Logging levels and browser elections

[David Dyer-Bennet]

For a set of systems that can almost talk to each other, I suspect I
have some really crazed underlying configuration glitches.  

I'm trying to add some samba servers (to make disks on my linux
servers visible to people on Windows boxes) to a small home network;
currently there is a Windows 3.11 box, two Windows 98 boxes, and an NT
Workstation (not server) box, in addition to the two Linux boxes with
Samba. Everything is on one IP net (10.x.x.x; in fact all assigned IPs
are in 10.0.0.x).  The wiring includes a compex 8-port dual speed
switch, and a d-link dsh-8 dual speed "hub with switch", which makes
me just *slightly* suspicious about broadcast propagation; though I
can't find any obvious patter to what doesn't get found.

I'm having troubles with boxes seeing each other in network
neighborhood.  I've been playing extensively, and at this point am
probably confused about what exactly I had set how when I saw which
results ;-) . 

I've tried forcing a Samba box to be master browser, which has
increased visibility a bit, but things aren't consistent or
successful.  (I have managed to transfer many hundreds of megabytes
through smb shares in this net, both between windows systems, and
between windows and Linux/samba systems).

At this point I'm looking for three fairly specific answers (though
more general observations would be welcome; I don't know that I've
provided enough data for any, though).

First, is there a defined procedure for what to do, including how long
to wait, after making a change to some config thing somewhere in the
samba network, before trying to test it, to make sure it's actually
had time to happen and propagate to the places I might be testing
from?  I really wonder if I'm testing what I think I'm testing,
sometimes. 

Second, what logging level is necessary to get browser elections to
show up in the debug log, and what entries should I grep for?

Third, if the Windows 98 client asks me for a password for service
\\gw\$IPC$, roughly, when accessing the samba 2.0.6 box (security =
user), shouldn't it want the password associated with a samba user of
guest?  (Which I have mapped to nobody in smbusers).  I haven't yet
found a password it'll accept.  In smbpasswd, would it be looking for
the unmapped name (guest) or the mapped name (nobody)?  (I'm using
encrypted passwords).
-- 
Photos: http://dd-b.lighthunters.net/ Minicon: http://www.mnstf.org/minicon
Bookworms: http://ouroboros.demesne.com/ SF: http://www.dd-b.net/dd-b 
David Dyer-Bennet / Welcome to the future! / dd-b at dd-b.net