database-migration

Rename a column with zero downtime using a multi-phase approach: add the new column, deploy code that writes to both columns, backfill existing data, switch reads to the new column, then drop the old one. Never rename directly on a live production table.

Implement the down function in every Sequelize migration to reverse the up operation, such as dropping a created table or removing an added column. Run npx sequelize-cli db:migrate:undo to execute the rollback for the most recent migration.

Sequelize uses JavaScript migration files with up/down functions, TypeORM uses TypeScript classes implementing MigrationInterface, and Prisma generates migrations automatically from schema.prisma changes. Choose based on your existing ORM and whether you prefer code-first or schema-first workflows.

Yes, but handle dialect differences explicitly, such as using JSONB for PostgreSQL and JSON for MySQL. Check the active dialect with queryInterface.sequelize.getDialect() and branch your migration logic accordingly.

Wrap migration steps in a transaction so failures trigger an automatic rollback to the pre-migration state. For large data changes, use checkpoint-based backups that create a backup table and restore from it if verification fails.