samba - Apr 2003 - 2.2.8 slows down to a crawl within a few days

We have been using samba for years now and are generally very happy.

But since a recent upgrade to 2.2.8 we have seen samba slowing down
periodically to a crawl every 2 or 3 days. BTW the start of these problems
coincided with the upgrade to samba 2.2.8 and a partial cabling upgrade to
100 MBit, which also coincided with the introduction of the first Win XP 
Pro
machines to our LAN. ;-(

We have about twenty users logged in at any one time and when it comes 
to
"the crawl", every two or three days, then everything on the samba
server
gets extremely slow. Documents that ordinarily open up inside 5-10 
seconds
take in excess of 5 minutes .... People start restarting their PCs and form
then on my phone starts ringing no end ;-)

We appear to be having lots of oplock timing out issues. It's probably best
if I don't say too much, except that we have set the "oplock break wait
time
= 20" with no improvement, but let the logfile speak:

/var/log/samba/log.smbd 
 
[2003/03/26 09:31:48, 0] smbd/oplock.c:request_oplock_break(1011) 
request_oplock_break: no response received to oplock break request to 
pid 
3092 
3 on port 33142 for dev = 815, inode = 32781, file_id = 1 
[2003/03/26 09:31:48, 0] smbd/open.c:open_mode_check(651) 
open_mode_check: exlusive oplock left by process 30923 after break ! 
For 
file 
logon.bat, dev = 815, inode = 32781. Deleting it to continue... 
[2003/03/26 09:31:48, 0] smbd/open.c:open_mode_check(655) 
open_mode_check: Existent process 30923 left active oplock. 
[2003/03/26 09:32:01, 0] smbd/oplock.c:oplock_break(824) 
oplock_break: no break received from client within 30 seconds. 
oplock_break failed for file logon.bat (dev = 815, inode = 32781, file_id 
= 1) 
. 
[2003/03/26 09:32:01, 0] smbd/oplock.c:oplock_break(869) 
oplock_break: client failure in oplock break in file logon.bat 
[2003/03/26 09:32:01, 0] tdb/tdbutil.c:tdb_log(531) 
tdb(/var/cache/samba/locking.tdb): tdb_lock failed on list 4 ltype=1 
(Resource 
deadlock avoided) 
[2003/03/26 09:32:01, 0] smbd/oplock.c:remove_oplock(272) 
remove_oplock: failed to lock share entry for file logon.bat 
[2003/03/26 09:32:01, 0] tdb/tdbutil.c:tdb_log(531) 
 
What's not shown in the samba logs is the msg which shows up in 
/var/log/messages 
 
kernel: lease timed out 
 
in the middle of the oplock_break msgs which tends to follow the 
"exclusive 
oplock left by active process" message 
 
I have also noticed that I have lots of "stale user sessions". What I
mean
is that according to "smbstatus -b" users, although logged out have up
to
twenty or more sessions open from the same machine. (this was especially 
bad 
for win xp pro machines) 
 
Any hints on how to track this down? Is this likely a hardware (i.e. cabling 
or nic), a samba software, or a client side (i.e. win xp pro) issue? Or is 
it a combination of any of these three? How would you go about clearing 
up 
an error happening about every two days when you can't really impact any 
of 
the 20 users doing it? 
 
Many thanks for any pointers. 
 

Urs RAU

Head of Information Services OM UK HeadQuarters