Theoretically you most probably can, but you take the risk of being vulnerable to split-brain.
Keep in mind that you can:
- update gluster to the same version that is used on oVirt 4.4
- update oVirt to 4.3.10 as this is the most tested migration scenario
- reduce the volume and remove that 3rd node (don't forget to peer detach)
- reinstall with EL8 and install gluster (same version) again. Add the brick and qait for the heal
- create a new volume for the new HE
- Stop the old HE and deploy the new one from the backup
- Add the hosts to the new HE and migrate some of the VMs to free one of the old hosts
- Reinstall the remaining hosts (one by one) to EL8
Best Regards,
Strahil Nikolov
On Thu, Jun 24, 2021 at 14:28, Fedele Stabile
<fedele.stabile@fis.unical.it> wrote:
I have a 3 node oVirt 4.2.1 cluster,
i'm using a self-hosted engine architecture and
storage-engine is on a glusterfs replica 3
of the cluster.
The oldest CPU is SandyBridge family (E5-2609)
and incidentally it has connectivity problems with IPMI NIC,
but Hosted Engine HA is enabled and working.
May I reduce the glusterfs replica to 2 and use one of the 3 nodes to begin the installation of ovirt 4.4 on that node?
Fedele
_______________________________________________