Kubernetes

Production-grade container orchestration using Kubernetes Deployments and Services.

The Kubernetes provider maps Ctrl Plane operations to native Kubernetes resources — Deployments, Services, ConfigMaps, and Secrets — for production-grade container orchestration.

Status

Interface defined — the package structure exists at provider/kubernetes and implements the provider.Provider interface. Full implementation is in progress.

Configuration

import "github.com/xraph/ctrlplane/provider/kubernetes"

prov, err := kubernetes.New(kubernetes.Config{
    Kubeconfig: "/path/to/kubeconfig",
    Namespace:  "ctrlplane",
    InCluster:  false,
})

Field	Env	Default	Description
`Kubeconfig`	`CP_K8S_KUBECONFIG`	—	Path to kubeconfig file
`Namespace`	`CP_K8S_NAMESPACE`	`default`	Target namespace for resources
`InCluster`	`CP_K8S_IN_CLUSTER`	`false`	Use in-cluster service account
`Labels`	`CP_K8S_LABELS`	—	Default labels applied to all resources

Capabilities

Capability	Supported
`provision`	Yes
`deploy`	Yes
`scale`	Yes
`logs`	Yes
`exec`	Yes
`volumes`	Yes
`gpu`	Yes
`rolling`	Yes
`blue_green`	Yes
`canary`	Yes
`autoscale`	Yes
`custom_domains`	Via Ingress
`tls`	Via cert-manager

Resource mapping

Ctrl Plane concept	Kubernetes resource
Instance	Deployment + Service
Environment variables	ConfigMap + Secret
Ports	Service ports
Resources (CPU/Memory)	Resource requests/limits
Custom domains	Ingress rules
TLS certificates	cert-manager Certificate
Health checks	Liveness/readiness probes
Scaling	HPA (Horizontal Pod Autoscaler)

Deployment strategies

The Kubernetes provider supports all three deployment strategies:

Rolling — maps to Kubernetes RollingUpdate strategy with configurable maxSurge and maxUnavailable.
Blue-Green — creates a new Deployment, verifies health, then switches the Service selector.
Canary — runs the new version alongside the old with weighted traffic splitting.

When to use

Production multi-node deployments
Environments that need auto-scaling, self-healing, and rolling updates
Teams already running Kubernetes clusters
Workloads requiring GPU access or persistent volumes