A managed cluster is for teams running multiple applications and environments. Running a single application that needs to be looked after? A CaffeineStack from €300/month is the better start.
Self-managed from €100 · managed for you from €300 · your own platform from €2,500.
We deliver Talos-based dual-cluster setups (staging and production) on a predictable monthly invoice. The package covers GitOps automation, observability, backups, and incident response, with a 99.9% monthly SLA. The Advanced service level is part of the cluster base: for P1 incidents on production we also respond outside office hours.
Need more capacity? The base grows with you. We absorb temporary spikes; structural growth we plan together.
Is Kubernetes right for you?
Kubernetes is the right choice when you run multiple apps and services that need orchestration, want container-native workflows, and need auto-scaling, rolling updates, and service discovery. Your team is ready to embrace cloud-native patterns and wants the flexibility containers offer.
Not there yet? For simpler setups with one or two apps, our Managed Servers from €300/month are often a better fit. They start faster and scale easily as you grow.
Want Kubernetes benefits without running a cluster? Pods from €100/month gives you a namespace on our managed cluster.
What you get
From €2,500/month you get a complete dual-cluster setup. Setup is a one-time €2,500: architecture session, cluster build, migration support, monitoring and backup onboarding, and a documented recovery test with runbook. The term is 12 months.
| Cluster | Nodes | Purpose |
|---|---|---|
| Staging | 3× control plane 3× workers |
Updates and changes are tested here first |
| Production | 3× control plane 3× workers |
Where your production workloads run |
The cluster base includes 72 vCPU, 176 GB of memory, 2 TB of NVMe, and 2 TB of protected capacity storage.
We absorb temporary spikes. The base grows with you; structural expansion is planned together.
What "managed" means here
Managed Kubernetes is a stretchy term. Some providers sell a managed cluster for under €100 per month. Read further and "managed" stops at the control plane. Patching worker nodes, setting up ingress and TLS, building the monitoring stack, arranging backups, waking up when something falls over: that's on you, or it costs separate consulting hours.
We draw the line a few layers higher:
| Layer | Control-plane-only providers | CoffeeSprout Managed Cluster |
|---|---|---|
| Control plane | Managed | Managed |
| Worker nodes | You patch, upgrade, and scale | We patch, upgrade, and scale |
| Platform (ingress, TLS, monitoring, backups) | Build and maintain it yourself | Included, managed by us |
| Incidents | Your pager | We respond; for P1 incidents on production also outside office hours |
That's the price difference. You're not comparing a cluster to a cluster, you're comparing a control plane to a team that keeps your platform running.
Without Kubernetes, too
The same dedicated platform also comes as a managed VM environment on Proxmox. Same base capacity (72 vCPU, 176 GB of memory, 2 TB of NVMe, and 2 TB of protected capacity storage), same Advanced service level, same price and setup. Only the shape differs: virtual machines instead of containers.
Kubernetes is the default for container-native teams. The VM variant fits estates that run on virtual machines: CI runners, GitLab, legacy services, or mixed workloads. Which shape fits your platform is something we settle together in the architecture session.
Predictable costs, no surprises
With AWS or Azure you pay separately for every NAT gateway, every gigabyte of bandwidth, and every load balancer. We work differently:
- Fixed monthly price: from €2,500/month for your dual-cluster foundation, Advanced service level and support included.
- One-time setup: €2,500 for the architecture session, cluster build, migration support, monitoring and backup onboarding, and a documented recovery test with runbook. It's right here on the page, no surprise afterwards.
- Load balancer and traffic included: 2 TB of NVMe and 2 TB of protected capacity storage come with the cluster base. Expansion at published prices: NVMe €25 per 100 GB, protected capacity storage €15 per 100 GB.
- 12-month term: The one-time setup and the hardware we commit for you justify a year.
- Work outside scope: A fixed price per job, up front, based on our €120 hourly rate. You know the amount before we start.
- Prices are on the site: What you read here is the starting point. Your workloads determine the final scope, but you always see the cost up front. No three-call quote process before you see a number.
How it works
Dual-cluster approach: Updates and changes go to staging first. When staging runs stable, we roll to production. How long we wait depends on the impact of the change. No surprises, no late-night incidents.
GitOps: Your infrastructure and deployments live in Git. We use Argo CD for automated rollouts. Every change is traceable, every deploy is repeatable.
Monitoring & observability: Prometheus, Grafana, and OpenTelemetry built in. You see what's running, where resources go, and where bottlenecks are. Dashboards for your team, alerts to whoever you want.
Support and management: We upgrade Kubernetes, patch nodes, monitor cluster health, and are there when something goes wrong. Advanced is part of the cluster base: for P1 incidents on production we also respond outside office hours. Routine requests are handled within a working day.
Why these technical choices?
Talos Linux: We use Talos as the OS for Kubernetes nodes. Talos is immutable, API-driven, and secure by default. Built specifically for Kubernetes without legacy OS baggage. That means smaller attack surface, predictable updates, and no manual SSH sessions.
Dual-cluster approach: Two identical clusters (staging and production) ensure updates are tested before hitting production. Staging runs the same config, same versions, same workloads. If it works there, it works in production.
Minimum cluster sizing: 3 control plane nodes and 3 worker nodes is the minimum viable HA setup. Add more workers as workload grows. This keeps costs predictable without compromising reliability.
What else is included
- Managed ingress & TLS: Certificate management, SSL termination, and routing to your services. Automatic cert renewal via Let's Encrypt or your own CA.
- Secure access: VPN or dedicated circuits to your cluster. RBAC configured, identity integration via Keycloak, and signed container pipelines.
- Disaster recovery: Backup strategies and tested recovery procedures, with runbooks for blue/green and canary releases. The measured recovery times are below, under Continuity.
- Capacity planning: Quarterly reviews with buffer for seasonal peaks. We plan growth together, no surprise charges.
- CI/CD integration: Buildpacks, Maven/Gradle pipelines, and artifact signing. We help integrate your pipeline with GitOps.
Continuity
- Remote recovery, included: Backups run twice a day and are automatically verified. Recovery is measured, not estimated: a full cluster restore in 49 minutes, individual VMs in 1.5 to 6 minutes. Read how we measured this.
- Standby in a second Dutch location: Available for platforms that need to keep running when a datacenter fails. We test failover at least once a year, and the test date goes in your annual report. The standby price depends on your environment; you get it up front as a fixed monthly amount.
- Fully dual-location: Your platform active in two locations, available as an option.
Java advantages
Java is where we're at home, so you get extras:
- JDK base images: Curated images with Temurin and OpenJDK. Automatic updates, triggered rollouts, tested in staging first.
- JVM tuning: Liveness checks and auto-scaling profiles tailored to Spring Boot, Quarkus, and Jakarta EE.
- Observability: JVM metrics, request traces, and log correlation. Dashboards with heap usage, GC stats, and thread dumps ready to go.
Who is this for?
Java teams without an SRE department: You want Kubernetes but not the hassle of cluster management. We run the infrastructure, you build features.
Agencies with multiple clients: One cluster for all your Java projects, with namespace isolation and predictable costs. No surprises when a client suddenly needs more resources. Building for clients? See how we work with shops.
Scale-ups outgrowing simple setups: You've outgrown managed servers but don't want to hire an SRE team yet. This is the middle ground.
