We have a machine:
OSF1 rigel.krl.caltech.edu V3.0 347 alpha
This machine is connected to our local network as a satellite machine, i.e.
a local hard disk containing the OS and the rest NSF-mounted from the server
machine, an OSF/1 machine running OSF/1 v2.1 (which doesn't matter :).
/usr/local/ is linked from a subdirectory of a remote-mounted disk. It
contains, among other things, a mail directory and most application programs
used (gcc, etc.) on the network.
The problem is, the machine hangs regularly 'cause the kernel grows too large.
I've isolated the problem to the linking and unlinking of /usr/local, which
is pretty weird.
If I 'rm /usr/local' and reboot the machine (with YP enabled but no user
accounts allowed on 'cause of /etc/nologin) it doesn't grow forever.
If I 'ln -s /disk/local /usr/local' and reboot the machine the kernel grows
forever.
DEC says, "we ain't heard of it before". We checked, and as far as I can
tell, it's unique with our configuration (the link of /usr/local). The only
thing I can think that's different from our other mounted disks (we mount
four other disks remotely to that machine!) is that /usr/spool/mail
probably has a lot of files that get locked pretty regularly, and we DO have
NFS locking enabled.
Help.
Please?
Thanks,
--titus
P.S. We currently have the machine "down" except for an XDM -query session
to one of our other satellites, so any crazy ideas people come up with
to test things can be set up easily. I'm planning to try disenabling
NFS locking and then rebooting, myself...
--
Titus Brown, brown_at_krl.caltech.edu.
Received on Sun Feb 26 1995 - 07:21:03 NZDT