Start a Conversation

Unsolved

This post is more than 5 years old

555

March 27th, 2009 08:00

EMC Devices automatically changed in Windows cluster

Hi Everyone,

I am having a very strange issue that I am hoping someone has seen before:

I have a Win 2k3 ENT sp1 64bit cluster - where it appears at some time, I am guessing during weekly reboot, that 2 identically sized devices 'swapped' and now contain incorrect information.

Example: Drive T contains sql log A data = device 123
Drive N contains sql log B data = device 456

After a reboot(?) - the configuration is as follows:

Drive T contains sql log A data = device 456
Drive N contains sql log B data = device 123

Notice that the drive letters and corresponding data did not change, only the devices - thus SQL did not complain. Since the data on the associated devices (STD) has changed, this affected our remote bcv which is mounted by another host which noticed that the log data was associated with the wrong drive letters on that host.

I have confirmed with support that replication is working properly and I am sure I can get the environment back by changing the drive letters and moving the data, but I am baffled by how this could happen.

Any insight will be greatly appreciated. Thanks in advance!!!

9 Legend

 • 

20.4K Posts

March 27th, 2009 12:00

very strange indeed ..how are cluster resource dependencies configured ? What do physical disk resources have for their dependencies ?

March 27th, 2009 12:00

The physical disk resources have no dependencies at all.
The cluster resources are set in a standard configuration ( I think)
Quorum (no dependencies), Cluster IP Address (no dependencies), Cluster Name (dependent on Cluster IP Address), MSDTC (dependent on all above)

This is a SQL database so the SQL Server has a dependency on all physical disk resources (except quorum of course), SQL name, SQL IP address

Thanks

86 Posts

March 30th, 2009 20:00

I don't see how the data can be swapped, unless someone has done a restore and done it to the wrong devices.
Check the host and backup hosts symapi logs, and possibly symevent.txt file from an EMCReport.

March 31st, 2009 08:00

I agree.. I could see a dba mistakenly restore to the wrong drive, but in this case, they would have also had to change the drive letter in Windows, which I can assure you they would not do...

I'm tending to lean towards the MSFT cluster service but haven't found anything concrete as of yet..

May 8th, 2009 06:00

Just wanted to update this -

So call were opened to both EMC and MSFT and no resolution was ever found...

I don't think we'll ever know - I'm just concerned that it happens again.
No Events found!

Top