Unanswered: installing SQL Server 2008 R2 in a Cluster
I have a couple of questions about installing SQL Server 2008 R2 Standard in a 2 node active\passive cluster
First, a little background. The Admin guys have not Clustered Windows yet. The hardware is due in next week so I'm just trying to get my notes together for the SQL install once the Cluster is available.
From what I'm reading, Failover clustering setup has a whole new work flow in SQL Server 2008.
These are the (very high level) steps as a see them to be:
- Run setup on the Active Node of the Cluster using the New Server Server Failover Cluster option
- Failover the resources and run set on the second (now active) Node using the Add Node to a SQL server Failover Cluster option
Is that correct?
Also, I'm not clear about how you would patch the installation and add Instances.
Would you run Service Packs and Cummulative Updates against both nodes? if so, in what order?
In order to add an Instance (I'll be adding 2 Named Instances), What setup option would you use? Would you run set up on the Active Node then failover and run it again to add an Instance?
Also about Named Instances... If I have a simple Active\Passive Cluster with one set of cluster resources and multiple instances, do I need seperate IP addresses for each Instance?
I apologize if the questions seem elementary but this is our first time with SQL Server 2008 and Clustering!
Any help or guidance would be greatly appreciated.
The Microsoft instructions for this are very good. The short of it is when you install, there is no "active cluster node", since there is no virtual server to install to. You will pick a node, install SQL Server on it, then move tot he next node, and install SQL Server there. In SQL 2008, there was a bug where if the license key was embedded in your installation media, it would not recognize it automatically on the second node, but you could manually enter it, and have no problems.
I have not done many service packs on SQL clusters, yet, but I believe you apply the service pack to the offline node, failover the instance, and complete the install on the newly offline node. This actually reduces the downtime immensely.
one thing to note, make sure that your server guys know to open the correct ports on the firewall for the two nodes of the cluster. These are listed in Microsoft's documentation on how to create a Windows cluster.
Each instance of SQL Server that you install on a cluster will need its own set of shared disks and TCP/IP ports. This is because you may sometimes run the instances on separate nodes to balance the load. When you do this, make sure that you do not overallocate the memory these instances can take up. Otherwise in those few times you need to run these instances all on one node, you may end up with very bad performance.
Yes, you would. The old setup routine (SQL 2005) would try to do all actions on all servers on the cluster. This would take some time in large clusters (4 - 8 machines), so they decided to break it down in this way. At least, that's the way I see it.