Consider the following small network:
--------- ---------- ---------- ----------
|IRX 114| |pm2e30-1| |pm2e30-2| |pm2e30-3|
--------- ---------- ---------- ----------
|A.1 |A.25 |A.26 |A.8
|-------------------------------------------------------------------|
|A.16
-----------
| Solaris |
-----------
In this example, all ethernet devices are on the same /24 network, shown
here as "A". All ComOS are version 3.7.2 and all portmasters (finally)
have 4 megs of ram, with exception of the IRX-114 which only has 1 meg of
ram (it has either 2 or 300k free on it).
Everything is kindof kludged right now because we primarily used to work
off of one network block, and as the number of virtual hosts we're using
and dialups we're adding is increasing, we've begun to break into another
/24 network block ("B"). We are preparing to put all our assigned address
pools on the B network and subnet the portmasters.
Portmaster 1 and 2 are using address pools in the A network and Portmaster
3 has its assigned address pool set to the first 30 addresses of our B
network. We're planning on subnetting the B network so that the first /27
blocks are for portmasters and the last block will be for static IP's
(note they can login to any given portmaster). For now, we've begun
assinging our static IP's from the last portion of that netork (ie, B.254,
B.253, etc).
I know that eventually, I'll have to go to OSPF, but I'm still
investigating it.
Consider the following scenario: Since portmaster 3 has the first block of
B in its address pool, it has sent a RIP broadcast to the IRX and the
other two portmasters saying that it has the whole B.0 network (I know
this is because RIP doesn't send a mask).
Now, when we dial into portmaster 1, with a static IP of B.253, something
wierd happens. A traceroute from the Solaris box shows it going to the IRX
(it's default route), over to portmaster 3, then to portmaster 1, then to
its destination. Looking at the route tables of the IRX, we still see our
RIP route of B.0 pointing to portmaster 3. Going into portmaster 2, we see
the same route. But looking at portmaster 3, we see a RIP route pointing
B.0 to portmaster 1. It looks as if portmaster 1 selectivly sent a RIP
update to portmaster 3, but not the router and the other portmasters. This
explains how portmaster 3 knew to send it to portmaster 1.
I expected this to break. I expected that portmaster 1 would have sent a
broadcast to all portmasters saying that it owned the entire B.0 network,
but it only sent it to portmaster 3. This way, packets made it to users on
portmaster 3 that were dialed in, but anything else in the B network got
bounced to portmaster 1. I didn't know RIP was selective, or is this a
feature of ComOS?
So what happens when someone dials into portmaster 2 with a B.252 static
IP with the B.1-30 assigned address pool on portmaster-3, and a user with
B.253 on portmaster 1. Will it break then? :-)
Thanks for any insight that can be offered.
-Jason
-
To unsubscribe, email 'majordomo@livingston.com' with
'unsubscribe portmaster-users' in the body of the message.