unable to start PKG

[srinimar]

I have umounted all the file systems and activated cluster lock with vgchange and tried to run the PKG

but i have file systems on server A in mounted position while i was doing this cmrunpkg on server B. Am I doing right...now i am unable to mount file systmes back on Server B

# cmrunpkg -n childps2 APPPKG
Running package APPPKG on node childps2
The package script for APPPKG failed with no restart. APPPKG should not be resta
rted
Unable to run package APPPKG on node childps2
Check the syslog and pkg log files for more detailed information
cmrunpkg: Unable to start some package or package instances

[IT_2007 · Accepted solution]

is it supposed to be up and running on node 2? If so then you need to unmount file systems on node1 and deactivate volume group.

Run package on node2. check syslog and package log for more information.

[srinimar]

no it was not running.
This is the first time i have done cmrunpkg..

i have two server's
A and B

all file systems are in mounted position in A while i was doing cmrunpkg on B.

but before doing cmrunpkg on B i have umounted file stsmes on B and did cmrunpkg.

i have pne PKG on A and other on B so i have tried to start PKG on B

i have found this in pkg log file


########### Node "childps2": Starting package at Tue Sep 19 17:25:30 CDT
2006 ###########
Sep 19 17:25:30 - Node "childps2": Activating volume group /dev/finvol with excl
usive option.
vgchange: Activation mode requested for the volume group "/dev/finvol" conflicts
with configured mode.
ERROR: Function activate_volume_group
ERROR: Failed to activate /dev/finvol
Sep 19 17:25:30 - Node "childps2": Deactivating volume group /dev/finvol
Volume group "/dev/finvol" has been successfully changed.

########### Node "childps2": Package start failed at Tue Sep 19 17:25:30
CDT 2006 ###########

[IT_2007]

########### Node "childps2": Starting package at Tue Sep 19 17:25:30 CDT
2006 ###########
Sep 19 17:25:30 - Node "childps2": Activating volume group /dev/finvol with excl
usive option.
vgchange: Activation mode requested for the volume group "/dev/finvol" conflicts
with configured mode.
ERROR: Function activate_volume_group
ERROR: Failed to activate /dev/finvol
Sep 19 17:25:30 - Node "childps2": Deactivating volume group /dev/finvol
Volume group "/dev/finvol" has been successfully changed.

============

Do

vgchange -a n /dev/finvol

and run package again.

[Rajeev Shukla]

Do one thing.
unmount the filesystems related to the package you are trying to run on both the servers and deactivate the VG using
vgchange -a n
then run
vgchange -c y /dev/
cmmodpkg -e -n serverB
then run the package
cmrunpkg -n serverB

Cheers
Rajeev

[srinimar]

Hey guys thnx alot that worked and PKG on node b is running and all the file systmes came back..

please help me on this too my cluster setup will be completed...

now i have umounted all the file systmes on node A to run the node A PKG.

i am unable to umount 1 filesystem rest all i have unmounted..

# umount /oracle/fin
umount: cannot unmount /oracle/fin : Device busy
umount: return error 1.
#

[IT_2007]

make sure that you are not sitting on the mount point.

fuser -cu /mountpoint

fuser -ku /mountpoint

[Rajeev Shukla]

Try seeing the proccess that is accessing this filesystem, use
fuser -cu to see the proccess or
fuser -ku to send kill to those process accessing this filesystem after that try umount command
use man fuser to learn more about it
PS: dont forget to assign points if your problems are resolved

[srinimar]

#cd /
# fuser -cu /oracle/fin
/oracle/fin: 4487mto(psoft) 5059mto(oracle) 1368mto(oracle) 1619mto
(oracle)
# fuser -k /oracle/fin
/oracle/fin:


I have tried fuser -ku as well same error comes up i am unable to umount


# umount /oracle/fin
umount: cannot unmount /oracle/fin : Device busy
umount: return error 1.

[Rajeev Shukla]

did you try
# fuser -cu /oracle/fin
/oracle/fin: 4487mto(psoft) 5059mto(oracle) 1368mto(oracle) 1619mto

kill -9 4487 5059 1368 1619 and then umount

[srinimar]

i could umount the file system

but again got the same error message

when i run vgchange -a n /dev/vgora
it

# vgchange -a n /dev/vgora
vgchange: Couldn't deactivate volume group "/dev/vgora":

FYI ..the same VG has othe lvm which has other mount points i did not umount that file systms as they are not part of package.



########### Node "childps1": Starting package at Tue Sep 19 19:44:08 CDT
2006 ###########
Sep 19 19:44:08 - Node "childps1": Activating volume group /dev/vgora with exclu
sive option.
with configured mode.
ERROR: Function activate_volume_group
ERROR: Failed to activate /dev/vgora
Sep 19 19:44:08 - Node "childps1": Deactivating volume group /dev/vgora
vgchange: Couldn't deactivate volume group "/dev/vgora":
Device busy
ERROR: Function deactivate_volume_group
ERROR: Failed to deactivate /dev/vgora
Sep 19 19:44:08 - Node "childps1": Deactivating volume group /dev/oradata
vgchange: Couldn't deactivate volume group "/dev/oradata":
Device busy
ERROR: Function deactivate_volume_group
ERROR: Failed to deactivate /dev/oradata

########### Node "childps1": Package start failed at Tue Sep 19 19:44:08
CDT 2006 ###########

[Rajeev Shukla]

Even if that LV is not part of the package you will need to unmount the LV to deactivate the VG. Otherwise you cant move this package to another server.
This kind of configuration doesn't suit the package or cluster requirement. For the package to be easily movable to other servers you should have one VG per package and all LV should be configured in the package otherwise you wont be able to move the package to another node

[Prashant Zanwar_4]

If is cluster enabled, you cannot have some file systems mounting out of package, best approach will be to control though package, then failover will be easy for you.
Also you cannot change VG attributes unless all LV's are unmounted..
if you have lsof, try

lsof +D /dev/vgXX it will give idea of what is there.

Hope it helps
Prashant

[Serviceguard for Linux]

To restate what Prashant said.

If you have a package, all VGs associated with the application(s) in teh package should ONLY be activated by teh package control scripts. Do not try to activate any VG manually if ti is associated with a package.

If you want to state the app on anotehr node, move the package by halting it on the node it is running on and start it on the other node.

[Calandrello]

Srinimar which SO this occurring this?
it verifies if the VG disactivated total in node A (vgdisplay /dev/vgxxx) it tries to activate manually in node B vgchange -a y /dev/vgxxx

[rvent]

I had the same problem happen to me last week, we dont use oracle, but unidata... When we stopped the package the scheduler for the databse didnt stop properley and that caused all sort of problem when trying to start the package, we were unable to start it on any other node (normally)..

We just rebooted the system and then we were able to run the packag with no problems...

i know rebooting ( <- noob ), but when everything else i did failed i just had to reboot since i couldnt think of anything else and the big boos was breathing on my shoulder...

If you havent rebooted give it a shot..