As I understand it, the traffic flows at both L2/L3from a VSS node will favour local connectivity to that of its peer and traversing the VSL. This is documented at a high-level in the Whitepaper and software documentation below.
http://www.cisco.com/c/dam/en/us/products/collateral/interfaces-modules/network-modules/white_paper_c11_429338.pdf
http://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst6500/ios/15-0SY/configuration/guide/15_0_sy_swcg/virtual_switching_systems.html#wp1054789
An abridged output of which includes:
Layer 3 Protocol Overview
IPv4
The VSS implements path filtering by storing only local paths (paths that do not traverse the VSL) in the FIB entries. Therefore, IP forwarding performs load sharing among the local paths. If no local paths to a given destination are available, the VSS updates the FIB entry to include remote paths (reachable by traversing the VSL).
So, the scenario is;
- Two DC's with a VSS pair in each. These are inter-connected with L3 OSPF peerings using 10Gbps links from node1 to node1 and node2 to node2.
- Each VSS pair has an access-layer switch connecting over a MEC.
- Multiple servers connected to the access-layer with multiple traffic flows of varying types.
The traffic flowing from a host connected to one access-layer switch to a remote host will be balanced to one of the MEC links and subsequently hit one of the VSS nodes. The VSS node will prefer its local CEF entry and forward the traffic over its L3 OSPF peering to the remote DC. If that link is removed the traffic is auto-magically routed over the VSL and forwarded over the peers L3 OSPF connection. Replacing the previously removed link and post convergence, the traffic flows over the locally connected L3 OSPF link.
However, if the primary VSS node is failed. All of the traffic is loaded to the single MEC uplink on the access-layer switch. The impact of which is that all of the traffic is the routed over a single L3 OSPF link. Upon recovery, the now standby VSS node will not process the traffic flows as the access-layer MEC is still loading the traffic to the current primary VSS node. All of the above makes some logical sense.
What I'm keen to understand is how the "LOCAL-CEF" entries work and what decision process takes place to ensure they are present in both of the Primary and Standby VSS nodes?
thoughts/comments would be welcome
Kind Regards
Ju
