agentleFS
Sign inStart free
Quad4-SoftwareQuad4-Software/skills/.agents/skills/k3s/SKILL.md

k3s

Skill1 starsChanged 43 days ago0BSDView on GitHub

This skill covers k3s, the lightweight CNCF-certified Kubernetes distribution from SUSE/Rancher, tracking upstream Kubernetes (v1.37.1+k3s1 current, Oct 2026). Use it for install flags and env vars, server/agent topology, the bundled components (containerd, flannel, servicelb, traefik, local-path-provisioner, helm-controller), datastore choices (sqlite vs embedded etcd vs external), registries.yaml mirrors, airgap installs, snapshots, and hardening.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read this file for the architecture and defaults.

    line 25
  2. Distinctive:

    Default CNI is not encrypted (vxlan plaintext).

    line 101

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

When to use this skill

  • You are installing or upgrading a k3s node or cluster.
  • You are choosing between embedded etcd, sqlite, and an external datastore.
  • You are configuring private registry mirrors, airgap images, or disabling bundled components.
  • You are debugging the kubeconfig, flannel, traefik, or the klipper load balancer.

How to use

  1. Read this file for the architecture and defaults.
  2. Load references/ops.md for install flags, config file, datastore, registries, airgap, backup, and upgrade procedures.
  3. Fall back to https://docs.k3s.io/ for field-level detail.

Examples

  • "Install a 3-server HA k3s cluster with embedded etcd."
  • "Point k3s at our internal registry mirror and disable traefik."
  • "Why does kubectl on my laptop fail after a k3s install?"
  • "Airgap-install k3s on an isolated node."

k3s

What it is

A single ~70 MB binary wrapping upstream Kubernetes plus the batteries most clusters actually run: containerd (2.x), flannel CNI, CoreDNS, servicelb (klipper LoadBalancer), traefik ingress, local-path-provisioner, metrics-server, helm-controller, and a network policy controller. CNCF certified, so manifests behave the same as upstream.

Versions track upstream Kubernetes directly: v1.37.1+k3s1 is the current release (Sept 30, 2026). 1.36.x, 1.35.x, and 1.34.x still get patch releases. The +k3sN suffix is the packaging revision. Upgrade one minor at a time. Do not skip.

Topology

  • k3s server - control plane (+ default worker). Multiple servers give HA.
  • k3s agent - worker, joins via K3S_URL=https://server:6443 and K3S_TOKEN from /var/lib/rancher/k3s/server/node-token on the server.
  • Single-server default datastore is sqlite. Two or more servers need embedded etcd (--cluster-init on the first) or an external datastore (--datastore-endpoint for etcd/MySQL/Postgres).
  • --docker is a hard error (dockershim died in k8s 1.24). k3s always runs its bundled containerd.

Defaults worth changing

  • traefik and servicelb are on by default. Disable with --disable=traefik --disable=servicelb if you run your own ingress/ LB. Same mechanism disables servicelb, coredns, metrics-server, local-storage, helm-controller.
  • kubeconfig lands at /etc/rancher/k3s/k3s.yaml, mode 600 root. K3S_KUBECONFIG_MODE=644 on install, or copy it and fix the 127.0.0.1 server address to the node's IP for remote kubectl.
  • flannel backend defaults to vxlan. --flannel-backend=wireguard- native, host-gw, or none (bring your own CNI, then disable the built-in policy controller too: --disable-network-policy).
  • Ports: 6443 apiserver (agents + kubectl), 8472/udp flannel vxlan, 51820/udp wireguard-ipv4 (51821 ipv6), 10250 kubelet, 2379/2380 etcd (embedded), 9345 for new-server join. NodePort range 30000-32767 handled by servicelb via host ports.
  • containerd config is managed by k3s. Edit via /var/lib/rancher/k3s/agent/etc/containerd/config.toml.tmpl, never the rendered config.toml. Registry mirrors and auth live in /etc/rancher/k3s/registries.yaml.
  • Config file /etc/rancher/k3s/config.yaml maps flags to keys (kebab to camelCase). Prefer it over editing systemd units.

Security posture

  • CIS benchmark: k3s ships hardened-by-default versus upstream (fewer ports, restricted component perms). The docs carry a CIS hardening guide and a self-assessment per release.
  • Secrets encryption at rest: k3s secrets-encrypt (AES-CBC default, supports rotation and disable).
  • SELinux supported via --selinux or the packaged policy. Rootless mode exists but is experimental.
  • The node-token is a cluster join credential: treat /var/lib/rancher/k3s/server/node-token as secret, and rotate via k3s token rotate.
  • Default CNI is not encrypted (vxlan plaintext). Use wireguard backend or an external CNI where east-west traffic needs protection.

Ecosystem position

  • vs kind/minikube: k3s is a real cluster (persistent, multi-node, systemd), not a dev-only sandbox.
  • vs k0s/MicroK8s: similar single-binary class. k3s is the most resource-light and airgap-friendly. MicroK8s uses snaps. k0s keeps a purer upstream layout.
  • vs RKE2: RKE2 is Rancher's hardened/FIPS-focused sibling. Pick it for compliance regimes, k3s for footprint.

Sources

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free