A MySQL replication topology can be very complex (never underestimate a DBA’s creativity). But it’s very frequent to use an asynchronous replica from your primary database to run reporting or logical backup… or any kind of read workload you need. It can also be used as delay slave for data restore purpose.
Once you decided to provide HA to your primary database by migrating to MySQL InnoDB Cluster, you can of course still need and use one or more asynchronous slaves.
You have then an architecture that looks like this :
So as you can notice, your asynchronous slave needs to pick one of the members of the cluster as master. However when this node that act as master crashes (or is stopped for maintenance) what’s happening ?
In fact, replication will break and a human intervention (or a script) will have to change the master on the replica and restart replication to keep the changes again:
As InnoDB Cluster is using GTIDs, this operation is easy, but it would be even better if this change could be done automatically more quickly and without human intervention.
Thanks to the router this is possible ! The plan is to run MySQL Router on the slave itself and connect the replication locally.
This is how it looks like :
The process is then fully automated and no human intervention is required. So the router is very useful to connect applications to the cluster but not only. Of course multiple replicas can use the same router or have their own router to replicate from the InnoDB Cluster.