Switching package prob.

[Chi Yi Lee]

Hi,

I have two node cluster
Nodes are scot1 and scot2.
Package pkg1 is running on scot1.
autorun is enable and so is node switching


I want to manually switch pkg1 from scot1 to scot2
#cmhaltpkg -v pkg1
#cmrunpkg -n scot2 pkg1

The error states that it's mount problem, but I checked scot1 and pkg1 is unmount already




----Error message from pkg1.sh.log
5æ 25 03æ 14å 23ç§ - Node "scot2": Mounting /dev/vg01/lvol_call1 at /home/fods/call
vxfs mount: /dev/vg01/lvol_call1 is already mounted, /home/fods/call is busy,
or allowable number of mount points exceeded
ERROR: Function check_and_mount
ERROR: Failed to mount /dev/vg01/lvol_call1
5æ 25 03æ 14å 23ç§ - Node "scot2": Deactivating volume group /dev/vg01
Deactivated volume group in Exclusive Mode.
Volume group "/dev/vg01" has been successfully changed.

########### Node "scot2": Package start failed at 2004å¹´05æ 25æ ¥ 03æ 14å 23ç§ #####

[Colin Topliss]

Check /home/fods/call to see if any processes are using this directory:

fuser /home/fods/call

failing that, try lsof.

Most likely its a shell process (so someone has logged in and gone to this directory to take a look around and left the session there).

We have had similar problems in the past.

Col.

[Mark Grant]

Make sure you're not in that directory yourself too!

[Shaikh Imran]

hi,
Try killing the process forefully by
fuser -ku /home/fods/call
Then try to unmount
Regards,

Imran

[Bharat Katkar]

Some check points:

1. # cmhaltpkg -v pkg1

2. Then check the status of vg01 volume group. Whether it is entirely used by pkg1 package or any LV from this is used by somebody else or any other application running aorund. See /dev/vg01/lvol_call1 is getting umounted properly. Check anybody is already there in the directory /home/fods/call. Also use #fuser -cu /home/fods/call to see who else is using that directory.
Unless and untill your LV gets unmounted and VG deactivates properly you won't be able to start it on other node properly. So once you sort out this issues then proceed.

3.# cmrunpkg -n scot2 pkg1

This should help

regards,

[Geoff Wild]

You don't happen to have vg01 in your /etc/fstab?

The message seem to say it was already mounted...

Another possibility - last time it was running on scot2, someone mounted a filesystem under /home/fods/call - and now the system is in a precarious state - do a mount on scot2 to see what is mounted....

Rgds...Geoff

[Kent Ostby · Accepted solution]

Chi -- When you do the cmhaltpkg -v pkg1, check the package log on that scot1 to make sure that the package halt has completed before doing the cmrunpkg -n scot2 pkg1.

It may be that you are typing a bit too fast and that the cmhaltpkg has not completed.

The other notes about checking for users accessing the directory is also a good one.

Best regards,

Kent M. Ostby

[Rita C Workman]

Well......wondering about a couple things.

Like your /dev/vg01 on scot1. Do you by any chance have a /dev/vg01 on scot2 ? Now I'm sure it has a unique minor number...but if pkg 2 on scot2 has the same vgname (and mountpoint) you could get a "already mounted".
I avoid letting the system pick my vg name and go for assigning it a 'name' like vggl or vgacpay to avoid repeating oneself. But alot folks let the next number have it.

So I might ask...have you ever mounted pkg1 on nocde scot2 before....? If you did was pkg 2 still up on node scot2 ? Or was it down, and then you didn't run into this...

Just a thought,
Rita

[Mohanasundaram_1]

HI Lee,

In addition to checking the points mentioned by others earlier, I suggest you check if the control file on scot2 is different from scot1.

I remember having this mounting problems if your indexing numbers are same for 2 LVs. for example,
LV[0]="/dev/vg00/lvol1"; FS[0]="/stand"; FS_MOUNT_OPT[0]="-o rw"; FS_UMOUNT_OPT[0]=""; FS_FSCK_OPT[0]=""; FS_TYPE[0]="vxfs"

LV[1]="/dev/vg00/lvol4"; FS[0]="/home"; FS_MOUNT_OPT[ ]="-o rw"; FS_UMOUNT_OPT[0]=""; FS_FSCK_OPT[0]=""; FS_TYPE[0]="vxfs"

if 2 LVs have the same Index number you will get this message. Correct this, or easier way is, copy the control file from scot1 provided you are sure that the package mounts properly there and it comes up.

Hope this helps.

Rgds,
Mohan.

[Edgar_16]

Some user cd to that directory (mount point) before package was started. Do fuser -ku on the mount point and try bring up the package again.