Local Kubernetes Cluster¶

Project: Arlyn Labs On-Premise Infrastructure
Date: April 2026

Overview¶

Arlyn Labs runs a local Kubernetes cluster on an on-premise Mac Studio to serve as the workspace backend for Coder. The cluster is provisioned using kind (Kubernetes IN Docker) and runs as a single-node setup inside Docker Desktop for Mac.

The Coder control plane runs natively as a macOS binary (not inside the cluster), backed by a PostgreSQL database managed via Homebrew. Kubernetes is used solely as the compute backend — Coder's Terraform templates provision workspace pods on the cluster on demand.

Hardware¶

Component	Specification
Machine	Apple Mac Studio
Chip	Apple Ultra
CPU cores	28
Unified memory	96 GB

Software Stack¶

Component	Detail
Container runtime	Docker Desktop for Mac (containerd image store enabled)
Kubernetes distribution	kind (Kubernetes IN Docker)
Kubernetes version	v1.34.3
Cluster type	Single-node

Cluster Configuration¶

The cluster runs as a single kind node named desktop-control-plane. All workloads run in the default namespace unless otherwise specified.

Setting	Value
Cluster type	kind
Nodes	1 (`desktop-control-plane`)
Kubernetes version	v1.34.3
Namespace	`default`

Docker Desktop Resource Allocation¶

Docker Desktop must be configured with sufficient resources to support the cluster and the workspace pods running inside it.

Resource	Recommended Value
Memory	16 GB
CPUs	8 cores

These values are set in Docker Desktop → Settings → Resources.

Coder Workspace Resource Limits¶

Each developer workspace runs as a pod on the cluster. The following resource requests and limits are applied per workspace pod via the Coder Terraform template:

Resource	Request	Limit
Memory	2 GB	6 GB
CPU	0.5 cores	2 cores
Storage	10–20 GB	—

The recommended auto-stop idle timeout is 30–45 minutes, which reclaims cluster resources from inactive workspaces while avoiding unnecessary cold starts.

Architecture¶

The system is split across two layers: the Coder control plane running natively on macOS, and the workspace pods running inside the Kubernetes cluster.

┌─────────────────────────────────────────────────────────────┐
│ Mac Studio (stargazer)                                      │
│                                                             │
│  ┌─────────────────┐    ┌────────────────────────────────┐  │
│  │ Coder Server    │    │ Docker Desktop                 │  │
│  │ (native binary) │    │                                │  │
│  │                 │    │  ┌──────────────────────────┐  │  │
│  │  ┌───────────┐  │    │  │ kind cluster             │  │  │
│  │  │PostgreSQL │  │    │  │ (desktop-control-plane)  │  │  │
│  │  │(Homebrew) │  │    │  │                          │  │  │
│  │  └───────────┘  │    │  │  ┌───────┐  ┌───────┐   │  │  │
│  │                 │◄──►│  │  │  Pod  │  │  Pod  │ … │  │  │
│  └─────────────────┘    │  │  └───────┘  └───────┘   │  │  │
│                         │  └──────────────────────────┘  │  │
│                         └────────────────────────────────┘  │
└─────────────────────────────────────────────────────────────┘

Useful Commands¶

Cluster health¶

kubectl get nodes
kubectl cluster-info

View all running resources¶

kubectl get all --all-namespaces

View pods¶

# All namespaces
kubectl get pods -A

# Default namespace only
kubectl get pods

Resource usage¶

kubectl top nodes
kubectl top pods -A

Deployments and services¶

kubectl get deployments
kubectl get services

Key References¶

Item	Value
Machine	Mac Studio (on-premise network)
Kubernetes distribution	kind
Kubernetes version	v1.34.3
Coder documentation	coder.com/docs
Docker Desktop	docker.com/products/docker-desktop