Continuous Data Protection in Snowflake

Official answer

Snowflake's Continuous Data Protection includes Time Travel and Fail-safe. They address different recovery windows and are not interchangeable with a configured cross-region disaster-recovery system.

Time Travel provides user-accessible historical queries and supported recovery operations within the object's effective retention window. Verify edition, table type and configured retention before relying on it. For an accidental update, identify the affected rows and a valid prior point, recover into an isolated target, and reconcile legitimate changes made afterward before restoring production data.

Fail-safe is a separate, best-effort Snowflake recovery service for eligible data after Time Travel. It is not ordinary historical-query access, and recovery can take hours or days. Temporary/transient and ingestion-specific limitations matter; do not assume every table has the same protection.

Replication/failover complements these controls when recovery must survive an account or regional problem. It needs supported object scope, destination configuration, freshness monitoring and a tested promotion/client-routing procedure. Replication can also propagate an unwanted change, so a current replica alone may not solve logical corruption.

Document recovery objectives, responsible operators and dependencies. Test accidental deletion, bad loads and regional recovery separately, verifying business totals and application access. Retained history and protected copies incur costs, and no single feature guarantees recovery from every failure.

Time Travel, Fail-safe, replication