curt@oce.orst.edu (Curt Vandetta) (07/11/90)
Hello folks, I'm having some very strange problems trying to export the platters of an HP Magneto Optical autochanger. I have 2 HP workstations, one is a HP370 it has the autochanger on it. The other is a HP835 I would like to give it access to the autochanger via NFS. I wrote a couple of C programs and a shell script to do the following: From the HP835 the user issues a command to mount an MO platter. The 835 remsh's the 370, askes the 370 to mount the correct MO platter. Control comes back to the 835 who then tries to NFS mount the platter from the 370. Then at another request be the user, every thing is unmounted. Everything worked great in my test cases. Which was only having one MO platter in the /etc/exports file. When I added ~16 platters to the /etc/exports file, the behavior became very sporadic. It always manages to get the mounts to work on the first request and generally fails on the second request (ie. can't get two NFS mounts at the same time.). Which ever platter I choose to mount first, is the only platter I'm allowed to mount, even if I unmount it and try another one. I'll get error messages that say "MachineName:/ac/slot23a access denied". If I go and modify the /etc/exports file, the process starts over ie. I can mount the next platter (different than before) but now that is the only one I can mount. At first I thought it was a problem with hostnames being mixed up. If I make the platters exported to the world I can mount about 4 platters. If I put hostnames in the /etc/exports file only about 1 platter (sometimes 2) can be mounted. Then I started to wonder if it was the length of the file /etc/exports. But know I'm wondering if a deamon (rexecd?) is looking throught the /etc/exports file and checking to see if the directories listed have something mounted under them, and if they don't then it takes them out of it's cache? Because none of the directories that are mount points for the platters of the autochanger are mounted until seconds before they get mounted via NFS, I would have to keep all 64 platters mounted to avoid this. Anyway, I hope someone understands what is going on. Thanks Curt curt@oce.orst.edu -------------------------------------------------------------------------- Curt Vandetta College of Oceanography curt@oce.orst.edu Oregon State University
harkin@hpindda.HP.COM (Art Harkin) (07/20/90)
Subject: Re: Problems with NFS and MO autochanger. Newsgroups: comp.sys.hp In-Reply-To: article <19285@orstcs.CS.ORST.EDU> of Wed, 11 Jul 1990 00:16:35 GMT I suggest trying to use the -e option in the mountd entry in the /etc/etc/inetd.conf file. mountd line was: rpc dgram udp wait root /usr/etc/rpc.mountd 100005 1 rpc.mountd mountd line becomes: rpc dgram udp wait root /usr/etc/rpc.mountd 100005 1 rpc.mountd -e This should fix the problem. Art Harkin Hewlett Packard