ClickHouse 26.10.3.7 slows on uncompacted Iceberg exports

DBTrail tested whether ClickHouse can read Apache Iceberg tables exported from MySQL every minute and found that while exact rows are returned and caches speed up reads, uncompacted delete and data files cause query times to degrade rapidly. The author demonstrates that compacting the tables using Amazon Athena successfully resolves the performance drop by rewriting data files and applying deletes. Additionally, Altinity's Project Antalya was tested and showed similar read behavior and compaction limitations. Open article →