My test cluster is now working, with one instance of SQL Server 2005 and two SQL Server 2008 instances. Yippee!
I had a couple of issues with clustering SQL Server 2008, and it is the purpose of this blog to outline these.
When you install SQL Server 2008 into a cluster, it does not automatically install the instance into all the nodes you specify in one installation. You install on the first node, then run setup again on any others, and join these installations into the cluster. After I ran the first 2008 install on the first node, the SQL Server Network Name would not come online. The following was logged into the System Event log:-
The computer account for Cluster resource ‘SQL Network Name (WZTEST)’ in domain our.domain.com could not be created for the following reason: Unable to create computer account.
The text for the associated error code is: Your computer could not be joined to the domain. You have exceeded the maximum number of computer accounts you are allowed to create in this domain. Contact your system administrator to have this limit reset or increased.
The Cluster Service Account may lack the proper access rights to Active Directory. The domain administrator should be contacted to assist with resolving this issue.
I googled this and there is lots to look through, and specifically I was suspicious of there being a problem with the cluster service account. The only rights an account needs to have by default in Active Directory to add computer accounts to the domain is to be part of the group ‘Authenticated Users’, which this account was. I tried putting it into ‘Domain Admins’ temporarily (and yes I did remember to remove it afterwards!). The Network Name still wouldn’t come online, but after I installed the instance into the second node and joined that to the cluster, it just started working okay. Mysterious, but I wasn’t complaining.
The second issue I had was when trying to run the install on the second node of the second 2008 instance, I got an error:
The current SKU is invalid.
SKU is jargon for ‘Stock Keeping Unit’, and usually refers to the version of SQL Server (e.g. Enterprise Edition 64-bit English .. etc). In this case that’s not relevant, it is a bug – see the relevant connect item here. The workaround is to run the setup on the second node from the command line with the /addnode switch. Hey presto, that worked!
So after that I have a spangly new test cluster which is working perfectly.
Needless to say when installing Windows 2003 clustering I had all the usual problems with clustering MSDTC. My boss thinks I’m just rubbish at it, she may well be right. It seems that I do exactly what I’m meant to, then try to cluster it, and it doesn’t work. So I remove it all and start again, and then it works. One learning point: enable network DTC as per this article, then before you install clustering, ensure the DTC service is stopped.
I also had a few weird issues with the quorum disk and with some of the other partitions, but I’m assuring myself these were simple glitches, everything seems stable now.