[ovirt-users] Re: [Help] The engine is not stable on HostedEngine with GlusterFS Herperverged deployment

It's an effect that also had me puzzled for a long time: To my understanding gluster volume command should only ever show peers that contribute bricks to a volume, not peers in general. Now perhaps an exception needs to be made for hosts that have been enabled to run the management engine, as that might require Gluster insights. But I have also seen that with hosts that I only added temporarily to oVirt as a compute node and I believe I have even seen really ill effects: Given G1+G2+G3 holding 3R or 2R+1A bricks for the typical oVirt volumes engine/data/vmstore, I had been adding C1+C2+C3 as mere hosts. When I then rebooted G3, while C1+C2 were also down, all of a sudden G1+G2 would shut down their bricks (and refuse to bring them back up) for lack of quota... I had to bring C1+C2 back up to regain quota and then delete C1+C2 from oVirt to take them out of the quorum for a volume to which they contributed no bricks. And then often enough, I had to actually detach them as peers via the gluster CLI, because the GUI didn't finish the job. Now of course, that only works when G1+G2+G3 are actually all up, too, because otherwise peers can't be detached.... I've just posted a query on this issue yesterday: Hopefully someone from the development team will shed some insights into the logic, so we can test better and potentially open an issue to fix.