to solve this (without it, it would be seriously problematic to start copr builders nowadays).
The per-flavor setup is unfortunate though, ... we'd like to have updated all those flavors to make everything work for copr instances:
Login to comment on this ticket.