samba - Nov 1999 - User's smbd processes don't get killed

Hi, all.

	Last saturday, we've done a cutover for Samba 2.0.5a to be our new
fileserver. All is going well, except for two network users that make me
reboot the server.

	What happens is as follows:

	* user logs on onto samba server;

	* user opens and closes files, etc, OK until in one of these
operations his/her machine hangs;

	* user reboots his/her machine to log on again. In this processes,
we see another smbd being spawned, and the first smbd doesn't get killed
at all. Even "kill -9" does not work. Meanwhile, the CPU usage of the
process grows without limits, and renice 20 <pid> has no effects.

	The strangest of all this behavior is that I have users that open
80+ files simultaneously (a Clipper application, actually), and they
don't have any problems at all.

	I've seen this happening both in a WinNT/Wks/SP4/Portuguese and in
a Win95/SP1/English, normally with Office applications and with just file
browsing.

	Any suggestions? Faulty network card, neccessary hotfixes... I'm
running Samba 2.0.5a (RPM) with RH6.0 and customized Linux Kernel 2.2.13
with e2comp patches, in a NetFinity 3500 Server with a Mylex DAC960 RAID
controller. My smb.conf is attached too.

	TIA,

	Paulo Fessel
-------------- next part --------------
# Samba config file created using SWAT
# from localhost (127.0.0.1)
# Date: 1999/11/08 18:31:53

# Global parameters
[global]
	workgroup = NETSOLUTIONS
	server string = Servidor de Arquivos Red Hat Linux 6.0
	encrypt passwords = Yes
	update encrypted = Yes
	passwd program = /bin/echo %u:%n | /usr/sbin/chpasswd
	passwd chat = .
	username map = /etc/users.map
	unix password sync = Yes
	log level = 0
	log file = /var/log/samba/log.%m
	max log size = 512
	deadtime = 120
	keepalive = 0
	socket options = SO_KEEPALIVE TCP_NODELAY SO_RCVBUF=12288 SO_SNDBUF=12288
	load printers = No
	logon path = \\%N\profiles\%U
	logon drive = H:
	domain logons = Yes
	preferred master = Yes
	domain master = Yes
	dns proxy = No
	create mask = 0640
	directory mask = 0750
	hosts allow = localhost, 192.168.1.
	printing = lprng
	print command = lpr -P%p %s; rm %s
	wide links = No

[homes]
	comment = Diret?rios pessoais dos usu?rios
	path = /home/%u
	read only = No
	create mask = 0600
	browseable = No

[comercial]
	path = /home/comercial
	read only = No
	browseable = No

[cdrom]
	path = /mnt/cdrom

[netlogon]
	path = /etc/netlogon
	read only = No
	browseable = No

[profiles]
	comment = Perfis ambulantes dos Usuarios
	path = /home/profiles
	read only = No
	create mask = 0600
	directory mask = 0700
	browseable = No

[tecnologia]
	comment = Diret?rio compartilhado - Departamento de Tecnologia
	path = /home/tecno/TECNO
	read only = No
	create mask = 0660
	directory mask = 0770

[PRINTER$]
	path = /etc/printer

[sistemas]
	path = /home/sistemas
	read only = No

[filialrj]
	comment = Diret?rio de Trabalho - Filial RJ
	path = /home/laierte
	read list = gnavas
	write list = @diretoria, @tecnlogia
	create mask = 0660
	directory mask = 0770

[listapre]
	comment = ListaPre1.xls - BLEAAAARGH!
	path = /home
	read list = dnakashi, cteles, ksantos
	write list = ebueno, gnavas, mraucci, dlima
	read only = No

[lp_comerc]
	path = /var/spool/samba
	print ok = Yes
	printing = bsd

I'd start by finding commonalities between the two users having the
problem, and the several users not having the problem. Exploit the
differences between the two groups. 

Temporarily put the 2 users on different boxes, and put other users on the
problematic boxes. Does the problem follow the users, or the box? Just keep
narrowing it down.

Steve Litt

At 09:32 AM 11/09/1999 +1100, Paulo Afonso Graner Fessel
wrote:
>	Hi, all.
>
>	Last saturday, we've done a cutover for Samba 2.0.5a to be our new
>fileserver. All is going well, except for two network users that make me
>reboot the server.
>
>	What happens is as follows:
>
>	* user logs on onto samba server;
>
>	* user opens and closes files, etc, OK until in one of these
>operations his/her machine hangs;
>
>	* user reboots his/her machine to log on again. In this processes,
>we see another smbd being spawned, and the first smbd doesn't get killed
>at all. Even "kill -9" does not work. Meanwhile, the CPU usage of
the
>process grows without limits, and renice 20 <pid> has no effects.
>
>	The strangest of all this behavior is that I have users that open
>80+ files simultaneously (a Clipper application, actually), and they
>don't have any problems at all.
>
>	I've seen this happening both in a WinNT/Wks/SP4/Portuguese and in
>a Win95/SP1/English, normally with Office applications and with just file
>browsing.
>
>	Any suggestions? Faulty network card, neccessary hotfixes... I'm
>running Samba 2.0.5a (RPM) with RH6.0 and customized Linux Kernel 2.2.13
>with e2comp patches, in a NetFinity 3500 Server with a Mylex DAC960 RAID
>controller. My smb.conf is attached too.
>
>	TIA,
>
>	Paulo Fessel
>
>Attachment Converted: "e:\eudora\attach\smb4.con"
>