It doesn't really have anything to do with etherchannel, that works just fine by itself. It has to do with the CEF load balancing algorithm being exactly the same as the etherchannel one.
This even propagates through to multiple switches, for instance I have tested it like: 3750 with 2 etherchannels, to two 2960 with an etherchannel link in between the 2960's CEF load balance to router on a single port on 2960 #2 (two /30's, two vlans) What happens is 3750 sends out both port channel interfaces to each 2960, and on port channel 1 it uses the 2nd port only for the traffic, and on port channel 2 it only uses the 1st port. Which means each 2960 only receives packets on 1 of the etherchannels effectively making it 1gbps instead of two. Further, this propagates to the link between the 2960's because the traffic is only going out 2960 #2 so what I see on the 2gbps port channel link between the 2960 is also that is only using 1 port of the etherchannel. If i take this a step further and put a router in between 3750 to router 1 to 2960 3750 to router 2 to 2960 for example it STILL has the same effect because it's only based on the source/destination ip or source or dst ip which the CEF load balancing is creating. A simple change to the etherchannel load balancing algorithm or the CEF one would clearly fix the problem. _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
