I recently began managing an Ovirt cluster. It has 5 cpu nodes and 3 gpu nodes in two
clusters. When I started it was at 4.2.8 with a standalone engine. I have henceforth
been steadily upgrading it to 4.5.4. I followed the upgrade guide and was successful in
upgrading all 9 nodes from 4.2 to 4.3.10. I then upgraded the engine to centos 8 with 4.4
engine. I then rebuilt each of the nodes in the two data clusters to centos 8 ovirt 4.4
using the ovirt node installer iso. They all joined the engine correctly. I then
upgraded the compatibility to 4.3 in both clusters and rebooted all the running VM's
to also be at 4.3 compatibility. I then upgraded the engine to 4.5.4 and proceeded to
upgrade two nodes to 4.5.4. They were successful and reattached to the engine correctly.
Here is where my issue arises. Any VM that I migrate to the 4.5.4 node from a 4.4.10 node
migrates correctly and then within 5 minutes will automatically migrate back. If I force
pin a VM it stays running.
I have searched the logs but not found anything to explain such behavior.
Any assistance or idea ?