Forum Discussion

chamindak_11539's avatar
chamindak_11539
Icon for Nimbostratus rankNimbostratus
Jan 21, 2014

Splitting HA across datacentres

Hi all,

 

We are looking at a typical datacentre design with geographically separated Production and DR datacentres, by ~100kms. Layer 2 connectivity is provided via 10Gb dark fibre so we have the option of cross-site VLANs to do HA.

 

Is there a limitation on doing HA across two boxes over 100kms apart? What is the best practice F5 architecture for such scenarios?

 

Thanks, CK

 

  • Are you referring to splitting LTMs in an HA pair between two physical sites? One LTM with the DR servers and one with the Production servers?

     

    In our environment, we keep HA pairs of LTMs in the same datacenter. We accomplish geographic redundancy using the GTMs ( one per data center ) on top of the LTMs for mission critical applications. We ensure each GTM can monitor all of the LTMs so we have no single points of failure.

     

  • Are you referring to splitting LTMs in an HA pair between two physical sites? One LTM with the DR servers and one with the Production servers?

     

    In our environment, we keep HA pairs of LTMs in the same datacenter. We accomplish geographic redundancy using the GTMs ( one per data center ) on top of the LTMs for mission critical applications. We ensure each GTM can monitor all of the LTMs so we have no single points of failure.

     

  • From my research so far it has been advised against doing active/passive HA across DC's. As per Jason's comment best option is to use GTMs to load balance across stand alone LTMs, however configuration can be synced.

     

    Alternately routing can be manipulated to choose the active LTM between 2 stand alone LTMs.