Skip to content
Beta

Managed Kubernetes Platform on STACKIT

Last updated on

Prodyna LogoProdyna Logo
PRODYNA

Managed Kubernetes Platform on STACKIT

Guided PRODYNA journey to a managed Kubernetes platform on STACKIT: discovery, landing zone, management cluster, fleet, workloads, day-2, and handover.

PLAN

Discovery and Target Architecture

Delivery roadmap across twelve weeks: discovery and target architecture, landing zone foundation, management cluster foundation, workload cluster fleet rollout, and handover, each with its deliverable
Delivery roadmap across twelve weeks: discovery and target architecture, landing zone foundation, management cluster foundation, workload cluster fleet rollout, and handover, each with its deliverable

A one-day discovery workshop with key stakeholders assesses the current infrastructure, workloads, and compliance requirements. Within the same week the target architecture and landing zone strategy are agreed.

  • Cluster topology, environment split, and landing zone placement.
  • Security, RBAC, and policy enforcement approach (OPA / Kyverno).
  • Observability, logging, and monitoring strategy.
  • Networking, ingress, and connectivity requirements.

Deliverable: Kickoff alignment, cloud strategy fit, and an assessment of the current platform landscape with the resulting target architecture decisions.

STEP

Management Cluster Foundation

Fleet architecture: a Git repository feeds a management cluster in the platform landing zone, which synchronizes add-ons, policy, configuration, and observability into the Dev, Test, and Prod workload clusters
Fleet architecture: a Git repository feeds a management cluster in the platform landing zone, which synchronizes add-ons, policy, configuration, and observability into the Dev, Test, and Prod workload clusters

The management cluster is deployed into a landing zone project and serves as the central control plane for the fleet. Expect roughly three to four weeks for this phase.

  • GitOps-based configuration and automated lifecycle management.
  • Distribution of Kubernetes add-ons, Helm charts, and standardized configuration.
  • CI/CD integration and governance tooling.
  • Policy enforcement and RBAC baselines applied fleet-wide.

Deliverable: Production-ready management cluster with CI/CD and governance tooling.

Choose the GitOps stack

Argo CD or Flux reconciles the desired state, and Crossplane extends the same loop to STACKIT resources so networks and databases are declared next to the workloads that use them. Which of the two engines you pick matters far less than committing to one and giving the whole fleet a single repository layout.

SAFE

Guardrails and Compliance Baseline

Define what every cluster inherits before the fleet exists: RBAC baselines, admission control through OPA or Kyverno, network policies, and the add-ons that carry ingress, secrets, backup, and observability.

Guardrails are distributed from the management cluster and reconciled continuously, so the baseline is identical in Dev, Test, and Prod and stays identical as the fleet grows. This is what makes BSI C5 and ISO 27001 operation an architectural property rather than an audit exercise.

Seven-layer cluster baseline from sovereign STACKIT infrastructure and the landing zone up through SKE, guardrails, platform add-ons, and golden paths to team workloads, with the guardrail and add-on layers enforced fleet-wide from the management cluster
Seven-layer cluster baseline from sovereign STACKIT infrastructure and the landing zone up through SKE, guardrails, platform add-ons, and golden paths to team workloads, with the guardrail and add-on layers enforced fleet-wide from the management cluster
LIFT

Workload Cluster Fleet Rollout

Initial workload clusters for Dev, Test, and Prod are provisioned, with add-ons and configuration synchronized from the management cluster. Expect roughly two to three weeks.

  • Automated cluster provisioning against the agreed baseline.
  • Configuration synchronization and drift detection from the management cluster.
  • Centralized observability wired up per cluster.
  • Optional: an Internal Developer Platform proof of concept validating portal-based onboarding.

Deliverable: Operational workload clusters with developer self-service enabled.

Developer Self-Service

A portal turns the platform into a catalog: a team picks a template and receives a repository, a pipeline, and a namespace with the baseline already applied, without filing a ticket. Worth proving with a small pilot during the rollout rather than deferring it to a later project.

AUTO

Land the First Workloads

A platform proves itself when workloads run on it. The first application walks the whole path end to end: built and tested in CI, scanned and signed, stored in the container registry, then delivered into the cluster by the same GitOps engine that governs the fleet.

The image tag is the handoff between CI and GitOps, which is why the pipeline never needs cluster credentials. The admission gate rejects anything unsigned or non-compliant before it reaches a node, promotion between Dev, Test, and Prod changes values rather than manifests, and a rollback is a reverted commit rather than a manual intervention. Once the first workload has walked this path, every team after it inherits the same route as a paved road.

Workload delivery in two lanes: the build lane runs from the source repository through the CI pipeline, scanning and signing, into the STACKIT Container Registry; the delivery lane commits the image tag to the config repository, reconciles it with GitOps on the management cluster, checks it at the admission gate, and runs it in Dev, Test, and Prod
Workload delivery in two lanes: the build lane runs from the source repository through the CI pipeline, scanning and signing, into the STACKIT Container Registry; the delivery lane commits the image tag to the config repository, reconciles it with GitOps on the management cluster, checks it at the admission gate, and runs it in Dev, Test, and Prod
OPS

Day-2 Operations and Fleet Lifecycle

Day-2 is where a platform is won or lost. The standing job is keeping the whole fleet current and provably healthy: Kubernetes versions, node images, add-on releases, certificate rotation, and the response window for new policies and CVEs.

Each of those changes is made once in the platform repository and rolled out in waves, Dev first, then Test, then Prod, with every wave gated on the previous one staying healthy. A wave that does not come up healthy stops the rollout, so the blast radius of a bad platform change is one environment rather than the whole fleet, and proving that every cluster runs the agreed version is a query against the fleet rather than a spreadsheet someone maintains by hand. Tuning an individual workload is deliberately not on this list. That belongs to the application team that owns the workload.

Day-2 in two sections: what the platform keeps current across the fleet, namely Kubernetes version, node pools and OS, platform add-ons, certificates and secrets, and policies and CVEs; and how such a change reaches the fleet, proposed in the platform repository, applied once on the management cluster, then rolled out in gated waves to Dev, Test, and Prod
Day-2 in two sections: what the platform keeps current across the fleet, namely Kubernetes version, node pools and OS, platform add-ons, certificates and secrets, and policies and CVEs; and how such a change reaches the fleet, proposed in the platform repository, applied once on the management cluster, then rolled out in gated waves to Dev, Test, and Prod
Trail historyLast updated Aug 26, 2026TMUpdatedNo updates · 1 bar = 1 week i
TMTobias M.Head of STACKIT Cloud Framework · STACKITOwnerActive 11 of the last 12 weeks · 196 updatesSTACKITwww.linkedin.com/in/tobias-müller-011304172??Name not publicThe Cloud Framework team knows who this is. The name is not shown on the site.Contributed in PRODYNA
  • Tobias M.Tobias M.Head of STACKIT Cloud Framework · STACKITOwnerActive 11 of the last 12 weeks · 196 updatesSTACKITwww.linkedin.com/in/tobias-müller-011304172 · Aug 26, 2026

  • Name not public?Name not publicThe Cloud Framework team knows who this is. The name is not shown on the site. · Aug 10, 2026