Maintaining Iceberg at Scale: Lessons from Slack

Apache Iceberg is a powerful and broadly applicable table format. But while Iceberg is generally applicable, maintaining Iceberg tables isn't. At Slack, our Iceberg adoption grew from a small number of curated datasets to tens of thousands of tables across many teams and workloads. We quickly found that maintenance-not table creation or query performance-was the real scaling challenge. Compaction, snapshot expiration, and metadata growth vary widely based on data volume, ingestion patterns, and business importance, making one-size-fits-all approaches ineffective. Just as importantly, the challenge was organizational. Expecting every data owner to understand Iceberg internals and reliably maintain their own tables didn't scale. This talk covers how we addressed both problems by treating Iceberg maintenance as a centralized, automated platform capability. I'll share how we classify datasets, apply tailored maintenance strategies, and operate Iceberg safely at scale-along with the lessons we learned along the way. Speakers: Lweikum (Speaker) - [email protected] --- Recorded at Iceberg Summit 2026. Learn more: https://www.icebergsummit2026.com/ #ApacheIceberg #IcebergSummit #DataLakehouse