Strange ARP behavior on 58xx Switches using multiple VPN-instances

[Joesi]

Hello all,

installed a 4-switch IRF using 3 VPN-instances, which works quite fine. Except one issue, i have no idea how to fix.

In addition within this switch there is Microsoft NLB cluster running.This cluster will be used from all of the 3 vpn-instances. This normally works.

But whenever there is a the move from one clustermember  to another, the the cluster is only reachable within the instance, it is installed.

Requests from the other instances will fail as long as the arp cache is cleared. 

The latest FW was installed yesterday (5.20-R1211-P01) but the effect stayed the same.

As a workaround i tried to install a job doing the "reset arp" every 10 minutes. Actually not sure, if it works because the reset has to be done in "user-view" and i have no idea how to test this job.

i know, there are some questions within one post, but maybe someone out there had a similar problem in the past.

any hint will be highly appreciated!!

br

Wolfgang 

[manuel.bitzi]

Test the arp entries in every vpn-instance (especially the egress-port). If it is different from the arp in the vpn-instance where the cluster is, then is a bug.

I had the same on 7500. R&D fixed it.

br

Manuel