Clusters

A cluster is the physical infrastructure on which your projects are deployed. Fransys provisions and manages Kubernetes clusters on European cloud providers — you only interact with a simplified configuration layer, without ever touching Kubernetes directly.

A cluster can host one or more projects. It is composed of node pools, each node having a defined amount of CPU and RAM.

Create a cluster

From the Clusters page of the dashboard, click Create cluster.

A wizard guides you through a few questions to optimally pre-configure your cluster.

Step 1 — Environment criticality

The wizard asks what your goal is for this environment:

• Minimize costs — Accept occasional interruptions (ideal for development or testing)
• Balance cost and reliability — Stable infrastructure at a reasonable budget (staging, pre-production)
• Maximize availability and performance — Even at a higher cost (critical production)

This choice influences the configuration recommendations offered afterwards (node size, number of replicas, etc.).

Step 2 — GPU requirements

The wizard asks if you need to run GPU workloads (machine learning, AI inference, etc.).

• Yes — Only providers and regions offering GPU instances will be displayed
• No — All standard CPU options remain available

tip

If you don't need GPU now, you can always add a GPU node pool later by modifying your cluster configuration.

Step 3 — Location

Choose the country where you want to host your cluster. You can select a specific country from the dropdown menu, or click No, I don't need a specific country to see all available options.

This choice is important for data sovereignty (GDPR compliance) and latency (proximity to your users).

Choose a provider and region

After the wizard, you access the interactive map displaying all available regions based on your filters.

What you see

• On the map — Available regions with their starting price, displayed by country
• In the left sidebar — The detailed list of hosting areas, each showing:
  • The datacenter country and city
  • The provider (Hetzner, Scaleway, OVH)
  • GPU availability
  • Starting price

Available filters

At the top of the screen, several filters let you refine your search:

Filter	Description
Criticality	Development, Stable, Production — filters regions suited to your criticality level
From / To (€)	Monthly price range
Provider	Filter by specific provider
Require GPU	Show only regions with GPU instances
Renewable energy	Coming soon
Certifications	Coming soon

Select a region on the map or in the sidebar, then click Continue.

Configure the cluster

Once the region is chosen, you configure the composition of your cluster. Two modes are available:

AutoPilot mode

AutoPilot mode is enabled by default. Fransys automatically manages the composition and scaling of your node pools based on load.

In AutoPilot mode, you only define a maximum monthly budget (in euros). Fransys handles the rest:

• Automatic selection of the most suitable node types
• Adding and removing nodes based on load
• Cost/performance ratio optimization

You can optionally click Edit node pool to customize the node types used by autopilot.

tip

AutoPilot mode is recommended for the majority of use cases. It saves you from manually sizing your infrastructure and adapts to load in real time.

Manual mode

Toggle the switch to Manual to configure the composition of your node pools yourself.

In manual mode:

1. Click + Add node pool to add a pool
2. Select the type of worker nodes — i.e., the hardware configuration of each node (CPU, RAM)

Available node types vary by provider and region. Each option shows:

• The number of shared or dedicated CPUs
• The amount of RAM
• The monthly price per node

The estimated total cost is displayed at the bottom left. Click Price details to see the breakdown per resource.

Deploy the cluster

Once the configuration is complete, click Deploy cluster. Deployment starts immediately.

Your cluster appears in the list with the Deploying status. Full deployment can take up to 15 minutes — this is normal. During this time, Fransys provisions the servers, installs and configures the entire Kubernetes cluster (network, ingress, monitoring, etc.) before bringing it online.

info

During deployment, the cluster status may show Degraded — this is expected behavior. The cluster is being configured and not all components are operational yet. Wait for the full 15 minutes before considering an issue. Once configuration is complete, the status automatically changes to Online.

Once the deployment is complete, your cluster is ready to host projects.

Manage an existing cluster

From the cluster list, a dropdown menu (icon ⋮) is available on each cluster. It offers:

• Edit config — Modify the cluster configuration (node pools, autopilot/manual mode, budget)
• Undeploy — Remove the cluster from production (associated projects go offline)
• Delete — Permanently delete the cluster

caution

Undeploying a cluster takes offline all projects deployed on it. Make sure no critical application depends on this cluster before proceeding.

Best practices

• Use AutoPilot mode to start — it adapts to your needs without complex configuration. Switch to manual mode only if you have specific sizing constraints.
• Choose the region closest to your users to minimize latency.
• Match criticality to your usage — don't over-size a development cluster, and don't under-size a production cluster.
• A cluster can host multiple projects — pool resources when appropriate to optimize costs.