On March 11, 2020 9:45:30 AM GMT+02:00, Nardus Geldenhuys <nardusg(a)gmail.com>
wrote:
Hi Ovirt Mailing-list
Hope you are well. We had an outage over several ovirt clusters. It
looks
like we had the same ISO NFS domain shared to all off them, many of the
VM's had a CD attached to it. The NFS server went down for an hour, all
hell broke lose when the NFS server went down. Some of the ovirt nodes
became "red/unresponssive" on the ovirt dashboard. We learned now to
spilt
the NFS server for ISO's and/or remove the ISO's when done.
Has anyone seen similar issues with an NFS ISO Domain? Is there special
options we need to pass to the mount to get around this? Can we put the
ISO's somewhere else?
Regards
Nardus
Hi Nardus,
The iso domain is deprecated, but there are some issues when uploading ISOs to
block-based data domains.Using ISOs uploaded to gluster-based data domain is working
pretty fine (I'm using it in my LAB), but you need to properly test prior
implementing on Prod.
As far as I know, oVirt is making I/O checks frequently , so even the hard mount option
won't help.
Is your NFS clusterized ? If not ,you may consider clusterizing it.
Actually, did your VMs got paused or completely crashed ?
Best Regards,
Strahil Nikolov