Handling failover with MySQL 5.6 and Global Transaction IDs

Replication and High Availability Strategies
4 April 11:00AM - 11:50AM @ Ballroom E

Experience level: 
50 minutes conference
Replication has been used for a long time to provide high availability. However the traditional position-based replication has a serious drawback: a same transaction will never have the same position on two different servers. Reconfiguring replication is then difficult and error-prone, which is bad when you want to make failovers as graceful and reliable as possible. Enter MySQL 5.6 and Global Transaction IDs (GTIDs): a transaction is guaranteed to have the same identifiers across all servers of a given replication topology. This opens up new ways to handle failover, for instance when used in combination with Oracle's MySQL Utilities. But are GTIDs ready for production usage? How do you set up GTIDs and the MySQL Utilities to provide high availability and automatic failover to your replication cluster? What are the limitations of the MySQL Utilities? How do they compare to existing tools, for instance with MHA which is a robust and widely used tool? Come to this session to get a sense of the benefits and challenges of the high-availability solutions available with MySQL 5.6.


Database Engineer, Slice Technologies
Stephane has worked for almost 15 years in open source environments, and has spent most of the last 10 years solving performance or high-availability problems with MySQL and more recently MongoDB. He is now working for Slice Technologies, a company that helps you keep track of your online purchases. His main role is to automate MySQL and MongoDB operations, and from time to time, he is also very excited to kill a nasty query or add a missing index on a 500GB table. When not in front of a computer, Stéphane likes spending time with his family. He also hopes to become a decent chess player someday.