Behind the Scenes: Making the 2016 Iowa Caucus App

Sign in to queue

The Discussion

  • User profile image
    smcgough

    Would have liked to hear more about the response to performance issues. Sounds like nothing was done to spin up more servers?

  • User profile image
    tommypatter​son

    @smcgough: The dev team said they were on the ground watching for any performance issues and instead of scaling on demand, they had the environment prestaged and ready.  With such a narrow window of operation, all systems were ready for maximum participation.  By all measurements the operation was a success.  I will ask the team to chime in if there are further details they can share as well. Thanks for watching the show! 

  • User profile image
    svarcoe

    @smcgough: @tommypatterson: When we started the investigation, resource utilization across our web server pool wasn't over any limits as this issue affected single servers independently at different times during the event, so we did not think that this would be mitigated by adding more servers. In addition, if we wanted to increase the number of instances we would have needed to deploy our solution to a separate data center due to the number of resources we were requesting, which would have taken too long. By the time we had a hypothesis for the root cause and were ready to deploy a fix the issue had been mitigated due to lower load.

  • User profile image
    smcgough

    Thanks for the additional details.

    It sounds like the reports of downtime were overstated in the media. Any chance of metrics? Something like x% of total requests were dropped due to peak load.

     

Add Your 2 Cents