May 2017 Summaries

In the continuation of their dogfooding series, Honeycomb shares insights on how they utilize their own platform to manage and assess changes to their API, focusing on two key episodes. In the first episode, they address the challenge of unpacking nested JSON objects by analyzing incoming data and opting for a per-dataset feature activation to avoid unexpected disruptions for users. This cautious approach involved using their dogfood cluster to evaluate the impact of potential changes before implementation. The second episode deals with improving rate limiting by transitioning from an in-process cache to a shared cache across their servers, ensuring consistent enforcement despite changes in cluster size or traffic patterns. By running the new system in parallel with the old, they were able to compare results and address any concerns with affected customers individually. These efforts demonstrate Honeycomb's commitment to using their tool for enhanced systems observability and maintaining a smooth user experience while rolling out updates.