Forum Discussion

Akki_133622's avatar
Akki_133622
Icon for Nimbostratus rankNimbostratus
Sep 13, 2013

F5 Load balancer - Failover time

Hi Experts,

 

I have a general query around how F5 load balancing happens with Application? My query is specifically around two applications which are in cluster and configured behind F5. When failover from App1 to App2 happens, it takes almost 45 second to switch and user needs to wait for 46th second. Some inputs I got is that this is due to LB configuration of retries around 15 second and 3 retries. How does it happen in real world as I believe there are lots of application behind LB and I haven't seen wait of 45 seconds. Any expert guidance. One more option was to reduce 15seconds but then one comment was it might impact Performance. What is best way to address this?

 

Akki

 

4 Replies

  • Hey. Could you provide a bit more detail please. Can I assume App1 and App2 are Pool Members? When you say they are clustered do you just mean in a Pool?

     

    Sounds to me like TCP is being relied upon to handle the switch, hence the delay. There's a few things that can be done to improve this depending on your set-up.

     

  • OK, thanks for that. So, you should initially look at the timers of the Health Monitor used to detect the failure. There's every chance they are set to monitor at a 15s interval and won't mark a Pool Member down until after 46 seconds? Can you check that and we'll take it from there.

     

  • OK, so that clarifies the delay, 3x 15s + 1 = 46. So, you should lower the timers accordingly to get to a time you find acceptable. You should probably also modify the Action on Service Down Pool setting from None to Reselect.

     

    Enabling OneConnect may also have a positive impact.

     

  • How about decrease interval monitor's timer into 5s and timeout into 16s ?