Forum Discussion

Charles_Harris's avatar
Charles_Harris
Icon for Nimbostratus rankNimbostratus
Mar 01, 2006

iControl Rules of Engagement...

Hi

 

 

Having written a few application specific iControl monitors for a service's backend servers, I was wondering if there were and F5 recommendations regarding the interval that iControl scripts should be run and what the maximum number of concurrent connections via iControl is ?

 

 

The Scenario:

 

 

5 Backend servers: A, B, C, D, E

 

 

Servers A, B & C have two associated Virtual Servers, but are also dependant on virtual servers associated with D & E.

 

 

As A, B & C need to know the status of D & E, the iControl scripts have to check their status before deciding if they need to take any BigIP action.

 

 

The application requirements we have mean that we need the fastest possible reaction to (application) failure so as a result, all nodes for the service check their BigIP status (and each others) every 15 seconds.

 

 

Operationally this is working well at the moment, but we have an additional three services that require a similar setup and I'm concerned that this could cause BigIP problems.

 

 

I apologise in advance if this sounds like a badly written test exam, but I'd appreciate anyone else's experiences / guidance here.

 

 

Thanks in advance!!

 

 

-=ChaZ=-

 

 

 

 

 

2 Replies

  • Great question ChaZ,

     

     

    The optimal frequency of calls is very dependent on what the calls are. If all you are doing is making status requests, then the overhead is very low and every 15 seconds should not cause any burdon on the system for your 5 servers (or additional three). If, on the other hand, you are making requests for large amounts of statistics, then you might want to space out the requests a bit longer.

     

     

    The BIG-IP serializes iControl requests into a simple FIFO queue so that only one is executed at a time. So, the more servers you have making iControl calls, the only side-effect will be that they have to wait a fraction of a second longer than they normally would if acting alone.

     

     

    Hope this helps...

     

     

    -Joe
  • Thanks for the reply Joe! - I'm glad to hear that we are not going to be causing any issues with the existing method and that requests are handled by a FIFO on te bigIP side, it will make things more simple for the backends.

     

     

     

    Best regards,

     

     

    -=ChaZ=-