MySQL Time Travel: Practical Row-Level Recovery With Open Source Tools

Recovering from a bad UPDATE or DELETE in MySQL is disproportionately expensive: restore a backup and replay binlogs, or hand-parse mysqlbinlog output and reconstruct the reversal manually. Both take hours for incidents that touched a handful of rows.

This session shows a faster workflow using dbtrail, an open source (Apache 2.0) tool that captures every row change from the binlog and makes it queryable.

Step by step:

• Identify which rows changed and when,
• Inspect full before and after images,
• Query any table AS OF a past timestamp with plain SQL,
• And generate the exact reversal statements for the incident, including changes propagated through FK cascades, which manual binlog parsing usually gets wrong.

Also covered: verifying that the captured history matches the source, a check you should run before trusting any of it for recovery.

Capture works over the replication protocol, so the same setup runs on-premises and on RDS or Aurora, where binlog files on disk are not accessible. Deployment is docker compose.

Everything in the session is reproducible on a test instance. Mostly live demo: real schema, real mistake, real recovery.