MySQL performance monitoring using Statsd and Graphite
This session will explain how Spil Games migrated from available MySQL performance monitoring systems to their new internal standardized system using Statsd and Graphite. It will cover its architecture and functional basics.
In the past few years we have been using and testing most MySQL monitoring systems available. While scaling up our infrastructure cross datacenter the monitoring systems we used did not scale in a similar way while we were suffering from serious lag and gaps in the metrics captured. Adding more monitoring instances and/or slaves could have helped but it would have complicated the infrastructure.
At the same time we were also facing scalability issues with monitoring the performance of other applications than MySQL and Spil Games decided to standardize a new monitoring system using Statsd and Graphite as components for its standardized performance monitoring and metrics system. This required (some) custom work from the Database department to send MySQL metrics through Statsd but the effort payed of as it allowed us to create our new dashboards with ease by using available frameworks for Graphite.
The new system also allowed us to increase the granularity of captured metrics from 3 minutes to 2 seconds. Apart from monitoring MySQL and operational systems we also use this system to monitor the performance of the various layers our new architecture built in Erlang.
Spil Games is a (social) gaming company that grew in a short time from an internet startup to a global online gaming leader.