06-20-2013 02:13 AM
I am posting here on behalf of a customer.
Scenario:
Solaris 10 SPARC with SF 6.0.1 Enterprise.
Storage luns are zoned to more than one system to accommodate snapshots, dg split and join.
Customer has found that a regular diskgroup (no CFS, no import with -t) can be imported on a 2nd system while imported and mounted on system 1 - no warning, no error message.
The problem obvious arises when the volumes are mounted and written to with resulting data corruption.
We have not had a chance to try and replicate the exercise in a lab environment.
Is there anybody out there with similar experience?
06-20-2013 03:39 AM
This should not be possible as the private region contains the host of the node that has the diskgroup imported, so you should only be able to import on a second node if you use the "-C" flag to clear the host name. When you deport a diskgroup the hostname is removed from private region so it can be imported elsewhere, so you should only use the -C flag in the case where you know the first node has crashed with diskgroup imported so host never got removed from private region.
Mike
06-20-2013 04:41 AM
Hi Marianne,
Are the hostid in /etc/vx/volboot file same for two nodes (perhaps truncated and rendered same)? I've seen this happen on a private dg with nodes having the same (truncated) hostid on HPUX.
06-20-2013 05:22 AM
Mike, that is my understanding as well.
rsharma1, I will ask the customer to check and let us know.
Thanks!
06-20-2013 07:25 AM
Hi Everyone,
I have forwarded this post to the customer and asked him to comment directly on the scenario observed and the exact sequence of events.
I know they did mention to me that they were able to reproduce the issue.
Thanks,
Rob.
06-27-2013 02:00 AM
Hi,
Did we hear back anything from the customer yet?
07-08-2013 01:57 AM
Hi, if it can be reproduced, why not ask CU to repro it and collect the VRTSexplorer for further diagnose?