Zero-downtime schema changes with expand and contract

Goal

Change database structure while the service keeps serving, with every intermediate state working for both the old and the new code version.

Prerequisites

Migrations versioned with the code, a deployment process that can release code and migrations separately, and knowledge of which statements take strong locks (the PostgreSQL ALTER TABLE documentation lists the lock level for each form).

Steps

1. Expand: add the new column, table or index in a way that is fast and does not block (nullable column without default rewrite, CREATE INDEX CONCURRENTLY). Old code ignores it.
2. Deploy code that writes both old and new representations and still reads the old one.
3. Backfill existing rows in small batches with pauses, monitoring lock waits and replication lag.
4. Deploy code that reads the new representation and verifies it against the old for a period (log mismatches).
5. Stop writing the old representation; deploy.
6. Contract: drop the old column or table in a later release once rollback to the previous code is no longer needed.

Expected result

Each deployment is individually reversible; no migration takes a lock long enough to be noticed.

Limits and test basis

The pattern multiplies steps and takes days rather than minutes. Renaming a column is done as add-copy-drop, not RENAME, if both code versions must run. Lock behaviour follows the cited documentation; timings are not claimed.