2019-06-20
A spike in workout-related API requests consumed a large amount of resources which resulted in an increase in latency and API error rates.
Website users experienced page load failures on pages that utilize these APIs, including the dashboard. Mobile and embedded console application users experienced page load failures on the library and dashboard screens.
Shortly after a code deploy occurred, we saw a large spike in API calls made to some of our Lambda workout APIs. These Lambdas get workouts from our core API. One particular route in the core API was experiencing especially high throughput and latency. We believe that the heavy load placed on this route was responsible for most of the issues.
Caching was added to the problematic route at the nginx layer. The MongoDB cluster that our production workout lambdas connect to was also scaled up to further boost performance.