Integration with Kubernetes Clusters

Cadence.CI integrates with Kubernetes so that Mélodium runners are deployed as pods directly within your cluster. Pipeline execution stays inside your infrastructure, using your existing networking, storage, and compute policies.

How It Works

A lightweight Kubernetes Controller is deployed in your cluster. It watches for runner provisioning requests from Cadence.CI and creates the corresponding pods. When a job completes, it tears down those pods. All interactions with your cluster go exclusively through the controller.

Setup Steps

Deploy the controller: install the Cadence.CI Kubernetes Controller into your cluster. It registers itself with Cadence.CI and handles communication between Cadence.CI and your cluster’s API server.
Configure cluster access: the controller uses a service account scoped to just the permissions needed to create and delete runner pods and their supporting ServiceAccount/Role/RoleBinding objects.
Runner deployment: when a pipeline runs, Mélodium runners are created as pods in your cluster, using the namespace and resource quotas you configure.
Direct execution: pipeline steps execute inside your cluster, with direct access to cluster-internal services and storage.

Benefits

Security: data remains within your Kubernetes environment throughout execution.
Cost efficiency: runners use your existing cluster capacity directly.
Compliance: compute stays within your designated cluster, meeting data residency requirements.
Scalability: Kubernetes’ scheduler handles pod placement and resource allocation automatically.
Self-healing: if a runner pod is evicted or crashes, Kubernetes reschedules it and Cadence.CI retries the affected work.

Self-Managed vs. Melodium-Operated Clusters

Clusters registered with Cadence.CI are one of three service modes:

Self-managed: a cluster you own and operate, registered with your own cluster token. This is the model described above, everything runs on your infrastructure.
Shared: a cluster operated by Melodium, available to organizations on eligible plans. Runners still execute as isolated pods with scoped, per-job service accounts, on Melodium infrastructure shared across organizations.
Dedicated: a cluster operated by Melodium and reserved for a single organization.

The pod-level architecture, controller, scoped service accounts, cluster access limited to the controller, is the same across all three modes. Self-managed clusters run entirely within infrastructure you control; shared and dedicated clusters run on infrastructure Melodium operates on your behalf.

Available Configurations

While Cadence.CI can run on any standard Kubernetes distribution, pre-built cluster configurations are available for specific providers. See the pages in this section for provider-specific setup guides.

Integration with Cloud Providers Build a CI/CD Pipeline