Track Errors First: The Most Valuable Observability Signal

2025-06-05
Track Errors First: The Most Valuable Observability Signal

This article argues that error tracking should be prioritized when building observability systems. The standard 'three pillars' model (logs, metrics, and traces) overlooks the most valuable signal: errors. Exceptions directly indicate failed assumptions in the code, offering more direct and valuable information than logs, metrics, or traces. The author advocates for deep error tracking, collecting full stack traces, local variables, request data, and user context, rather than simply counting errors. He criticizes the abstraction of errors as generic events and the trend of full-stack APM tools marginalizing error tracking, emphasizing that error tracking should be paramount, not an afterthought.

Read more
Development error tracking

Why I Abandoned Self-Hosted Sentry: 16GB RAM and a Complex Installation Were the Dealbreakers

2025-04-18
Why I Abandoned Self-Hosted Sentry: 16GB RAM and a Complex Installation Were the Dealbreakers

The author recounts their experience abandoning self-hosted Sentry. Initially, due to work requirements, they successfully self-hosted Sentry. Years later, attempting to set up self-hosted Sentry for a colleague, they encountered numerous warnings in Sentry's documentation about the risks of self-hosting, along with demanding resource requirements (at least 16GB RAM and multiple cores). This proved to be costly and incredibly difficult to maintain, with the installation process involving hundreds of lines of scripts. Online user feedback confirmed the difficulty of maintaining self-hosted Sentry. Ultimately, the author gave up on self-hosting Sentry and decided to develop a more lightweight alternative.

Read more
Development