Which practice best supports rollback of database migrations?

Which practice best supports rollback of database migrations?

• A. Use breaking changes to schemas.
• B. Do nothing and wait for automatic fix.
• C. Rebuild the database from scratch during rollback.
• D. Use backward-compatible changes, feature flags, and separate rollback procedures.

Answer: (D)

Rollback-ready database migrations hinge on designing changes that don’t break existing code if you need to revert. By making changes backward-compatible, you avoid forcing applications to handle a suddenly vanished or altered schema while the system remains operational. Pair that with feature flags to enable or disable the new behavior without rolling back the database itself, and with explicit, tested rollback procedures that include down migrations and data adjustments. This combination provides a safe, controllable rollback path, quick recovery, and predictable steps to return to the previous state. Other approaches fall short: breaking changes can disrupt live systems if you try to revert them; waiting for an automatic fix isn’t a reliable plan; and rebuilding the database from scratch is typically impractical and risky in production.

Rollback-ready database migrations hinge on designing changes that don’t break existing code if you need to revert. By making changes backward-compatible, you avoid forcing applications to handle a suddenly vanished or altered schema while the system remains operational. Pair that with feature flags to enable or disable the new behavior without rolling back the database itself, and with explicit, tested rollback procedures that include down migrations and data adjustments. This combination provides a safe, controllable rollback path, quick recovery, and predictable steps to return to the previous state. Other approaches fall short: breaking changes can disrupt live systems if you try to revert them; waiting for an automatic fix isn’t a reliable plan; and rebuilding the database from scratch is typically impractical and risky in production.