SQL Server 2008 Clustering is better than ever thanks to the new installation process that supports the “rolling upgrade.” When this was announced at the PASS conference, I remember there was a huge cheer, so obviously the feature is well appreciated by SQL cluster administrators. Maybe there was an element of “it’s about time” to the cheer, but a cheer is a cheer.
The rolling upgrade means that when performing an upgrade of SQL Server, you do not need to bring down the cluster itself. You can upgrade the passive node(s) first, then request failover and upgrade the formerly active node when it’s passive. Then failback and you’ve upgraded the failover cluster with minimal downtime. This also applies to services packs and cumulative updates. Yay!
The new installation process means that SQL Server 2008 is installed on each node separately instead of via remote install from one node as with SQL 2005. The new install architecture enables the rolling upgrade feature but raises an interesting anomaly; you can start your SQL failover cluster with only one node. A one-node cluster does not make much sense but it is a temporary situation as you install SQL on the other node(s). This is called the integrated install. If you want to avoid the one-node cluster situation you can go with the advanced/enterprise install where all the nodes are installed first then the cluster is brought online as a unit.
You can find more details on the Microsoft SQLCAT web site. They have lots of useful whitepapers, tech notes and top ten lists based on real customer experiences. In this case, they could only come up with 6 in the top ten list but who’s counting? All I can hear is the cheering.
Six Failover Clustering Benefits Realized from Migrating to SQL Server 2008:
Author: Brian Egler