2013-03-07 08:16 AM
Several of our clusters (Red Hat cluster, Veritas cluster, XenServer pools) have a quorum disk that lives on the same aggregate as the data in that cluster. We have concerns about performance impact that the data volumes might have on the quorum disks. We have observed that occasionally, high load has saturated the heads and aggregate, thus causing the quorum disk to time out and fence the database/XenServer nodes. Is there any way/best practices for using a quorum disk on a shared disk array like this? Would it be best to put the quorum disks on their own spindles or is a head saturation going to bring them down, no matter what?
Should we just get a separate storage array to handle these quorum disks separately from the rest of our data?
2013-03-07 08:22 AM
having a dedicated aggregate for quorum only would cost you a few disks. If system is overloaded, it impacts all other aggregates as well.
Might be a good idea to have quorum on the partners disks so there is at least a chance if one head is jammed that the other one still reacts.