Core concepts

Clusters

A Kubernetes Cluster is a set of nodes on which you deploy your applications. One of the nodes is the Control Plane, and the remaining nodes are workers.

Deckrun creates managed clusters, which means that the Control Plane is fully managed by the Cloud provider and is transparent to the user. This brings multiple advantages since there is no need to manage how the installation or upgrade processes of the Cluster are performed.

How to create a Cluster in Deckrun

Create a new Cluster using the CLI with the following command:

deck cluster create

Cluster name

Choose a name for the new Cluster. How you name your Clusters is a decision for each team.

One option would be to have a cluster for each project. Another option, recommended if you want to separate resources even further, would be to have a cluster for each environment as well.

The Cluster name could include both the project name and the environment name. If your team also operates in multiple geographic locations, you could add the region and/or zone name to the cluster name.

Some examples of names could be:

  • myproject-prod
  • myproject-staging
  • myproject-dev
  • myproject-nyc-prod
  • myproject-ams-prod

Provider

Choose one of the Providers you have created previously. Deckrun will use the associated credentials to create a Cluster in this Provider.

Region and zone

Depending on the project’s geographic requirements, select the region (and, depending on the specific provider, the zone) where Deckrun will create the Cluster.

Number of nodes

This number determines how many nodes will make up the Kubernetes Cluster. Initially, it is advisable to start with a small number and increase it as needed. The minimum number of nodes to create a Cluster is 2.

Node size

The node size determines the type of CPU, the number of vCPUs, and the amount of RAM that the node has. Also, depending on the provider, it may determine the disk size and speed.

Disk type

For providers where the disk is independent of the node size, this parameter determines the disk’s read and write speed.

Disk size

For providers where the disk is independent of the node size, this parameter determines the disk's capacity.

Cluster scaling

Configure your cluster's capacity in two ways:

  • Fixed node count: Set a specific number of nodes that remains constant
  • Auto-scaling: Define minimum and maximum nodes, and let the cluster scale automatically based on workload

Auto-scaling adjusts the number of nodes up or down based on resource demands, helping optimize costs while ensuring capacity during traffic spikes.

Creating a Cluster without prompts

Every question deck clusters create asks can be answered with a flag, so you can create a Cluster from a script, CI or an AI agent in one command:

deck clusters create myproject-prod --provider <provider-id> --region fra1 --size s-2vcpu-4gb --nodes 2 --yes
  • --nodes <n> creates a fixed-size Cluster. --min-nodes <n> --max-nodes <n> creates an auto-scaling one; --auto-scale is implied, and you can also pass --auto-scale or --auto-scale=false explicitly.
  • --custom-vpc=false creates a new VPC without asking. --vpc-id <id> uses an existing one (and, on Scaleway, --private-network-id <id>).
  • --yes (-y) skips the final confirmation in a terminal. Without a terminal there is no confirmation.
  • --no-wait returns as soon as the creation is accepted, instead of waiting for the Cluster to be ready, which can take several minutes. Follow its status with deck clusters list.

When a required value is missing and there is no terminal, the command fails and names the flag to pass.

Private networking

Clusters are provisioned with private networking enabled. This keeps internal traffic between your applications and services secure and isolated from the public internet. Communication between nodes happens over a private network that is not accessible externally.

Previous
Providers