Originally posted by: zhuli_pl

I directly connected two p520 servers to DS3400 using FC. HACMP 5.4.1 is installed on the two P520. I defined the concurrent resource group in HACMP. when I start HACMP, I can see the concurrent VG is active on both node. When I do #lsvg concurrent_vg on one node, it works. Then I do the #lsvg concurrent_vg on the other node, it hangs.
Any suggestion?

Originally posted by: Casey_B

Hello Zhuli

First thing: There is a developer works forum dedicated to PowerHA.
It's location is here: http://www.ibm.com/developerworks/forums/forum.jspa?forumID=1611
You can place any future HA discussion in that forum. (This is more of an LVM question, but wanted to let you know. :) )

I don't know for sure what the issue is from your description, but I just did a web search for "lsvg hang" and came up with an APAR:
http://www-01.ibm.com/support/docview.wss?uid=isg1IZ02543

Some questions to help you define the problem a little more:
Does lsvg rootvg work?
(Seems basic, but it might be that actually running lsvg hangs for reason other than LVM...for instance, broken NFS mount in PATH on node B...Low probability, but possible)
Does the hang always happen on the same node?
If so, do you start the cluster in the same way every time...(for instance, A before B always...)
Or does it hang (for instance) on the second node started, or second node running lsvg?
A very specific definition of the issue is always helpful.
Does lsvg -L seem to work?
This might mean the above APAR is involved, or one like it.
Are there any error logs on either node indicating temporary errors on the disks?
I don't think this could cause lsvg hangs, but this could indicate reservation errors on your machines.
If this is the case, then an application is trying to access the disk and vg through direct open calls, and setting a reservation on a disk.

If you have trouble, you can always open a PMR with IBM support...They will be able to look at your software levels, and LVM logs to see what is causing
the hang, and see if there are any APARs that should be installed to resolve the issue.

Hope this helps
Casey

Originally posted by: zhuli_pl

Thank you for your reply.
The PTF http://www-01.ibm.com/support/docview.wss?uid=isg1IZ02543 is:
bos.rte.lvm 5.3.7.0

My OS level is: AIX 5.3.8-07
bos.rte.lvm.5.3.8.6
bos.clvm.enh.5.3.8.4

All HACMP 5.4.1 latest PTF has been applied.
The lsvg rootvg works. lsvg -L works.
Here is the scenario:
1.start HACMP concurrent on node1, all vg commands work on node1
2.start HACMP concurrent on node2, all vg commands work on node2. But
they take 2-3 minutes on node1 for the first time. After that, they work well on node1. Then they take 2-3 minutes on node2. After that they work well on node2 and they are slow on node1.
This repeats no matter on which node I start HACMP first.
It seems if one node has the lock, the other node will take a few minutes to get.

I checked the DS3400 and FC adapter firmware. They are up-to-date.
If only one node uses the DS3400, there are no errors.
If the other node tries vg commands, there are errors in errpt.

The FC adepter is 2GB, The DS3400 is 4GB. They connect directly.
Does this matter?