An intro to observability

Are you tired of managing three separate systems for logs, metrics, and traces? This video explores the traditional three-pillar approach to observability and introduces a simpler, more unified alternative.

We start with the classic observability setup - logs stored in Elasticsearch with search UIs, metrics aggregated in Prometheus with Grafana dashboards, and traces collected in systems like Zipkin with Cassandra backends. While this approach works, it creates artificial boundaries that complicate debugging and increase operational overhead.

• Traditional observability challenges - Managing separate storage systems, UIs, and collection agents
• The debugging problem - Manually stitching together data from disconnected systems during incidents
• Wide events approach - Ivan Bermiskov's insight that unified event data can replace the three-pillar model
• Simplified architecture - Single column store, unified UI, and consolidated data collection

Instead of maintaining three separate systems, we explore how a column store that supports semi-structured data can handle logs, metrics, and traces together. This unified approach reduces complexity while maintaining the observability insights you need.

We also tease the ClickStack as one opinionated, open-source solution that provides all these components in a cohesive package.

• https://clickhouse.com/engineering-resources/observability