Friday, July 23, 2010 3:52 PM
We have a HPC R2 Beta 2 cluster in a POC environment. we wanted to test head node clustering. I have read through the documentation regarding head node clustering but wanted to confirm a few things.
The documentation states that we need at least 2 LUNs. One for the witness and the other for the SQL db's. In our case we do not plan to host the SQL cluster on the head node and instead want to use remote DB's. Do we need the second LUN in our situation? Is it hosting or is needed for something else? Secondly, we don't have a shared LUN available in the lab to use as a witness disk. Can we use Node Majority and File Share Witness instead or is that not supported?
Monday, July 26, 2010 11:13 PM
I attempted the install and I was able to setup a majority node and file share witness based cluster without problems.
Now when I try to install HPC it comes up with an error message saying configure a file server resource to continue. How do i do this? Microsoft's documentation suggests to setup a full clustered file server on the head node with shared storage which we have no need for and do not want to create. Anyway i can get around this like pointing both nodes to a remote file server or share?
Secondly, how to do i point the clustered head nodes to the remote SQL db's?
Tuesday, July 27, 2010 9:15 AM
I did pretty much this with the HPC R2 beta but using iSCSI LUNs and a remote (non HA) SQL server.
I needed a LUN for the quorum and then you need a second one to create the HA Fileserver (HPC itself is not HA aware) and then added the HPC packs on top as per the docs. You select the remote SQL db instance during the HPC pack install(s).
Tuesday, July 27, 2010 5:56 PM
You will need to install SQL on a machine where the HN can connect to it. It is recommended that this network connection (HN to Remote SQL machine) be made through your Enterprise network and I would also recommend that you make sure that the SQL machine is not on the private network of your cluster. You can call the SQL db instance whatever you like but for installation purposes it is easier to use the default naming. That way in the HPC Setup GUI you just put in the name of the remote DB machine and things should work (instead of having to specify both the remote DB machine name and the instance name).
The remote DB docs should have the appropriate setps for setting up the required components of the DB (Managment, Scheduler, Diagnostics and Reporting).
Monday, September 13, 2010 10:20 PM
I still am unable to get this to work.
Here are the steps I took to try and set this up.
1: Setup two servers with 2008 Ent Ed R2
2: Setup Db's with default names on a remote SQL server. The Db names are the recommended defaults.
3: Exposed 2 LUN's to both servers and formatted to NTFS on the first server
4: Installed the file services role on both servers
5: Installed the failover clustering feature on both servers
6: Ran the clustering setup wizard to create a cluster and added an generic app instance and named it HPC (dont see the need for the instance but it said in the step-by-step guide to do it so i did)
7: Made the current node the owner of the instance
I ran the installer but i get the same Error Message:
"HPC has detected that this node is a member of a Microsoft Windows Server 2008 R2 Failover Cluster. Please configure an online file server resource with the failover cluster and make current node owner of the clustered instance to continue setup"
So i setup a File server cluster on the server. But i still get the same error. The only part i am skipping is the SQL setup part. Can somebody provide me with the steps i need to perform to get this setup? Micorosft's step-by-step guide is terrible and pretty limited to describing setting up everything on the same machine.
Wednesday, September 15, 2010 12:03 AM
Which guide were you using so I can check what it says? The R2 guide is different than the Server 2008 guidance due to the addition of remote DB support.
Creating a file server doesn't actually create the file server resource which is what HPC setup keys off of. Go back to your file server group and click on "Add a shared folder" in the Actions pane. This will cause a file server resource to be created in the group and it is that resource that is used by HPC setup.
HPC setup will then talk about a clustered instance and displays one or more names. These names will correpond to the Server and Application names show in the failover cluster manager.
- Marked As Answer by Kamran Niazi Monday, September 20, 2010 6:17 PM
Monday, September 20, 2010 6:17 PM
Yep that worked. Thanks!. That was the problem. Creating the share created the resource resolving the issue.