On Sat, Jul 6, 2019 at 4:58 AM Richard Chan <richard(a)treeboxsolutions.com>
Did a 4.1 - > 4.2.8 -> 4.3.4 upgrade in two hops. Here is my experience,
hope it can save you some pain.
1. For shutdown VMs, their disks were deactivated (though still attached)
2. For shutdown VMs, their NICs were "Up" but "Unplugged"
3. VM Custom Compatibility Version is hard to find. (If it is too old say
3.6, 4.1) they won't start up. The location is: VM -> Edit -> System ->
Advanced Parameters (click this)
"Advanced Parameters" is an expandable widget that is quite small when
collapsed. When it is expanded then you will see "Custom Compatibility
4. Storage Domain migration to V5: I use a NFS storoage domain. This is
migrated to the "V5" format and a lot of cleanup is done. If you have stray
files or images not owned by 36:36 (vdsm:kvm on my NFS server), the V5
upgrade will fail, there will be no SPM, and the Data Center will oscillate
between Non-responsive and Contending.
Why would there be non 36:36 uid/gid files? This is a long story, but from
the older 3.x days when there was less functionality, we would sometimes
directly manipulate the filesystem and/or qcow2 images using guestfish or
equivalent tools which would change the owner to qemu:qemu. Forgetting to
chown back to vdsm:kvm is a recipe for pain.
Bad permission are expected to cause failures when converting a domain,
since we need
to modify domain and volumes metadata. This will also cause failures to use
or in some other storage operations.