[Samba] Random "File Not Found" errors Using Samba, Works on NetApp

[Clay Jordan]

I am experiencing a strange problem where I am randomly getting "file not
found" error messages on a Samba share when they are accessed via a series
of tcl scripts running from a Win2K server.  Some of the scripts are located
on the 2K server and some reside on the Samba share. Unfortunately, this is
a third party application and I do not know enough tcl to usefully
troubleshoot it from that direction; however we also have a second site
using a similar setup except that the share resides on a NetApp Filer and
there they do not experience this problem.

Nonetheless, I am fairly sure that the problem resides with the tcl scripts
possible passing additional control characters when trying to access files.
In addition to this problem, occasionally trash files are generated with
control characters such that I have to do a find by inode and delete them
from the UNIX side. I am using Samba 2.0.9 as provided by HP on an HP
9000/RP7410. I have also compiled and tried later Samba versions ( 2.2.5
compiled by me and 2.2.3a as provided by the latest HP package) and the
problem gets worse, zeroing out the contents of one of the initial tcl
scripts used from the Samba share.

The scenario for the application is this. A series of twenty files are
generated and placed on the share from the UNIX side. The NT based tcl app
is supposed to process these files and return a single file to the share
which is then picked up from the UNIX side.  The NT App gets the files from
the Samba share that resides on this UNIX box. In the NetApp scenario, the
difference is that the share is mapped via NetApp CIFS to the Windows server
and NetApp NFS to the UNIX server.

I've also tried another route, using Windows 2K UNIX services "Client for
NFS" and I get similar random occurrences of the "file not found" error
message. I''ve logged in detail some of the problems from the log.smb and
they are below.

My question is, has anyone seen this before or can you think of a reason it
does not happen with the NetApp filer ?


Here it looks like it get's the call for the correct name but ends up
loosing it after mapping the name to UNIX.

[2002/08/16 11:12:08, 3] lib/util.c:(522)
  unix_clean_name [/SiteList.csv]
[2002/08/16 11:12:08, 3] lib/util.c:(522)
  unix_clean_name [SiteList.csv]
[2002/08/16 11:12:08, 3] lib/util.c:(522)
  unix_clean_name [./]
[2002/08/16 11:12:08, 3] smbd/dir.c:(490)
  creating new dirptr 256 for path ./, expect_close = 1
[2002/08/16 11:12:08, 3] smbd/process.c:(618)
  Transaction 146 of length 560
[2002/08/16 11:12:08, 3] smbd/process.c:(448)
  switch message SMBnttrans (pid 14587)
[2002/08/16 11:12:08, 3] lib/util.c:(522)
  unix_clean_name [U/]
[2002/08/16 11:12:08, 3] lib/util.c:(522)
  unix_clean_name [U]
[2002/08/16 11:12:08, 3] lib/util.c:(522)
  unix_clean_name [U]

Here's a more detailed log of the error when it works:

2002/08/16 15:04:45, 10] lib/util.c:(2926)
  [020] 00 00 00 40 00 00 00 90  01 00 00 00 00 00 00 1A  ... at .... ........
[2002/08/16 15:04:45, 10] lib/util.c:(2926)
  [030] 00 00 00 02 00 00 00 03  00 5C 00 53 00 69 00 74  ........ .\.S.i.t
[2002/08/16 15:04:45, 10] lib/util.c:(2926)
  [040] 00 65 00 4C 00 69 00 73  00 74 00 2E 00 63 00 73  .e.L.i.s .t...c.s
[2002/08/16 15:04:45, 10] lib/util.c:(2926)
  [050] 00 76 00 01 00 04 80 58  01 00 00 74 01 00 00 00  .v.....X ...t....
[2002/08/16 15:04:45, 10] lib/util.c:(2926)
  [060] 00 00 00 14 00 00 00 02  00 44 01 04 00 00 00 00  ........ .D......

and when it doesn't work, I get this

[2002/08/16 15:03:44, 10] lib/util.c:(2926)
  [020] 00 00 00 40 00 00 00 90  01 00 00 00 00 00 00 1A  ... at .... ........
[2002/08/16 15:03:44, 10] lib/util.c:(2926)
  [030] 00 00 00 02 00 00 00 03  64 5C 00 53 00 69 00 74  ........ d\.S.i.t
[2002/08/16 15:03:44, 10] lib/util.c:(2926)
  [040] 00 65 00 4C 00 69 00 73  00 74 00 2E 00 63 00 73  .e.L.i.s .t...c.s
[2002/08/16 15:03:44, 10] lib/util.c:(2926)
  [050] 00 76 00 01 00 04 80 58  01 00 00 74 01 00 00 00  .v.....X ...t....

  -----Original Message-----
  From: Larry McClellan [mailto:mcclellan at sugar-land.oilfield.slb.com]
  Sent: Saturday, August 17, 2002 11:41 AM
  To: Clay Jordan; Jeffrey Willett
  Subject: Tested Direct Samba extract from Srv252sgl-failure attached


  Hi Clay,
  Got your message, and tested a direct extract from SRV252SGL. I'm afraid I
gout the usual error. Attached is a screen shot of the UCL command window.

  I was able to move the same files to the D drive, and run without error.

  Sounds like you are on the trail, however.

  Thanks,
  Larry



  Larry McClellan
  Schlumberger Oilfield Services
  WCP- Advanced Manufacturing Systems
  7030 Ardmore Street
  Houston,Tx 77054
  713-715-2251
  mcclellan at sugar-land.oilfield.slb.com



-------------- next part --------------
HTML attachment scrubbed and removed