Problems with SAMBA 2.2.2 Oplocks

[David Collier-Brown]

Drew.Zeller at statcan.ca wrote:
> [2002/01/10 11:27:06, 0] smbd/oplock.c:(1026)
>   request_oplock_break: no response received to oplock break request to pid
> 29920 on port 52339 for dev = 40050001, inode = 31477  for dev = 40050001,
> inode = 31477, tv_sec = 3c18bba8, tv_usec = e2dca

> [2002/01/10 11:27:06, 0] smbd/open.c:(554)
>   open_mode_check: exlusive oplock left by process 29920 after break ! For
> file windows/extend.dat, dev = 40050001, inode = 31477. Deleting it to
> continue...
> [2002/01/10 11:27:06, 0] lib/util.c:(1055)
>   PANIC: open_mode_check: Existant process 29920 left active oplock.

	This is usually a symptom of a client
	machine disappearing, leaving a samba
	child process listening to a dead 
	connection, holding locks and oplocks.
	When resurreected, the client tries to
	reconnect and the logs show similar
	entries.


> Also, when I checked, there are a number of smb processes left running on
> the server that I am unable to kill.

	While it doesn't help with then real problem
	(an in fact tends to hide it!), set
	keepalive = 3600

	This will cause samba to send a "are you alive"
	packet after 10 minutes of inactivity, and
	if the client is down, shut down and clean up
	the samba process serving it.

	The REAL problem is what's making a client
	need to disconnect/reconenct, and secondarily 
	why it's producing a PANIC...  can you look 
	at what machine it was, and see if it's
	the same one or the same subnet each time?
	The common cause is a bad NIC or hub!

--dave
-- 
David Collier-Brown,           | Always do right. This will gratify 
Performance & Engineering      | some people and astonish the rest.
Americas Customer Engineering, |                      -- Mark Twain
(905) 415-2849                 | davecb at canada.sun.com