(howto-shutdown)=
# How to shut down a machine

## Stop or live-migrate all instances on the cluster member

To shut down a machine that is a MicroCloud cluster member, first ensure that it is not hosting any running LXD instances.

You can stop all instances on a cluster member using the command:

```
lxc stop --all
```

Alternatively, for instances that can be {ref}`live-migrated <lxd:live-migration>`, you can move them to another cluster member without stopping them. See: {ref}`lxd:move-instances` for more information.

You can also temporarily move all instances on a machine to another cluster member by using cluster evacuation, then restore them after you restart. This method can live-migrate eligible instances; instances that cannot be live-migrated are automatically stopped and restarted. See: {ref}`lxd:cluster-evacuate` for more information.

## Enforce services shutdown and restart order

During the shutdown process of a MicroCloud cluster member, the LXD service must stop _before_ the MicroCeph and MicroOVN services. At restart, the LXD service must start _after_ MicroCeph and MicroOVN. This order ensures that LXD does not run into issues due to unavailable storage or networking services.

To enforce this shutdown and restart order, create a configuration file in each cluster member's `/etc/systemd/system/snap.lxd.daemon.service.d` directory to override the behavior of `snap.lxd.daemon.service`. To simplify creating the directory and configuration file, you can copy and paste the following commands into each cluster member:

```
# Create the directory if it doesn't exist
sudo mkdir -p /etc/systemd/system/snap.lxd.daemon.service.d

# Create the configuration file
cat << EOF | sudo tee /etc/systemd/system/snap.lxd.daemon.service.d/lxd-shutdown.conf
# Makes sure the LXD daemon stops before Ceph/OVN and restarts after Ceph/OVN
[Unit]
After=snap.microceph.daemon.service
After=snap.microovn.daemon.service
EOF

# Reload systemd daemon
sudo systemctl daemon-reload
```

You only need to perform this step once for each cluster member. Afterwards, the `snap.lxd.daemon.service` respects this configuration at every shutdown and restart.

### Shut down

Once you have completed the steps above, you can safely shut down and restart the machine as normal.