How to migrate the quorum to a standalone machine ?? 2

[gqma0]

Hello MSCS gurus.

I really need your expertise on this case.
One of my colleague is trying to achieve a task I can't help with.

He has an Active/passive cluster and he wants to migrate the data and quorum to a new drive.
However, the new storage is connected to another standalone server because of HBA driver issue with new storage. The HBA that connects to the old storage is incompatible with the new storage and vice-versa for the new HBA. Also they don't have enough physical PCI slots to add another HBA in the nodes.

We know how to migrate the shared data between the cluster and the standalone system but he's trying to migrate the quorum disk to the standalone server.
Once the migration is completed, they disconnect the old disk and connect the new disk with data and quorum replicated to the cluster.

What's the best approach to achieve this task ??

One of my suggestion was to add an old HBA (compatible with the old storage) on the standalone server, make it join the cluster, move the groups to that 3rd node, add the new storage as disk resource, migrate the data locally and move the quorum disk, evict the 1st 2 nodes, change the old HBA to a new HBA, make the 1st 2 nodes join the cluster and move back the groups. I'm not candidate to mess around with the quorum.


Thanks For your suggestions
Regards
Gaetan

 

[GiaBetiu]

Your idea is good.
But, don't be so afraid of the quorum information, It is stored also in the registry in a hive (so, it is also found in the \winnt\cluster as clusdb).

The cluster parameters (resources, groups) is not such a big problem you can always import them loading the right clusdb file. The problem that I see it is the applications that are running. If there is an SQL server, then, I think that all that you can do is to reinstall it in the new cluster. So, this trick will just help you to have same resources (objects) but you don't know how many of them will be able to come online (without to have the real associated object).
With my knowledge, but I have also to test it (if yo give me time), the procdure could be:
1. In the old cluster:
- stop cluster service on both nodes
- go to regedit, HKLM/Cluster
- choose from menu: unload hive
- save winnt/custer/clusdb
- run dumpcfg and note the disk(s) signature

2. On the new cluster (just one node)
- disable cluster service (via services)
- disable cluster disk driver (via device manager/show hidden devices...)
- using dumpcfg write the disk signature saved above to the new disk(s)
- copy clusdb saved above in the winnt/cluster
- use regedit and load hive (the new clusdb file)
- start your cluster service manually with the option –ResetQuorumLog. this will reset the quorum log from the quorum disk
- at this moment the setings for the cluster should be the same like the old cluster
- the new server has to be in the same network, same address
- you try then to start the resources one by one

Well, I'm not sure if it will work, as I said, I have also to test it. However it should be a logical procedure.

Your method is more hardware oriented, but it will sure work.

Gia Betiu
gia@almondeyes.net
Computer Eng. CNE 4, CNE 5, MCSE Win2K
new:

http://www.almondeyes.net

(just started)

 

[gqma0]

Gia,

Thanks for your reply. Much appreciated.
The second machine will never have MSCS installed, therefore it won't even be a single node cluster. The data and quorum disk is migrated via a replication software using IP.

Once the data is replicated on the new disk, he wants to swap the old storage (on the cluster) with the new storage (on the standalone server). No changes will be made to the replicated quorum disk.

If you could test this procedure, that would be great. I want to make it as simple as possible with minimum disruption and propose the best solution.

Thanks again
Regards
Gaetan

 

[GiaBetiu]

If you will use the same computers, and you want to change just the shared device, then the above procedure is more simple, being a clasical procedure for replacing a shared disk.
Is like I said above, just that you don't have to migrate any clusdb. The node wil remain there, so the registry information will be there.
Shortly (as i said in previous message):
- backup!
- disable cluster service
- disable cluster disk driver
- use dumpcfg to extract the disk signature and remember it
- replace the disk
- add the new disk
- write the signature with dumpcfg: dumpcfg.exe /s signature disknumber
- start the cluster disk driver
- start the cluster with the option –ResetQuorumLog.

And, that's all. How safe it is? Is Microsoft receipe.

Gia Betiu
gia@almondeyes.net
Computer Eng. CNE 4, CNE 5, MCSE Win2K
new:

http://www.almondeyes.net

(just started)

 

[gqma0]

Thanks again.

I'm already testing the procedure but I have issues with the HBA and the new shared storage. Will let you know the results.

Cheers

BTW, nice site.

Gaetan

 

[gqma0]

OK.

I tried

- backup!
- disable cluster service
- disable cluster disk driver
- use dumpcfg to extract the disk signature and remember it
- replace the storage array with the new one
- write the signature with dumpcfg: dumpcfg.exe /s signature disknumber
- start the cluster disk driver
- start the cluster with the option –ResetQuorumLog.
and I keep getting

A system error has occured
System error 1067 has occured
The process terminated unexpectedly.

Tried to apply the KB257905 which does not solve the issue.

- Disk signature is correct
- Tried to copy back the clusdb in its original location
- Tried to start with -noquorumlogging
- Tried to reload the hive

the system log shows
It has done this 11 time(s). The following corrective action will be taken in 960000 milliseconds: Restart the service.
The Clustering Service could not join an existing cluster and could not form a new cluster. The Clustering Service has terminated.
The Cluster Resource Monitor could not load the DLL clusres.dll for resource type Physical Disk.

Any ideas ??
Regards
Gaetan

 

[gqma0]

Gia,

Figured out what was wrong.
I forgot to set Cluster disk driver back to system and enabling it wasn't sufficient enough

explained in

http://support.microsoft.com/default.aspx?scid=kb;en-us;299041

Thanks for your help
Much appreciated.
Cheers
Gaetan

 

[GiaBetiu]

Good!
In my explaination it was that step too. Is good that everything is OK. If you have questions in the future, don't hesitate to contact me.

Success!

Gia Betiu
gia@almondeyes.net
Computer Eng. CNE 4, CNE 5, MCSE Win2K
new:

http://www.almondeyes.net

(just started)

 

[ShackDaddy]

All this was a big help to me, thank you, both of you. I'm upgrading my SCSI shared array with larger drives and it seems like you are describing the steps that I need to take as well.

- backup data and quorum volumes
- stop SQL service
- stop cluster service
- stop cluster disk driver
- use dumpcfg
- replace all drives
- recreate RAID volumes
- write signature with dumpcfg
- start cluster disk and cluster with -ResetQuorumLog

Will backing up the quorum drive be a problem if I haven't stopped the cluster service yet? I mean, will critical files be locked or anything? Is the disk signature only important for the quorum drive? Why do I need that particular disk signature?

ShackDaddy

http://david.shackelford.org

 

[GiaBetiu]

The backup proceddure HAS to be done with the cluster service running, and from a node that is the owner in that moment of the shared disk (check to be sure in Computer Management/ Disk Management, and be sure that the shared disk is there).
As I said in above messages, the quorum data is an information that is mirrored between the two nodes and the shared quorum disk (there is a copy of it in the registry of every node).
AT the end, after restore, the option –ResetQuorumLog will recreate the quorum on the shared drive used as quorum drive.

The Disk Signature is part of the private properties of the disk resource.
Try this command:
cluster cluster_name resource "Disk Q:" /priv
and you will see the disk signature.

Gia Betiu
gia@almondeyes.net
Computer Eng. CNE 4, CNE 5, MCSE Win2K
new:

http://www.almondeyes.net

(just started)