💡Logs Are Out, Events Are In for Observability
Why logs are the new junk drawer of observability
TL;DR
Logs are being phased out in favor of events for observability. Events provide metadata for slicing and dicing data, making them the foundation for metrics and observability. Logs are now seen as clutter.
Logs are being phased out in favor of events for observability. Logs are intended to diagnose issues but often capture unnecessary success cases, leading to clutter. Events, on the other hand, are data points with metadata that enable detailed analysis and are the foundation for metrics. This shift means developers can now derive metrics from events more easily, improving observability. The article emphasizes the importance of starting with events for better observability practices and highlights the limitations of metrics-only approaches.
Key Points
Logs are no longer recommended for observability as they capture unnecessary success cases (February 28, 2024).
Events provide metadata for slicing and dicing data, making them essential for observability (February 28, 2024).
Metrics are derived from events, not logs, for better observability practices (February 28, 2024).
Starting with events makes it easy to derive metrics, improving observability (February 28, 2024).
Starting with metrics makes it impossible to work with events effectively (February 28, 2024).
Why It Matters
If you're managing observability in a complex system, focusing on events over logs can drastically improve data analysis. For instance, a team using Prometheus for monitoring might find that switching to event-based data collection simplifies metric aggregation and anomaly detection. However, this shift requires a fundamental change in how data is collected and stored, which can be challenging for existing systems.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,567 builders reading daily.