Here's an example with one shared Storage Domain that works. (local.ca is an internal
domain)
/etc/exports (n1.local)
-----------------------
/data/images/ovirt/n1_data_storage *.local.ca(rw,async,wdelay,root_squash)
/data/images/ovirt/n1_data_storage 192.168.1.0/24(rw,async,wdelay,root_squash)
/etc/exports (n2.local)
-----------------------
/data/images/ovirt/n2_data_storage *.local.ca(rw,async,wdelay,root_squash)
/data/images/ovirt/n2_data_storage 192.168.1.0/24(rw,async,wdelay,root_squash)
NFS mounts (n1.local)
---------------------
Server
n1.local.ca:/data/images/ovirt/n1_data_storage
Mount point
/rhev/data-center/mnt/n1.local.ca:_data_images_ovirt_n1__data__storage
Server
n2.local.ca:/data/images/ovirt/n2_data_storage
Mount point
/rhev/data-center/mnt/n2.local.ca:_data_images_ovirt_n2__data__storage
NFS mounts (n2.local)
---------------------
Server
n1.local.ca:/data/images/ovirt/n1_data_storage
Mount point
/rhev/data-center/mnt/n1.local.ca:_data_images_ovirt_n1__data__storage
Server
n2.local.ca:/data/images/ovirt/n2_data_storage
Mount point
/rhev/data-center/mnt/n2.local.ca:_data_images_ovirt_n2__data__storage
Like I said, local storage is not ideal, because you do lose some functionality. Also, If
you lose a node, you also lose it's storage and its VMs, but at least this way (using
NFS), you can keep it all within the same Data Center which gives you more flexibility by
not isolating each node's local storage to its own Data Center.
Once I move that /data directory (where the VMs sit) off the root filesystem "/"
to it's own LV, I'll try the upgrade again. I did find examples on how to shrink
an NSF, but I'm not sure how to handle oVirt's onn-pool00-tpool.
I did try reinstalling a node from scratch, but the custom disk partitioning during setup
doesn't seem to work. It would be nice if oVirt automatically partitioned the /data
directory to its own LV while sizing it with the remaining disk space.