John Haxby wrote:
Steven Moix wrote:
I'm surprised that it works at all -- except that I know that the
routing stuff occasionally seems sentient :-) If you put eth0 and
eth1 on separate networks, e.g. 192.168.1.100/24 and
192.168.2.100/24 (and change other machines correspondingly) then
it should all work properly as their won't be any routing ambiguities.
I was surprised too, but it works...the problem is that it needs to
be on the same network :/
Ok, I investigated a bit further...in fact all the traffic for eth0
is redirected to eth1! Where can I change this?
I tried this by pinging with a large packet size. I can ping
192.168.1.100 (eth1) and 192.168.1.200 (eth0) and the traffic on eth0
isn't increasing, but the traffic on eth1 is the sum of the 2
interfaces :-/
Well, it's behaving as advertised then. Given a choice of two
equally good routes to use, the routing algorithm is using one of
them. You don't get much influence over which one -- at least I
never do :-)
You say you need them to be on the same network -- is this that you
need them to be on the same physical network or the same logical
network. If you need them to be on the same physical network then
you can quite cheerfully run multiple logical networks on one physical
network (I do it all the time). If you need a single logical network
on two physical networks then I think you're out of luck -- you can
use two NICs and two networks to increase bandwidth, but even then I
think you have two distinct logical networks as well as two distinct
physical networks. Must go and read TCP/IP books though ...
jch
They need to be on the same logical network...I found that it is a
routing problem..I'm playing with it.
--
Shrike-list mailing list
Shrike-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/shrike-list
