Adding a new column sounds trivial until it breaks production. Performance drops. Queries time out. Data skews. The solution is not just to alter the table, but to do it in a way that scales, rolls back cleanly, and preserves uptime.
A new column changes your schema, which can cascade into ORM models, API contracts, and ETL jobs. Before you run ALTER TABLE, define exactly what the column will store, its type, default value, and nullability. If the dataset is large, consider adding the column without a default first, then backfilling in batches to avoid locks.
Use transactional DDL when supported. If not, create a migration plan that includes failover and rollback paths. Test with real workloads, not sample data. Run explain plans before and after to confirm index usage remains optimal. For critical paths, measure query latency across replicas before merging to main.