Session Title: Geographically Based Cloud Scaling Using Open Source Monitoring, Configuration and a DNS as a Service
Moderated By: John Willis
Time: 7:30 pm (PDT)
Room: Napa 3, Santa Clara Convention Center
At the BoF session OpsCode, Dynect and Zenoss (monitoring) will be demonstrating a simulated datacenter failure. Each companies tools will be used in order to detect and recover from this disaster scenario.
We have ‘reserve.dyntini.com’ running (on EC2 fwiw) and set up in using the Dynect Platform’s Traffic Management service. The Web Apps are structured with half running in the East coast datacenter and half in the West coast datacenter.
We will simulate a server failure for *some* of the servers in the West
coast datacenter. Zenoss monitoring will send an alert that the traffic in the West is going 100% to one single server, and then use Chef recipes to modify the Dynect service via the Dynect API to add some of the East coast servers into the West region. Zenoss will then bring up more servers in the West, verify they are up, then reconfigure the Dynect TM service to again have only West coast datacenter servers in the west region DNS.
Next, we’ll do a simulation of the entire West coast datacenter failing. In this scenario we’ll allow the Dynect TM service to do its thing and swap out the DNS for the West region to the remaining available servers. In the mean time, since the West coast datacenter is not available, and 100% of our traffic is going to three servers, we’ll fire up three more instances in the East data center, and use the Dynect API to add those new servers into the Traffic Management service.
It’s going to be AWESOME!!