(howto-remove)=
# How to remove a machine

If you want to remove a machine from the MicroCloud cluster after the initialisation, use the {command}`microcloud remove` command:

    sudo microcloud remove <name>

Before removing the machine, ensure that there are no LXD instances or storage volumes located on the machine, and that there are no MicroCeph OSDs located on the machine.

See {ref}`how to remove instances <lxd:instances-manage-delete>` in the LXD documentation.
See {doc}`how to remove OSDs <microceph:how-to/remove-disk>` in the MicroCeph documentation.

```{note}
If local storage was created, MicroCloud will have also added some default storage volumes that will need to be cleaned up:

    lxc config unset storage.images_volume --target <name>
    lxc config unset storage.backups_volume --target <name>
    lxc storage volume delete local images --target <name>
    lxc storage volume delete local backups --target <name>

Any additional storage volumes belonging to this machine must also be deleted before removal without the `--force` flag.
```

If the machine is no longer reachable over the network, you can also add the `--force` flag to bypass removal restrictions and skip attempting to clean up the machine. Note that MicroCeph requires `--force` to be used if the remaining cluster size will be less than 3.

```{caution}
Removing a cluster member with `--force` will not attempt to perform any clean-up of the removed machine. All services will need to be fully re-installed before they can be re-initialised. Resources allocated to the MicroCloud like disks and network interfaces may need to be re-initialised as well.
```

## Reducing the cluster to 1 machine

When shrinking the cluster down to 1 machine, you must also clean up the Ceph monmap before proceeding, even when using the `--force` flag.

    sudo microceph.ceph mon remove <name>
    sudo microceph cluster sql "delete from services where member_id = (select id from core_cluster_members where name='<name>') and service='mon'"
    sudo microcloud remove <name> --force

If the machine is no longer reachable and Ceph is no longer responsive, see the [Ceph documentation](https://docs.ceph.com/en/squid/rados/operations/add-or-rm-mons/#removing-monitors-from-an-unhealthy-cluster) for more recovery steps.