From: Fernando Serto (fernando_at_serto.com.br)
Date: Thu 03 Jul 2003 - 06:16:38 BST
> > skywalker is a slackware 9 and fenestra is a redhat 8. both of them with
> > ctx17 patch applied. and the vserver packages, I compiled and installed
> > same source tarball... I tried the rpms, but as I was having problems, I
> > decided to remove them and use the same files... sorry for my stupidity!
> what about the network configuration?
> is this also equivalent or at least compareable?
yes, both have an external interface (eth0) and an internal one (eth1). and
even with the wrong broadcast address, my setup on slackware is running, and
on redhat (with all your instructions to Matthew, ie correct broadcast) it's
> > how would I set this up? isn't it automatic? in my box at home, the
> > broadcast is from eth0 (external), but it's still working...
> look, the linux ip stack is very powerful, the
> _usual_ interface utilities are from 2.0.x ...
> you'll have to use iproute2 to unleash all the
> features ...
> In addition to that complexity, the ctx-17 network
> code seems to do some things wrong (or at least in
> a very strange way ...)
but why only in redhat? I have the same kernel, patch and vserver source
tarball on both...
> if you attach a packet logger (eg tcpdump), you'll see that
> for example a ping to an aliased interface, will result in
> packages with 'wrong' source addresses leaving the 'right'
> interface ...
> as far as I know (didn't investigate yet) this is _not_ an issue
> of the kernel. ping simply uses the first address of an interface
> if no -I <address> is specified ...
> so if the alias interface is configured correctly, you'll
> still need to specify -I 192.168.10.111 for ping to do the
> 'right' thing ...
> from within a vserver this is not required, because the
> virtual server only 'sees' the aliased interface, so ping
> does what you expect ...
if I use the "ping -I 192.168.10.111 192.168.10.142" from the vserver, I can
ping my desktop, but, without the -I (still from the vserver) I can't.