Forum Discussion

RKallas_77708's avatar
RKallas_77708
Icon for Nimbostratus rankNimbostratus
Feb 17, 2014

RRT for monitoring pool members?

My LTM has a pool of servers that from time to time, one server seems to be unreachable for extended periods of time. My application team would like to monitor them on the LTM, based on the rrt between the LTM and the servers. So for instance, if it takes 30ms for one server to respond to the LTM, it would be marked down. I was thinking of using a TCP half-open monitor to do this.

 

Has anyone tried this? and are there any suggestions for a better way?

 

Thanks, Ray

 

availability
ihealth
management
TMOS
TMSH

3 Replies

Sort By
  • Philipp_Stadler's avatar
    Philipp_Stadler
    Icon for Nimbostratus rankNimbostratus
    Feb 17, 2014

    Hi,

    do you really want to know the RTT, in my opinion this is only a TCP-related metric and this might not be useful for you (because maybe only the server is overloaded and TCP RTT will not increase very much).

    So maybe you want to measure response time:

    when HTTP_REQUEST_SEND {
    set time_start [clock clicks -milliseconds]
}

when HTTP_RESPONSE {
    set server_latency [expr {[clock clicks -milliseconds] - $time_start}]
    
     do anything you want with $server_latency
    
}

    If server load increases also the $server_latency will increase and you can make comparison to max allowed value.

    Philipp

    • Philipp_Stadler's avatar
      Philipp_Stadler
      Icon for Nimbostratus rankNimbostratus
      Feb 17, 2014
      oh, I forgot to mention that you can check the selected pool member with LB::server (https://clouddocs.f5.com/api/irules/LB__server.html) within HTTP_RESPONSE.
  • RKallas_77708's avatar
    RKallas_77708
    Icon for Nimbostratus rankNimbostratus
    Feb 17, 2014

    While this is a good point although I don't know if it applies here. The Clients are doing SFTP to these servers, not http. I think that's why the apps team was asking for RRT as a measurement.

     

    Ray