Forum Discussion

matt_w_53423's avatar
matt_w_53423
Icon for Nimbostratus rankNimbostratus
Jul 19, 2017

Delay Failover for 5 minutes on GTM wide IP with an irule

I want to be able to delay the failover on a GTM device for a particular VS with an irule. 2 Data Centers, Pools are setup as active standby on LTM. 2 servers in a Pool, Server1 active, Server2 standby. DC1 server1 fails, it will take 5 minutes for server2 to come online. In that 5 minute period, we do not want to failover to DC2. After 5 minutes, failover to DC2.

 

4 Replies

  • Can you post your GTM and LTM configurations, Your configuration explanation is unclear.

     

  • Thanks for the response...

     

    basically, I want to just delay the failover from DC1 to DC2 for a particular virtual server.

     

    On the LTM, there are 2 servers in the pool. They are in an active/standby state. Active server receives all connections unless the HM fails. However, the standby server, does not come online immediately (anywhere from 5secs to 5mins depending on the amount of data transferred between the failed server and the newly appointed active server). GTM on the other hand, see both servers are in a failed state and it sends traffic to DC2 because that's what its supposed to do.

     

    The application owners want to delay this GTM failover, because they want the DC1 standby server to have enough time to come online (go active) and receive the connections. This could take 5secs to 5mins.

     

    In the case DC1 fails completely, then the failover to DC2 would be acceptable.

     

    I hope this is a better explanation. If this is unclear, I will post LTM/GTM configs

     

  • How's the DC1 virtual server on the GTM pool added. Is your GTM using Iquery to get the LTM's data or have you put a monitor to the VS on the GTM pool, as in like a generic host setup. If you have put like basic tcp monitor on the GTM pool, once the DC1 VS fails, its marks it unavailable and sends the traffic to the DC2 VS on the GTM pool.

    Meaning what ever effort you take on the LTM to make wait for 5 mins has no effect, since GTM has already marked the DC1 VS down. Thus the GTM setup has to be understood first.

    Also when you say on the LTM pool, you have active/standby end servers. Hope you meant priority group activation here.

    The problem you are facing is your DC1 LTM's pool marking down noth members on same interval. And requirement is to have to wait for 5-5mins for standby server on the same pool. As a solution, you can have pool-member specific monitor assigned to the standby server in the ltm pool. Something like below, where http_1.1_monitor will have larger timeout value ranging till 300 secs.

    pool my_pool {
    members {
        1.1.1.1:80 {
            monitor http_1.1_monitor
            session monitor-enabled
        }
        1.1.1.2:80 {
            session monitor-enabled
        }
    }
    monitor http
    }