Summary: Advanced cluster preparation install failed.
Details:

• Successfully installed advanced cluster preparation on first node.
• Copied configuration file to secondary node and attempted “install based on configuration file”.
• Install failed: “The setting ‘AGTDOMAINGROUP’ is not allowed when the value of setting ‘ACTION’ is Install.”
• Tried several times, always resulting in the same error. Unable to find anything relevant online.
• At this point, decided to simply uninstall and start over.
• Attempted uninstall on first node. Failed: Object reference not set to an instance of an object.
• Ok great, I can’t uninstall … what to do. Figured I’d try to do the advanced cluster preparation on secondary node without the configuration files. It worked.
• Alright, let’s try to complete the cluster “advanced cluster completion” option.
• Failed: The instance prepared for clustering that was installed on computer XXX failed during installation and cannot be clustered. To continue, remove the failed cluster-prepared instance and run the cluster-prepare Setup operation again.
• Fantastic. I now have two preparded nodes, but can’t complete the cluster.
• Well, let’s try the uninstall again, tried on both nodes. Failed: Object reference not set to an instance of an object.
• This just keeps getting better. Tried to do the repair installation on either node. Failed The SQL Sever feature is not in a supported state for repair, as it was installed in preparation to be clustered. Cluster prepared features cannot be repaired. To continue, uninstall the specified SQL Server feature.
• At this point, I am at a loss. Unable to find anything relevant online. Decided to open a Microsoft premier support case.
• They connect to my work computer via Easy Assist and spend a good hour or so recreating exactly what I had been experiencing. They were at a loss as well, and decided the next best route at this point was to attempt to uninstall via registry.
• Almost 2 hours pass and no progress has been made. After manually searching and removing every entry related to SQL on this node, they were unable to even launch the install as it was erroring out with a new error (did not record, but it was basically the same error you get when you attempt to uninstall an app from add/remove programs (appwiz.cpl) when the program doesn’t exist.
• At this point, Microsoft asked if they could continue to work on the problem. I said that we had already spent to much time and even had they got this working, I had little faith that I’d have a clean install moving forward. I made the decision to have my IT guys simply reformat the servers.
• A few days later, Microsoft deemed the problem a bug and credited our hours back. While it’s not neccessarily a bug, the feature to install from config file in advanced cluster prepare should not be available. The error is expected if the config file is provided through the GUI. The correct way to prepare the remote nodes [in case we use the config file] is to provide the config file at the command prompt.

What did I learn?
In looking back, yes, this was my fault. I should have followed the instructions, but honestly … why would they give you the option to fail? Why if you’re not supposed to use the GUI, is it an option!? AHH, oh well, done and over with. Moving on …

Summary: Validate SCSC-3 Persistent Reservation failure
Details:

• So at this point, we formatted the servers and started over, hooray.
• Worked out well because at this point we now have 4 nodes of the 6 available to install on.
• The pre install validation checks failed due to the rule “Validate SCSC-3 Persistent Reservation” failing.
• This is a windows cluster validation check and if windows fails it, so does SQL. Found the log file in C:\windows\cluster\reports in the Validation Report 2009.08.19 At 07.42.56 – Validate SCSI-3 Persistent Reservation
• Found a decent article on the matter, but again, this was above me http://www.servercare.nl/Lists/Posts/Post.aspx?ID=71
• Sent it over to my systems guys and they worked on the issue for a week or so with HP and this was their response to the matter:
  Somehow, one of the machines created a persistent reservation on a disk. Magically, the HP tool (hpprutil) couldn’t see the reservation. HP support couldn’t remote control the machine, as the install apparently wasn’t completed. Once they had that set up, they got in and saw that the array said there was still a reservation on the disk. They had to create a single-use password to force the machine to drop the paths that it wasn’t letting go of. It hasn’t been explained to me farther how in the world it would allow me to remove 6 paths, but the remaining 2 it held on to.

  In the middle of it all, machines were swapped out. The WWN should have stayed the same on all the machines because we pay an exorbitant amount of money for virtual connect, but of course VC didn’t work right and all the WWN’s changed. Then once the machines were firmwared-yet-again, they all changed again. Everything ended up being rezoned about 20 times each, along with all the time wasted trying to figure out just what the WWN’s should and shouldn’t be, why the array wouldn’t let the damn LUN go (but had the previous 6 times), etc.

What did I learn?
This one was out of my league. I didn’t understand it then, and unfortunately I still don’t understand it.

Summary: The credentials you provided for the sql server service are invalid & The current SKU is invalid
Details:

• Ok, disk problems resolved, resuming the install (not doing advanced prepare this time!)
• Successfully installed first single node cluster, hoooooray!!!!
• When attempting to add node to the cluster, it failed stating “the credentials you provided for the sql server service are invalid”. It Failed on SQL Server / SQL Server Agent. I attempted to change the password to no avail, failed every time.
• Found this article: http://support.microsoft.com/default.aspx/kb/955948/ which states “To work around this issue, type the domain account manually in the Account Name text box.” This is not true, as the boxes are grayed out and do not allow you to change the account.
• Unsure to as where I got the information from, but I had read somewhere that CU1 would resolve this specific issue (trying to find link).
• Ok, installed SP1 on the node which has the single node cluster setup.
• Attempted to add node, “Failed: SQL 2008 Cluster – The current SKU is invalid.”
• Found these articles: http://support.microsoft.com/kb/957459/ & https://connect.microsoft.com/SQLServer/feedback/ViewFeedback.aspx?FeedbackID=363036&wa=wsignin1.0M
• Install SP1 on a node that doesn’t have SQL installed huh? Interesting. The 2008 cluster install doesn’t use the task scheduler any more, and the nodes have to be manually installed, so why do I need to patch a server that doesn’t have any trace of SQL on it yet?
• Low and behold, when I went to install SP1 on each of the other nodes, it patched the “shared features” successfully.
• Tried to add node to cluster, successful on each node!
• Had to install SP1 again on each node after I completed the install.

What did I learn?
You need to install SP1 (or at least CU1) after installing your first single cluster node. After that, you have to install SP1 (or CU1) prior and post to adding any other nodes on the cluster. Crazy, but at least I know now.

Recap
1) Don’t slack. When you’re installing a new technology for the first time, do your homework and read the documentation.
2) Install SP1 before and after adding a new node for SQL 2008
3) Did Microsoft test a 2008 cluster install? To many oddities here that seem like some further testing would have easily resolved. I have not heard one way or the other if they will remove the feature to perform an advanced cluster installation from a configuration file or not. If I hear any more on the matter, I will post about it.