brandon/oam
1
0
Fork 0
mirror of https://gitea.com/mcereda/oam.git synced 2026-02-09 05:44:23 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
881360d5bc77ac354e4eda2cfa7d5bcd647a11cd
oam/knowledge base/kubernetes/k3s.md
Michele Cereda 855aa588ba chore(k3s): expand notes and snippets
2024-09-14 23:12:08 +02:00

5.6 KiB
Raw Blame History

K3S

Lightweight Kubernetes distribution built for IoT and Edge computing.

  1. TL;DR
  2. Packaged components
  3. Split roles on different nodes
  4. Further readings
    1. Sources

TL;DR

Starting the server with the --cluster-init option will start all control-plane components (including the apiserver, controller-manager, scheduler, and etcd).
When using the embedded etcd, one will be able to disable specific components to split the control-plane and etcd roles onto separate nodes.

Setup 
# Install as single-node server.
curl -sfL 'https://get.k3s.io' | sudo sh -
curl -sfL 'https://get.k3s.io' | sudo sh - server --token '12345'
# Install as agent node and add it to an existing cluster.
curl -sfL 'https://get.k3s.io' | K3S_URL='https://server.fqdn:6443' K3S_TOKEN='node-token' sudo sh -
# Install as node dedicated to the control plane.
curl -sfL 'https://get.k3s.io' \
| sudo sh -s - server --token 'node-token' --disable-etcd --server 'https://server.fqdn:6443'

# Disable the firewall (recommended).
systemctl disable firewalld --now
ufw disable
# Or open at least the required port and networks:
# Port 6443 --> apiserver, network 10.42.0.0/16 --> pods, network 10.43.0.0/16 --> services.
firewall-cmd --permanent --add-port '6443/tcp' \
&& firewall-cmd --permanent --zone='trusted' --add-source '10.42.0.0/16' \
&& firewall-cmd --permanent --zone=trusted --add-source=10.43.0.0/16 \
&& firewall-cmd --reload
ufw allow '6443/tcp' && ufw allow from '10.42.0.0/16' to 'any' && ufw allow from '10.43.0.0/16' to 'any'

# Uninstall.
/usr/local/bin/k3s-uninstall.sh
/usr/local/bin/k3s-agent-uninstall.sh
Usage 
# Check the configuration.
k3s check-config

# Use the provided `kubectl`.
k3s kubectl get pods

# Restore clusters from snapshots.
k3s server --cluster-reset \
  --cluster-reset-restore-path="/var/lib/rancher/k3s/server/db/etcd-old-${BACKUP_DATE}"

Packaged components

Refer Managing packaged components.

Any file found on server nodes in /var/lib/rancher/k3s/server/manifests will automatically be deployed to the cluster both on startup and when any file is changed on disk.
Deleting files from this directory will not delete the corresponding resources from the cluster.

Manifests are tracked as AddOn custom resources in the kube-system namespace.
Use kubectl describe on the AddOn resource to see errors or warnings encountered when applying the manifest files.

K3s comes with packaged components, deployed as AddOns via the manifests directory:

  • coredns
  • traefik
  • local-storage
  • metrics-server

The embedded servicelb LoadBalancer controller does not have a manifest file, but can be disabled as if it was one.

Manifests for packaged components are managed by K3s, and should not be altered.
These files are re-written to disk whenever K3s is started to ensure their integrity.

Split roles on different nodes

This can only be done when using the embedded etcd component.

Refer Managing server roles.

Procedure:

  1. Dedicate the first server to etcd by starting k3s with all the other control plane components disabled:

    curl -sfL 'https://get.k3s.io' \
| sh -s - server --cluster-init --disable-apiserver --disable-controller-manager --disable-scheduler

    This first node will start etcd, then wait for additional control-plane nodes to join.
    The cluster will not be usable until one joins an additional server with the control plane components enabled.

  2. Create a server with only the control plane, by starting k3s with etcd disabled:

    curl -sfL 'https://get.k3s.io' \
| sh -s - server --token 'node-token' --disable-etcd --server 'https://etcd-only.server.fqdn:6443'

  3. Check the nodes have the correct roles:

    $ kubectl get nodes
NAME           STATUS   ROLES                       AGE     VERSION
k3s-server-1   Ready    etcd                        5h39m   v1.20.4+k3s1
k3s-server-2   Ready    control-plane,master        5h39m   v1.20.4+k3s1

Add roles to existing dedicated nodes by restarting k3s on them without the disable flags.

One can disable components in the /etc/rancher/k3s/config.yaml file instead of passing the options as CLI flags:

cluster-init: true
disable-apiserver: true
disable-controller-manager: true
disable-scheduler: true

Further readings

Sources