File systems could not unmount, package came down, but did not move

[chisle]

We executed a move of a package from nodeA to nodeB in the following manner:

# cmhaltpkg -v SG_PACKAGE
# cmrunpkg -v -n SG_PACKAGE
# cmmodpkg -v -e SG_PACKAGE

Operation normally takes about 12 seconds. The package contains a VG with three file systems. In the process of doing cmhaltpkg, it eventually failed. I checked, and only one of the three file systems was unmounted.

fuser -c showed that there was a process on one of the remaining file systems.

Is there some flag I can use for the unmount option of the LVM file systems to force the umount, even if a process is still 'locked' on it?

In this scenario, due to the nature of the file systems, our start/stop script for the package includes a loop to do an fuser -ck on the file systems twice before it then goes to the package shutdown. However, the particular process in question was rapidly being respawned by a parent process elsewhere on the system.

A force umount would be good, if it exists.

[Shibin_2]

>> the particular process in question was rapidly being respawned by a parent process elsewhere on the system.

Is this process belongs to application that use cluster FS or system process ?

[chisle]

it was a user process from elsewhere that was trying to reload a ksh script from the SG FS each time it was killed with the fuser -ck command.

We were finally able to unmount once we killed the PPID, then varied off the vg, etc.

However, what we really needed, and would have saved time, would be a force option, if it exists.