samba - Jan 2001 - Runaway nmbd processes

I have runaway nmbd processes. It's the second time I have this
problem in a few weeks. The first time I noticed it when I started
having 'cannot fork' errors. This time I'm more lucky and I noticed
it
earlier. Currently I have `ps aux | grep nmbd | wc -l` = 120 nmbd
processes. My log says:


[2001/01/05 19:09:40, 1] nmbd/nmbd.c:main(757)
  Netbios nameserver version 2.0.7 started.
  Copyright Andrew Tridgell 1994-1998
[2001/01/05 19:09:40, 0] nmbd/asyncdns.c:start_async_dns(150)
  started asyncdns process 20533
[2001/01/05 19:09:41, 0] lib/pidfile.c:pidfile_create(86)
  ERROR: nmbd is already running. File /var/samba/nmbd.pid exists and
process id 21248 is running.
[2001/01/05 19:09:41, 1] nmbd/nmbd.c:main(757)
  Netbios nameserver version 2.0.7 started.
  Copyright Andrew Tridgell 1994-1998
[2001/01/05 19:09:41, 0] nmbd/asyncdns.c:start_async_dns(150)
  started asyncdns process 20535
[2001/01/05 19:09:41, 0] lib/pidfile.c:pidfile_create(86)
  ERROR: nmbd is already running. File /var/samba/nmbd.pid exists and
process id 21248 is running.
...

   I got 39 messages like this in just two seconds. All these processes
are still running. An strace tells me they are blocked in a read:

   read(6,

   Then it stops and starts again later at 23:44. I don't know what's
so special about these times. Actually it seems to happen when I
boot a Windows machine on the network.
   This is with the Samba from the Debian distribution: 2.0.7-3.

   What's happening? I still have all the processes running so if you
tell me what to do maybe I can help.


--
Francois Gouget         fgouget@free.fr        http://fgouget.free.fr/
                  Hiroshima '45 - Czernobyl '86 - Windows '95

On Sat, 6 Jan 2001, Mark Symonds wrote:
>
> From: "Francois Gouget" <fgouget@free.fr>
[...]
> >    Then it stops and starts again later at 23:44. I don't know
what's
> > so special about these times. Actually it seems to happen when I
> > boot a Windows machine on the network.
> >    This is with the Samba from the Debian distribution: 2.0.7-3.
> >
[...]
> Only solution I've found is to revert to the old slink package,
> or upgrade to 2.2.19.  For other reasons I had to upgrade the 
> kernel from 2.2.17 to 2.2.19 ... after that, I upgraded the 
> samba packages and haven't had any problems.
> 
> If you get any further information on this it would be great 
> to hear.
   It happened again!

   It all started when I did the following:
 - a Win2000 laptop connected to my machine (traces below)
 - I also connected to a network which has a primary domain controller
for the domain for which Samba is usually a master browser
 - the laptop connected to a machine in that domain

   Now I don't know which operation specifically triggered the problem
because I've redone them all since then and had no problem. I suspect
more some sort of a race condition. The problem is my Samba server does
not get much exercise (used once/twice a week) so problems don't happen
very often.

   It created something like 219 nmbd processes and this time I ran out
of filedescs. The runaway processes were blocked in a "read(6, ..."
(probably not of much help). Here's what I have in syslog:

Feb 11 11:04:16 amboise nmbd[21783]: connect from 192.168.1.201
Feb 11 11:04:16 amboise nmbd[21783]: [2001/02/11 11:04:16, 0]
nmbd/asyncdns.c:start_async_dns(150) 
Feb 11 11:04:16 amboise nmbd[21783]:   started asyncdns process 21784 
Feb 11 11:04:16 amboise nmbd[21783]: [2001/02/11 11:04:16, 0]
lib/pidfile.c:pidfile_create(86) 
Feb 11 11:04:16 amboise nmbd[21783]:   ERROR: nmbd is already
running. File /var/state/samba/nmbd.pid exists and process id 8040 is
running. 
Feb 11 11:04:16 amboise inetd[1002]: /usr/sbin/tcpd: exit status 0x1
Feb 11 11:04:16 amboise nmbd[21785]: connect from 192.168.1.201
...

   And here's the corresponding section of the nmb log:

[2001/02/11 11:04:16, 1] nmbd/nmbd.c:main(757)
  Netbios nameserver version 2.0.7 started.
  Copyright Andrew Tridgell 1994-1998
[2001/02/11 11:04:16, 0] nmbd/asyncdns.c:start_async_dns(150)
  started asyncdns process 21784
[2001/02/11 11:04:16, 0] lib/pidfile.c:pidfile_create(86)
  ERROR: nmbd is already running. File /var/state/samba/nmbd.pid exists
and process id 8040 is running.
[2001/02/11 11:04:16, 1] nmbd/nmbd.c:main(757)
  Netbios nameserver version 2.0.7 started.
  Copyright Andrew Tridgell 1994-1998
...


   Since the last report I upgraded to the latest Debian samba packages
from unstable):
samba:        2.0.7-4
samba-common: 2.0.7-4
libcupsys2:   1.1.4-3 
(that last one is probably irrelevant, I don't do printing)


   Let me know if there's any other information that I can provide to
help.


--
Francois Gouget         fgouget@free.fr        http://fgouget.free.fr/
                              145 = 1! + 4! + 5!