A routing loop prevention scenario
Posted on 2006-11-15
Hi, we have got around this problem now, but I wanted to ask this question for future reference.
We have a cab in an ISP's co-lo and they have provided us with 2 ethernet cables from their core Internet network which we originally wanted to plug in to 2 Cisco Catalyst 3560 switches that we have in the cab.
The switches are clustered, and run an internal VLAN (default), and Vlans named "Internet 1" and "Internet 2". The default VLAN spanned the cluster whilst the two Internet Vlans were tied to one or other of the switches. This was to prevent a routing loop when plugging the feeds in to the switches.
Whilst the ISP agreed that this config would work, they basically told us that they wouldn't allow us to plug their core in to our switches because - potentially - we could reconfigure the port assignment and create a routing loop (which we wouldn't do but hey ho....i get their point).
So anyhow - what they suggested we do is basically issue the "no switchport" command on both of the ports used for the Internet subnet feeds (one on each of our switches) however I think I am right in saying that this would have basically rendered the port useless and we couldn't then switch the traffic to the other 3 ports on the same switch that were configured for the Internet 1 (or 2) Vlan.....I hope I'm not losing anyone.
So - basically - is there a way to make this work. I'm good with IP routing but no good with switching (yet) and so hence my question.
Oh - yeah - the underlying reason for needing this config was that we wanted to create a "no single point of failure" route in to a number of servers running off of the Switch Cluster (dual NICs that are teamed) and were using 2 2811 routers with HSRP and a 2 PIXs with failover to do this.
Thanks in advance.