Hi Thomas, On Fri, Jan 21, 2022 at 11:16 PM Thomas Hoberg <thomas@hoberg.net> wrote:

In the recent days, I've been trying to validate the transition from CentOS 8 to Alma, Rocky, Oracle and perhaps soon Liberty Linux for existing HCI clusters.

I am using nested virtualization on a VMware workstation host, because I understand snapshoting and linked clones much better on VMware, even if I've tested nested virtualization to some degree with oVirt as well. It makes moving forth and back between distros and restarting failed oVirt deployments much easier and more reliable than ovirt-hosted-engine-cleanup.

I agree - and IIRC the documentation does recommend reinstalling hosts on failed deployments rather than using ovirt-hosted-engine-cleanup. If you have a polished PXE framework, reinstalling a host should not take much clock time and close to zero actual-work time.

Installing oVirt 4.10

4.4.10?

on TrueCentOS

What's that?

systems, which had been freshly switched to Alma, Rocky and Oracle went relatively well, apart from Oracle pushing UEK kernels, which break VDO (and some Python2 mishaps).

I'm still testing transitioning pre-existing TrueCentOS HCI glusters to Alma, Rocky and Oracle.

While that solves the issue of having the hosts running a mature OS which is downstream of RHEL, there is still an issue with the management engine being based on the upstream stream release: It doesn't have the vulnerability managment baked in, which is required even for labs use in an enterprise.

Not sure what you mean here, exactly.

So I'd like to ask our Redhat friends here: How does this work when releases of oVirt transition to RHV?

RHV is built and supported only on RHEL, and RHV and RHEL releases are coordinated. So e.g. if in the next release of oVirt/RHV we want to use some feature that will only be included in the next RHEL release, we make sure to release RHV slightly after that. For oVirt, this is in principle irrelevant, because these new features are already in CentOS Stream by the time of oVirt's release. Sometimes we do have to manually help with this, e.g. because the "Advanced Virtualization" part of RHEL is not maintained in the main CentOS Stream process, but in the CentOS Virt SIG. I think all of this is quite clear for anyone following the relevant release announcements and the occasional discussions about missing parts.

Do you backport oVirt changes from Stream to RHEL?

Not sure what exactly you mean here. oVirt is upstream of RHV, not RHEL. RHEL is not related to oVirt. Upstream of RHEL is CentOS Stream.

When bugs are found in that process, are they then fed back into oVirt or into the oVirt-to-RHEV proces?

Again, not sure what you mean. It sounds to me like your question(s) are purely about RHEL/CentOS, unrelated to oVirt/RHV. Please clarify. Thanks. The only thing in oVirt related to your question(s) AFAIU is the ovirt-engine-appliance and ovirt-node images. These are built/published/announced as needed if important CentOS security bug fixes are released. Again, this should be obvious to anyone following our announcements. Best regards, -- Didi