2 messages in net.nether.puck.cisco-nsp[c-nsp] 3750 CEF and equal cost multi...
FromSent OnAttachments
Matt BazanJan 6, 2005 12:03 pm 
Oliver Boehmer (oboehmer)Jan 6, 2005 12:20 pm 
Actions with this message:
Paste this link in email or IM:
Paste this link in email or IM:
Atom feed for this thread
Paste this URL into your reader:
Subject:[c-nsp] 3750 CEF and equal cost multihomingActions...
From:Matt Bazan (Mba@onelegal.com)
Date:Jan 6, 2005 12:03:28 pm
List:net.nether.puck.cisco-nsp

Have a mystery I'd like to solve: On one of my 3750 switches I've setup two static routes with equal metrics (on two separate routed ports) via their respected interfaces (gi1/0/19, gi1/0/21) to a /27 network. So, I've got two routes in my routing table that look like this:

XX.XXX.XXX.XXX/27 is directly connected, GigabitEthernet1/0/19 is directly connected, GigabitEthernet1/0/21

Problem is, when I try to ping one of the machines on this block half of the pings timeout. From looking at a sniff of the traffic I'm seeing that both interfaces will arp for the address in question and when the reply comes back the router seems to hang for a couple seconds (corresponding to the ping timeout). Finally a couple ping requests will go through. At some point in this process the arp entry will be linked to one of the interfaces. What appears to happen, however, is that the router, seeing it has two routes for this block and that one of the ints does not have an arp mapping for this address, will proceed to arp down the interface that doesnt have the mapping. This again causes the router to hang.

In looking at the FIB I see that there are two entries, one for the interface routes and one for the host address. Is this causing a problem? I know this is not the preferred way of going about this but I'd like to get this working, if possible.

Is it possible to have the router only use the interface that has the arp entry mapped to it?

Thanks for the help.

Matt