Re: inet_interfaces for outbound smtp traffic

a8505970unet.univie.ac.at
Sun, 28 Nov 1999 23:04:57 +0100 (MEZ)

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
• Next message: Vlad Skvortsov: "Re: relaying rejected"
• Previous message: Jim Seymour: "Re: Newbie question: .forward, luser_relay and postmaster"
• Maybe in reply to: Rask Ingemann Lambertsen: "Newbie question: .forward, luser_relay and postmaster"
• Next in thread: Wietse Venema: "Re: inet_interfaces for outbound smtp traffic"
• Next in thread: Wietse Venema: "Re: inet_interfaces for outbound smtp traffic"
• Reply: Wietse Venema: "Re: inet_interfaces for outbound smtp traffic"

> Did you route the traffic through that virtual interface? Something like
> 'route add -net a.b.c.0/24 dev eth0:x' (or the corresponding command on your
> system) should do that. Didn't test it, but it works with most software I
> know.
>
> Remco

Remco,

this would route *all* outgoing traffic through this interface (if you
have no routing based on destination protocol/port). You would need the
equivalent of routing only traffic for destination protocol/port tcp/25
through this virtual interface. Otherwise I can achieve the same result by
specifying that one virtual IP as the canonical and leave out the routing
table entry.

What I'd like to get right is a "real" virtual smtp server, both incoming
and outgoing smtp traffic using the same xxx.xxx.xxx.xxx IP address which
is a virtual IP interface. "Normal" outgoing traffic should be routed
through the canonical address.

That way outgoing smtp should resolve in reverse DNS to the same as the
machine claims to be when connecting to the IP on port 25 which it
currently does not do: currently if if you connect to the host you get a
"220 virthostname.domain ESMTP Postfix" greeting, which is correct. But
on the outgoing mail smtp claims to be "virthostname.domain" too, which is
not correct since the source IP of this connection resolves to
"canonicalhostname.domain"...

Maybe there is no easy way to get around this.. but something like the -i
parameter for traceroute to specify the source IP just for this one
program looks like what would be a solution. Of course if you have a real
multihomed host (versus aliased only) and no ip_forwarding specified then
things do get tricky in routing-land..

-Michael

• Next message: Vlad Skvortsov: "Re: relaying rejected"
• Previous message: Jim Seymour: "Re: Newbie question: .forward, luser_relay and postmaster"
• Maybe in reply to: Rask Ingemann Lambertsen: "Newbie question: .forward, luser_relay and postmaster"
• Next in thread: Wietse Venema: "Re: inet_interfaces for outbound smtp traffic"
• Next in thread: Wietse Venema: "Re: inet_interfaces for outbound smtp traffic"
• Reply: Wietse Venema: "Re: inet_interfaces for outbound smtp traffic"

This archive was generated by hypermail 2.0b3 on Sun Nov 28 1999 - 16:22:41 CST