Deploying Aqueduct on Kubernetes with Helm Chart

You can use this Helm chart to deploy Aqueduct on a Kubernetes cluster. Please follow the steps below.

Prerequisites

• You need to have access to an existing Kubernetes cluster. This means your local machine needs to have a kubeconfig file (typically located in ~/.kube/config) that points to the cluster you want to run Aqueduct on.
• Make sure kubectl is installed on your local machine.
• Follow the instructions here to install Helm on your local machine.

Setting Up

1. Add the repository to your local Helm manager.

   helm repo add aqueduct https://aqueducthq.github.io/aqueduct-helm/
   

2. Sync versions available in the repo to your local Helm cache.

   helm repo update
   

3. Search for the latest stable versions of the charts found.

   helm search repo aqueduct
   

4. Install the Helm chart.

   helm install aqueduct/aqueduct --generate-name
   

Customizing Install Options

The installation command above deploys the latest version of Aqueduct in the default namespace with a Helm-generated release name. By default, the server runs with Python 3.8. You can customize all of these by running the following:

helm install <RELEASE_NAME> aqueduct/aqueduct --namespace <NAMESPACE> --version <VERSION_NUMBER> --set image.repository=<REPO_NAME>

• RELEASE_NAME: the name of the release.
• NAMESPACE: the namespace under which the Aqueduct release will be created.
• VERSION_NUMBER: the version number of Aqueduct (e.g., 0.1).
• REPO_NAME: the name of the image repository that indicates which Python version the server runs with.

Supported image repositories are:

• aqueducthq/aqueduct-py37
• aqueducthq/aqueduct-py38
• aqueducthq/aqueduct-py39
• aqueducthq/aqueduct-py310

By default, aqueducthq/aqueduct-py38 is used.

Accessing the Server

After the installation, you can get the server endpoint by running the following:

export SERVICE_IP=$(kubectl get svc <RELEASE_NAME> --template "{{ range (index .status.loadBalancer.ingress 0) }}{{.}}{{ end }}" -n <NAMESPACE>)
echo http://$SERVICE_IP:8080

The API key is printed to the log of the Aqueduct server pod:

kubectl logs <AQUEDUCT_POD_NAME> -n <NAMESPACE> | grep "API Key"