On Wed, Mar 12, 2008 at 11:31:22PM -0400, Jason Berenson wrote: > We have 3 7206's used as edge routers. PA-MC-T3 in from our DAX and > ethernet out to our transport. So there are a few adjacencies along > with iBGP and eBGP. It seems like the router that goes down (flaps > OSPF/BGP instance 1) is the only one that takes a hit out of all of them > connected to the switches.
I've seen one scenario where one router claims it can't hear any OSPF neighbors on an Ethernet, even though the other neighbors issue no complaints. In that case, the router was no longer receiving Ethernet packets, apparently due to an IOS bug. The router was still sending its HELO packets, and the other routers heard them, so they had nothing to complain about. I've seen HSRP (another broadcast-based protocol) get into a similar stuck state. However, you should not see behavior like that with BGP, which uses a TCP session. Both sides should issue alarms when the session flaps, though one side might take longer to report the event (i.e. it may not sense the event until a timeout occurs). > I'm leaning away from thinking this is an ethernet issue, but I > definitely could be wrong. Based on what you said above, I would say your problem router's Ethernet transmitting is OK (at least for broadcast packets). If you've got Netflow on the box, that would give you some hints as to how receiving is working. And of course, you need to check the switch port statuses. This is where something like Cricket/MRTG/Cacti comes in handy - if they show you that you've had zero errors on those ports going back the last 3 months, then you know the physical medium is probably OK. _______________________________________________ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/