Le vendredi 05 avril 2024 à 15:09 +0000, luc.lalonde(a)polymtl.ca a écrit : Hi that's it I successfully accomplished migration to centos 9 stream 4.5.5 in my production env, and all is ok thanks to live migration from centos stream 8 to centos stream 9 and so on. Not a very complicated path, but you have to reinstall from scratch each of your node considering that some hardware (like some HBA's qlogic controller) may be unsupported in el9. Good luck. -- Nathanaël Blanchet Administrateur Systèmes et Réseaux Service Informatique et REseau (SIRE) Département des systèmes d'information 227 avenue Professeur-Jean-Louis-Viala 34193 MONTPELLIER CEDEX 5        Tél. 33 (0)4 67 54 84 55 Fax  33 (0)4 67 54 84 14 blanchet(a)abes.fr

HCI has been deprecated years ago, but somehow the code survived until oVirt 4.5.5 or so. Which means it's still present in Oracle's 4.4 derivative. but not in their 4.5 release. On that base (make sure to use to switch to the Redhat kernel on all hosts and the management engine to avoid troubles) you can at least still build a HCI implementation that is in support (EOL not announced by Oracle) and much less beta than oVirt releases. When it comes to do an in-place migration of a HCI setup, open heart surgery may have more predictable results. Essentially you'd have to treat it like a disaster recovery and follow the path of restoring a management engine from backup. I've gone the migration way, created a new cluster and used a backup domain to transfer all VMs. Gluster in itself is quite independent, which might be a bonus here, while I mostly suffered from how it was never really integrated and had many mismatches in design philosophy. It's the storage domains on top that are at risk, but those again are relatively independent of the storage underneath, since they were designed for SAN, NFS and whatnot. Again, in theory, they can be imported elsewhere, because they have metadata describing them. I've done all my testing using play HCI clusters on VMware Workstation using nested virtualization, before I tried things on "live patients". Whatever you're planning, HCI has basically expired with oVirt 4.4 on EL8. If you're happy to loose everything, you may keep trying. I went with Proxmox using HCI on Ceph and while it's much more manual in many ways, it has a future, is way faster (no Ansible) and rock solid. Getting VMs migrated is a lot of bother, because none of these orchestrators were ever keen on making it easy for customers to defect. But since it's all KVM/QEMU underneath, tools that operate at that level will work. Now you'll just have to learn them or rebuild those VMs: that's "the cloud path" anyway...

I have QLA2692 HBA's that are no longer maintained... In the RHEL documentation it says that support may be removed in 'future' versions of RHEL. In my EL8 cluster I have this type of message: [ 3.290628] qla2xxx [0000:84:00.0]-00fb:5: QLogic QLE2562 - PCI-Express Dual Channel 8Gb Fibre Channel HBA. [ 3.290634] qla2xxx [0000:84:00.0]-00fc:5: ISP2532: PCIe (5.0GT/s x8) @ 0000:84:00.0 hdma+ host#=5 fw=8.07.00 (90d5). [ 3.290712] Warning: Deprecated Hardware is detected: qla2xxx:1077:2532 @ 0000:84:00.1 will not be maintained in a future major release and may be disabled I'm hoping that they still work when I upgrade to EL9. Otherwise, I'll have to buy new HBA.