7.4.6 Using MySQL Clone with InnoDB Cluster

7.4.6.1 Working with a Cluster that uses MySQL Clone

In MySQL 8.0.17, InnoDB Cluster integrates the MySQL Clone plugin to provide automatic provisioning of joining instances. The process of retrieving the cluster's data so that the instance can synchronize with the cluster is called distributed recovery. When an instance needs to recover a cluster's transactions we distinguish between the donor, which is the cluster instance that provides the data, and the receiver, which is the instance that receives the data from the donor. In previous versions, Group Replication provided only asynchronous replication to recover the transactions required for the joining instance to synchronize with the cluster so that it could join the cluster. For a cluster with a large amount of previously processed transactions it could take a long time for the new instance to recover all of the transactions before being able to join the cluster. Or a cluster which had purged GTIDs, for example as part of regular maintenance, could be missing some of the transactions required to recover the new instance. In such cases the only alternative was to manually provision the instance using tools such as MySQL Enterprise Backup, as shown in Using MySQL Enterprise Backup with Group Replication.

MySQL Clone provides an alternative way for an instance to recover the transactions required to synchronize with a cluster. Instead of relying on asynchronous replication to recover the transactions, MySQL Clone takes a snapshot of the data on the donor instance and then transfers the snapshot to the receiver.

Warning

All previous data in the receiver is destroyed during a clone operation. All MySQL settings not stored in tables are however maintained.

Once a clone operation has transferred the snapshot to the receiver, if the cluster has processed transactions while the snapshot was being transferred, asynchronous replication is used to recover any required data for the receiver to be synchronized with the cluster. This can be much more efficient than the instance recovering all of the transactions using asynchronous replication, and avoids any issues caused by purged GTIDs, enabling you to quickly provision new instances for InnoDB Cluster. For more information, see The Clone Plugin and Cloning for Distributed Recovery

In contrast to using MySQL Clone, incremental recovery is the process where an instance joining a cluster uses only asynchronous replication to recover an instance from the cluster. When an InnoDB Cluster is configured to use MySQL Clone, instances which join the cluster use either MySQL Clone or incremental recovery to recover the cluster's transactions. By default, the cluster automatically chooses the most suitable method, but you can optionally configure this behavior, for example to force cloning, which replaces any transactions already processed by the joining instance. When you are using MySQL Shell in interactive mode, the default, if the cluster is not sure it can proceed with recovery it provides an interactive prompt. This section describes the different options you are offered, and the different scenarios which influence which of the options you can choose.

In addition, the output of Cluster.status() for members in RECOVERING state includes recovery progress information to enable you to easily monitor recovery operations, whether they are using MySQL Clone or incremental recovery. InnoDB Cluster provides additional information about instances using MySQL Clone in the output of Cluster.status().