I see the openshift CentOS CI supports VMs, but it doesn't look like I have permission to do that. Could we get the necessary access please? Specifically, I'm interested in spinning up VMs for doing smoketests of the builds we produce (these VMs would just run for a few minutes for the duration of the test). Thanks!
Metadata Update from @siddharthvipul1:
- Issue tagged with: centos-ci-infra, medium-gain, medium-trouble
We have a way to provide baremetal (using rest api or cico-client which is a wrapper and you get ip address of node). The node has your ssh keys deployed in there from a database so that you can ssh or automate workflow
We have plans to add vm checkout capability in there but it will be a work I will do with Fabian in coming weeks. Would you like to use that for now and once we have vm checkout capability ready, it will be just small modification in API request?
The reason we don't want to get into kubevirt is because while discussing the CI workflow consistency, updating duffy to support it made more sense to us. This way we can also have better control over the resources. I will discuss with the team on what's needed for kubevirt but wanted to let you know of our current workflow
So, the reason I was looking into VMs is that the testing we're interesting in doing is somewhat disruptive. In the case of systemd, for example, one thing we'd want to do is to upgrade it, reboot, validate that the host came back with the right version and everything worked properly, then reset to the original state (so we can do this again for the next build). Likewise, for testing the Hyperscale kernel we'd need to reboot multiple times (to test both EFI and non-EFI scenarios).
This kind of thing isn't a good fit for bare metal in my experience (both because the "reset" is expensive, and because constantly rebooting physical hosts puts non-trivial stress on the hardware and will eventually lead to premature failures).
The work I want to do around Btrfs-based atomic updates basically imposes similar requirements. Since the transactional update software forces us to do reboots for applying updates, I need to do this regularly. Combined with things like kernel and software management changes that we will be doing in the SIG, we need to be able to test them holistically and completely.
Metadata Update from @arrfab:
- Issue marked as depending on: #4
The #4 was closed in favor of CPE initiative https://pagure.io/cpe/initiatives-proposal/issue/15
The https://pagure.io/cpe/initiatives-proposal/issue/15 is still in progress, when it will be ready, you will be able to request VM through Duffy.
Metadata Update from @zlopez:
- Issue close_status updated to: Duplicate
- Issue status updated to: Closed (was: Open)
to comment on this ticket.