Baidu Cloud Redis Data Flashback: Hybrid RDB+AOF Persistence for Precise Point‑In‑Time Recovery

In database operations, DBAs often rely on Redis's native persistence mechanisms—RDB snapshots or AOF logs—to archive data and restore it after failures, but these methods have drawbacks in high‑traffic scenarios such as games and e‑commerce.

RDB cannot restore to arbitrary points, losing data between snapshots, while AOF records every write command, resulting in large files and slow recovery times.

Baidu Cloud Redis introduces a “Data Flashback” feature that uses a hybrid persistence model combining RDB and AOF, adding timestamps to AOF entries, which enables fast, space‑efficient point‑in‑time recovery from seconds to hours.

The solution stores a baseline RDB snapshot for quick baseline recovery and incremental AOF logs with timestamps for detailed command‑level restoration; AOF files are retained for one week before being cleared and a new RDB+AOF cycle begins.

When a rollback request is submitted, the platform clones the original cluster, locates the appropriate RDB and timestamped AOF files for the specified time, loads them into memory, and restores the cluster state precisely to that moment.

This approach overcomes the limitations of traditional RDB/AOF, providing near‑second precision and reduced storage overhead for Redis data recovery.