we have a cluster using 2 building blocks, each one containing 2 servers and 2 disk cabinets. The servers of each BB can only access de disks inside their BB.Thus, if 2 servers from different BB stop, GPFS works as usual, but if the 2 servers of the same BB stop, the GPFS cluster will not have access to half its NSDs and will have to stop.
How would you setup the quorum in this scenario? Does it even make any sense?
Thanks in advance,
Pinned topic Quorum and Building Blocks
ezhong 060001JWNJ33 Posts
Re: Quorum and Building Blocks2013-06-13T02:56:30ZThis is the accepted answer. This is the accepted answer.
You'd need to have 5 quorum nodes which would enable the GPFS cluster to stay up after 2 servers fail. On the non-server quorum node, you should have a desc-only disk to help maintain disk quorum.
It's better to replicate both metadata and data and put disks into different failure groups to enable a filesystem to stay up even after 2 servers for the same BB fail.