[Samba] Problems with one thread excessive CPU usage.

[Kasper Sacharias Eenberg]

Good morning.

I'm having some problems with one thread of samba, taking excessive CPU
usage.
This is one single thread, that i can kill with 'kill -9 pid', and the
server will keep other connections.
This thread will severely slow down the server, and clients notify me
within minutes of this happening.
In essense the server becomes unuseable, because it's too slow.

This is happening with 2 different servers now.
Both of them are CentOS:
4.7 (fully upgraded) 
	samba-common-3.0.28-0.el4.9
	samba-3.0.28-0.el4.9
4.3 (not fully upgraded)
	samba-3.0.10-1.4E.12.2
	samba-client-3.0.10-1.4E.12.2
	samba-common-3.0.10-1.4E.12.2
	samba-swat-3.0.10-1.4E.12.2

The 4.7 was updated because of the problem, but it did not solve it.

The 4.7 company, that has the problem alot more than the 4.3 company,
are architects, and as such work with big CAD files.
The other company consists of average users, with text documents and
such.

Here are the rundowns of the configs. They're not complete, and i've
removed unneeded options. They are to give a general overview of what
options are in use.

---------------------------------------
Config 4.7:
[global]
        load printers = yes 
        socket options = TCP_NODELAY SO_RCVBUF=8192 SO_SNDBUF=8192
        domain master = yes
        encrypt passwords = yes
        wins support = true
        dns proxy = no 
        os level = 33
        printcap name = /etc/printcap

[sags_nr]
        directory mode = 777
        vfs objects = recycle

[profiles]
        store dos attributes = yes
        browseable = no

---------------------------------------
Config 4.3:

[global]
        keepalive = 30
        socket options = TCP_NODELAY SO_SNDBUF=8192 SO_RCVBUF=8192
        os level = 30
        dns proxy = No
        ldap ssl = no
        cups options = raw
        deadtime = 30
        use client driver = Yes
[c5]
        oplocks = no
        level2oplocks = no
        sync always = yes
        veto oplock files = /*.dat/*.DAT/*.usr/

[gruppe]
        vfs objects =recycle:repository /data/data/gruppe/.recycle
        vfs objects =recycle:keeptree
        vfs objects =recycle:versions
---------------------------------------

As you might see, both of them utilize vfs objects. With which i have no
experience.
Could this be the root of our problem, and is there a solution?


With regards, and thanks for your help,

________________________________________________________________________

Kasper Eenberg

HOVMARK DATA
Ravnevej 13   
dk-6705 Esbjerg Ø
tlf:   +45 76 12 59 04
mobil: +45 40 70 69 63