route add default does not take for second vlan

[Joy Grammatica]

I have two front end networks and one backend. The backend of course works for all as it is one large broadcast domain.

The second front end is pingable but routes do not take. I remember a few years ago something about an hp-ux bug (ok it wasnt a bug but some arguement about rfc conformaty) that prevented this scenerio from working but that there was a work around.


Web/App network: 10.81.46.1/26
Web server1: 10.81.46.5 (Works)
Web server2: 10.81.46.11 (Works)
App server1: 10.81.46.13 (Works)


DB Network: 10.81.46.64/26
DB Server1: 10.81.46.69 (Front end does not work)

The web/app servers work just fine as their gateway is 10.81.46.1 and hp-ux likes that.

I can ping 10.81.46.64 all day long just fine, but can not add that as my default route.

I have tried using different metrics, netmasks (though only 255.255.255.192 should work), different incantations of the route add command and no luck.

I can ping the gateways just fine. In both vlans the gateway is pingable from the servers within the respective vlan.

Destination Gateway Flags Refs Interface Pmtu
127.0.0.1 127.0.0.1 UH 0 lo0 4136
10.81.46.69 10.81.46.69 UH 0 lan0 4136
192.168.0.10 192.168.0.10 UH 0 lan1 4136
10.80.16.252 10.80.16.252 UH 0 lan5 4136
10.81.46.64 10.81.46.69 U 2 lan0 1500
10.80.16.0 10.80.16.252 U 2 lan5 1500
127.0.0.0 127.0.0.1 U 0 lo0 0

Any thoughts?

This is HP-UX 11.11 with latest patch bundle applied.

[Sridhar Bhaskarla]

Hi,

The bug or the compliance was that first and last subnets as a result of subnetting were being considered as "network" and "broadcast" addresses by old routing protocols. So, they wouldn't work if you used those IP addresses for hosts.

You can turn off checking it by setting the ndd parameter 'ip_strong_es_model' to 0

ndd -set /dev/ip ip_strong_es_model 0

You may also want to consider setting 'ip_check_subnet_addr' to 0.

-Sri

[Ron Kinner]

I think you have made a mistake in your IP plan or perhaps in your thinking. You have a network DB Network: 10.81.46.64/26. You can not use 10.81.46.64 (nor 10.81.46.127) as a host address since that is the network address (broadcast address) and it is no wonder that the route won't take. In many older unix systems the network address is also responded to as if it were a broadcast so pinging it may work work even if there is no host on that address.

The default route should point to a router in the range 10.81.46.65 to 20.82.46.126 (and use a metric of 1). The router should know how to get to other subnets. Is there such a router? Is there really a router at 10.81.46.64?

Assuming there really is a router at 10.81.46.64 it cannot have a mask of 255.255.255.192 (/26). This is what happens if I try to assign that ip address to a cisco router:

(config-if)#ip add 10.81.46.64 255.255.255.192
Bad mask 255.255.255.192 for address 10.81.46.64.

Ron