Disk issue in DL380 G4 Packaged Cluster

[Paul Snook_2]

Hello experts

I'm currently setting up a new DL380 G4 Packaged cluster with MSA1000. Both nodes run Windows Server 2003 Enterprise R2. For maximum redundancy, I've put in dual MSA1000 controllers and SAN switches. Each node contains two FCA2214 adapters and one adapter is connected to one SAN switch and the other adapter is connected to the other SAN switch. There is an MSA30 unit connected to the SAN via VHDCI SCSI.

I've just set up the logical drives using ACU. When using the Windows Disk Management tool however, Windows sees two copies of every logical drive I've created. I've created six logical drives and Windows sees twelve disks that it wants to write a signature on. This doesn't seem right.

Any suggestions as to why Windows sees double the number of disks it should? Is it because of the dual controllers and SAN switches? I've got HP SecurePath software. Should I install this before setting up the logical drives?

Thanks

[pabel]

Hi Paul, Windows sees every disk through 2 paths (one for FCA2214). This is a normal situation; you must install secure path before setting up the logical drives.

After secure path installation you'll see only 6 drives, in case of switch/fiber/hba/controller fail, it manages failover through alternate path.

Regards

PD. Sorry for my poor english.

[Paul Snook_2]

Thanks for the confirmation and quick response. I understood your English no problem.

[pabel]

Ok, good luck.

[Paul Snook_2]

OK, I've now installed the SecurePath software and it's fixed the problem with Windows seeing double the number of disks, but now I have another problem :-(

I've formatted the logical drives in Windows and installed cluster services on both nodes and all went well. I've done several reboots on both nodes and cluster resouces fail over with no problem. All of a sudden now though I can't start either node up even with the other one turned off. The GINA never appears and Windows sits for ages displaying "Running Startup Scripts". I can't logon locally but I can logon over the network. What I see in the event log is:

Cluster service is requesting a bus reset for device \Device\ClusDisk0.

and:

The driver for device \Device\Scsi\ql23001 performed a bus reset upon request.

and also:

Cluster resource Disk K: timed out. If the pending timeout is too short for this resource, consider increasing the pending timeout value.

I'm running the latest ProLiant Support Pack (PSP) - 7.50.

Could SecurePath not be configured correctly? Or could this be a disk signature problem?

Thanks

[pabel]

What is your cluster failback policy? Allow failback or prevent failback?

Regards.

[Paul Snook_2]

The cluster failback policy is set to default settings, i.e. Prevent Failback for all groups.

I've managed to get the cluster online, but only while both nodes are running. The cluster service won't start on only one node by itself. It is always in a "starting" state.

I'm also getting this message but I'm not sure it is related:

MS DTC has determined that the cluster service is configured but not running when MS DTC was started. MS DTC will continue to startup but it will not failover. Additionally, if the cluster service is started then please shut down this instance of MS DTC as it will become unusable. Error Specifics: d:\nt\com\complus\dtc\shared\mtxclu\mtxclusetuphelper.cpp:514, Pid: 1660, CmdLine: C:\WINDOWS\system32\msdtc.exe

[pabel]

Hi Paul, is MS DTC configured as a cluster resource? It should only be running on one node of the cluster at a time.

Regards.

[Paul Snook_2]

MSDTC is not configured as a cluster resource. I have not modified MSDTC since I installed the server.

Additional info: I've confirmed that there is not a disk signature or drive letter problem on either node by running the dumpcfg utility and comparing this information with that in the registry whilst the cluster disk driver is disabled.

Thanks for your continued help

[pabel]

Hi again Paul, I think your storage it's ok and you're suffering a windows cluster configuration problem.

In a Win2k3 cluster, MSDTC must be configured as a resource. Take a look at this page, it may help you:

http://support.microsoft.com/?id=301600

Regards.

[Paul Snook_2]

The cluster will be used as a file/print/FTP/DHCP server. Will MSDTC be needed in this case? I have another cluster with SQL and that has MSDTC configured as a cluster resource, but I didn't think it would be needed in the case of file/print/FTP/DHCP. Should I turn off MSDTC completely?

[pabel]

I've seen MSDTC installed only in Exchange & SQL Server clusters. You can disable MSDTC and try the failover again.

[Paul Snook_2]

I've trashed and rebuilt the entire cluster. I did a system erase on both nodes and I cleared the configuration of the MSA1000 and started from scratch.

I rebuilt both nodes and the SAN LUNs and started the Cluster Administrator on node 1 with node 2 off. The wizard completed successfully with no errors or warnings. I rebooted node 1 and I've got exactly the same problem :-( Windows fails to complete the boot process and displays "Running Startup Scripts" indefinitely. The logon dialog box never appears and I can't log on locally. However I can log on from over the network and I can see that no cluster resources apart from the Quorum resource is brought online.

However, I have identified the cause of the problem. We're running McAfee VirusScan Enterprise 8.0.0

Setting the Network Associates McShield service to Disabled and rebooting fixes the problem. Possibly the McShield service has the quorum resource locked, preventing the other resources from being brought online.