Hi Adam, The arbiter holds only metadata for gluster to be able to resolve any potential split brains. It can run on another host (not in the 2 replicas) and the only requirements are: 1. Low network lattency - you don't want to put it in Chona, unless your replica servers are also there :) With remote arbiter - the lattency will affect you only when 1 of the replicas is down. 2. Fast storage that can easily create the files' inodes. I would preffer a SSD , but even high speed rotational disks are OK In other words, you should create gluster bricks like this one: server1:/gluster_bricks/myvolume/brick1 server2:/gluster_bricks/myvolume/brick1 arbiter:/gluster_bricks/myvolume/arbiter_brick In this case, if a node dies - the other node + arbiter will have quorum and will serve any I/O for oVirt. The arbiter_brick does not require much space, as only metadata will be written. Actually with hyperconverged setup you have 2 clusters on the same nodes - storage (gluster) & oVirt (virtualization). Best Regards, Strahil NikolovOn Jun 17, 2019 15:06, adam.fasnacht(a)gmail.com wrote: