Hello:
I've supported F5 BIG-IP LTM appliances for a while. After a few incidents in a production network, I've started to think through different failure scenarios and want to ascertain exactly what is supposed to happen under certain situations. I have a lab using the VMs so I can't test scenarios involving the failover cable. If anyone knows what happens under these scenarios, please comment.
Topology 1: * lb1 and lb2 have a single physical interface connected for local traffic management * they use a hardwire failover cable and no network failover
Scenario: The active LB loses its network connectivity
Result: The active LB will not failover b/c it will continue to send heartbeats over the hardware failover cable
Topology 2: * lb1 and lb2 have a single physical interface connected for local traffic management * they use a hardwire failover cable AND network failover using a dedicated, directly-connected VLAN
Scenario: The active LB loses its network connectivity
Result: I don't know. Does anyone know whether the results are identical to topology 1 b/c of the hardwired failover cable?
Topology 3: * lb1 and lb2 have a single physical interface connected for local traffic management * they use network failover using a dedicated, directly-connected VLAN AND the mgmt IP as the secondary interface
Scenario: The active lb loses its network connectivity for local traffic management
Result: I will lab this and report back. Does anyone know whether the active LB will make itself go on standby via network failover after it recognizes it's lost connectivity to the network via its local traffic interface?
The moral of the story seems to be that deploying an F5 active/standby HA pair without port channels (trunks in F5 parlance) is equivalent to having a single point of failure in most failure scenarios. If one doesn't have the ability to create a port-channel, the next best scenario is probably independent interfaces, one of which is blocked by STP until it's needed.
No comments:
Post a Comment