From: Greg Schwimer (schwim@xxxxxxxxxxxxxxx)
Date: Sat Apr 08 2000 - 03:17:57 GMT-3
Title: BGP Intelligence??
Jamie
I believe your issue has nothing to do with BGP in any way.
Were these EBGP peers connected directly via serial links? What
encapsulation was being used? What were their IP addresses?
By default, any Cisco router will try to get a network config if it
finds the proper set of circumstances. In your case, I am guessing
that you are using and HDLC connection between the routers in question
and thier EBGP peers. What happens in this case is the router uses a
protocol called SLARP to find out it's peer's address, and determines
it's own address based upon that discovery. I believe in order for
this to work the nodes need to use the first two host IDs of a given
subnet (1 and 2 in many lab scenarios - NOTE: I've seen documentation
stating that it MUST be either 1 or two).
Once SLARP determines the local router's IP address for the serial
interface, it will try to load the normal series of config files,
beginning with network-confg. If this is successful, it will find out
it's host name based on the contents of this file and the IP address
it discovered with SLARP. Once its hostname is discovered, it will
broadcast for a TFTP server. If if finds a TFTP server (either
locally or via "ip helper-address") it queries for a file named
<router's hostname>-confg. Obviously, if it finds it, it loads it up
and thats that. If it doesn't find it, it tries several iterations
and variations on the file name and will ultimately fail with the
serial link up and with an IP address. See Cisco documentation for
further info.
For more info, check this out:
http://www.cisco.com/univercd/cc/td/doc/product/software/ios103/acscg1
03/78621.htm
Hope this helps!
Greg
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
Price, Jamie
Sent: Friday, April 07, 2000 10:00 PM
To: ccielab@groupstudy.com
Subject: BGP Intelligence??
Hi,
I was working on a BGP lab this afternoon - no big deal - but once
I'd finished with that I moved on to playing (or at least intending
to) with dialer profiles, callback, etc. I have 5 x 2600's. In
the BGP lab I was working on, the 2 2600's (routers 3 and 4) that
have BRI interfaces both had EBGP peers. Obviously these were the
ones I needed to work on.
Anyway - without thinking anything of it I just did a "wr er" and
"reload" on the 2 routers with the BRI's in them. When they came
back up however they immediately started looking for network config
files. I hadn't touched the config registers and I was a bit
stumped as to why they would do this all of a sudden. Once they
had both timed out I looked over the config and found that
everything but the IP address on the serial interface that
connected the routers to the EBGP peers had been erased. Those
interfaces were still active too. So I did a "wr er" and "reload"
again.....and again same deal.
Only when I had erased the configs of the EBGP peers (routers 2 and
5) that were still active in the lab and reloaded everything at the
same time did routers 3 and 4 boot normally (when I say normally I
mean the way I have them - to go straight to setup). I'm assuming
the same would have happened if I powered off routers 2 and 5. So,
to investigate further I threw a bunch of configs on all the
routers in the lab that I saved from some OSPF stuff that I'd done
previous. Recreating the scene I erased and reloaded routers 3 and
4 but this time they booted normally.
Am I right in thinking that the router/IOS is intelligent enough to
realize that it is an edge device in a BGP autonomous system even
after an erase and as a last ditch attempt to retain functionality,
considering that routers in this role would be pretty mission
critical, it does what I saw?? That, or the keepalives from the
active router make an effort to get the inactive peer on line. Has
anyone experienced this or actually knows of it? My theory,
without delving too deeply into it yet, is that on reboot the BGP
keepalives were saying to the peer (ex-peer actually) "this is who
you should be" and those keepalives somehow forced the interface to
come up with the intended address. The router then seeing it had
no config did what it should do and went to the network looking for
one. But how does the interface come up when default is for it to
be in a shutdown state? Or are all interfaces actually live on
initial boot and then go into shutdown??
If what I think was happening then I think that's a really cool
feature. I could see the admins of both AS's agreeing to store
configs for each other in case of emergencies such as a router
failure/config loss.
Or have I simply been studying too much and the hermit lifestyle I
have adopted of late is granting me hallucinations and causing my
mind to wander on spectacular tangents?? Maybe I'm simply fried
and need to go to bed.
Jamie
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:23:13 GMT-3