MySQL Enterprise High Availability

MySQL has always been famous for its very flexible Master/Slave replication. However, MySQL users have been looking for a high availability solution built into the MySQL Server. Unlike other products, MySQL Group Replication delivers native, built-in HA for your MySQL databases:

• Core part of the standard MySQL Server
• Leverages proven MySQL features: InnoDB, GTIDs, binary logs, multi-threaded slave execution, multi-source replication, Performance Schema
• Supported on all MySQL platforms

Group Replication provides native high availability with built-in group membership management, data consistency guarantees, conflict detection and handling, node failure detection and database failover related operations, all without the need for manual intervention or custom tooling.

MySQL Group Replication delivers native, built-in HA for your MySQL databases

High Availability with MySQL Group Replication

Group Replication implements both a single-primary mode with automatic leader election and a multi-master update everywhere mode. By using a powerful new group communication system, which provides an in-house implementation of the popular Paxos algorithm, the group automatically coordinates on data replication, consistency, and membership. This provides all of the built-in mechanisms necessary for making your MySQL databases highly available.

Elasticity

With Group Replication, a set of servers coordinate together to form a group. The group membership is dynamic and servers can leave, either voluntarily or involuntarily, and join at any time. The group will automatically reconfigure itself as needed and ensure that any joining member is synchronized with the group. This makes it easy to quickly scale your total database capacity up and down as needed.

Failure Detection

Group Replication implements a distributed failure detector to find and report servers that have failed or are no longer participating in the group and the remaining members in the group coordinate to reconfigure the membership.

Fault Tolerance

Group Replication builds on an in-house implementation of the popular Paxos distributed algorithm to provide distributed coordination between servers. In order for a group to continue to function, it requires a majority of the members to be online and for them to form an agreement on every change. This allows your MySQL databases to safely continue to operate without manual intervention when failures occur, without the risk of data loss or data corruption.

Self-Healing

If a server joins the group, it will automatically bring itself up to date by synchronizing its state from an existing member. In the event that a server leaves the group, for instance it was taken down for maintenance, the remaining servers will see that it has left and will reconfigure the group automatically. When that server later rejoins the group, it will automatically re-synchronize with the group again.

Monitoring

Performance Schema tables provide clear and detailed information and statistics on individual members and for the group as a whole.

Additional Resources

• Documentation
• Presentation
• Forum
• Team Blog