I'm setting up three VSS system with OSPF as IGP and MPLS ontop.
In normal operation the setup work as expected with VRF's, traffic is going between test PC's.
But when I try to do a VSS failover, power off on the active switch, I get about 30sec of downtime in the VRF.
First I was thinking of OSPF, that's because I use SVI for L3 interface and not a L3 MEC. In the 'Quick Reference to Best Practices guide for IOS on 6500 it refers to that as a configuration that can cause extra downtime. (Hello/Dead timer)
But trying this out, using OSPF instead of MPLS in testing betweent the end systems, there is minimal impact in a VSS failover. Maybe about a second downtime.
Anyone have experience with MPLS and VSS? Ideas?
Im using nsf and tried mpls ldp graceful-restart and bgp graceful-restart.
I have tested the same type of design with fewer VSS domains. It works fine when the active switch is turned off. We actually tested this with multicast from one VRF to another over an MPLS backbone and the fail over of the VSS pair from active to passive was instant. If possible try to use routed ports instead of switch ports with SVIs. This may cut down on convergance time and you don't have to worry about spanning tree conergance and ports going from blcking to forwording.
I need to use trunks between our systems because of L2.
What's annoying but interesting is the failover with traffic in OSPF is instant, but through the vrf its 30sec downtime.
I just did a failover in the core vss pair and got instant failover. This one is only P, no BGP.
So MP-BGP might be what needs a bit of tuning?