On Sun, Mar 17, 2002 at 06:49:26PM -0800, David Talkington wrote: > Ben Logan wrote: > > >At first I wasn't 100% sure that the problem was caused by NFS, but > >I'm sure now. > > Convince us.
The reason I say I'm convinced now is that I go for so long without having any problems, but then *always* have strange problems shortly after running NFS. The first time or two, I thought it was coincidence, but it has happened too many times for that to be the case in my mind anymore. Not exactly proof...just strong evidence. > When you say "within hours of using NFS", does that mean within hours of > a) starting specific daemons (which ones?), b) mounting specific > filesystems (which ones, with what options?) on the problem machine (at > what mount points?), c) sharing specific filesystems (which ones, with > what options?) from the problem machine? What's in /etc/fstab on the > clients, and /etc/exports on the server? a) The problem doesn't seem to be caused by the NFS-related daemons (all the daemons started by going to /etc/init.d and starting nfs, nfslock, and portmap--rpc.nfsd, rpc.mountd, rpc.rquotad, rpc.lockd, rpc.statd, portmap) running, but rather exporting/mounting one or more filesystems. b) The problem doesn't seem to be filesystem specific. I have exported different filesystems on different machines and it (almost) always causes a problem. Remember, this has happened on three different machines, both exporting and mounting. The mount point is usually /mnt/cdrom, but that's not always the case. c) Since I don't need the filesystems mounted all the time, I don't edit fstab. Instead, I use a mount command that looks like this # mount -t nfs -o wsize=512,rsize=512 mach1:/filesystem /local/filesystem I've tried different values for [rw]size in an attempt to solve the problem, but it hasn't worked. The largest value I tried was 8192, and 512 was the lowest. /etc/exports is usually empty, but when I need to export a filesystem, it looks similar to this /mnt/cdrom mach3(ro) At different times I have used other options including rw, no_root_squash, anongid, and anonuid. After editing /etc/exports, I either restart nfs from /etc/init.d or (more commonly) use exportfs -r. I have also skipped editting exports altogether, and just used the appropriate syntax to the exportfs command. > Interesting puzzle, but you haven't given us many clues. I hope I've given a few more clues, although I doubt they will help much. One of the problems is that there are never any error messages in the logs--none pertaining to nfs anyway. One thing I did remember this morning: Many months ago, I had a certain filesystem mounted/exported all the time. It was mounted via fstab at boot, and was always in /etc/exports. It was also accessed every day. I don't recall there ever being any problems on either of the two machines. As I recall, that was back when both machines were running the stock 2.4.2-2 kernel. Since then I have mostly run kernels which I compiled and which have an Alan Cox patch applied. The patch was necessary to get hiddev support for my USB UPS. I'm told that I don't need the patch anymore (past 2.4.16, I think), so I am going to install 2.4.17 as soon as I build an updated version of e2fsprogs required by the kernel. Thanks, Ben -- Ben Logan: ben at wblogan dot net OpenPGP Key KeyID: A1ADD1F0 Be sure to evaluate the bird-hand/bush ratio. _______________________________________________ Redhat-list mailing list [EMAIL PROTECTED] https://listman.redhat.com/mailman/listinfo/redhat-list
