I have 2 CE6810 stacks and a CE5800 stack, as core switches in three different buildings on one site. Each stack consists of 2 units.
I don't want to stack all 6 devices because they're in different buildings - if a stack member (building) is isolated and rejoins, it tends to reboot (which means downtime for that building).
One of the CE6810 stacks is currently acting as a site router, using static routing (vlanif 's in each vlan configured for each active subnet on that vlan).
The stacks are connected together using 2 * 10Gb links, LACP, with spanning tree enabled across them.
In order to create a mesh connection between the buildings and make better used of the interbuilding links (LACP is rather inefficient and spanning tree would shut a redundant path down completely), I want to replace STP+LACP with TRILL.
The edge switches within each building are likely to continue running STP in isolation but keeping them within one location helps reduce the size and convergence time of trees ( It is worth noting that if edge equipment such as servers are connected to a STP switch via LACP and that equipment reboots, a complete spanning tree rebuild is triggered, even if the edge equipment is not participating in the spanning tree.)
Here is my problem:
After replacing one of the LACP/STP links with a TRILL one, routed L3 traffic is not going across the TRILL link.
Anything staying within the same subnet passes across the TRILL link without any problems. It's only routed traffic which breaks.
Ideally, L3 routing should be distributed across all 3 cloudengine stacks - as described in https://tools.ietf.org/html/draft-hao-trill-irb-04 - but at this point I'm unsure how to proceed. (VRRP isn't a distributed routing solution. one member of the group ends up as a designated router)
Am I forced to have a separate router, or can the cloudengines run TRILL and L3 routing simultaneously and if so, how do I achieve this?
Thanks