When starting a package, error activating volume group

[Paresh Parikh]

Greetings,

When I run cmrunpkg command on a package I get the following error "vgchange: Activation mode requested for the volume group conflicts with configured mode."
I always need to do "vgchange -c y /dev/vg" then run the cmrunpkg command to start a package. This only happens on some packages and not all on the same node.
Any ideas why I get into this mode?
Thanks,
Paresh

[Rainer von Bongartz]

check your package control file
you should have an entry

VGCHANGE="vgchange -a e"

[melvyn burnard]

I would suggest you start by checking the package logs after shutting down the pacage, then check the halt portion of the p[ackage scripts. The next thing is to see if someone is doing vgchange -c n via the command line when they should not be.

[Paresh Parikh]

Yes, the control file has vgchange -a e as the default.

And no one else is running vgchange -a n on the command line.

[Rainer von Bongartz]

OK, as Melvyn suggested:

check your package log file after package shutdown and
check the status of your vg when package is down.

can you do vgchange -a e without starting package ??

[Kenneth Yap]

Hi,
When the cluster up....try to run #cmgetconf. Check the results. Is the Volume Group included in the cluster aware disk ? The VG must include in the cluster aware.

[Tim D Fulford]

Hi

If you need to manually make the VG clustered then the SG probably does not think it is part of the cluster! Probably the easiest way to fix it is to re-do cmapplyconf AFTER doing a cmdeleteconf. This obviously means an outage! Maybe someone else can think of a better way of getting the correct info into the VGRA but I do not know of one! (I have tried in the past)

Good Luck

Tim

[Mark van Hassel]

Hi,

It could be that your volume groups are activated during boot, so check the /etc/lvmrc file and make sure that the AUTO_VG_ACTIVATE is set to 0.

[Varghese Mathew]

Hi,

You need to see no one else(or any script) activates the vg in normal node in the server. In the package control log file can you see the specific volume group --> how it is made down (when the cluster made down), it should be in the exclusive mode.

If you feel some thing is fishy, i will go in this way --
1) verify the entry of volume group name in the cluster ascii file.
2) verify the logical volumes name etc which has to be mounted
3) if you feel everything is Ok, you can write a small script which is for making all the Volume groups active (manual, not in cluster).
Assume your vg's are "vgcluster1" and "vgcluster2". Then you manual activation script looks like this:
----
vgchange - c n /dev/vgcluster1
vgchange -c n /dev/vgcluster2
vgchange -a y /dev/vgcluster1
vgchange -a y /dev/vgcluster2
----

One other script for making the manually made up Volume groups deactivate back to cluster aware mode - the same mode in which all the cluster volume groups have to be kept.
----
vgchange -a n /dev/vgcluster2
vgchage -a n /dev/vgcluster1
vgchange -c y /dev/vgcluster2
vgchange -c y /dev/vgcluster1
----

Run these scipts For making the vgs activate first, and the second one for deactivating the manually activated vg's and keep all the volume groups cluster aware.

4) Do a #cmgetconf -v -C cluster config file -P package config file and see everything is fine or not. Please keep a copy of all the cluster configuration files in the server.
5) If no errors are found , then go for applying it by using the command
#cmapplyconf -v -C cluster config file -P package config file

Then check for the errors (hopefully it won't come).

Hope this helps..
Cheers !!!
Mathew

[Laurent Paumier]

To check nobody is changing VG mode, you can grep /var/adm/syslog/syslog.log for vgchange. All LVM commands are logged in syslog.