05 · Kubernetes on your laptop¶
Outcome: a three-node Kubernetes cluster (RKE2: one control plane, two workers) on private VMs behind a hardened bastion, on your own machine; your kubeconfig pointing at it; cs kubectl, cs helm and cs k9s working against it; and a first workload running. Scenarios 06 to 10 and 13 build on this cluster.
Verified live on VMware Fusion 13.6
tests/scenarios/05-local-kubernetes.sh runs every command below with CLOUDSEED_LIVE=1. Without it, it renders and validates the cluster (RKE2 and kubeadm), and checks that each cluster command says the cluster is not created yet.
| Time | Cost | Level | Runs on |
|---|---|---|---|
| ~35 min (the cluster itself ~25 min) | $0, local | Intermediate | ~8 vCPUs, 14 GB RAM, 60 GB disk free |
What you'll build¶
%%{init: {"flowchart": {"nodeSpacing": 30, "rankSpacing": 38, "padding": 12}}}%%
flowchart TB
laptop["Your laptop<br/>cs kubectl / helm / k9s<br/>~/.kube/config"]
subgraph nat["VMware NAT network"]
bastion["Bastion<br/>gateway + NAT"]
end
subgraph private["Private host-only network"]
cp["cloudseed-lab-cp1<br/>RKE2 server"]
w1["cloudseed-lab-wk1<br/>RKE2 agent"]
w2["cloudseed-lab-wk2<br/>RKE2 agent"]
end
laptop -- "SSH" --> bastion
laptop -- "Kubernetes API :6443" --> cp
bastion --> cp & w1 & w2
cp --- w1 & w2The VMs are named <name>-<env>-cp1, -wk1, ... (with the default name: cloudseed-lab-cp1). They sit on VMware's host-only network, which your machine reaches directly, so the API needs no tunnel.
Before you start¶
- Everything from 01 · Your first lab, and no other VMware environment holding VMs: every environment without
--cidrshares the built-in host-only network, one at a time. - Resources: bastion 2 vCPU / 2 GB, each node 2 vCPU / 4 GB / 40 GB (thin-provisioned). Change them with
--var kubernetes_cpus=4 --var kubernetes_memory_mb=8192.
Step 1: Install the Kubernetes tools¶
install kubernetes is kubectl + helm. Anything missing later is also offered on first use.
Step 2: Preview the cluster¶
Expected output (abbreviated)
╭─ Environment vmware-lab ───────────────────────────────────────────────╮
│ enable_kubernetes true │
│ kubernetes_distro rke2 │
│ kubernetes_version (not set) │
│ kubernetes_cis_profile false │
│ kubernetes_control_planes 1 │
│ kubernetes_workers 2 │
│ kubernetes_cpus 2 │
│ kubernetes_memory_mb 4096 │
│ kubernetes_disk_gb 40 │
╰─────────────────────────────────────────────────────────────────────────╯
$ terraform validate -no-color
Success! The configuration is valid.
✔ Dry run complete. Rendered root(s): ~/.cloudseed/envs/vmware-lab/stack
Step 3: Build the cluster¶
cs setup vmware -y --env lab --var enable_kubernetes=true --var kubernetes_distro=kubeadm --dry-run
cs setup vmware --env lab --var enable_kubernetes=true --var kubernetes_distro=kubeadm
Upstream kubeadm with containerd and Flannel (it needs 2 vCPUs and 2048 MB per node). Choose the distro when you create the cluster.
Terraform creates the bastion and the node VMs; then Ansible, run from your machine (ansible/kubernetes.yml, installed into ~/.cloudseed/venv-ansible on first use), installs RKE2 on the control plane, joins the workers, writes the kubeconfig to ~/.cloudseed/envs/vmware-lab/k8s/kubeconfig and waits until every node is Ready (at most 10 minutes), so the cluster is usable when the command ends. RKE2's stable channel is used unless you pin --var kubernetes_version=v1.36.4+rke2r1; --var kubernetes_cis_profile=true turns on RKE2's CIS hardening profile.
Step 4: Make it the current environment¶
Cluster commands (node, platform, kubectl, helm, k9s, dr, chaos, scan) now act on this cluster without vmware --env lab. cs env clear forgets the choice.
Step 5: Connect your tools¶
cs k8s info vmware --env lab
cs k8s kubeconfig vmware --env lab
kubectl get nodes -o wide
cs kubectl get pods -A
cs helm list -A
cs k9s
k8s kubeconfigmerges the cluster into~/.kube/configand makes it the current context, so plainkubectlworks too (cs undotakes the entries out again and restores your previous context).cs kubectl/cs helm/cs k9salways use the environment's own kubeconfig, whatever your current context is.
Expected output of kubectl get nodes (the version is RKE2's current stable release)
Step 6: Run a first workload¶
This small web app is what scenarios 08 (backup and restore) and 09 (chaos against your own workload) use.
cs kubectl create namespace shop
cs kubectl -n shop create deployment web --image=nginxinc/nginx-unprivileged:1.27-alpine --replicas=3 --port=8080
cs kubectl -n shop expose deployment web --port=8080
cs kubectl -n shop rollout status deployment/web --timeout=180s
cs kubectl -n shop get pods -o wide
Everything after cs kubectl goes to kubectl as typed. Mutating calls are recorded as undo points: a create is undone by deleting exactly what it created.
Use an agent, MCP or the UI¶
Follow the same numbered steps and verification/cleanup conditions through your chosen interface. Start with the interface setup and coverage guide; replace account/project/subscription and SSH placeholders before any live request.
Agent prompt: “Follow the local Kubernetes walkthrough for vmware-lab. Preview the selected RKE2 or kubeadm setup, verify nodes and kubeconfig, then show the Helm release list. Do not start an interactive terminal on my behalf.”
MCP starter: cloudseed_k8s with:
Use the matching tool for each remaining step in this page; the command-to-tool map lists the tool family. Keep vmware-lab selected. Preview first; add confirm:true only to the specific change you have authorized. Host bootstrap, provider login and interactive applications retain their documented human steps.
UI: Create → VMware creates the selected lab cluster. Select vmware-lab, then Environments → Kubernetes for info/kubeconfig. All actions → kubectl/Helm runs the listed noninteractive checks. k9s and interactive SSH remain terminal experiences; their readiness checks are available in the UI.
Verify it worked¶
node listshowscloudseed-lab-cp1,-wk1,-wk2as Ready, with their roles and private IPs.kubectl -n shop get podsshows threeweb-...pods Running, spread over the workers.troubleshootends without red findings.
Clean up¶
Keep the cluster if you continue with scenarios 06 to 10 and 13. When you are done:
Unattended: cs destroy vmware -y --env lab --purge --auto-approve.
What just happened¶
enable_kubernetes=trueadded thekubernetesmodule ofterraform/vmware(control-plane and worker VMs at fixed private IPs: control planes from.20, workers from.40).- Ansible's Kubernetes play hardened the nodes and installed RKE2 (Canal CNI) or kubeadm (containerd + Flannel). Nodes are deliberately not auto-updated: upgrades are yours to schedule.
- One kubeconfig per environment, merged on request: several clusters never step on each other.
- Learn more: Platform guide · VMware reference · CLI guide · Explain index
Next steps¶
- Kubernetes access and bastions: keep the workstation's admin kubeconfig on the controller; explicitly authorize any separate bastion user.
- 06 · A production platform in one command on this cluster.
- 13 · Day-2 operations: add and remove nodes.
- 03 · Private GKE: the same commands against a managed cluster.