Improving slow box-to-box failover time

[bensmith]

I have a pair of 11506's configured for box-to-box redundancy. Failovers appear to happen as quickly as expected (picked up in around 3 secs), but it takes a further 30-40secs for the 12 scripted keepalives we use to be started (the scripts run and complete in < 1 sec).

Is there a way to improve this fail-over time or should I look as using VIP redundancy or ASR instead?

Thanks,

Ben

[d.parks · Accepted solution]

This is the same behavior I see in box-to-box with scripted and non-scripted keepalives.

VIP/Interface redundancy is definitely the way to go if you want faster failover. In this configuration, the backup CSS is already sending keepalives before the VIP failover so convergence is much faster.

[bensmith]

Many thanks - any idea why this is the case? It seems a shame that I have to double the 'keepalive' load on my server farm to achieve an acceptable fail time...

[d.parks]

My best guess is that this is a spanning tree thing. I currently have CSS's in a bridged environment which would definitely involve spanning tree. Even in a routed configuration, the CSS is still an ethernet switch, so I assume that when an interface comes on-line, it has to go through the ST phases prior to forwarding traffic.

I believe I saw a command at one point that allowed for turning off Spanning Tree. That may or may not be an option depending on how your network is setup.