Deleting Data at Organizational Scale
summary
Stripe shows a working deletion pipeline that removes many billions of documents deleted across hundreds of thousands of datasets and hundreds of petabytes while reporting zero wrongful deletions, zero high‑severity incidents. Implementation details: denormalized affiliation columns for discoverability, a central inventory for the legacy Ruby monolith, per‑service delici endpoints for orchestration, a temporary access lock plus vaulted backups for safe rollback, and age/retention limits that solve roughly 80% of offline lake cases.
AI-written summary · about
chapters (13)
- 0:00Data deletion at scale
- 1:06Stripe’s scale and motivations
- 2:34Account lifecycle triggers
- 3:38Data distribution complexity
- 5:02Identifying target data challenges
- 6:25Exploration and solutions
- 7:08Patterns and learnings from deletion
- 9:15Complex data affiliations
- 9:37Online orchestration
- 11:25Safe deletion with authorization
- 12:08Results and outcomes
- 12:52SOA-friendly endpoints
- 13:36Offline propagation challenges
related talks
19mData etiquette in a regulated world
43mThe Death of Data: Risk, Retention, and Rot
16mHow Canva Built Simple, Auditable, and Maintainable Data Retention
31mHow I Deleted Clickstream and It Was Restored from Oblivion
27mGDPR Data Cleaner: Mutating Immutable Data
37mData Swamp to Data Swimming Pool: Cleaning up the Mess
25mDealing with data in dynamic systems
25mDelete This: Decommissioning Servers at Scale