Closed intentions closed 6 years ago
In this case, you would create multiple groups and associate a fence device with each pool.
You can set resource co-location constraints to pin each pool to the node you want, allowing failover of that workload to the other node.
fence-vol1 (stonith:fence_scsi): Started zfs3-node1
Resource Group: group-vol1
vol1 (ocf::heartbeat:ZFS): Started zfs3-node1
vol1-ip (ocf::heartbeat:IPaddr2): Started zfs3-node1
fence-vol2 (stonith:fence_scsi): Started zfs3-node2
Resource Group: group-vol2
vol2 (ocf::heartbeat:ZFS): Started zfs3-node2
vol2-ip (ocf::heartbeat:IPaddr2): Started zfs3-node2
Thanks!
Hello,
I've got a question about the fence_scsi, I have a similar setup to yours but instead of everything being in one big pool I have two pools that are supposed to be in an active-active configuration (so one pool being exported from each head unless there is a problem with one of the heads).
for the fence_scsi do I create two different stonith objects (one for each pool) or one big stonith object for both pools?