THUFIR HAWAT wrote:
On 7/6/05, Lai Zit Seng <lzs@xxxxxxxxx> wrote:
How are you selecting to ping via eth0 or eth1? It seems to be that in
both cases your ping would have chosen eth0. Yet the 2nd case fails.
Are you selectively pulling out physically the network cable to "select"
the network interface to use? This is not going to work!
...
if you understand me and I understand you, then yes ;)
when the cable is plugged into eth0, and the network script on arrakis
is restarted, then arrakis can ping caladan.
when the cable is plugged into eth1 and the same procedure is followed
then the ping fails. if the cable is then plugged into eth0 and the
network script restarted then the ping succeeds again.
what's wrong with the "experiment"? it seems to me that there's a
problem with eth1.
the immediate goal is to ping from arrakis eth1 to caladan.
Nothing's wrong with the experiment. You just can't do it like this :)
Ok, I'll cut short my reply since many others have followed-up.
It looks to me like you're trying to have "redundant" network
connections for, perhaps, high-availability reasons. I.e. you still have
network access should one link fail.
Also as others have mentioned, you'll need to bind the interfaces
together, and needs similar support on the switch side. E.g. this is
called EtherChannel in Cisco switches.
An alternative is to setup bridging on your systems. This is even better
actually since now each system can connect to two separate switches, all
hooked up in a redundant configuration, and running spanning tree
protocol to eliminate forwarding loops. Then your IP interface itself
will be on the bridge device.
Yes all these start to get more complicated :) But there should be
HOW-TOs you can google for :)
Regards,
.lzs
--
http://zitseng.com/
thanks,
Thufir
