[cobalt-users] tail of woe: eth0 as ethI, ipchains, and dhcpd

[Josh Kuperman <josh@xxxxxxxxxxxxxxxxxx>]

I tried to get my server into production this weekend. I'm pretty
annoyed about what really made it impossible.

I realized some parts were going to be time consuming and awkward. I
would either have to forward e-mail from some old accounts or create
accounts for all the users first. It would be nice if I knew of a
utility where I could simply clone accounts from another redhat
server.

I do think that it should be stated in bold letters that the SSH
Package at package master has tcp wrapper support compiled in. And it
was very annoying after I got a vt510 hooked up to the serial port to
find that in addition to not supporting Mozilla with the web, the RaQ
interface won't support the provided Lynx broswer either. [ I suppose
some compainies just want everyone to use Microsoft Product!]

Has anyone configured DHCPD on a RaQ XTR or another system with two
ethernet interfaces? Where do you change the script so that DHCP is
only done for one interface?

There seems to be an IPCHAINS script that initiates a lot of
accounting chains. Where should I be adding my own. Is there something
to edit -- there's a warning that /etc/ipchains.conf is created by
log_traffic. I couldn't find anything on where to put my chains. Can I
just run ipchains-restore at the end of /etc/rc.d/rc.local? If I can
I'd like to do a lot of this before next weekend. Is there an easy was
to port forward mail, dns requests, etc over to the old server and
then bring each service up one at a time on the the new one. [I stuck
an old DSL/Cable router on my connection to deal with some problems
and I don't think it is working well enough, i.e. it can't handle the
traffic.]

My major problem was the labeling of the ethernet ports on the RaQ
itself. ON the web page they are labled eth0 and eth1 for setup and on
the front of the RaQ they are labelled 0 and 1. Sadly the connection
in the back are labeled 1 for 0 and II for eth1, I took the II as a
zero -- the print is small -- and thus was never connecting the inside
and outside to the right network.


-- 
Josh Kuperman                       
josh@xxxxxxxxxxxxxxxxxx