Windows 2008 Clustering – New and Improved
I finally got a working Windows 2008 cluster for SQL 2005. I figure one new set of problems at a time is enough. I iwl go for SQL 2008 on a Windows 2008 cluster later. One thing I noticed right away; clustering no longer requires a domain account for the cluster service. Everything runs on Local Service on each node. Kudos to the Cluster development team on this simplification. One less domain account to track.
The new terminology also takes some getting used to, but I can learn to call a "Resource Group" a "Service or Application". Installing SQL is pretty much straight up, except for needing to create an empty Resource Group Service or Application and put the correct disks in it. The Create Empty Service or Application command is not on the same menu and level as the Create New Service or Application command, causing a minor annoyance.
I did get to test out heterogeneous clustering with two different systems for the host nodes. Same brand of hardware, but very different models. We also used different brand HBAs in each machine. The validation wizard noticed but didn't really care.
Another very useful feature is the ability to change dependencies WITHOUT taking the SQL Service offline. After you add a disk to the Service or Application (Resource Group. I still want to call it a Resource Group), you can make SQL dependent on it without taking SQL offline as long as the new resource is online. This is great for adding new LUNs to an existing clustered instance in a non-disruptive manner. And yes, SQL sees them and will allow you to store SQL database files on these new disks.
Windows 2008 is rapidly becoming my platform of choice for a Clustered SQL Server.