Best Practice for HSRP Timers

[ayo.monehin]

Can anyone advise what the best practice is for HSRP Timers when you're supporting VoIP on your enterpise network?

The design guys in my organisation have recommended 200 msecs (Hello) and 750 msec (Dead) because of the Voice, but I suspect this is causing issues as we're seeing HSRP state changes on the standby, however the active still remains "active"

My guess is that the standby doesn't receive any Hellos for 750 msecs, then it tries to become active, then oops, it receives the hello timer.

So my question is what are good values for HSRP timers if supporting Voice on your network.

NOTE: We've investigated physical issues and spanning-tree issues and we're sure that's not the cause.

Also, both switches are connected to each other via Gigabit links.

[Ryan Carretta]

200/750msec is really aggressive. I am not surprised that you have some HSRP flapping with these timers, since if the CPU is the least bit taxed we'll either delay or send a hello just a bit late. I think a hello interval of 1 second with a dead interval of 3 seconds should be good. Those timers are what we recommend in the SRND as a starting place for those who want more aggressive failover times. Here's the SRND:

http://www.cisco.com/en/US/solutions/ns340/ns517/ns224/ns304/net_design_guidance0900aecd800e4d2e.pdf

I would think that during a critical network failure 3 seconds of lost voice would be the least of anybody's worries. ;)