Zero-Downtime Schema Migrations

A schema migration is any change to the structure of a database: adding a column, renaming, dropping a table, changing a type, adding an index. On a small project, it is one command and a restart. On a production system with live users, the same change is one of the most operationally fraught events in the calendar.

The database is shared, persistent, and under load. The application has many instances running mixed versions of the code, all reading and writing the same tables. A migration that locks the database brings the application down; a migration that changes a column's meaning out from under running code corrupts data; an unindexed update on a billion-row table can lock for hours. Zero-downtime schema migration is the practice of making these changes without taking the system down, breaking the application, or putting data at risk.