From Chaos to Clarity: Scaling Observability at Dropbox with Centralized Logging and Metrics Solutions

At Dropbox, managing observability for systems producing terabytes of logs daily presented a unique challenge. Initially, developers needed to log into individual servers to view logs, a process that was cumbersome and posed security risks. With our shift to containerized environments, the challenge only grew as short-lived containers led to log data disappearing upon termination. We recognized the need for a scalable, centralized observability solution leveraging open-source tools.

In this session, I’ll detail our journey to implement a robust observability framework, focusing on our decision to use Loki as our primary logging solution. Deploying Loki at Dropbox’s scale required extensive optimization to handle high data volumes and deliver reliable, efficient query responses. I’ll share insights into our approach to deployment, the challenges we faced, and the technical strategies we employed to achieve consistent, high-performance logging.

Additionally, we integrated Grafana for a unified view, combining logs and metrics for a comprehensive observability solution. This transformation not only improved troubleshooting efficiency but also enhanced security by offering a centralized, accessible platform for monitoring. Join us to learn how these efforts impacted Dropbox’s operations and the lessons we gained in scaling observability with open-source solutions.