Create a cluster
A cluster is where your deployments run. Ownkube provisions two shapes in your AWS account: K3s (Starter) (K3s on one AWS instance, cheap and fast) and EKS (Production) (managed EKS, highly available across availability zones).
Start at your dashboard.
Before you start
Section titled “Before you start”- A connected AWS account. You can’t create a cluster without one.
- An idea of what you’ll run on it. Dev, staging, or a side project? Starter. Production with real traffic? Production.
The two flows
Section titled “The two flows”A single EC2 instance running lightweight Kubernetes. Free Ownkube platform fee. You only pay AWS for the EC2 instance, and AWS Activate or other startup credits cover that in full. Provisions in about 5 minutes. Best for dev, staging, side projects, and small production workloads.
-
Start the wizard
Open the create-cluster wizard from your dashboard and select Deploy on EC2.
-
Name and target
Pick a cluster name (e.g.
side-project,dev), your AWS account, and a region. Names are lowercase, alphanumeric, with dashes. -
Pick an instance type
Instance vCPU Memory ~EC2 cost Good for t3.medium2 4 GiB ~$30 / mo First apps, light prototypes t3.large2 8 GiB ~$60 / mo Small production web + db t3.xlarge4 16 GiB ~$120 / mo Multi-service workloads t3.2xlarge8 32 GiB ~$240 / mo Busy workloads on one node No Ownkube platform fee on top, you pay AWS only. AWS Activate or other startup credits cover these EC2 costs in full.
-
Review and create
Starter clusters skip the separate networking and compute steps. Confirm the summary and click Create cluster. Provisioning takes about 5 minutes.
Managed Kubernetes with node autoscaling, highly available across availability zones. $5 / vCPU + $1 / GB RAM per month, metered per second on requested resources. Provisions in about 10 minutes.
-
Start the wizard
Open the create-cluster wizard from your dashboard and select Deploy on EKS.
-
Details
Pick a cluster name, your AWS account, region, and Kubernetes version. Defaults are typically fine.
-
Networking
Accept defaults unless your VPC has conflicts.
- Pods CIDR, default
10.244.0.0/16 - Services CIDR, default
10.96.0.0/12
Change these only if they’d clash with something you’re already running.
- Pods CIDR, default
-
Compute
Drag Workload vCPUs and Memory to your starting size. Defaults: 8 vCPU / 16 GiB. Autoscaling adds and removes nodes automatically once the cluster is live. You’re picking the starting capacity.
-
Review and create
Confirm the review summary with estimated monthly cost and click Create cluster. Provisioning takes about 10 minutes.
What happens during provisioning
Section titled “What happens during provisioning”-
Account setup runs once per AWS account
On your first cluster, Ownkube creates shared IAM roles and a regional S3 bucket for state. This takes about a minute and only happens once per AWS account.
-
VPC and networking
A new VPC, subnets (public and private across AZs for EKS), NAT, and security groups are created for the cluster.
-
Cluster control plane
For EKS, the managed control plane is provisioned. For K3s, the EC2 instance is launched and bootstrapped.
-
Platform components
Ownkube installs the internal components that make the platform work: deployment pipeline, public routing, TLS, metrics collection.
-
Cluster becomes active
The cluster status flips to active on the dashboard. You can start creating deployments.
After creation
Section titled “After creation”Once active, the cluster detail page in your dashboard lets you:
- Review telemetry: node count, vCPU, memory, active deployments
- Adjust compute capacity on EKS
- View the cluster’s IAM role and VPC details
- Upgrade the platform version when new versions are available
- Delete the cluster when you’re done with it
See Clusters for the full reference.
Limits and constraints
Section titled “Limits and constraints”- One region per cluster. Clusters can’t span regions. Create separate clusters in separate regions.
- Cluster types can’t convert. You can’t turn a K3s cluster into EKS. Create a new EKS cluster and move deployments.
- Plan limits cap the number of clusters and total vCPU / RAM. See Pricing for tier limits.