ClickHouse Observability: ClickStack Unifies Logs, Traces & Metrics at Scale (Open House)

Mike Shi, founder of HyperDX, introduces ClickStack: the high performance open-source observability stack. Lightning-fast queries and powerful aggregations on logs, metrics, traces, session replays and errors with unmatched resource efficiency for even your highest-cardinality data.

Session recording from Open House, the ClickHouse user conference.
https://clickhouse.com/o11y
https://clickhouse.com/openhouse

ClickStack is an open-source observability platform that unifies logs, metrics, and traces in ClickHouse, enabling real-time analysis of high-cardinality, unsampled data at petabyte scale. It provides a comprehensive solution for unifying telemetry by centralizing all data in a single database, eliminating the data silos common in traditional observability tools. This unified architecture supports seamless correlation across back-end error logs, application traces, Kubernetes infrastructure metrics like pod CPU and memory, and front-end user browser sessions. The platform is architected for high performance, leveraging ClickHouse to process petabytes of data with real-time results for both queries and ingest. This allows for fast aggregations on unsampled, high-cardinality events, a critical capability for true observability that enables engineers to drill down from aggregate charts to individual source events without performance degradation. For cost efficiency, ClickStack utilizes ClickHouse's high data compression ratios and compute-storage isolation, which reduces hardware footprint and makes it affordable to retain all telemetry data without sampling. Built on open standards, the platform uses OpenTelemetry for collecting and pipelining telemetry data via the OTEL exporter. It is designed to monitor Kubernetes environments, with deployment options including a Helm chart, Docker Compose, and a single container that can be running in minutes. The system simplifies the query experience by translating a user-friendly, Lucene-like search syntax into optimized SQL that runs directly on ClickHouse. Key features include an event patterns view that automatically clusters similar logs to quickly identify new or trending errors from billions of events and native support for schemaless data ingestion via JSON, allowing it to handle arbitrary telemetry formats performantly.