A Table Had No Clear Owner

The Anatomy of an Orphan Table

An orphan table is a database structure that remains active in production despite its original context and steward having disappeared. In our latest audit, we discovered a table named agg_order_summary_cache occupying 1.2 TB of storage. No documentation explained its purpose, and the Git history pointed to developers who had left the organization four years ago. This discovery triggered a site-wide investigation into data governance.

The Hidden Costs of Unclaimed Data

Operational risks escalate quickly when data ownership is fuzzy. Without a clear owner, critical maintenance tasks are deferred, and security audits become nightmares. We identified several specific dangers inherent in this scenario:

• Storage Bloat: The table was still growing, consuming expensive high-speed storage for potentially obsolete records.
• Migration Paralysis: The Database Reliability team could not upgrade the underlying PostgreSQL version because they couldn't verify if a schema change would break legacy systems.
• Security Gaps: Automated compliance scans often overlook orphaned tables, leading to potential PII leakage in unmonitored columns.

Tracing Silent Dependencies

We solved the mystery by enabling detailed session logging for that specific table. By monitoring the incoming requests, we traced the traffic back to a long-forgotten microservice that was part of a decommissioned logistics portal. The service was still periodically writing heartbeat data into the table. It was a technical ghost—still functioning but forgotten by the current engineering teams.

Resolution and Prevention

We moved the table to a 'Restricted' state, eventually archiving the data and deleting the table after 90 days of zero traffic. To prevent a recurrence, the Engineering Council established a strict stewardship policy. Every new schema object now requires a mandatory metadata tag identifying the owner team. Without this tag, the CI/CD pipeline automatically rejects the migration. This simple enforcement ensures that every byte in our database has a living human or team responsible for its lifecycle.