Forum Discussion

Allen_Shatzer_6's avatar
Allen_Shatzer_6
Icon for Nimbostratus rankNimbostratus
Dec 29, 2009

GTM Health Monitor fails to remote site even though LTM vip is still available

We are using LTM and GTM 9.4.3.

 

 

We have Custom HTTP health monitors with a 60 second interval and 121 second timeout on both the LTM and GTM. Probe timeout on GTM is 5 seconds. We have two nodes in our primary site. One is a "primary" server while the other is secondary server via use of separate priority groups for the two servers; When the LTM monitor detects that one of the nodes is down via the HTTP monitor, it will correctly mark the primary node down and direct traffic to the secondary node.

 

 

However, the GTM also appears to be detecting a "down" state on the primary server at the primary site and failing over to our remote site even though the service on the LTM "eventually" fails over to the secondary node at the primary site.

 

 

Any ideas? We have tried increasing the timeout on the GTM to 181 seconds to allow for three successive failed montor attempts by the GTM before having the GTM mark the primary site down so that the LTM VIP will respond with a success within the GTM's 3 attempts, but that did not fix the problem. Is it simply a matter of increasing the timeout even more on the GTM?
management
monitoring
  • The_Bhattman's avatar
    The_Bhattman
    Icon for Nimbostratus rankNimbostratus
    Dec 30, 2009
    Are you monitoring the primary server directly or are you doing it via LTM to GTM communication?

     

     

    Bhattman
  • Allen_Shatzer_6's avatar
    Allen_Shatzer_6
    Icon for Nimbostratus rankNimbostratus
    Dec 30, 2009
    I am assuming that we are monitoring it directly since we are sending an HTTP request from the GTM using an HTTP monitor and checking the response for a specific string, just like on the LTM.

     

     

    If we were doing the monitoring via LTM to GTM communication, where would this be configured in the GTM and what would the settings be? Would this be a "BigIP" monitor? How would this be associated with the GTM Pool or WIP?

     

     

    Thanks,

     

    Allen
  • The_Bhattman's avatar
    The_Bhattman
    Icon for Nimbostratus rankNimbostratus
    Dec 30, 2009
    Hi Allen,

     

    There is a big3d_install script on the GTM which allows the GTM and LTM to communicate metrics. Basically you would add the LTM if you haven't already as a server where you would define the virtual server within that server. The monitor would be BigIP. You also need run big3d_install script on the GTM so that the GTM and LTM exchange keys. Once that is done you don't need a direct health check. The LTM and GTM will communicate whether the VIP is up or down based on the LTMs health check of the servers.

     

     

     

    I hope this helps

     

     

    Bhattman
  • Allen_Shatzer_6's avatar
    Allen_Shatzer_6
    Icon for Nimbostratus rankNimbostratus
    Dec 30, 2009
    Thanks. We removed the health monitors from the pool on the GTM and I tested bringing down our services, and they worked as I would expect by just using the bigip monitor between the GTM and LTM. It no longer causes the GTM to fail the service to our remote site when only the primary server is taken down at our local site. It also correctly detected that the local site was unavailable when both the primary and secondary servers were taken down and failed to the remote site.

     

     

    Thanks for the info.