4500 VSS Issue/Question

[Ryan Curry]

So I've run into an issue where I have two 4500 VSS pairs attached to the same shared network segment (a MPLS/Metro Ethernet mesh).  Aside from these two devices, I have 8 other routers also on that segment which don't seem to be affected by this anomoly.  When I bring up the interfaces connected to that shared segment on both VSS switches, both sites start to drop packets and connectivity is horrible after the OSPF adjacency completes; when I shut down the link to the shared segment on either of the VSS switches everything goes back to normal.  I've also tested puting an internediary switch between the shared segment and one of the VSS switches and everything works as expected in that scenario.

What I realized is this: I have both of these VSS switches on virtual domain 1 and was wondering if this could be the cause of the issue.  Part of me thinks it shouldn't since the link to the shared segment is a L3 interface and, IMO, the VSS domain shouldn't play a part in the OSPF process (the other part of me says I shouldn't have two domain 1s on the segment).

My other thought is that I'm running into a bug and may need to update the code on one or both of the 4500 pairs.

Any insight into this matter is much appreciated!

[Jon Marshall]

Ryan

How does the topology look ? Are you using etherchannels on the VSS switches to connect to the shared segment. If so then see this section from the 6500 VSS Campus design guide where it states that connecting VSS pairs together requires a separate domain ID -

http://www.cisco.com/en/US/docs/solutions/Enterprise/Campus/VSS30dg/VSS-dg_ch2.html#wp1055911

If this applies to your topology i would look to change the domain ID before upgrading the IOS.

Note - i appreciate it is the 6500 VSS guide but i suspect the same applies to the 4500.

Jon

[Ryan Curry]

Thanks for the reply Jon, these two VSS switches are attached to the PE switches via a single physical L3 copper interface so there is no MEC in that mix.  I'll check out the design guide and see if something jumps out at me.  I'm currently going through my inventory to see if I can find a few 4500s to conduct some testing with as well.

[Jon Marshall]

Ryan

these two VSS switches are attached to the PE switches via a single physical L3 copper interface so there is no MEC in that mix.

Yes that was what i was wondering. There may well be other protocols  that Cisco haven't mentioned in the design guide though, but they do seem to be mainly concerned with MEC.

Jon

[Jon Marshall · Accepted solution]

Ryan

It may be a problem with the virtual mac address allocation. Have you got -

"mac-address use-virtual"

in your config. If you have the domain ID is used to generate unique mac addresses per VSS pair. If you do, see this link for workarounds -

http://www.labminutes.com/blog/2012/12/cisco-6500-vss-domain-id-and-virtual-mac-address

Jon

[Ryan Curry]

Jon, you are a God among men - this is exacly my issue.  I validated by looking at the arp table from neighboring devices and sure enough they have the same address.  This totally explains the symptoms I was seeing.  I'm going to go into the non-critical site and change the domain number.

I GREATLY appreciate your input since I've been working this issue for the past two weeks.

[Jon Marshall]

Ryan

I think changing the domain ID is the right way to go because you never know what other problems it could cause down the line.

Glad to have helped.

Jon