Locking Problems?

[Peter Schinagl]

Hi List,

the following strange things happen at a customer of us.

We have succesfull migrated an NT4 PDC to an Samba 2.2.7a PDC
All tests were also succsessfull, but one thing we can't
test...the amount of users, because of a small test network.

The customer uses a 3D-CAD System on WinNT/WinXP Workstations
with a central NT4 (now Samba) Fileserver, for constructing 
his machines. There are about 50 people who work with this system.

You must immagine that 50 Users open a three dimensional drawing
if a machine with about 400 or more pieces. For example think of
a screw...there are many screw's in a big machine...and all single
pices references a file.

With NT4 they have sometimes some error messages on some well
used pieces of the machine (e.g. screw's). The message says
there are too many open files and so the piece can't be loaded.

They think with going to linux and samba there problem goes away...
NOW with Samba 2.2.7a this is even more worse :(( We also tried
Samba 2.2.8a with the same effect.

With debug 10, i can see this snipplet raises the error message. 
But what happens AND we can we solve it? Any ideas?

 --snip--
   [2003/06/16 12:04:00, 5] lib/util.c:show_msg(286)
     smb_vwv[7]=1 (0x1)
   [2003/06/16 12:04:00, 5] lib/util.c:show_msg(291)
     smb_bcc=20
   [2003/06/16 12:04:00, 10] lib/util.c:dump_data(1531)
     [000] FF FE 00 00 00 00 00 00  93 FF FF 7F 00 00 00 00  ........ ........
   [2003/06/16 12:04:00, 10] lib/util.c:dump_data(1539)
     [010] 01 00 00 00                                       ....
   [2003/06/16 12:04:00, 3] smbd/process.c:switch_message(685)
     switch message SMBlockingX (pid 1261)
   [2003/06/16 12:04:00, 4] smbd/uid.c:change_to_user(119)
     change_to_user: Skipping user change - already user
   [2003/06/16 12:04:00, 10] smbd/reply.c:reply_lockingX(4712)
     reply_lockingX: lock start=2147483539, len=1 for pid 65279, file SNCF44/50636286_Zahnstange~00.SLDPRT timeout = 0
   [2003/06/16 12:04:00, 10] locking/locking.c:do_lock(113)
     do_lock: lock type WRITE start=2147483539 len=1 requested for file SNCF44/50636286_Zahnstange~00.SLDPRT
   [2003/06/16 12:04:00, 10] locking/locking.c:do_lock(113)
     do_lock: lock type WRITE start=2147483539 len=1 requested for file SNCF44/50636286_Zahnstange~00.SLDPRT
   [2003/06/16 12:04:00, 10] locking/locking.c:do_lock(113)
>|   do_lock: lock type WRITE start=2147483539 len=1 requested for file SNCF44/50636286_Zahnstange~00.SLDPRT
>| [2003/06/16 12:04:00, 3] smbd/error.c:error_packet(113)
>|   error packet at smbd/reply.c(4757) cmd=36 (SMBlockingX) NT_STATUS_LOCK_NOT_GRANTED
   [2003/06/16 12:04:00, 5] lib/util.c:show_msg(275)
     size=35
     smb_com=0x24
     smb_rcls=85
     smb_reh=0
     smb_err=49152
     smb_flg=136
     smb_flg2=16385
 --snip--

What's happen here? Could also be the wrong snip...


>From the application developers from the CAD-System we got the info this
behavior is an restriction of the NT4-SMB-Protokoll

 --snip--
 Win Server Communications Limits

 For Windows 2000 Client to Windows 2000 Server communications the limit is 
 16384.

 For Legacy Clients to Windows 2000 Server communications the limit is 8192.

 For all Clients to Windows NT 4.0 Server communications the limit is 2048.
 --snip--

If this is right, can we act like a Win2000 Server?

Has someone some ideas what we can do/fix ...?

Please email me directly for Log/conf/...


Thanks a lot
-- 
Peter Schinagl                      email: Peter.Schinagl at suse.de
-----------------------------------------------------------------
                      - SuSE Linux AG -                         
     Divison ES, Deutschherrnstr. 15-19, 90429 Nuernberg
-----------------------------------------------------------------