This is an unusual question that I'm hoping someone can offer advice.
I have a SQL server 2008 cluster setup between 2 VMware ESXi hosts (4.1) with shared FC LUNs for the data, tempdb and any required cluster devices (ie. quorum). The SQL server is part of our active directory infrastructure.
Unfortunately, we have been experiencing issues with our shared storage; so I would like to dismantle the cluster and make use of vmdks instead. Ideally, we remove the clustered nodes and go back to a single node.
However, when removing the cluster setup, the IP of the database changes. So the cluster IP will be removed and the single node IP will be used. Unfortunately, a lot of our apps point to the database via IP (yes, I realize this was a mistake and will be fixed to use hostnames in the future).
To get around the IP problem, I'm thinking of doing the following:
1) remove the 2nd node from the cluster. Now I have single node in a cluster.
2) create the appropriate number of vmdk disks and move the databases (user, master, model, msdb) and any other cluster devices (quorum) to the vmdks
Since I'm keeping the single node in a cluster, I don't need shared storage anymore (since there will never be a failover to a second node) and I'm assuming the newly created "clustered" vmdks SHOULD work since the SQL Server cannot differentiate
between local vmdks or shared disks.
Now I have a cluster with a single node and no dependencies on shared storage. But most importantly, I can keep the cluster IP intact.
Does this approach work? Is there an easier solution? This is just temporary until we can correct all of our apps to replace the IPs with hostnames. then the cluster will be completely uninstalled.