Having issues with connecting a CE STP network redundantly to our FabricPath domain.
we have 2 N7K as spine and 4 N5K as leaf nodes.
all leaf nodes are connected to both spine switches with fabricpath ports.
the 4 leafe nodes are devided in 2 vpc pairs.
all N5K's have the same spanning-tree priority for all Fabric path VLANS.
spanning-tree ID is not configured, that is default configuration. (not sure if this is required)
connecting an existing stp CE network redundantly to each of the vpc pairs creates bridging loop.
this stp CE network runs PVST N5K's run RPVST.
When connected only to the second pair of VPC N5K's, the CE switch does not update stp with that pair as root, even though it has lower pri number that any of the CE switches and it is receiving bpdu's on the port-channel connected to the N5K's.
When connected only to the first pair of VPC N5K's, the CE switch updates stp with that pair as root.
This should work, as the config is identical om both vpc pairs.
Any clues on what could be wrong?
Is there a config guide or whitepaper that describes this particular setup? seems like it should be a common setup with location redundancy for datacenters and existing CE networks.