Signature Error with Cluster Disk..... 2

[sarav2k]

Hi

Recently We restored the full data(OS & App) for both cluster nodes.

But now we have problem with starting the cluster service. The error is the expected signature was not found on the cluster disk. The signature was changed because we replaced a failed disk , but it not the Cluster Disk. Still all the signatures were changed.

Event ID is 1034 for Clussvc.

There is a Reskit dumcfg.exe is there for W2k to write a signature on disk. Is there anything like this for Windows NT?


When I followed KB article for Event ID 1034 Clussvc. It corrupted the Cluster DB. Any other way?

Any Help is very much appreciated.

Thanks
Sarav

 

[Killman]

Is the procedure in this link the one you followed?

http://support.microsoft.com/default.aspx?scid=KB;en-us;243195&

We used this same procedure for a disk replacement on our NT 4 cluster and it worked.

 

[sarav2k]

Hi Killman,
Thanks for the response.

Initialy we got error message as cluster DB corrupted after I followed all the steps in the URL.

Then next time I changed the registry only .. I didn't delete the MSCS files in quorum drive as mentioned in the link....this time quorum Drive itself failed. So I am trying to bring up the node 2 alone in the anticipation of cluster server service might run on other node.

Any help?

Thanks
Sarav

 

[Killman]

Here are a couple of things to try:

Quorum Resource Fails

If the device that holds the quorum resource fails and cannot be brought online, the Cluster Service will not start. It can be started with a special parameter that starts the Cluster Service without a quorum resource. Then the administrator can use the Cluster Administrator utility to select a new quorum resource. To correct a quorum resource failure, implement the following:

1. Shut down one cluster member. Only one node should be running.

2. Use the Services option from Control Panel to stop the Cluster Service if it is running.

3. In the Startup Parameters box, enter "-fixquorum," then start the Cluster Service.

4. Use the Cluster Administrator utility to modify the properties of the cluster and select a new quorum resource.

5. Use the Services option in Control Panel to stop and restart the Cluster Service. This clears the fixquorum parameter that was passed. It is not necessary to clear anything from the Startup Parameters box, because anything entered is not saved.

6. Reboot the second cluster member.

This works as long as there is more than one physical disk on the shared SCSI bus. The fixquorum parameter does not bring the quorum disk online. Therefore, it is not possible to move the quorum resource from one partition to another on the same disk, since the disk is offline.



Quorum Disk or Quorum Log is Corrupted

If the quorum disk or quorum log becomes corrupted, the cluster server software will attempt to correct the problem by resetting the log file. This can be determined by examining the Window NT event log and looking for the message, "The log file quolog.log was found to be corrupt." The source of the message is the Cluster Service. If the quorum log cannot be reset, the Cluster Service will fail to start. If the Cluster Server software fails to determine that the quorum log is corrupt and starts, the message, "ERROR_CLUSTERLOG_CORRUPT," will be entered in the cluster log. To correct this problem, do the following:

1. Use the Service option from Control Panel to stop the Cluster Service if it is started. Do this on both cluster members.

2. On one node, enter "-noquorumlogging" in the Startup Parameters box for the Cluster Service and start the service. This starts the Cluster Server software without quorun logging, which means that the cluster files on the quorum disk will not be open.

3. Run a disk repair utility, such as CHKDSK, against the quorum disk. If the disk shows errors, allow CHKDSK to fix them. If CHKDSK reports no errors, the quorum log itself is probably corrupted. Delete the file quolog.log and any temporary files from the MSCS directory on the quorum disk.

4. Use the Services program to stop and restart the Cluster Service.

The only potential problem with the above procedure is that the quorum log stores cluster configuration changes until they can be communicated to all nodes. When the Cluster Service is configured to start without a quorum log, it is possible that recent configuration changes to the cluster could be lost. But, since the quorum log is corrupted anyway, starting the cluster with a quorum log is the best solution.

Hope this helps....

 

[sarav2k]

Hi Killman,

Thanks. I am able to open the Cluster administrator after I started the cluster service using -fixquorum parameter.

After that I noticed, all the cluster disks have failed due to mismatch of expected signatures(Event ID 1034 ).

So I used FTedit to get the correct signatures, But as soon as I change it in the registry . It is inaccessible\failed.
Can you tell me why this is happening?

Is there any way to write the Disk signatures (in WinNT) instead of changing it in the registry as suggested by KB article for Event ID 1034?

Any help is very much appreciated.

Thanks once again
Sarav

 

[sarav2k]

Killman ...Thanks. I changed all the signatures and all resources are working.

By mistake I evicted the 2nd node how to bring it back. Do I need to uninstall cluster server and install back on evicted node, When I tried to do this also....It is asking to remove the installation before uninstalling the cluster service.

Hoe can I join back the evicted node safetly?



Thanks a lot.
Sarav

 

[Killman]

Yes, if you evict a Node you need to uninstall and re-install MSCS.

 

[Killman]

PS: On the evicted Node. Im sure you already knew this but i just wanted to be clear on the above statement.

 

[sarav2k]

Yes killman.....everything working fine....


Thanks a lot
Sarav