> Suddenly in the past week we have been having users who share excel
> files via our Samba complain that the file has become corrupt, and also
> some database files (Orcad .tdb database files). Fortunately there has
> in each instance been a good copy in the .recycler... so far.
> I came across some docs and threads in the net about oplocks. I did find
> an error in the samba log:
> [2008/04/14 10:19:05, 0] smbd/oplock.c:oplock_timeout_handler(351)
Were there any other adjacent messages relating to the lock?
I'd suspect a basic network problem before I suspected Samba's locking.
> and there are some other errors that have been in the logs for a long
> time that we've been ignoring (because we were ignorant & nobody
> complained of any problems)
> Apr 13 07:14:04 aa-samba smbd[10083]: [2008/04/13 07:14:04, 0]
> lib/util_sock.c:get_peer_addr(1229)
> Apr 13 07:14:04 aa-samba smbd[10083]: getpeername failed. Error was
> Transport endpoint is not connected
> Apr 13 07:14:04 aa-samba smbd[10083]: [2008/04/13 07:14:04, 0]
> lib/util_sock.c:write_data(562)
> Apr 13 07:14:04 aa-samba smbd[10083]: write_data: write failure in
> writing to client 192.168.1.46. Error Connection reset by peer
> Apr 13 07:14:04 aa-samba smbd[10083]: [2008/04/13 07:14:04, 0]
> lib/util_sock.c:send_smb(769)
> Apr 13 07:14:04 aa-samba smbd[10083]: Error writing 4 bytes to client.
> -1. (Connection reset by peer)
These errors are meaningless 99.999% of the time. Lots of discussion of
these in the list archives.
> Should we add a line disabling oplocks in smb.conf, and if so, what
> would be the recommended approach?
No, I think it is premature [given just this information] to blame Samba
or oplocks.
I'd root around in dmesg and the interface counters first. And make
sure you don't have a dying disk (is PFA enabled, if so has it
flipped?).
--
Consonance: an Open Source .NET OpenGroupware client.
Contact:awilliam@whitemiceconsulting.com
http://freshmeat.net/projects/consonance/