Why we're moving our warehouse off DuckDB
Summary
PostHog explains why they moved their data warehouse from DuckDB to a distributed architecture using Trino and HogLake. The piece details scale-related limitations of DuckDB in cloud environments, the drawbacks of DuckLake, and the decision to switch to a Trino-based stack to handle multi-petabyte workloads and complex queries. It also shares lessons, upcoming improvements, and invites potential design-partner discussions.