# Migrating to v3.9 HA

Memgraph 3.9 includes a fix for a bug in how durability is handled on
coordinators in a high-availability (HA) setup. Because of this fix, **all
coordinator data directories must be cleaned and coordinators must be started
fresh**.

Data instances are not affected and can stay unchanged.

> **Warning**
>
> Upgrading to 3.9 **cannot be performed without downtime** if you are using a
> routing connection to connect to the currently writable instance. You can,
> however, keep using direct Bolt connections to data instances during this
> maintenance period.

### Clean coordinator data directories

Remove the data directories of all coordinators. If you are running Memgraph HA
in Kubernetes, delete the coordinators' PVCs and PVs before the restart.

### Restart coordinators

Start all coordinators with the new Memgraph 3.9 version.

### Reconnect the cluster

After the coordinators are running, reconnect your cluster by executing
`REGISTER INSTANCE` and `ADD COORDINATOR` queries for each data instance and
coordinator respectively.