Monday, April 27, 2009

Windows 2008 Failover Clustering

I just taught the Windows 2008 clustering course last week and figure I better get some notes related to it before I forget. It seems that every course I teach no matter how much I think I know, I have new info highlighted by students concerns or just plain old figure out something new. This course was no exception:
  • Clustering in Windows 2008 is much easier than previous versions of Windows. The wizardized process is very easy. You barely need to understand what clustering is to get this up and running.
  • Printer clusters now store their drivers as part of the printing cluster on shared disk. You no longer need to worry about synchronizing drivers on different cluster nodes.
  • File share clusters are accessible only by name and not IP address. It seems like they are using a technology similar to host headers on an IIS Web site.
  • iSCSI shared storage can be access two different ways. Both nodes in the cluster can share the same target (the way I've always done in the past) or two targets can be created that point to the same LUN on the SAN, one target for each cluster node. Both function fine. However, having two targets on two separate HBAs in the SAN can provide some redundacy and opportunity for higher performance.