When somebody wants to migrate from a single MySQL instance to a full HA solution using MySQL InnoDB Cluster, the best solution to reduce the downtime is to use asynchronous replication and switch database only once at a certain point in time when everything is ready. This is almost what I explained already in this post.
The most difficult part was related to the provisioning of the existing data to the new cluster members. A backup (physical or logical) was required. It should have been restored on every nodes and we had to be sure to not mess up with the GTIDs.
This is not more the case since MySQL 8.0.17 ! Now we can use the CLONE plugin to start the cluster provisioning too.
The current situation
…
[Read more]