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 — in a mostly live demo, on a real schema, with a real mistake and a real recovery.
Step by step, Daniel shows how to:
- identify which rows changed and when
- inspect full before-and-after images of each change
- query any table as of a past timestamp, using plain SQL
- generate the exact reversal statements for the incident — including changes propagated through FK cascades, which manual binlog parsing usually gets wrong
He’ll also cover verifying that the captured history matches the source — a check worth running 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 aren’t accessible. Deployment is Docker Compose, and everything shown is reproducible on a test instance.