|
Good morning,
Thanks for your reply!
Stephen Rowles wrote:
> >something to be desired). The important part is that my setup didn't
> >work for clients outside the VIP's netmask (the VIP is one of the
> >external routable IP's). If it'd help, I can draw the topology on
paper
> >and scan it in..
>
> I have had the same problem when using an ATM backbone which was
provided
> by a company called newbridge. As you have mentioned VLANS I assume
that you
> might be running into the same problem....
>
> In my case the ATM backbone routing software will not route IP packets
to a
> MAC address
> that has not replied to the ARP request for that IP, and it will not
route
> IP packets claiming to be
> one IP address from multiple different MAC addresses. This breaks
direct
> routing, or any other
> routing which requires one IP address to come from multiple MAC
addresses.
I'm still learning what is what and how this terminology affects me, but
overall what you said makes sense. However, I'm not sure whether it
applies to me. When I said "vlan," I was referring to the terminology
this switch manual uses when breaking the ports up of the switch into
various chunks (or mini-hubs, or whatever it's called) -- be it the same
thing?
In simplified terms, I used two nics in all machines (directors and real
servers alike) to maximize throughput by having each nic going one way
-- I used the switch with two vlans to split into an external and
internal network. In theory, it sounded good, but, unforunatley, it
didn't work, and only the ip's on that network's netmask could actually
see the web site *shrug*
<snip>
> If the problem is the same as above then you have to route back
through the
> director,
and this, of course, then becomes a NAT method instead of a DR, right?
This makes more sense now. But I'd still like to hear an answer to my
first question: is it possible to use DR with only one "real" (i.e.
routable) IP? By what you've written, my guess is "no" because that
would imply the real-servers then have the ability to spoof. Right?
Then the NAT method would be the right choice for my setup. But then
why am I getting all those collisions, as shown by ifconfig (or sporadic
blips of yellow lights on the switch -- which DON'T appear in my now
defunct DR setup)? I don't think I can ignore it, because the
observable performance with NAT is somewhat disappointing (long delays
to access the mock website.. and there's no traffic on my testbed to
conflict etc etc).
Thanks again for your reply,
Jano
|
