Replication is used everywhere with MySQL. But applying writes in a single thread is often a bottleneck: replicas have a hard time keeping up with the master that can execute writes in parallel. Enter MySQL 5.6 and you can use several threads to apply writes on replicas as long as you have several schemas. MySQL 5.7 goes even further: by introducing a logical clock and by changing the scheduling logic, you can apply writes in parallel in a single schema.
Come to this session to learn all you need to be comfortable with multi-threaded replication.
Replication is used everywhere with MySQL. But applying writes in a single thread is often a bottleneck: replicas have a hard time keeping up with the master that can execute writes in parallel. Enter MySQL 5.6 and you can use several threads to apply writes on replicas as long as you have several schemas. MySQL 5.7 goes even further: by introducing a logical clock and by changing the scheduling logic, you can apply writes in parallel in a single schema.
The result is of course a much better replication throughput, but this feature also raises a lot of questions: - How does it work internally? - Does it play well with GTIDs? - How can you monitor replication status? - How do you recover from a replication error? - If you want to take a backup from a replica, how can you make sure you will get a consistent snapshot?
Come to this session to learn all you need to be comfortable with multi-threaded replication.
Speakers: Stephane Combaudon