Replay Safety at Datadog

Abstract

At Datadog, Temporal is a key building block that supports many customer-facing products. As a result, workflow replay safety incidents often have customer impact and are operationally intensive to recover from.

In this talk, we cover how the team has worked towards reducing reliance on human factors in order to safely modify workflow code, especially since maintaining replay safety can be tedious and error-prone for even experienced Temporal developers.

We’re excited to share our approach to reducing the frequency and impact of replay safety incidents, focusing on Replay Tests, Dynamic Breaking Change Detection, and Deployment Gates.

About the Presenter

Jing Yi Wang is a software engineer at Datadog here to share her experience in enhancing developer tools and ensuring system reliability. After graduating from the University of Waterloo, Jing Yi has worked on ensuring replay safety with Temporal, a key part of the developer experience for many at Datadog.