I’ve been meaning to put this post up. Anyhow, I put together a SQL cluster build guide here on a project. We had to rebuild the cluster and Paul was tasked with the rebuild. He followed my build guide but was not able to install the SQL cluster. When the install goes out to install SQL on the other node, it would fail without giving a meaningful error message.
Why? Good question. Did you know while creating a SQL 2000 cluster on Windows 2003, you have to configured named-pipes communication because Windows 2003 is aware the SQL 2000 base install (without SP3) has some security issues making the box vulnerable to attacks so… the OS automatically blocks certain communication to SQL until you install SP3.
This is a problem… you need this communication enabled to install a SQL cluster. To fix this, you have to enable a named-pipe for the install. Click the following link for the details:
http://support.microsoft.com/?kbid=815431
Well… I knew about this issue and it was documented in my guide. Using the same guide, I was able to successfully create a cluster the first time. Poor Paul rebuilt the cluster at least 5x. What stinks was our cluster (two Unisys ES7000) takes 10+ mins or so just for a reboot. This does not take into account how long the SQL install takes. Ugh… poor Paul.
He did figure it out though… I usually map a drive to our software share. He was using a UNC path to the share (\\ourserver\sql2000\xxx). Yup… that was the problem. Don’t you love it?
So what is the point of the story?
Always use a mapped drive. Avoid using UNC paths whenever possible.
And… Paul is one of the most patient people I know. But I’m still the best looking… at least that is what my mom tells me.