lan0 lan1 and lan2 problrm

[j773303]

I meet a networking problem and confused me for a long time. I have 2 machine, one is L3000 with HPUX 11.00 , antoher is RP7410 with HPUX 11.22; Both machine have 3 lan card, lan0 / lan1 / lan2.

Below are the configuration of rp7410 (HPUX11.22) /etc/rc.config.d/net.conf

INTERFACE_NAME[0]=lan0
IP_ADDRESS[0]=10.28.66.7
SUBNET_MASK[0]=255.255.255.128
BROADCAST_ADDRESS[0]="10.28.66.127"
INTERFACE_STATE[0]=""
DHCP_ENABLE[0]=0

INTERFACE_NAME[1]=lan1
IP_ADDRESS[1]=10.28.66.207
SUBNET_MASK[1]=255.255.255.128
BROADCAST_ADDRESS[1]="10.28.66.255"
INTERFACE_STATE[1]=""
DHCP_ENABLE[1]=0

INTERFACE_NAME[2]=lan2
IP_ADDRESS[2]=10.28.67.7
SUBNET_MASK[2]=255.255.255.0
BROADCAST_ADDRESS[2]="10.28.67.255"
INTERFACE_STATE[2]=""
DHCP_ENABLE[2]=0

ROUTE_DESTINATION[0]="default"
ROUTE_MASK[0]="255.255.255.128"
ROUTE_GATEWAY[0]="10.28.66.126"
ROUTE_COUNT[0]="1"
ROUTE_ARGS[0]=""

ROUTE_DESTINATION[1]="default"
ROUTE_MASK[1]="255.255.255.128"
ROUTE_GATEWAY[1]="10.28.66.254"
ROUTE_COUNT[1]="1"
ROUTE_ARGS[1]=""

When I using "traceroute -i lan0 ip-address" and "traceroute -i lan1 ip-address", it is ok.
But failed in "traceroute -i lan2 ip-address", that's means it is can't find the router. Why lan2 got this proglem. However, in the L3000 (hpux 11.00), the traceroute are ok on the 3 lan. Does hpux 11.22 need to configue for lan2?

[j773303]

NOTE: I've 5 rp7410 server was installed on hpux 11.22. They also has 3 lan, and lan2 can't traceroute from router. Is it possible 11.22 need to configue something?

[Hoefnix]

Add another part in your /etc/rc.config.d/netconf
For example like the next part:

ROUTE_DESTINATION[2]="default"
ROUTE_MASK[2]="255.255.255.0"
ROUTE_GATEWAY[2]="10.28.67.254"
ROUTE_COUNT[2]="1"
ROUTE_ARGS[2]=""

[Hoefnix]

You have to activate the netconf file using:
/sbin/init.d/net start

[Steven E. Protter]

What you need to remember is unless you have auto port agregation you can not bring up two machines on the same network.

The network portion of the IP address is defined by how you've done subnetting. This case is a good example for the HP Certification test. Its early here, but I'll give it a hack.

With a standard netmask 255.255.255.0 you have 253 possible addresses the lowest is the network, the highest is the broadcast address but those two are not usable.

You have divided that address space into to networks with your netmask. You network portion of your address extends into the forth octet, the last part of the network address.

There are two networks available on your machine but three network cards. Because of the way HP-UX networking is done, one of your three cards will not come up properly.

Network1 address: 10.28.66.0
Network1 broadcast: 10.28.66.127

Network2 address: 10.28.66.128
Network2 broadcast: 10.28.66.255

There is no room in this configuration for three active network cards.

SEP

[Thayanidhi]

Hi SEP,

But his third lan card (lan2) is using diffent network 10.28.67.x which doesn't clash with either lan0/lan1.
lan0/lan1 uses 10.28.66.x
Will it not work this way?

TT

[Ron Kinner · Accepted solution]

I would also check the subnet mask on the routers to make sure they are correct. It really should have failed for lan1 too but the router could have a route back that fixed the problem or it could have decided to use the second default route. What you are doing when you say traceroute -i lan2 is telling it to use the IP address of lan2 as the source address for the traceroute packets. You are not telling it to go out on lan2. The routing table controls which interface it uses based on what gateway it thinks will get you to the destination. (I think the -g option may force it to use a particular gateway but I'm not sure.) Since this fails then what it is telling you is that the gateway does not know how to get back to the source address (that of lan2). The gateway is probably the default one so I would check it to see if it has a route pointing back to 10.28.67.0 via the 10.28.66.7. If you add the route to the router then it will probably work (at least to the first step whether it works any further depends on whether the other end knows how to get back to that subnet). The reason that the -i lan1 works is that the gateway it uses knows how to get back to the ip address of lan1 or it may have the wrong mask and not realize that they are two different subnets or if an early cisco it may not have ip classless turned on. Without seeing your full netconf, netstat -rn and the results of your traceroute I can't really tell you.

Unless you have turned on the strong end-system model in ndd (and it could be on by default in 11.22 for all I know) it is wrong to have two default routes with the same metric. The default route is to be used when it knows no other route. How is it supposed to know which to use if you give it two? (with the strong end-system model it assumes that traffic coming in on one interface will be replied to using the same interface in which case it makes sense to use a default route for each one.) The answer is it just uses the first one it comes to or it flops between them or it may have some other way you just don't know and can't rely on it. The normal rule for the default route is point the default to the router which knows how to get to the internet (or the most subnets). In order to use the other router you then add routes pointing to specific networks using the second router as a gateway. Note that the numbers in the []'s in the route statements do NOT correspond or magically link to the numbers in the []'s for the interfaces. They are just used to group the five ROUTE_... statements together.

I'm not sure about 11.22 but in 11.0 there are two other possible problems. First ip check subnet is turned on and will not allow you to use a 128 mask. This enforces an obsolete RFC so they should have stopped making it the default but who knows. Check it in ndd to see.

ndd -get /dev/ip ip_check_subnet_addr
if it comes back with 1 you need to change it.

The other problem is the dead gateway detection. If it can't ping the router it will not use it. This is turned off in ndd by setting ip_ire_gw_probe to 0.

ndd -get /dev/ip ip_ire_gw_probe

will tell you what it is set to.

You change both the same way:
ndd -set /dev/ip ip_ire_gw_probe 0
ndd -set /dev/ip ip_check_subnet_addr 0

Then edit /etc/rc.config.d/nddconf to make sure it will not go away after a reboot.

You may need to remove and re-add the route after making the change in order to make it use it again.

Ron

[rick jones]

First things first - how is it possible that you are running HP-UX 11.22, which is an IPF-only release, on a PA-RISC system (iirc, the rp7410 is PA-RISC yes?)

Second - if you want to have more than one default route, you need to be setting ip_strong_es_model to a value of one (1). That will bring source IP addresses into the routing lookup. Otherwise, only one of the default routes will be used. It is also necessary (or you have to do a lot of static route table additions) if you want to configure more than one interface ("physical" as far as the Transport is concerned) in the same IP subnet. ftp://ftp.cup.hp.com/dist/networking/briefs/annotated_ndd.txt


Third - the traceroute in HP-UX tries entirely too hard to be clever and guess interfaces itself. If the other devices on each network will respond to the packets, you might verify connectivity with linkloop(1m). Not all non-HP-UX systems will respond to a linkloop(1m) frame.

[John Dvorchak]

Am I missing something here or do you NOT have any gateway (router) defined for the network on lan2. You have lan2 set with a class C netmask defining the network as 10.28.67.0 but there is no router definition stanza for that network. If the default gateway on the 10.28.67 network is say 10.28.67.1 then you need a stanza with that information:

ROUTE_DESTINATION[2]="default"
ROUTE_MASK[2]="255.255.255.0"
ROUTE_GATEWAY[2]="10.28.67.1"
ROUTE_COUNT[2]="1"
ROUTE_ARGS[2]=""