samba - Feb 2007 - name query returning all ip addresses for server instead of primary one

Our server is multihomed, and I've been finding that sometimes clients 
doing a NetBIOS Name Service (NBNS) query to our samba server sometimes 
end up getting and using one of the non-primary addresses. I used 
ethereal to look at the actual NBNS response from the server and notice 
that all of its ip addresses are listed, with the primary one first but 
all the others as well. I guess the client somehow picks randomly among 
them? Clients picking the wrong one seems to happen particularly often 
on vpn connections via the pptpd server running on the same machine.

I've perused the samba o'reilly book, online documentation, and the 
mailing list archives, and I had thought that the "interfaces" and
"bind
interfaces only" commands in smb.conf should restrict the samba server 
to only vending the ip address of the primary interface eth0, but that 
doesn't seem to work. Does anyone know how to get samba to only reply 
with one of its ip addresses to NBNS queries?

Thanks!
Antony


System info:
  CentOS 4, Linux 2.6.9-42.0.3.ELsmp

Samba rpm's installed:
  samba-common-3.0.10-1.4E.9
  samba-3.0.10-1.4E.9


Here is our smb.conf:

(I've changed the actual server name changed to "servername", and 
workgroup changed to "serverworkgroup")

[global]
   netbios name = servername
   server string    workgroup = serverworkgroup
   encrypt passwords = yes

   domain master = yes
   local master = yes
   preferred master = yes
   os level = 33

   security = user
   domain logons = yes

   # We set these to empty so that we don't use roaming profiles
   logon path    logon home 
   time server = yes

   admin users = root

   wins support = yes

   # only bind to the loopback and first ethernet interfaces
   interfaces = lo eth0
   bind interfaces only = yes

   socket options = TCP_NODELAY SO_RCVBUF=8192 SO_SNDBUF=8192

[home]
   path = /home/
   veto files = /lost+found/
   writeable = yes
   browseable = yes
   guest ok = no

On Friday 09 February 2007, Antony Sargent wrote:
> ? ?# only bind to the loopback and first ethernet interfaces
> ? ?interfaces = lo eth0
> ? ?bind interfaces only = yes
Back on 10/15/06 I posted about an issue - titled:
"nmbd problems with secondary lo".
No responses were received.

It's not the same problem you have but they might be related.

The workaround I posted then was to use an IP address in place of an interface 
name. You might want to try that.
Ex:
-----------------------------------------
interfaces = 192.168.1.2, 127.0.0.1
bind interfaces only = Yes
hosts allow = 192.168.1. 127.
-----------------------------------------

Chris

On Friday 09 February 2007, Antony Sargent wrote:
> ? ?# only bind to the loopback and first ethernet interfaces
> ? ?interfaces = lo eth0
> ? ?bind interfaces only = yes
And one more thought. Always use testparm to see what's really being set.
One
time I had made some changes but further down in the conf file the change was 
being reversed by the distros default conf entry that I missed when skimming 
the file. So if you had:
-----------------------------------
	bind interfaces only = yes
...
...
	bind interfaces only = no
-----------------------------------

The "no" entry is coming after the "yes" entry and therefore
"no" will take
precedence, which testparm would confirm, eliminating a skimming error and 
guesswork.

Chris