I’m using linstor on two locations (let’s call them datacenters ) in active/active mode. Performance is great. I tag nodes with aux properties to get resources deployed to both availability zones (datacenters/locations). Everything works just fine, except when I move a vm to the other site, I have to move the tiebreaker for its disk resource manually to the other site also in order to preserve the quorum and be able to continue to run the vm in case of inter-site connection failure. Not a big deal, I made some cron scripts for that. But it would be great, if linstor would have a feature to move the tiebreaker together with the primary to the node with the same aux property (parameter) when the primary moves.
Unless you are automatically failing over VMs between sites you could safely disable quorum all together. The problem you’re describing boils down to the two sites being unable to provide quorum in a meaningful way. Generally, you want quorum enabled to prevent split-brains in HA clusters, but if there’s no automated failover, then an admin would have to promote a peer (start a VM) while the DRBD devices were disconnected in order to introduce a split-brain.
Yes I have tested it on a few resources. I have set auto-quorum=disabled,quorum=off and auto-add-quorum-tiebreaker=false on resource definition. Everything works as expected. The only thing I find strange is that after every migration rd property auto-add-quorum-tiebreaker get set back to True. Doesn’t affect the function though. I’m using the linstor-proxmox plugin.