RE: [nsp] CGMP leave vs. HSRP

From: David Sinn (dsinn@cisco.com)
Date: Tue Dec 19 2000 - 13:07:38 EST


CGMP entries are excluded for the multicast mac range that is used by HSRP
(and in effect other multicast streams that use the same MAC address). Thus
it is supposed to flood to every port, not just the router ports.

You can confirm this with the "sho multicast group". If you see a entry
here then it means that CGMP has populated the CAM table with a MAC address
that will only go to the string of ports listed. If you have no entry in
the CAM table, then the Cat will do it's default action when there isn't a
entry (uni- or multi-cast) and flood.

Now, I'm not 100% sure, but I believe that certain versions of CatOS/IOS
didn't operate this way, and would generate CGMP entries for the multi-cast
address that is used by HSRP, and thus when the client left that was using
that multicast address the router would prune, and HSRP would break.

David

-----Original Message-----
From: Dana McIntosh [mailto:dana@nimbus.skycache.com]
Sent: Monday, December 18, 2000 2:38 PM
To: cisco-nsp@puck.nether.net
Subject: [nsp] CGMP leave vs. HSRP

Greetings,
Ive recently run into the problem of flooding caused by using both cgmp
leave-processing and HSRP on the same switched segment.. However, I dont
quite understand why this happens, and the cisco documentation is
especially vague on this. The best explination I can find is at:
http://www.cisco.com/warp/public/cc/pd/iosw/prodlit/cgmp_an.htm
but it still leaves me with a couple questions.. The HSRP stand-by address
uses a MAC that is not the router multicast mac... how does this
interruption of router to router HSRP packets by CGMP affect the
reachability of the standby address by the remote host? Can anyone flesh
out my knowledge of this "feature"?

TIA,
Dana McIntosh Network Engineer
Cidera, Inc. (301)598.0500 ext 2017



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:23 EDT