Fail-over/Fail-back

In combination with the Spanning Tree Protocol (STP) enabled on the physical bridges, MidoNet VABs are able to provide fail-over capabilities by forwarding Bridge Protocol Data Unit (BPDU) frames across their trunk ports.

Assuming that both physical switches belong to the same bridged network, as a result of the STP, both devices detect a loop through MidoNet’s VAB and one switch chooses to block its trunk. For example, let’s assume the left switch blocks. The VAB only sees ingress traffic from the right trunk, and thus associates all source MAC addresses seen in those frames to the right trunk.

A variety of events, including failures in the network, may result in the switches deciding to invert the state of the trunks. An example could be MidoNet losing connection to the left switch, and thus stop forwarding BPDUs to/from the right bridge and undoing the loop.

In such a fail-over scenario, traffic would start flowing from the other switch. With this change, MidoNet now detects ingress traffic on a new port, and thus updates its internal MAC-port associations. If the former state of the topology is restored (that is, MidoNet recovers connectivity to the left switch), MidoNet will again react and update its MAC-port associations.

The fail-over/fail-back times depend on the STP configuration on the switches, mainly the "forward delay," and the nature of the traffic. With standard values, and continuous traffic ingressing from the trunks, fail-over and fail-back cycles should be completed in 50 seconds, plus MAC learning time.

Questions? Discuss on Mailing Lists or Chat.
Found an error? Report a bug.


loading table of contents...