Why Is My 5406zl Using Different Gateways for the Same Subnet?

[iglablues]

Hi all,

I have two networks: one is running HP, and uses derivatives of 10.100.x.x/16 and is pretty much one flat network, and the other runs Extreme and uses derivatives of 10.1.x.x/24, with a bunch of vlans. The Extreme network has no outside connection so it needs to go through the HP core, a 5406zl in order to get to the WAN and to access some resources on the HP network like DNS, file server, etc. To faciliate this I did the following: 

-created a vlan just for interconnecting the two networks. On the HP side that part of the config looks like this:

vlan 12
name "crossconnect"
ip address 10.1.1.1 255.255.255.0
tagged D1

-edited the routing table on the HP to point traffic destined for 10.1.0.0/16 out the interconnect link. My routing table now looks like this:

Destination            Gateway       VLAN      Type      Sub-Type      Metric    Dist.
------------------ ---------------   ----      ---------   ----------    ----------   -----
0.0.0.0/0            10.100.1.2    1             static                                1            1
10.1.0.0/16         10.1.1.2       12           static                                 1           1
10.1.1.0/24      interconnect        12        connected                        1           0

The Extreme switch on the other side has an IP of 10.1.1.2 and is in the same vlan, tagged. As you can see the second entry is the static one I created with the Extreme IP as the gateway, and the third entry was automatically created because routing is turned on. I can ping between the two with no problem. The Extreme also has several other networks/vlans behind it, differentiated by the 3rd octet. 

So here's what happens:

-if I ping the interface for another vlan on the Extreme network, like 10.1.3.1, I get a response. The HP appears to send the packet to that vlan interface directly, i.e. it sees it as the next hop for that destination. Here's the traceroute:

traceroute 10.1.3.1
traceroute to 10.1.3.1 ,
1 hop min, 30 hops max, 5 sec. timeout, 3 probes
1 10.1.3.1 3 ms 1 ms 1 ms

-if I try and ping a host behind the Extreme, i.e. an actual host on the vlan and not the vlan interface itself, it gets routed to the IP of the interconnect vlan instead. 

traceroute 10.1.3.33
traceroute to 10.1.3.33 ,
1 hop min, 30 hops max, 5 sec. timeout, 3 probes
1 10.1.1.2 2 ms 1 ms 1 ms
2 * * *

Why, if I have a route entry that says send all traffic destined for 10.1.0.0/16 to the gateway of 10.1.1.2, would traffic destined for the same vlan/subnet get sent to two different gateways? I understand I have a whole different problem of why 10.1.1.2 isn't forwarding that traffic on to the vlan, but for example, if somehow HP is putting traffic for vlan x right into the appropriate subinterface on the Extreme switch (as it appears to be doing when pinging 10.1.3.1, that would be fine. I just don't know how/why it's doing that and how to control it, i.e. get it to send traffic for 10.1.3.33 the same way. 

[paulgear · Accepted solution]

Hi iglablues,

What you are seeing is the Extreme switch responding to pings to its own address according to the "weak host model".  This means that it responds on any interface to requests for any of its interfaces.  Linux also uses the weak host model, as do Windows XP & 2003.  BSD uses the strong host model.  See http://en.wikipedia.org/wiki/Host_model and the references there for more info.

[iglablues]

Thanks!