[Samba] Intermittent file server problems

[Jim]

Hi - We have been running into some intermittent problems with Samba in
a configuration where it is employed as a file server for several Windows
Server 2003 boxes. This problem generally occurs only under heavy load
when running some I/O intensive programs on the Windows side. Another
common element seems to involve working with files that remain open for
a long duration and are heavily updated.

On the Windows side the failure is typically preceded by a series of event
log messages similar to the following.

   MRxSmb Warning Event ID: 50

   {Delayed Write Failed} Windows was unable to save all the data for the
   file \Device\LanmanRedirector. The data has been lost. This error may be
   caused by a failure of your computer hardware or network connection.
   Please try to save this file elsewhere.

   (This message is the one that includes the c000020c status code)

   Application Popup Information Event ID: 26

   Application popup: Windows - Delayed Write Failed: Windows was unable to
   save all the data for the file \\Server\mount\path\to\file.dat. The data
   has been lost. This error may be caused by a failure of your computer
   hardware or network connection. Please try to save this file elsewhere.

At the time of the failure there is another pair of messages that are
identical to these, except that the path in the Information event changes
to \path\to\file.dat (i.e. the mount point information drops out). I
believe I have now read all relevant MS knowledge base articles on these
events and none of them seem to be directly applicable to our situation.

On the Samba side at the time of failure, with the logging level bumped
up to 3, the following appears in the log.

   [2005/04/28 17:43:05, 3] smbd/process.c:process_smb(1091)
     Transaction 333515 of length 76
   [2005/04/28 17:43:05, 3] smbd/process.c:switch_message(886)
     switch message SMBtrans2 (pid 4787) conn 0x88c0a60
   [2005/04/28 17:43:05, 3] smbd/trans2.c:call_trans2qfilepathinfo(2291)
     call_trans2qfilepathinfo: TRANSACT2_QFILEINFO: level = 1005
   [2005/04/28 17:43:05, 3] smbd/error.c:error_packet(129)
     error packet at smbd/trans2.c(2326) cmd=50 (SMBtrans2) NT_STATUS_INVALID_HANDLE

This sequence repeats four times and then it appears that Samba proceeds
to start closing a lot of files that were opened by the application on
the Windows side.

We are using a minimal Samba configuration intended only to support basic
file serving. Everything seems to work reasonably well except for these
occasional failures. Unfortunately such failures often cost us hours of
work.

We are running Samba under Red Hat Enterprise Linux AS 4. We are using
the current Samba RPM's provided by Red Hat for this platform (samba,
samba-client, and samba-common all of version 3.0.10-1.4E). We have
checked switches, cables, NICs, and drives and haven't found any obvious
problems there.

If anyone could provide ideas about how to solve or further debug this
problem it would be greatly appreciated.

Jim