why Vg can't be deactived!

[leyearn]

#cmruncl
my Cluster starts up successfully
and all the services are available
when i test the cluster
on node1:
#cmhaltpkg sybase
the package Sybase can't be halted successfully
and the corresponding vg vg_db can't be deactived successfully
what's wrong with my cluster?
thanks

[Jeff Schussele]

Hi,

Number 1 reason is you haven't successfully stopped all the PIDs accessing the VG.
So take a long/hard look at your halt commands in the halt function of your SG control scripts. You may need to "hammer" some of the PIDs as a last resort if they won't stop gracefully.
MC/SG, by default will attempt looping graceful kills, but if that doesn't do it - it gives up. And then the umount/deactivate fails.

Rgds,
Jeff

[V.Tamilvanan]

Hi,
The above error comes when there is active processes and the system is unable to umount the Filesystem at the time of halting the package.
Check for proper shutdown config of sybase database on the package config. You may increase the FS_UMOUNT_COUNT=10 parameter on your package control file which tries 10 times to unmount the FS after some interval (meanwhile all the active process might be dead on the FS by the halting of package )
In my environment this parameter helped me a lot.

[leyearn]

there is no fileSytem on VG_DB

how can i examine which Process(es) is standing on VG_DB?

[Mobeen_1]

Hello there,
The reason you are unable to deactivate the vg is due to the fact that some processes are still using the vg. I would advise like others have done that you relook at the commands you used to stop the processes. I think it will be a good idea if you could confirm that the processes have been stopped after you issue the commands.

Also make sure that there is not anything that is accessing the db vol by using the command fdisk

Let us know if you have any questions.

regards
Mobeen

[Radhakrishnan Venkatara]

check and process for sybase is running
some may gone in zombie state.

Check any body has done "cd" to the corresponding mount points of the vg.

[generic_1]

Get some downtime and kill processs until the culprit becomes evident. Then add a cleanup script to your package prevent it.

fuser and lsof might be useful.

[Mobeen_1]

I am sorry, please read fdisk in my previous post as fuser :-)

Jeff,
Thanks, when i read your post i realised my mistake. I think its time i go to bed :-)

regards
Mobeen