diff --git a/README.md b/README.md
index b4fb445..c8ca1d5 100644
--- a/README.md
+++ b/README.md
@@ -108,42 +108,56 @@ Before testing the backups, you may want to shut `yggdrasil` down for extra conf
 not being accessed/modified during this process. It is easy to access `yggdrasil` by accident if
 `/etc/hosts` is not modified in the test VM, something that is easy to forget.
 
+### Baldur on Scaleway
+
 1. Create `baldur` by running:
    ```sh
    python scripts/scaleway/baldur.py create --volume-size <size-in-GB>
    ```
    Pick a volume size that's larger than what `yggdrasil` estimates for
    `rpool/var/lib/yggdrasil/data`.
-2. Provision `baldur` by running
+2. When done destroy `baldur` by running:
+    ```sh
+    python scripts/scaleway/baldur.py delete
+    ```
+
+### Baldur on Yggdrasil
+
+1. Create a VM on `yggdrasil`.
+  - Install the OS on a zvol on `rpool`.
+  - Prepare a zvol on `hpool` of size that's larger than what `yggdrasil` estimates for
+    `rpool/var/lib/yggdrasil/data` and mount at `/var/lib/baldur/data`.
+  - Create non-root user `wojtek` with `sudo` privileges.
+2. Configure SSH to use `yggdrasil` as a jump server.
+
+### Test
+
+1. Provision `baldur` by running
    ```sh
    ansible-playbook --vault-id @vault-keyring-client.py -i inventory/baldur_production playbooks/baldur.yml
    ```
-3. Restore all the backups by ssh'ing into `baldur` and running (as root):
+2. Restore all the backups by ssh'ing into `baldur` and running (as root):
    ```sh
    /usr/local/sbin/restic-batch --config-dir /etc/restic-batch.d restore
    ```
-4. Start all the pod services with:
+3. Start all the pod services with:
    ```sh
    ansible-playbook --vault-id @vault-keyring-client.py -i inventory/baldur_production playbooks/services_start.yml
    ```
    Give them some time to download all the images and start.
-5. Once the CPU returns to idling check the state of all the pod services and their `veth`
+4. Once the CPU returns to idling check the state of all the pod services and their `veth`
    interfaces. If necessary restart the affected pod. Sometimes they fail to start (presumably due
    to issues related to limited CPU and RAM).
-6. Boot into a test VM. Ideally, one installed onto a virtual disk since the live system might not
+5. Boot into a test VM. Ideally, one installed onto a virtual disk since the live system might not
    have enough space. A VM is used to make sure that none of the services on the host workstation
    connect to `baldur` by accident.
-7. Modify `/etc/hosts` in the VM to point at `baldur` for all relevant domains.
-8. Test each service manually one by one. Use the Flagfox add-on to verify that you are indeed
+6. Modify `/etc/hosts` in the VM to point at `baldur` for all relevant domains.
+7. Test each service manually one by one. Use the Flagfox add-on to verify that you are indeed
    connecting to `baldur`.
-9. Stop all the pod services with:
+8. Stop all the pod services with:
    ```sh
    ansible-playbook --vault-id @vault-keyring-client.py -i inventory/baldur_production playbooks/services_stop.yml
    ```
-10. Destroy `baldur` by running:
-    ```sh
-    python scripts/scaleway/baldur.py delete
-    ```
 
 ## Music organisation