Cluster
Kubernetes cluster configuration — version, instance types, node scaling, Karpenter, and cluster admins.
Cluster
The cluster section configures the Kubernetes control plane and worker nodes.
Fields
| Field | Type | Description |
|---|---|---|
| K8s Version | Select | Provider-specific versions (e.g., 1.32 for EKS, 1.31 for GKE/AKS) |
| Instance Types | Multi-select badges | Up to 5 types, with hourly cost shown per type |
| Min Nodes | Number | Minimum node count (default: 2) |
| Desired Nodes | Number | Target node count (default: 3) |
| Max Nodes | Number | Maximum node count for autoscaling (default: 5) |
| Autoscaler | Toggle | Provider-specific autoscaler (Karpenter/Autopilot/Cluster Autoscaler) |
| Cluster Admins | Array | IAM users/emails with admin access |
Kubernetes Service
- Elastic Kubernetes Service with managed node groups
- Versions: 1.31, 1.32
- Autoscaler: Karpenter — just-in-time node provisioning based on pod requirements
- IRSA (IAM Roles for Service Accounts) configured automatically
- OIDC provider created for workload identity
- Google Kubernetes Engine with node pools
- Versions: 1.30, 1.31
- Autoscaler: GKE Autopilot — fully managed node provisioning
- Workload Identity enabled by default
- Azure Kubernetes Service with node pools
- Versions: 1.30, 1.31
- Autoscaler: Cluster Autoscaler — scales node count based on pending pods
- Azure AD integration for RBAC
Instance Types
Instance types are selected as badges with pricing shown. Up to 5 types can be selected — when using Karpenter (AWS), multiple types allow the autoscaler to choose the best fit for each workload.
| Category | AWS | GCP | Azure |
|---|---|---|---|
| General | t3.medium, t3.large | e2-medium, e2-standard-2 | Standard_B2s, Standard_D2s_v5 |
| Compute | c5.large, c5.xlarge | c2-standard-4 | Standard_F2s_v2 |
| Memory | r5.large, r5.xlarge | n2-highmem-2 | Standard_E2s_v5 |
| GPU | g4dn.xlarge | a2-highgpu-1g | NC4as_T4_v3 |
Pricing is fetched from the AWS Pricing API (or equivalent) and shown per hour next to each badge. See Cost Estimation.
Node Scaling
Three numbers control the node group size:
- Min: Floor for the autoscaler (nodes will never scale below this)
- Desired: Initial node count at creation
- Max: Ceiling for the autoscaler (nodes will never scale above this)
Validation: min ≤ desired ≤ max.
Cluster Admins
An array of identities with admin access to the cluster:
IAM user ARNs or IAM user names. The form can look up IAM users from the cached resources (discovered via resource refresh). Admins are added as EKS access entries with AmazonEKSClusterAdminPolicy.
Google email addresses granted roles/container.admin.
Azure AD Object IDs granted cluster admin RBAC.