samba - Mar 2002 - oplock_break and Excel data corruption with Samba 2.2.3a

Hi everyone,

Just today one of our users informed me that a file she was working on in
Microsoft Excel 97 got corrupted. She was working on the file directly on
our server, which is running Samba 2.2.3a compiled using gcc-3.0.4 on
Linux kernel 2.4.17-xfs and XFS-based filesystems.

I checked the Samba logs for her machine and found the following in a
search for the filename of the Excel file she was working on:

[2002/03/01 10:52:42, 0] smbd/oplock.c:oplock_break(758)
  oplock_break: receive_smb error (Success)
  oplock_break failed for file Gov't Agencies/PAG-IBIG/PAG-IBIG LOAN
LEDGER.xls (dev = 80a, inode = 9379268, file_id = 1).
[2002/03/01 10:52:42, 0] smbd/oplock.c:oplock_break(843)
  oplock_break: client failure in break - shutting down this smbd.
[2002/03/01 10:52:42, 1] smbd/service.c:close_cnum(653)
  payroll (192.168.0.20) closed connection to service ap
[2002/03/01 10:52:42, 1] smbd/service.c:make_connection(615)
  payroll (192.168.0.20) connect to service ap as user milanie (uid=1047,
gid=50) (pid 26341)

There were about 19 occurences of this series one after the other, until I
guess she hit that corruption (whatever actually caused it) and obviously
gave up.

I double-checked my smb.conf and there are no special lines to modify any
of the oplock settings. I double-checked the cable using a UTP cable
tester and it's all okay. The machine she was using was running Microsoft
Windows 95 with a 10Mbps NIC, a combination that's been stable so far.

Is there anything else I should check? I've heard she's not the only one
who ran into Excel data corruption problems lately, except the others
didn't report the incidents (grumble).

Is there anything I can do (even perhaps at the expense of performance) to
make things more robust? I'm afraid our data might be going down the drain
and I do not want that to happen. Yes, we have backups, but restoring
yesterday's data everyday isn't our view of productivity. ;>

I would appreciate any thoughts or follow-up questions on this.

Please be kind enough to cc me replies to this thread as I am not on the
Samba mailing list. I will continue to check the web archives, though.

Tanks a lot in advance.

 --> Jijo

--
Federico Sevilla III  :: jijo@leathercollection.ph
Network Administrator :: The Leather Collection, Inc.
GnuPG Key: http://jijo.leathercollection.ph/jijo.gpg

On Fri, 1 Mar 2002 at 21:31, Federico Sevilla III wrote:
> I checked the Samba logs for her machine and found the following in a
> search for the filename of the Excel file she was working on:
I did some follow-up scans of the Samba logs and found out the following
statistics:

1. Most of the "oplock_break failed" entries are for Excel files but
are
not limited to this.

   a. There are 97 logged oplock_break failures. Logs are exclusive to
Samba 2.2.3a installed 2002/02/25, since I flushed the logs during the
upgrade.

   b. Of these, 79 are with Excel files.

      1) 15 are with an installation of Microsoft Excel 2000 on a
Microsoft Windows ME machine, with complete driver updates.

      2) The rest are with installations of Microsoft Windows 97 on
Microsoft Windows 95 OSR2 and 98, also updated, with some almost fresh
installs on new machines.

   c. The remaining 18 are distributed among Word and PowerPoint files.


2. I am not certain if the corruption is caused by the oplock_break
failures. Not all of the files whose modify operations ran into
oplock_break failures are corrupt.


3. Aside from oplock_break, the related "client failure in break -
shutting down this smbd", and authentication failures, the most frequent
errors are related to tdb:

[2002/02/27 14:17:59, 0] smbd/oplock.c:oplock_break(758)
  oplock_break: receive_smb error (Success)
  oplock_break failed for file 2002 JO Masterfile/ProdSched2002Q1.xls (dev =
80a, inode = 62920619, file_id = 2).
[2002/02/27 14:17:59, 0] smbd/oplock.c:oplock_break(843)
  oplock_break: client failure in break - shutting down this smbd.
[2002/02/27 14:17:59, 1] smbd/service.c:close_cnum(653)
  boss (192.168.0.3) closed connection to service jomaster
[2002/02/27 14:17:59, 0] tdb/tdbutil.c:tdb_log(475)
  tdb(/var/run/samba/connections.tdb): remove_from_freelist: not on list at
off=9376
[2002/02/27 14:17:59, 0] tdb/tdbutil.c:tdb_log(475)
  tdb(/var/run/samba/connections.tdb): tdb_free: left free failed at 9376
[2002/02/27 14:17:59, 1] smbd/service.c:make_connection(615)
  boss (192.168.0.3) connect to service jomaster as user boss (uid=1001, gid=50)
(pid 16809)


   a. This is only an example, though. Not all of the tdb_free errors are
preceded by oplock_break errors.

   b. The oplock_break with the reported data corruption that I sent in my
first post did not have tdb_free errors directly before or after it, or
anywhere relatively near timestamp-wise.

   c. Up to Samba 2.2.3 we ran into severe problems related to tdb.

      1) Something that the logs could not explain (at least to me, at log
level 1) would make Samba grind to a halt, rejecting all new connections
because of max_connection problems.

      2) My latest report on this went unanswered. It had the subject
"rec_free_read bad magic and tdb corruption with Samba 2.2.3":
<http://lists.samba.org/pipermail/samba/2002-February/067485.html>.


That's all I've got for now. I have done my best to detail the whole web
of problems we've been experiencing lately, in the hopes that somehow the
little details will help the developers see the bigger picture and
hopefully pinpoint the root.

Again I request that I be cc'd replies to the thread.

Thank you very much in advance.

 --> Jijo

--
Federico Sevilla III  :: jijo@leathercollection.ph
Network Administrator :: The Leather Collection, Inc.
GnuPG Key: http://jijo.leathercollection.ph/jijo.gpg

Rudolf,
(cc Samba list)

On Fri, 1 Mar 2002 at 16:03, Rudolf Kliemstein wrote:
> had the same problem except my data didnt corrupt but i couldnt open
> excel97 files.
It's interesting how despite the variety of files with oplock_break
errors, it's only the Microsoft Excel 97 files that ran into these
"corruption" problems. By "corruption" I mean suddenly the
file couldn't
be opened because Microsoft Excel 97 would complain that the file was of
an invalid format. The rest of the Excel files are okay, though.
> my problem was solved by setting "nt acl support = no" for that
share.
> since you have win95 and samba2.2.3a enables acl by default this could
> be the problem.
It sounds wierd, but I'm jumping at it anyway. "nt acl support" is
now
formally a per-share setting. Will it work fine as a global directive as
well?

Thanks. :)

 --> Jijo

--
Federico Sevilla III  :: jijo@leathercollection.ph
Network Administrator :: The Leather Collection, Inc.
GnuPG Key: http://jijo.leathercollection.ph/jijo.gpg

If you do not care about the preformance you can turn of oplocks. 
Just set "oplocks = no" in smb.conf

Christian

> Hi everyone,
> 
> Just today one of our users informed me that a file she was working on in
> Microsoft Excel 97 got corrupted. She was working on the file directly on
> our server, which is running Samba 2.2.3a compiled using gcc-3.0.4 on
> Linux kernel 2.4.17-xfs and XFS-based filesystems.
> 
> I checked the Samba logs for her machine and found the following in a
> search for the filename of the Excel file she was working on:
> 
> [2002/03/01 10:52:42, 0] smbd/oplock.c:oplock_break(758)
>   oplock_break: receive_smb error (Success)
>   oplock_break failed for file Gov't Agencies/PAG-IBIG/PAG-IBIG LOAN
LEDGER.xls (dev = 80a, inode = 9379268, file_id = 1).
> [2002/03/01 10:52:42, 0] smbd/oplock.c:oplock_break(843)
>   oplock_break: client failure in break - shutting down this smbd.
> [2002/03/01 10:52:42, 1] smbd/service.c:close_cnum(653)
>   payroll (192.168.0.20) closed connection to service ap
> [2002/03/01 10:52:42, 1] smbd/service.c:make_connection(615)
>   payroll (192.168.0.20) connect to service ap as user milanie (uid=1047,
gid=50) (pid 26341)
> 
> There were about 19 occurences of this series one after the other, until I
> guess she hit that corruption (whatever actually caused it) and obviously
> gave up.
> 
> I double-checked my smb.conf and there are no special lines to modify any
> of the oplock settings. I double-checked the cable using a UTP cable
> tester and it's all okay. The machine she was using was running
Microsoft
> Windows 95 with a 10Mbps NIC, a combination that's been stable so far.
> 
> Is there anything else I should check? I've heard she's not the
only one
> who ran into Excel data corruption problems lately, except the others
> didn't report the incidents (grumble).
> 
> Is there anything I can do (even perhaps at the expense of performance) to
> make things more robust? I'm afraid our data might be going down the
drain
> and I do not want that to happen. Yes, we have backups, but restoring
> yesterday's data everyday isn't our view of productivity. ;>
> 
> I would appreciate any thoughts or follow-up questions on this.
> 
> Please be kind enough to cc me replies to this thread as I am not on the
> Samba mailing list. I will continue to check the web archives, though.
> 
> Tanks a lot in advance.
> 
>  --> Jijo
> 
> --
> Federico Sevilla III  :: jijo@leathercollection.ph
> Network Administrator :: The Leather Collection, Inc.
> GnuPG Key: http://jijo.leathercollection.ph/jijo.gpg
> 
> 
> -- 
> To unsubscribe from this list go to the following URL and read the
> instructions:  http://lists.samba.org/mailman/listinfo/samba
> 

               _(_)_                          wWWWw   _
   @@@@       (_)@(_)   vVVVv     _     @@@@  (___) _(_)_
  @@()@@ wWWWw  (_)\    (___)   _(_)_  @@()@@   Y  (_)@(_)
   @@@@  (___)     `|/    Y    (_)@(_)  @@@@   \|/   (_)\
    /      Y       \|    \|/    /(_)    \|      |/      |
 \ |     \ |/       | / \ | /  \|/       |/    \|      \|/
jgs|//   \\|///  \\\|//\\\|/// \|///  \\\|//  \\|//  \\\|// 
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Pearse,
(cc Samba list)

Quoting Pearse Stokes <info@ecobaby.ie>:
> Hi, from Pearse Stokes at Ecobaby Limited in Dublin,
> 
> We've been having increasing problems with TAS accounting software that
> uses Btrieve databases on a Linux RH7.0 server and old Samba.
> 
> Because of that, we are installing a new server with Mandrake Pro Suite 8.0
> and Samba 2.2.1a. However, our problems have increased. These don't
appear
> to affect software other than the Btrieve based stuff.
> 
> Noting your problem with Excel etc., have you found a solution? Did you try
> the oplocks change suggested, and if so, did it work?
Using Samba 2.2.3a with oplocks completely disabled, our corruption problems
with Microsoft Excel have gone away completely. Thankfully with the acceptable
loads in our environment (vis a vis the capabilities of the server) there has
been no noticeable slowdown in operations even with oplocks disabled.
> Any ideas would be appreciated. Am downloading Samba 2.2.4 right now.
If you are not under intense pressure to fix things, you may want to upgrade to
Samba 2.2.4 while keeping oplocks enabled, and then only disable oplocks if
things still break.

Hopefully the other people on the Samba list will have more tips to share with
you. Good luck with your quest to fix things!

 --> Jijo

-- 
Federico Sevilla III   :  <http://jijo.free.net.ph/>
Network Administrator  :  The Leather Collection, Inc.
GnuPG Key Fingerprint  :  0x93B746BE