101TECH

I have recently built a DRBD + Linux HA NFS cluster , part of the structure was to create a drbd disk resource and mount it on /var/lib/nfs. This allows to switch nfs over to the secondary node without getting “stale nfs handle” errors. Heartbeat was having a hardtime doing a clean umount of the file system and ended up rebooting the machine.

Apr 20 06:47:07 fs Filesystem[9823]: [9887]: INFO: No processes on /var/lib/nfs were signalled
Apr 20 06:47:08 fs Filesystem[9823]: [9890]: ERROR: Couldn’t unmount /var/lib/nfs; trying cleanup with SIGKILL

After some poking around the systems It looked like nfs was not showtdown cleanly. looking into the rc script i changed the signal from 2 to 9

From :

stop)
# Stop daemons.
echo -n $”Shutting down NFS mountd: ”
killproc rpc.mountd
echo
echo -n $”Shutting down NFS daemon: ”
killproc nfsd -2
echo

To :

stop)
# Stop daemons.
echo -n $”Shutting down NFS mountd: ”
killproc rpc.mountd
echo
echo -n $”Shutting down NFS daemon: ”
killproc nfsd -9
echo

and now, all is well

Apr 20 06:52:49 fs Filesystem[8696]: [8754]: INFO: unmounted /var/lib/nfs successfully

searching the web shows this is a somewhat known issue