The highs and lows of semi-synchronous replication

High Availability
22 September 3:10PM - 4:00PM @ Matterhorn 1

Duration: 
50 minutes conference
MySQL semi-synchronous replication is a powerful feature allowing very fast failover with no data loss. However managing it operationally is tricky as you move in to an environment where replication stopping will cause writes to stop too. Any implementation needs to ensure the reduction in downtime from faster failover is not offset by an increase in downtime caused by stalls or failures in your replication topography. I will talk about the challenges Facebook faced as we rolled out semi-synchronous replication across our different services, discussing our initial approaches that did not work out, as well as describing our current solution.


Speakers

Production Engineering Manager, Facebook
Simon has been working with MySQL for most of his career, starting from 2 servers built out of spare parts and moving through to one of the largest deployments in the world.

Slides