Simultaneous file copying failing on Samba 2.2.2

[Tim Allen]

Hi

I've been running a number of tests associated with the problem I described
below in the hope of finding something which will help track it down, which
I'll just recap:

In three windows on a NT4 server, start copy of three large directories to
three separate shares on Samba server.
On W2000 workstation, start copy of fourth NT4 server directory to separate
share on Samba server.
After a few minutes, one or more copies from NT4 server stops with "Unable
to copy file... Session cancelled".

Further tests:

1. Replaced NIC in Samba server, still fails.
2. Switched to different network cabling on Samba and NT4 server, still
fails.
3. Repeat test, but replace Samba server with another W2000 machine. Test at
100Mb and 10Mb network speed, no failure.
4. Put back Samba server. Instead of doing 3 simultaneous copies from NT4
server, just do one from NT4 server with simultaneous copy from W2000
machine. No failure.
5. Now the interesting bit. The W2000 workstation happens to be
dual-bootable with NT4. Testing with the three copies from NT4 server plus
one from NT4 workstation (instead of W2000) results in No failure.
6. Contrary to what I said in my first email, I think the failures are prone
to occur on larger file transfers (around 2M).


So it looks like there's a problem in Samba if multiple sessions from an NT4
machine are interleaved with one from a W2000 machine, but the problem
disappears if the W2000 machine is replaced with NT4.

I'm now pretty convinced this is a Samba bug. I also think it's the sort of
problem that's not going to be very easy to replicate. Our setup here is
such that I can readily reproduce it, but I need some help from you Samba
gurus. Firstly, can you suggest a log level appropriate to tracking down the
cause of this, and the sort of things I should be looking for in the logs?

Tim Allen




> Hi
>
> I've run into a problem when simultaneously copying directories onto our
> Samba server (RH7.1, Samba 2.2.2, 10Mb ethernet):
>
> From our NT4 server, started 3 simultaneous copies of different
directories
> (each a few hundred Mb) onto different Samba shares. From a W2000
> workstation, started another copy of similar directory onto another Samba
> share. Everything runs fine for 5-10 minutes, then one of the NT4 copies
> hangs with "Unable to copy file ... Session cancelled" or "Unexpected
> network error occurred". Typically one or other of the other NT4 copies
also
> then fails. The W2000 copy always continues OK.
>
> This failure is repeatable, but not on any specific file.
>
> Swapped out NIC but get the same symptoms.
>
> Trawling back through the archives, I see similar reports, particularly
when
> copying large files, although file size doesn't seem to be a factor here.
>
> The NT4 event log reports "The redirector has timed out a request to
Golux"
> (the Samba server).
>
> With Samba log level at 3, the last lines report:
>
> [2001/12/23 10:34:54, 3] smbd/process.c:process_smb(860)
>   Transaction 389 of length 46205
> [2001/12/23 10:34:54, 3] smbd/process.c:switch_message(667)
>   switch message SMBwriteX (pid 1143)
> [2001/12/23 10:34:54, 3] smbd/reply.c:reply_write_and_X(2807)
>   writeX fnum=8845 num=46138 wrote=46138
> [2001/12/23 10:34:54, 3] smbd/process.c:process_smb(860)
>   Transaction 390 of length 46
> [2001/12/23 10:34:54, 3] smbd/process.c:switch_message(667)
>   switch message SMBclose (pid 1143)
> [2001/12/23 10:34:54, 3] smbd/reply.c:reply_close(2997)
>   close fd=24 fnum=8845 (numopen=1)
> [2001/12/23 10:34:54, 2] smbd/close.c:close_normal_file(206)
>   root closed file Admin/Set1/ASANALH.DBF (numopen=0)
> [2001/12/23 10:35:54, 3] smbd/sec_ctx.c:set_sec_ctx(320)
>   setting sec ctx (0, 0) - sec_ctx_stack_ndx = 0
> [2001/12/23 10:36:54, 3] smbd/sec_ctx.c:set_sec_ctx(320)
>   setting sec ctx (0, 0) - sec_ctx_stack_ndx = 0
> [2001/12/23 10:37:54, 3] smbd/sec_ctx.c:set_sec_ctx(320)
>   setting sec ctx (0, 0) - sec_ctx_stack_ndx = 0
> [2001/12/23 10:38:54, 3] smbd/sec_ctx.c:set_sec_ctx(320)
>   setting sec ctx (0, 0) - sec_ctx_stack_ndx = 0
>
> etc.
>
>
> My smb.conf is as follows:
>
> # Global parameters
> [global]
>  workgroup = DCAE
>  server string = Samba Server
>  encrypt passwords = Yes
>  username map = /etc/samba/smbusers
>  log level = 3
>  syslog = 3
>  log file = /var/log/samba/log.%m
>  socket options = TCP_NODELAY SO_RCVBUF=8192 SO_SNDBUF=8192
>  local master = No
>  dns proxy = No
>  hosts allow = 192.168.1. 127.
>  printing = lprng
>  oplocks = No
>
> [print$]
>  path = /usr/local/samba/printers
>  write list = administrator,root
>  guest ok = Yes
>
> [hp_lj3p]
>  path = /shared/printers/spool
>  printable = Yes
>  print command = lpr -Fb -r -P %p %s
>  printer name = lp
>
> [faxlp]
>  path = /shared/printers/spool
>  printable = Yes
>  print command = lpr -r -P %p %s
>  printer name = faxlp
>
> [Data]
>  comment = Development data
>  path = /shared/Data
>  read only = No
>  inherit permissions = Yes
>
> [Admin]
>  comment = Share for Admin group
>  path = /shared/Admin
>  read only = No
>  inherit permissions = Yes
>
> [Apps]
>  comment = Applications
>  path = /shared/Apps
>  read only = No
>  create mask = 0644
>
> [homes]
>  comment = Home Directory
>  path = /home/%S/pchome
>  read only = No
>  create mask = 0700
>  directory mask = 0700
>  browseable = No
>
> [Common]
>  comment = Common directory for file sharing
>  path = /shared/Common
>  read only = No
>  create mask = 0777
>  directory mask = 0777

>