Originally posted by: SystemAdmin

Hi,

I'm building two nodes(active and standby) HACMP environment when strange thing happens as belows:

there's two shared disks: hdisk2 and hdisk3, upon which I build datavg1 and datavg2 as shared VG seperately using cluster LVM, but I find that both node1 and node2 are able to vary on the two shared non-concurrent VG at same time whether or not HACMP is running, and I don't know why and how to resolve this problem, so please help, any clue would be appreciated, thanks.

Originally posted by: pvaughan

It's been a while since I used non-concurrent VGs, use enhanced-concurrent VGs if at all possible. But, I think that's correct. Either node can varyon the VG, it's a valid VG on both nodes. The only thing that keeps them from being varied on is that the HACMP config only does a varyon on one node at a time. I know with my concurrent VGs with HACMP down I can vary the same VG on non-concurrent manually on both nodes at the same time. I wouldn't try using it that way , though.

I think part of the reasoning behind it is that if one nodes crashes hard, the surviving node has to be able to access the VG. If there were a VG locking mechanism used, the surviving node would have to negotiate the lock in order to access the VG. There are already enough problems with SCSI reservations in that event.

This is why shared VGs should always be set to not auto varyon, and should be controlled only by HACMP.