Skip to main content
Token App runs in your Kubernetes cluster, or as one container if you do not run Kubernetes. The image and the Helm chart are published together to a private Docker Hub organization. Cantor8 gives you a read-only organization access token for both. The chart version and the image version are one release. Chart 1.2.3 runs cantor8technologies/token-app:v1.2.3. Install the version Cantor8 names. Set the image tag yourself only when Cantor8 asks you to. Cantor8 runs the Registry. You receive a base URL, an organization prefix, and a client credential for the admin API.

What you have before install

Pull access

The Docker Hub user is the organization name cantor8technologies.
The chart expects a pull secret named cantor8-dockerhub. If you mirror the image into your own registry, set image.repository and imagePullSecrets to match that mirror.

Secrets

External Secrets is fine; the secret names and keys must match credentials.existingSecret, database.existingSecret, and database.dsnKey.

Values

Create values.yaml. The config block is the Token App configuration Cantor8 walks through with you at launch. You name the party for each role. The example below uses one id throughout; use a different id wherever your arrangement does.
Token App runs the admin module only. Rewards need no Token App configuration. Rewards covers them. config.scanApiBaseUrls overrides the Canton Scan endpoints used to price traffic. The chart’s defaults are the public mainnet endpoints. Override them for a test network.

Install and upgrade

The chart creates a ClusterIP Service on port 8002. It does not create an Ingress. Add a private route only for callers inside your network, and leave apiAuth.enabled on.
helm uninstall removes the workload. The database and the secrets remain.

How the process behaves

The chart requests 2 CPU and 10 GiB of memory, and limits the pod to 4 CPU and 16 GiB. Keep these defaults unless Cantor8 gives you different figures. Observability is what to alert on.

Log shipping

The chart can send Token App logs to Cantor8. It is off by default, and Token App does not depend on it. Metrics stay on GET /metrics. When it is on, the chart adds one Grafana Alloy pod, a ServiceAccount, and a Role. The Role can read pods and pods/log in the release namespace only. Alloy reads the logs of the token-app container through the Kubernetes API. It needs no host path and no root. It needs outbound HTTPS to the push URL. The chart has no proxy setting. Alloy does not ship /healthz request lines. Install Token App in a namespace of its own. The Role cannot be limited to one pod, so Alloy can read every pod and pod log in that namespace.
The logs include party ids, amounts, command ids, and the error bodies the participant returns. When you turn shipping on, Alloy also sends the lines the container has already logged, not only new lines.
Cantor8 gives you the push URL, the install id, and the credentials. Create the secret:
Add this to your values file:
Keep installId unique across all installs. Do not change it on upgrades. Alloy reads the credentials when it starts. If they change, update the secret and restart Alloy:
Shipping is best effort. If a push still fails after retries, Alloy drops that batch, and the Token App pod stays healthy. To stop shipping, set observability.enabled: false and upgrade. This removes the Alloy pod, the ServiceAccount, and the Role. Log shipping does not replace your monitoring. You still monitor Token App. Observability lists the signals.

Manual install

Use the same image, the same configuration, and one process. The chart is the supported packaging of the variables below; a manual run sets them in the environment.
Run one container, publish port 8002 on your private network, and set: Leave the poll intervals at the chart defaults unless Cantor8 gives you different ones. Do not run a second container against the same database. Log shipping is a chart feature. A manual run has no Alloy pod. Point a private probe at http://<container>:8002/healthz and a private scraper at /metrics.