Virtual Connect and HP A-Series switches (A5820) IRF Integration Guide

26
Switch failure
VM1 has a continuous ping to its default GW 192.168.1.1. Under normal conditions, vSwitch
hashes the traffic from this VM to the vmnic3, which is mapped to VC2 and then enters the Bridge-
Aggregation3 interface in the IRF logical switch.
The test issues a reboot command on Switch1 A5820. Switch 2 takes over as the new master and
any interface related to 1/y/z is shut down.
Test Result:
Switch1 down: Ping packet loss did not occur, so the convergence time was less than one second.
Switch1 up: Ping packet loss did not occur, so the convergence time was less than one second.
Note:
The convergence time remained less than one second because the traffic flow did not switch over to
the other path. It still used int b3 because even with switch 1 and all 1/y/z interfaces down, int b3
still had the other interface ten2/0/26 up. So, the convergence time is the result of LACP rehashing
the traffic to the other remaining link, which is typically less than one second.
For this scenario, IRF does not change the traffic flow path, even when losing one switch. The two
uplinks operate at 10G each.
After switch 1 comes back up, it remains the slave to prevent traffic switch-over again, even though it
has higher priority.