samba - Aug 2000 - samba runaway processes

[This is a discussion that started earlier about automounter]

John Posenau wrote:
>     What I've found so far is that the windows client (any machine,
roving
> profile) with this user's account authenticates in a samba domain on
machine
> A. His files are located on  machine B (not running Samba). Machine A
> automount his nisplus_home. A little indirection but hey. 
	As always, mixing NFS and SMB is immensely useful,
	but sometimes drives you mad [It's what I'm using
	at work, to reduce load on the sysadmin team: we
	no longer have a large number of NT guru-level
	developers, so we can afford the performance hit]

>					After increasing the
> log level, it looks like when he works on an Office 2000 document it
doesn't
> matter whether its word, excel or access, it appears that the oplocks do
not
> always work correctly. I'll see multiple requests for breaks without a
> repsonse from the machine holding the lock.
	I'm CCing this to the samba@samba.org mailing list,
	as the team is watching out for incompatibilities
	with late-model MS products, including win2k and 
	office2k.


[paraphrase] The user in question often accesses the same
file from any one of a pool of machines, and it looks like
the hand-off from machine to machine isn't being dome cleanly.

	That makes me nervous... anyone know if this is likely
	to be an interaction with NFS???

> What I'm currently trying as of 10:00P.M. last night is to disable
oplocks on
> that share. I think I can live with the performance hit if it solves the
> machine locking up problem.
	This is a good idea, especially if it's this user's
	private directory, so that other users still get oplocks.

	If it's [homes] instead, You might try adding 
		include = /opt/samba/lib/extra.%m
	to the homes share and and creating a extra.<the name of 
	the user> file  which contains "oplocks = no".

	That way only that particular user gets his oplocks turned off.

> While 
> I've got your attention is there a simple system way (Solaris) to kill
all of
> a user's processes or do I just need to write a script?
	pkill will whack his normal processes, but for samba you
	may need to write a script that looks at smbstatus output
	to see which pid to kill.

--dave
-- 
David Collier-Brown,  | Always do right. This will gratify some people
185 Ellerslie Ave.,   | and astonish the rest.        -- Mark Twain
Willowdale, Ontario   | //www.oreilly.com/catalog/samba/author.html
Work: (905) 415-2849 Home: (416) 223-8968 Email: davecb@canada.sun.com

[This discussiuon redirected to samba@samba.org, until we find 
something germaine --dave]

I wrote:
> >   Gordon Lack just found an interesting symptom for us: if I
> > interpret this properly, it looks like Samba has problems
> > when the underlying filesystem locks (in this case NFS) were
> > wonky
Jie J Zhang wrote:
> Yep, I've experienced the same thing.  We had samba running on a local
dir
> and accessing local files with no prob.  Second we moved it to another
> machine access samba and its shared dirs thru NFS, [...]
> Particularly in the fact that users spawned millions of
> processes. [about 5 times the normal number]
	I just built a dead-stock 2.0.7 samba and brought
	it up on my workstation, looking at both NFS-mounted
	and local disks.

	1) With lockd running on Solaris 8, everything is fine.
	   smbclient //elsbeth/davecb returns normal results
	   and there are only two smbds running.

	2) When I kill lockd with smbd running, nothing changes.
	
	3) When I start smbd with no lockd, it starts normally,
	   and I get only the expected two smbds.


	There is clearly a problem, but I can't reproduce it on 
	Solaris 8 and Samba 2.0.7
	
	Can anyone try this experiment on a late-model Linux?

--dave
-- 
David Collier-Brown,  | Always do right. This will gratify some people
185 Ellerslie Ave.,   | and astonish the rest.        -- Mark Twain
Willowdale, Ontario   | //www.oreilly.com/catalog/samba/author.html
Work: (905) 415-2849 Home: (416) 223-8968 Email: davecb@canada.sun.com