We've made the switch from PostgreSQL to ClickHouse for our event data storage in Sentry, driven by performance and scalability issues with our existing PostgreSQL-based solution. We prioritized fast aggregations over large numbers of rows, bulk insertion of large amounts of data, and predictable performance. ClickHouse's immutable data approach allowed us to simplify our database design and reduce the need for transactions, but also required us to adapt to its unique features, such as the MergeTree storage engine and lack of DELETE statements. We've developed custom solutions, like using the ReplacingMergeTree engine and tracking recently deleted issues in Redis, to handle mutations and deletions, and have learned to be more creative with our database interactions. By stripping away unnecessary features and leveraging ClickHouse's strengths, we've achieved significant performance improvements and reduced the complexity of our system.