Forum Discussion

Kibao_130897's avatar
Kibao_130897
Icon for Nimbostratus rankNimbostratus
Oct 20, 2014

Failover Configuration Between Datacenter and Data Recovery Center using F5 BIG-IP GTM Feature

I want to deploy the F5 BIG-IP 6900 GTM to do fail over between DC and DRC. Traffics from remote site usually go to the DC. When the DC/services at the DC is down, LoadBalancers should automatically redirect traffics to the DRC. I have attached setup which I think will serve purpose. Is that setup correct? If not, Can you advice the correct setup?

 

Thanks in Advance, Saidi

 

9 Replies

  • The GTM feature uses DNS to re-direct traffic. You can set it up to monitor health of the servers behind it. You can also have it set one set of servers as the primary, and the other as backup.

     

    You'll need to tune your DNS record TTL appropriately though. If you go with a 24 hour TTL of 86400... the GTMs would immediately start serving out the Data Recovery Center IP address right away when the Data Center is unavailable, but anyone that has connected will have the old IP address cached for the next 24 hours. If you tune your DNS TTL down to 0... you'll be serving out the IP address from your GTM's on every single request. This is doable, but doesn't scale for large environments.

     

    Jason

     

    • Kibao_130897's avatar
      Kibao_130897
      Icon for Nimbostratus rankNimbostratus
      Thanks Very Much Jason, What do you mean when you say large environments? We have 214 sites and each site has around 20 users. Will this solution work perfect with our infrastructure? Saidi
    • Jason_40733's avatar
      Jason_40733
      Icon for Cirrocumulus rankCirrocumulus
      Turning your DNS TTL for things being load balanced by GTM means that every connection from every client will always do a DNS lookup. If you have a very busy application or one that does not make use of persistent connections you can flood your GTMs with DNS requests by setting the TTL to 0. Determining an acceptable amount of "lag" between switching from a failed primary to a backup and then setting your DNS TTL appropriately is advised. I've had some applications that useda 300 second TTL, and some that used a 10 second TTL. Jason
  • Thanks Very Much Jason,

     

    What do you mean when you say large environments? We have 214 sites and each site has around 20 users. Will this solution work perfect with our infrastructure?

     

    Saidi

     

  • Hi Kibao,

     

    How your remote sites connects to your DC is it over the internet with some sort of VPN to the main site also what are the services that the users on remote site access is it all web based services or different??

     

    Regards,

     

    • Kibao_130897's avatar
      Kibao_130897
      Icon for Nimbostratus rankNimbostratus
      Currently all remote sites connects to the DC through VPN connection over the WAN but later there will be some sites which will connect to the DC over the internet. All users access DC services via web based services
  • Hi Kibao,

     

    How your remote sites connects to your DC is it over the internet with some sort of VPN to the main site also what are the services that the users on remote site access is it all web based services or different??

     

    Regards,

     

    • Kibao_130897's avatar
      Kibao_130897
      Icon for Nimbostratus rankNimbostratus
      Currently all remote sites connects to the DC through VPN connection over the WAN but later there will be some sites which will connect to the DC over the internet. All users access DC services via web based services
  • With respect to your 214 sites and 20 users. It will depend on the application and how it does requests. 4280 users is not a lot for doing DNS requests even concurrently. It is unlikely your application is written poorly enough to overwhelm your GTM even with a TTL of 0.

     

    Anycast may also be a different non-GTM solution for you. Compare the two and choose the one that best fits your needs.

     

    Jason