Forum Discussion

type_ra_126722's avatar
type_ra_126722
Icon for Nimbostratus rankNimbostratus
Feb 05, 2014

High availaibility for highly transactional applications

I have a question re the best practice for ensuring the high availability of a few applications that are highly transactional by nature.

 

If I have a virtual server in DC1, that is constantly writing info to a back end DB, with a secondary server available in a secondary DC for DR purposes only, how can the BIG-IP ensure high availability in the event of the primary VM going offline? I'd plan to use the same VIP to advertise the service from both locations, just not load balance across DCs due to a few constraints in the DR DC I currently have? Storage replication is synchronous.

 

Any best practice/deployment guidelines? Is this a realistic requirement?

 

Thanks

 

availability
deployment

3 Replies

Sort By
  • Cory_50405's avatar
    Cory_50405
    Icon for Noctilucent rankNoctilucent
    Feb 05, 2014

    Using the same VIP at your backup DC would likely require some kind of supporting network routing change to send requests there vice your active DC. If you used a different VIP at your secondary DC and setup load balancing to always prefer the VIP at your primary DC unless it was down, and then and only then to send traffic to your backup DC, that may be cleaner.

     

    Of course, you're probably going to encounter at least a little disruption in service between the time your server goes down and the time the BIG IP marks it down and starts sending traffic to the backup DC. You can better control this through the interval and timeout values on monitor you set.

     

  • type_ra_126722's avatar
    type_ra_126722
    Icon for Nimbostratus rankNimbostratus
    Feb 05, 2014

    I've facilitated clever routing to ensure the traffic is routed to the appropriate DC as required. I'm more concerned around the backend. The particular VM is highly transactional, constantly writing to storage. Although the replication is synchronous, if the primary VM fails and the DR VM is then used, whats the deal around the storage element?

     

    • Cory_50405's avatar
      Cory_50405
      Icon for Noctilucent rankNoctilucent
      Feb 05, 2014
      Sounds like you have the storage element covered with synchronous replication. As long as that replication is reliable and has the bandwidth to do what it needs, that part should be good. You'll probably just want to set the interval and timeout values relatively low on the monitor for your VM.