Thanks to the following for their speedy replies:
Sheila Hollenbaugh
Dr. Tom Blinn
The consensus is that it's safe to purge stale entries from /var/adm/mountdtab
and that it's most likely a good idea to modify /sbin/init.d/nfs to take care
of this automatically.
============================== Original question ==============================
We have a Digital Unix NFS server whose /var/adm/mountdtab contains lots of
vestigial information. Is there any harm in purging this file the next time
(or every time) we reboot, or even doing it on the fly for the entries that we
know are no longer valid? Does anyone have any ideas why these entries were
never commented out or deleted from the file? I'd suspect that this was caused
by the NFS clients not unmounting the exported filesystems cleanly the last
time they had mounted them. It would be nice if 'showmount' displayed reliable
information once again.
============================== Specific replies ===============================
>From Sheila Hollenbaugh <shollen_at_valhalla.cs.wright.edu>:
We added the following line to the /sbin/init.d/nfs script, and it seems to
work fine. We added it in the 'stop' section after mountd is killed.
cp /dev/null /var/adm/mountdtab
-------------------------------------------------------------------------------
>From Dr. Tom Blinn <tpb_at_zk3.dec.com>:
I don't know why it isn't automagically removed during system startup, maybe
as part of the script that starts the mountd. It might make sense to remove
it when you restart the mountd (or shut it down), too.
In any case, I just cleaned out the version on my V3.2G system without any
negative side effects. It gets updated when a remote system unmounts or
mounts an NFS file system, but if the server system panics or maybe when
the client goes away unexpectedly, it doesn't get updated, which is why it
has stale entries.
Received on Thu Feb 26 1998 - 23:10:08 NZDT