Samba screwed up when trying to save a word97 document

[Chris Stacey]

Dear All,

We are in the process of evaluating Samba to provide SMB access to unix
file storage for a few hundred PC's.

One of our tests is to save a Word97 document - something that has
caused various PCNFS clients problems in the past.  While saving the
document the client stopped and said something like connection lost, or
seesion closed, and the drives mapped from the samba box disappeared.  
Unfortunately, I wasn't doing the test and I can't remember exactly what
the error dialog said.

The tester remapped the drive and saved the document without problem.

I then had a look at the samba server and saw:

serenity# smbstatus

Samba version 1.9.18p8
Service      uid      gid      pid     machine
----------------------------------------------
bramsons     bramsons its      25917   ukswi1000 (175.12.50.218) Fri Jun
19 15:51:05 1998
its          bramsons its      25917   ukswi1000 (175.12.50.218) Fri Jun
19 15:51:19 1998
its          bramsons its      26270   ukswi1000 (175.12.50.218) Fri Jun
19 16:02:26 1998
its          bramsons its      26308   ukswi1000 (175.12.50.218) Fri Jun
19 16:03:41 1998
its          bramsons its      26325   ukswi1000 (175.12.50.218) Fri Jun
19 16:04:56 1998

Locked files:
Pid    DenyMode   R/W        Oplock           Name
--------------------------------------------------
25917  DENY_NONE  RDWR       EXCLUSIVE+BATCH  samba-tests/W97pure.doc  
Fri Jun 19 16:01:33 1998

Share mode memory usage (bytes):
   101984(99%) free + 336(0%) used + 80(0%) overhead = 102400(100%)
total
serenity# ls
locks          log.nmb        log.smb        log.ukswi1000 
log.ukswi3101
serenity# tail log.ukswi1000 
1998/06/19 17:23:20 request_oplock_break: no response received to oplock
break request to pid 25917 on port 3273 for dev = ff00061a, inode =
3b20d
1998/06/19 17:23:30 request_oplock_break: no response received to oplock
break request to pid 25917 on port 3273 for dev = ff00061a, inode =
3b20d
1998/06/19 17:23:41 request_oplock_break: no response received to oplock
break request to pid 25917 on port 3273 for dev = ff00061a, inode =
3b20d
1998/06/19 17:23:52 request_oplock_break: no response received to oplock
break request to pid 25917 on port 3273 for dev = ff00061a, inode =
3b20d
1998/06/19 17:24:02 request_oplock_break: no response received to oplock
break request to pid 25917 on port 3273 for dev = ff00061a, inode =
3b20d
1998/06/19 17:24:13 request_oplock_break: no response received to oplock
break request to pid 25917 on port 3273 for dev = ff00061a, inode =
3b20d
1998/06/19 17:24:24 request_oplock_break: no response received to oplock
break request to pid 25917 on port 3273 for dev = ff00061a, inode =
3b20d
1998/06/19 17:24:34 request_oplock_break: no response received to oplock
break request to pid 25917 on port 3273 for dev = ff00061a, inode =
3b20d
1998/06/19 17:24:45 request_oplock_break: no response received to oplock
break request to pid 25917 on port 3273 for dev = ff00061a, inode =
3b20d
1998/06/19 17:24:56 request_oplock_break: no response received to oplock
break request to pid 25917 on port 3273 for dev = ff00061a, inode =
3b20d
serenity# ps -ef | grep 25917
    root 25917 25831  0 15:30:06 ?         0:06
/usr/local/samba/bin/smbd
    root 27407 26968  0 17:25:16 ttyp1     0:00 grep 25917
serenity# 

Other info:

serenity# uname -a
HP-UX serenity B.10.20 A 9000/821 2002731580 two-user license

Client was an NT 4.0 SP3 box.

I tried killing the offending process (PID 25917) but it will not die. 
I stopped and started the samba daemons again, and at first glance they
appear to be providing service.

Has anyone seen this before?  

Why did it happen?  

How can I stop it from happening again?  

Is the unkillable process going to cause me problems?  I had been
telling people that one of the things I liked about samba is that it
doesn't force me to reboot the server - so much for that idea. 


On a slightly different note.  There seems to be very little impact to
connected clients when I restart the samba daemons.  What problems might
I cause for the clients by restarting samba?  It would be nice if I
could add and remove shares "on the fly."  Is that planned?

Thanks in advance for your help.

Cheers,
-- 
Chris Stacey
Unix Systems Administrator, Motorola ECID
Tel: +44 (0)1793 565142  Fax: +44 (0)1793 565419
mailto: cstacey1 at email.mot.com