Forum Discussion

Ossar_178453's avatar
Ossar_178453
Icon for Altostratus rankAltostratus
Dec 15, 2015

Active-active and RHI (BGP) failover

Hello,

I successfully managed to set up the functionality I am looking for but I am lacking the speedy failover that is required.

Two F5 LTMs (VE edition), in an active-active configuration, i.e. two traffic groups. One primary on each LTM.

Each LTM is connected with BGP to separate routers. I am running eBGP LTM<->router and iBGP router<->router. Each LTM communicates with a its respective router over a link net (LTM endpoint as self IP, no floating self IPs due to L3 separation of the two LTMs)

Each LTM is situated on separate L3 segments and all VIPs are announced successfully via RHI.

Traffic groups fail over based on a gateway failsafe that icmp monitors an interface on its router.

It all works beautifully in every failure scenario I have tested so far but, failover takes around 10-20 seconds. I have tweaked lots of parameters in the LTMs but none of them improve the situation.

Is there a way to come down to less than a five second failover time?

R1-----R2
|      |
F5     F5
application delivery
bgp
BIG-IP

4 Replies

Sort By
  • PeteWhite's avatar
    PeteWhite
    Icon for Employee rankEmployee
    Dec 18, 2015

    You need to do some testing to see what the delay is - from the monitor triggering or the BGP update once it's triggered. Might be the BGP holddown timer delaying the route being used.

     

  • fire12_313356's avatar
    fire12_313356
    Icon for Altocumulus rankAltocumulus
    Mar 28, 2017

    Hello there is a protocol in the F5 LTM that's named BFD, In general, BFD detects connectivity changes more rapidly than the endpoints' standard Hello mechanisms.

     

    i think that will help you

     

    best regards

     

    • ingard's avatar
      ingard
      Icon for Nimbostratus rankNimbostratus
      Jul 05, 2017

      Did you get to the bottom of this? We're in the process of switching to an active/active bgp setup and was wondering if we should go with a separate link net for each ltm<->router pair or stick them on the same subnet. Does it even matter if you have a floating ip or not when bgp deals with the failover?

       

  • SWJO's avatar
    SWJO
    Icon for Cirrostratus rankCirrostratus
    Jan 31, 2019

    please send me piece of zebos configuration.

     

    I could be help you.