Synthesize a cluster

This explanation goes into the details that are omitted in synthesizing a cluster.

Creating a new cluster on

There’s quite a few fields to fill out when creating a cluster.


The cluster ID is auto-generated if the field is left empty. Optionally, you can provide a unique ID for the cluster. If the field is filled in, the value will override the auto-generated ID. This field is limited to lower-case characters a-z, numbers 0-9 and -. The length is limited to 1-61 characters.

This field cannot be changed after creation, it’s used as the Kubernetes object name for the cluster by Lieutenant.
Display Name

Choose a descriptive name. Check first on the cluster listing overview how other clusters of the tenant are named.

Lieutenant Tenant

Choose the tenant to which this cluster belongs

Service Level (Fact: service_level)

Check with Service Manager which Service Level applies to this cluster


Check folder cloud in the Commodore defaults repo for already defined clouds.

Use value none for on-premise setups.

If you set this field to <name>, there must be a file cloud/<name>.yml in the Commodore defaults repo. See the how-to on creating a missing cloud for some notes on adding a new cloud.


Assuming that the value of field "Cloud" is <cloudname>, check folder cloud/<cloudname> in the Commodore defaults repo for already defined regions for the cloud.

Field "Region" is not required to be set for "Cloud" none.

If you set this field to <region>, there must be a file cloud/<cloudname>/<region>.yml in the Commodore defaults repo. See the how-to on creating a missing cloud region for some notes on adding a new region to a cloud.


Check folder distribution in the Commodore defaults repo for already defined K8s distributions.

If you set this field to <dist>, there must be a file distribution/<dist>.yml in the Commodore defaults repo. See the how-to on creating a missing distribution for some notes on defining a new K8s distribution.

Rancher Cluster ID (Fact: rancher_cluster_id)

If it’s a Rancher cluster, add the cluster ID from the Rancher Management Server to this field

Repository Type

Leave at "auto" to have Lieutenant manage the cluster catalog repo on

GitRepo URL

Leave empty if "Repository Type" is auto. Check with the Project Syn IG, if you have special requirements for the cluster catalog repo.

Not all combinations of attributes work smoothly. If in doubt, seek help from the Project Syn IG.

Running Commodore and adding the initial cluster configuration

After running Commodore you have a local environment where you can edit your cluster’s configuration. You can add initial configuration in the <cluster-id>.yml file in the tenant’s Project Syn configuration repository (often called the "Syn tenant repo"). The Syn tenant repo is cloned to inventory/classes/<tenant-id> by Commodore.

See for a list of tenants and their associated configuration repositories.

Adding required tokens to Vault

The "Clusters" view on has a button which directly links to the cluster’s secrets in Vault on the right-hand side.

This section breaks down sources which may inform you about required tokens for the cluster you’re synthesizing.

In general, we recommend that you first check the following documentations for required secrets.

  • The installation documentation for the K8s distribution of your cluster

  • Tenant-specific documentation for the cluster you’re synthesizing


All distributions require tokens for K8up and the cluster-backup component. Make sure you’ve got the Vault CLI available locally and run the following commands to create those tokens.

export VAULT_ADDR=
export TENANT_ID=<tenant id> (1)
export CLUSTER_ID=<cluster id> (2)

vault login -method=oidc

# Configure global restic password for K8up
vault kv put -cas=0 clusters/kv/${TENANT_ID}/${CLUSTER_ID}/global-backup password=$(pwgen -s 32 1)

# Configure restic password for cluster backups
vault kv put -cas=0 clusters/kv/${TENANT_ID}/${CLUSTER_ID}/cluster-backup password=$(pwgen -s 32 1)
1 You can find the tenant ID on
2 The cluster ID which you chose in step 1a

List all secret references present in the cluster catalog

You can find all the secrets referenced in the cluster catalog with the command below. This command assumes that you’re currently in a directory which holds the cluster’s catalog repo in subdirectory catalog. This will be the case if you’re currently in the directory in which you’ve ran Commodore to compile the cluster catalog.

find catalog/refs/ -type f -printf "clusters/kv/%P\n" | sed -r 's#(.*)/#\1 #'

The output of the command is formatted to match the location of the secrets in the Project Syn Vault. The first value on each line indicates the secret in Vault, while the second value on the line indicates the key in that secret.

If you ensure that all the secrets and keys listed by that command are present in the Project Syn Vault, there should not be any unwelcome surprises when installing Steward on the cluster.

Installing Steward on the cluster

The kubectl create command given in the how-to will install Steward on the cluster in the namespace syn and Steward will:

  • Generate SSH key which is then being used as a deploy key to access the Git repositories managed by the GitRepo objects. This SSH key will be sent to the Lieutenant API once generated.

  • Retrieve Git repository URL from Lieutenant

  • Bootstrap Argo CD and generate the "root" Application definition to connect to the GitOps repository

The "bootstrap" Argo CD instance will reconfigure itself and also take over management of the Steward deployment. In parallel, Argo CD will deploy all the manifests which are present in the cluster catalog repo onto the cluster.

Some Argo CD applications will fail to deploy until Argo CD has reconfigured itself. This happens because the bootstrap Argo CD deployment doesn’t support revealing secret references from the Project Syn Vault.