On Fri, Nov 25, 2011 at 05:12:18PM +0000, Ken Moffat wrote: Following hasn't made it to the list, and the exportfs part doesn't help - addition at end.
> On Fri, Nov 25, 2011 at 07:34:56AM +0000, Ken Moffat wrote: > > On Thu, Nov 24, 2011 at 10:18:34PM +0000, Ken Moffat wrote: > > > In fact, the longest part was banging my head against the wall > > while testing :-( It's working now, but along the way the results > > were very intermittent and iffy - a lot of the old 'getfh failed' > > messages in the server logs (i.e. "something is wrong somewhere"), > > mostly fixed by rebooting the server. This is with an old client on > > an LFS-6.7+ system using portmap. > > > > ISTR that my previous tests of nfs setups tended to be equally > > problematic, so perhaps it's all as good as ever. However, there > > are two points worth noting here: > > > Realised I'd left in my sed for the server bootscript, to take out > the "only in 2.6" nfsd. Since everyone with LFS-7.0 is using a 3.x > kernel, I tried using the original script and *enabling* this part - > as expected, error messages about nfsd. Oh, and back to the 'getfh > failed: function not implemented" message from mountd. > > This was NOT fixed by stopping nfs-server, correcting the script, > and (re)starting nfs-server. What did fix it was running 'exportfs > -ra' and then bouncing the server. So, I'm wondering if it will > work better with 'exportfs -ra' as the first stage of starting the > server ? In theory, that should only be necessary during changes... > > But, since we *are* using 3.x kernels (and in any case the nfsd > part doesn't work on recent 2.6 - see my earlier comments on > bringing up my LFS-6.8 server in -support) I think the parts that > mount and umount /proc/fs/nfsd should be dropped. > > Comments welcome. > Since then I've been cosmetically cleaning up one of my own bootscripts, and several of the BLFS scripts, to use the lsb stuff and thus get prettier output. That mostly works now - nfs-server messages from the various start processes within it still run across the screen, but no worse than the mount message from LFS for ext4 (messages where the kernel tries ext2 then ext3 but rejects both because of unsupported features before using ext4), nor the kernel messages when my ethernet module is loaded. So, I'll put those cosmetic cleanups in soon. As part of this, I was testing the nfs server again - running exports when first bringing it up, whether as -a or -Ra, does NOT help, it still has to be bounced before the client can connect. Without the exportfs in bringing it up, I had to reload it (i.e. run exportfs -Ra only), and then bounce it. Looks like a race somewhere, and rather similar to vague messages on debian lists. For the moment, I'll leave this to Bruce, if he wants it. Supposedly, debian has fixed this sort of problem - dunno. ĸen -- das eine Mal als Tragödie, das andere Mal als Farce -- http://linuxfromscratch.org/mailman/listinfo/blfs-dev FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
