What disk could be FIRST_CLUSTER_LOCK??

[Carme Torca]

Hi,

I have to put a disk like FIRST_CLUTER_LOCK, but I don't know what disk I have to put....
Has it to be a disk that see the 2 servers in cluster?...

Could I no configure CLUSTER_LOCK??

I only have 2 internal disk in each server,
and after, all other are SYMMETRIX disks that have different patch in each server.

I have FIRST_CLUSTER_LOCK_VG /dev/vgges
too. Maybe I could eliminated it?? This /dev/vgges its only activated in one node

02:/# vgdisplay -v /dev/vgges |more
vgdisplay: Volume group not activated.
vgdisplay: Cannot display volume
group "/dev/vgges".

01:/#
-- Logical volumes ---
LV Name /dev/vgges/gestion
LV Status available/syncd
LV Size (Mbytes) 8628
Current LE 2157
Allocated PE 2157
Used PV 1


--- Physical volumes ---
PV Name /dev/dsk/c14t1d0
PV Name /dev/dsk/c17t1d0 Alternate Link
PV Status available
Total PE 2157
Free PE 0
Autoswitch On

PV Name /dev/dsk/c17t1d1
PV Name /dev/dsk/c14t1d1 Alternate Link
PV Status available
Total PE 2157
Free PE 2157
Autoswitch On

Maybe I could eliminated
FIRST_CLUSTER_LOCK_PV if I have FIRST_CLUSTER_LOCK_VG??

Maybe I put one disk of this /dev/dsk/c14t1d0 like FIRST_CLUSTER_LOCK_PV, but this disk it's no the same in the other server....

What could I do?

Thanks!,
Carmen.

[RAC_1]

You can not replace FIRST_CLUSTER_LOCK_PV with FIRST_CLUSTER_LOCK_VG. or vice a versa.

Put first pv in FIRST_CLUSTER_LOCK_VG as
FIRST_CLUSTER_LOCK_PV.

[Rainer von Bongartz]

I recommend you take a good look at the
HP-UX software recovery handbook. Here's the link

http://www5.itrc.hp.com/service/iv/docDisplay.do?docId=/DE_SW_UX_swrec_EN_01_E/ServiceGuard.pdf

In your case look out for the chapter 19 (Cluster lock disk initialization)

Regards
Rainer

[Geoff Wild]

Regen a new ascii file:

cmquerycl -v -C /etc/cmcluster/new.ascii -n node1 -n node2

That will tell you what to use for a cluster lock.

Here's some other sites that may be of interest:

http://docs.hp.com/hpux/onlinedocs/ha/haFAQindex2.html


http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B3936-90073/B3936-90073_top.html&con=/hpux/onlinedocs/B3936-90073/00/00/1-con.html&toc=/hpux/onlinedocs/B3936-90073/00/00/1-toc.html&searchterms=mc/serviceguard%7cmanaging&queryid=20031020-063907

Rgds...Geoff

[Gary L. Paveza, Jr.]

You can choose to use any disk within your FIRST_CLUSTER_LOCK_VG for your lock disk. You need to pick the same disk (though it can have different device files) for each node. The purpose of this disk is to prevent a "split-brain" from occuring when both nodes attempt to take over. Whoever is able to put a lock on the lock disk wins. It doesn't matter if the volume group is only active on one system at a time, just that both systems have access to the drive (so it cannot be an internal drive to either system).

[Uday_S_Ankolekar]

It's nice to have Lock disk in a two node cluster and it should be an external disk (Both node should be able to see it) You need to select a disk from your EMC.

-USA..

[Stephen Doud]

When creating a cluster, the admin must perform the command:
# cmquerycl -C -n -n ...

The command causes each node in the list to pass information back to the originating node in order to build an acceptable file.
Part of the process requires looking for a shared volume group which does not already have a cluster ID marked on it. (A cluster ID is installed on each VG listed in the bottom of the cluster ASCII file when 'cmapplyconf' is run on the cluster configuration file)
So normally, the admin will not need to create a cluster configuration file by hand.

If the cmquerycl command has found a shared VG that is not already owned by a previous cluster, the file created will contain a FIRST_CLUSTER_LOCK_VG reference as well as a FIRST_CLUSTER_LOCK_PV reference in each node section.

If the cmquerycl command cannot locate a shared VG that is not already assigned to an older cluster, it may be necessary to perform a "cmdeleteconf" to dissolve the old cluster already built on the servers in order to start over with a new cluster.
"cmdeleteconf -f" will not remove any configuration files. It only unclusters shared VG's and zeroes out /etc/cmcluster/cmclconfig (the cluster binary file which makes up the current cluster).

If you wish to change the cluster lock VG and PV, each reference must be updated in the cluster ASCII file.
# strings /etc/lvmtab
on each server will show the names of the VGs and each /dev/dsk path that goes with that VG. It may be possible that the /dev/dsk paths do not match between servers. This is okay, as long as they reach the same disks for that VG.

If you already have a cluster and wish to change the cluster lock VG and PV, you must halt the cluster prior to performing 'cmapplyconf' on the cluster configuration file.

-s.d.