vgstatus server on node2

[Victor_101]

hi
im setting up MCSG , with shared vols for Oracle RAC in a test environment.
the thing is that when i start my cluster , node 2 always takes the shared vgs as server status and node 1 takes client status on the vgs..
is this abnormal ?? if so .. how do i correct this. i can manually start the package only on the node 1 and that solves my prob.. but i want to know why...
my production does not behave like this..

cheers..
vic

[Steven E. Protter]

Shalom Vic,

Is the volume group activated in shared mode?

SEP

[Prashant Zanwar_4]

Whichever VG's activate first, will be showing as server and other as client..
this should matter as it is on OS level.. As far as oracle instance is concerned, it will be running on both hosts, and is upto listener to direct traffic to one of these..

Hope it helps
Thanks
Prashant

[Victor_101]

Hi Prashanth ,
I realised that and hence I let that setup run and Oracle installation is goin on ...
I just want to verify why ?? since this being my test setup .. all the VGs are going to Server status on Node2 whereas on my production setup its the opposite which is what I expect to be standard.

Steven..
yes its in shared mode
for my shared VGs , this is what I did

cmruncl
vgchange -a n vgxx (on both nodes)
vgchange -S y -c y vgxx (on primary node only)
vgchange -s y vgxx (on primary node 1st)
- sets up as server status on node
vgchange -s y vgxx (on primary node 2nd)
- sets up as client status on node
and now all is fine..

then when i restart the cluster
cmhaltcl -f
cmruncl
all the shared vgs go to the 2nd node as Server status and the 1st node as client...
I believe that there will be something to control this like maybe the 2nd node is getting the lock disk or something of the sort...

I can always control this manually .. but I shouldnt have to..

Thanx..

[Stephen Doud]

use 'cmviewconf' to determine which node is listed first in the cluster binary file - this may be the reason why the second server is the "shared vg" server and the other node the client.

[Prashant Zanwar_4]

You got to check what time VG's getting activated what time.. u will come to know what is happening then
Prashant

[John Bigg]

Why do you consider it an error that node 2 is the server node and node 1 the client? This is not wrong and does not require "fixing". It's just timing which determines which gets to be the server rather than the client. I suggest that your node 2 happens to activate the volume groups first.

Unless you have a specific reason for wanting one node rather than another to be the server (and I cannot think of one) I would not worry. Do you worry which node in your cluster is the coordinator node? Probably not since the Serviceguard commands do not report this. If a vgdisplay did not show client/server status you probably would never have considered this.

[Victor_101]

well i didnt actually consider this an error.. like i asked in my first post .." is this abnormal?" i just wanted to know how we could control it if at is is possible...
i guess what most of u are saying that it is not possible!!
will close now.. thanx guy