I have two nodes which boots from iscsi to implement windows 2008 cluster. And I'm using disk majority option as quorum over iscsi. But when the quorum's iscsi connection failed(May be san server reset), the failover cluster is failed too. If I reset one of the nodes, it can open, but its system disk goes offline. I cant change its status as online, because it says that its reserved by failover cluster(disk is on iscsi, beacuse iscsi boot). And this disk works as readonly. Anything on it cant be deleted or written. So, I cant rejoin the node to the cluster again. I have to reinstall windows. So, what I'm asking is, how can I implement more quorum backup? I mean, can I use both disk majority and file share majority at same time? AFAIK, every nodes also keep the quorum's copy too. But I don't know sometimes san servers goes offline. And quorum's iscsi connection and nodes' iscsi connections get lost. So, nor the quorum that is kept in the nodes neither the quorum iscsi disk is not enough to start the cluster again. I want to use both disk majority and file share majority at the same time. Can I do this? Have you any other suggestion? Regards.
this is an issue with any Microsoft failover cluster using Server 2012 or below using either disk majority or node and disk majority. The Disk witness is a point of failure for the cluster. Even if all nodes are up, if the disk witness fails (we loose access to it) the cluster goes down. so what to do?
1) Make sure your disk witness is highly available so is less likely to go offline 2) Consider using a file share and alternate file share witness instead, select node and file share witness as your Quorum type 3) Use Windows Server 2012 R2 for your cluster. Server 2012 r2 uses a dynamic Quorum model and doesn't suffer from the same issue.
I think you have to review your architecture. You actually have 1 component that cause 2 fault at the same time: you SAN server will cause 2 link failure. This scenario is usually not handled by failover cluster if you don't take extra coutermeasures. If you use a SAN server this have to be fault tolerant and never resetted if not for maintenance and you know that you will have to stop the cluster. I have written a failover cluster software in POSIX shell and tested it on windows too with CYGWIN, it can work but since it's classic unix script it use lot of fork and so on windows it consume much more cpu than on linux. If you can be interested in exploring new solution: https://github.com/nackstein/back-to-work/