Clustering

[Ayman Altounji]

I have had the same forementioned problems with access to a 3500 controller with a 2 node cluster. Node 1 was complete including cluster configuration and access to the shared storage. When installing the 2nd node the server hangs accessing the 3500 until node1 is shutdown. Consequently, the second node installed the boot partition to 'E' and shared storage to various other drives. Has anyone successfully completed a cluster configuration under this scenario. I have loaded support paq's and SP1.

[Ayman Altounji]

I'm running into the same problem with my cluster. I haven't found the resolution either. If anyone solves this it the info would be greatly appreciated.

I'm thinking of running the system on Node 1 for now, until the problem is solved. Has anyone done this as a temporary fix to maintain uptime?

[Ayman Altounji]

For Node 1:
1) Run SmartStart on Node 1 to configure it and load the System Configuration Utility,
2) Load Windows NT Enterprise Edition, Service Pack 3, and the Support Software for Windows NT a.k.a. NT SSD.
3) Install the CR3500 Configuration Utility into the new Windows NT installation from the [CD-ROM drive letter]:\CR3500\ENG directory of the SmartStart CD.
4) Run the CR3500 Configuration Utility from the Compaq System Tools folder and configure the array as desired.
5) Run Disk Administrator, create volumes out of the logical drives, and be sure to format them NTFS and assign drive letters to them. NTFS is the only file system supported for the shared storage.
6) Install and configure Microsoft Cluster Server.

For Node 2:
1) Run SmartStart on Node 1 to configure it and load the System Configuration Utility.
2) On this node, change the Ultra2 SCSI Adapter ID from ID 7 to ID 6
3) Load Windows NT Enterprise Edition, Service Pack 3, and the Support Software for Windows NT a.k.a. NT SSD
4)Run Disk Administrator to set the drive letter to match the drive letter for the shared storage area which was created on Node 1
5) Install and configure Microsoft Cluster Server.

It sounds as though you have not configured the 2nd controller as SCSI ID 6 as indicated in step 2 of Node 2.

[Ayman Altounji]

Lee N - thanks for the prompt response, but I'm running Windows 2000 Advanced Server.

To add further definition to the problem set. I had experienced a problem this morning. Both nodes were running as follows:

NODE1: Exchange Server 2000 (preferred owner)
NODE2: File services (preferred owner)

Node 2 reported an error in its event log stating that it was unable to open connections; at the same time users starting complaining that they were unable to access mapped drives and file shares.

I opened the Cluster Administrator and it hung on Node 2. So I tried to open the Cluster Administrator on Node 1, it also hung the application. So I powered off Node 2 and subsequently was able to access the Cluster Administrator from Node 1. As expected Node 1 took control of all the applications.

Currently, Node 1 is running on its own. Tomorrow I'm going to power-up Node 2, fail-over all the applications to it and shutdown Node 1. I want to unit test each Node, then if they pass bring both nodes online and see if the problem can be recreated.

In the past, have you seen the problem I'm encountering? Any guidance would be greatly appreciated.

[Ayman Altounji]

We had the same problem the solution we found was to disconnect Node one while we were configuring node two

[Ayman Altounji]

I have been experiencing similar problems. I have followed the compaq documentation, which boils down to this.

install win2k and sp1 on node 1, install support paq on same. No problem.

no mention is made of shutting it down, so i didn't.

Install win2k and sp1 on node 2, install support paq on same. No problem.

Again, no mention is made of shutting anything down.

Config the shared storage from node 1. Done, no problems.

That was as far as i went (taking me to 8pm on friday night, alone in the datacenter!

Now, if you read microsofts step by step guide to installing clusters, i've already messed up, as both nodes are up, with the shared storage powered on, but i'm going with compaq for the moment!

The plan now is to shutdown node 2, then install cluster services on node 1 (ie, following the microsoft way from here on in). Anyone got a better idea?

I've been using smartstart 4.9 if that helps.

[Ayman Altounji]

Right, I've got it sorted on a test box. Proceed as follows.

Power on shared storage and node 1, boot from smartstart, setting the id of the scsi card to #6, install win2k, sp1 and support paq on node 1.

While you are there, config the internal network card (I used 192.168.0.1)

Shutdown node 1.

Power up node 2, boot from smartstart, setting the id of the scsi card to #7, install win2k, sp1 and support paq on node 2. While you are there, config the internal network card on this one (I used 192.168.0.2).

Shutdown node 2.

Power up node 1 again, and run the CR3500 config utility, setup the disks on the shared storage as you require. quit the utility, then create partitions and volumes as required using logical disk manager. (Don't be surprised if it takes a few minutes before you can see the disks in disk manager after using the CR3500 utility).

If you get prompted to write signatures to the disks and convert to dynamic, do so, then just right click on them and select 'revert to basic'.

Assign drive letters as required.

Shutdown node 1.

Power up node 2, and check that you can see the shared disks from it. you will probably have to change their drive letters so that they match those assigned on node 1.

Shut down node 2.

Boot node 1 again, and install cluster services.

Leaving node 1 up, boot node 2.

Install cluster services on node 2 (joining an exisiting cluster).

Done, hopefully.

Regards,

Rob Ellis
robert.ellis@ntl.com

[Ayman Altounji]

For what it's worth, I've found it's easiest to setup the first node using SCSI id 6... then when you setup the 2nd node, and node 1 is already working, you won't run into issues with it freaking out before you've managed to switch that SCSI id on the second node.

Works great... just whichever node you setup first, make it use SCSI id 6 for the connection to that cluster drive.