Hazardous Defaults: Managing Cardinality and Performance for Your Logging Stack

This talk addresses challenges in managing cardinality and performance within logging stacks, particularly for those new to observability. The speakers detail their experience with the PLG stack (Promtail, Loki, Grafana), explaining how default configurations can lead to unexpected performance issues and data loss. They share a case study of missing logs and their journey to optimize the system by reducing label cardinality through Promtail pipelines, demonstrating significant improvements in query speed and data efficiency.