On Sun, Jan 30, 2011 at 02:27:34PM +0100, Jan Engelhardt wrote: > > On Sunday 2010-11-28 19:19, J. Bruce Fields wrote: > >>>On Sun, Nov 28, 2010 at 03:23:23PM +0100, Jan Engelhardt wrote: > >>>> Sometime between 2.6.33 and 2.6.36 it seems, knfsd has become really > >>>> unreliable, having clients start to emit messages like > >>>> > >>>> (nfsv3) > >>>> [42306.521225] NFS: server nova error: fileid changed > >>>> [42306.521226] fsid 0:10: expected fileid 0x1b007be, got 0x1b006b0 > >>>> > >>>> On nfsv4, it is even worse, as updates on the server are not > >>>> immediately reflected on the client like it was the case on > >>>> nfsv3. Unfortunately, that is all the information I currently > >>>> have. (Server is on 2.6.36-rc8, client on 2.6.37-rc1.) > >>> > >>>Have you really being changing only the server, or have you been > >>>changing the client at the same time? > > Fast forward. I have now a 2.6.38-rc1+ client with a 2.6.37 server, > using nfs4 with ext4. The "fileid changes" messages seem to be gone. That leaves in the dark about that particular problem. > I change a file on the server, compile it there, and then would expect > the client to pick up the change, but some awkward internal caching > mechanism pins the old copy and I don't get, say, new debug messages I > have added to my program that was just recompiled. Remind me what the backend filesystem is? Also, does echo 0 >/proc/sys/fs/leases-enable on startup fix the problem? --b. -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
