Update arch_install.sh
This commit is contained in:
parent
394dc98229
commit
967d168b1f
267
arch_install.sh
267
arch_install.sh
|
|
@ -4,51 +4,53 @@ set -Eeuo pipefail
|
|||
###############################################################################
|
||||
# Arch Linux Kubernetes bootstrap script
|
||||
#
|
||||
# Purpose:
|
||||
# Fully automate a single-node Kubernetes control-plane install on Arch Linux
|
||||
# using:
|
||||
# - containerd
|
||||
# - kubeadm
|
||||
# - kubectl
|
||||
# - Flannel CNI
|
||||
#
|
||||
# What this script does:
|
||||
# 1. Updates the system
|
||||
# 2. Replaces legacy iptables with iptables-nft in one transaction
|
||||
# 3. Installs Kubernetes packages and containerd
|
||||
# 4. Configures containerd to use systemd cgroups
|
||||
# 5. Enables required kernel modules and sysctl settings
|
||||
# 6. Disables swap now and on reboot
|
||||
# 7. Enables containerd and kubelet
|
||||
# 8. Initializes a single control-plane node with kubeadm
|
||||
# 2. Replaces legacy iptables with iptables-nft in the same pacman transaction
|
||||
# 3. Installs Kubernetes packages and dependencies
|
||||
# 4. Configures containerd for systemd cgroups
|
||||
# 5. Enables required kernel modules and sysctl values
|
||||
# 6. Disables swap now and on boot
|
||||
# 7. Enables and starts containerd and kubelet
|
||||
# 8. Initializes the cluster with kubeadm (if not already initialized)
|
||||
# 9. Configures kubectl for the invoking user
|
||||
# 10. Installs Flannel CNI
|
||||
# 11. Optionally allows scheduling pods on the control-plane node
|
||||
# 11. Optionally removes the control-plane taint for single-node use
|
||||
#
|
||||
# Notes:
|
||||
# - This script is intended for a fresh/single-node lab setup.
|
||||
# - Re-running is mostly safe; it skips kubeadm init if already initialized.
|
||||
# - Run it as:
|
||||
# sudo ./arch_k8s_bootstrap.sh
|
||||
# Usage:
|
||||
# chmod +x arch_install.sh
|
||||
# sudo ./arch_install.sh
|
||||
#
|
||||
# Re-run behavior:
|
||||
# This script is written to be mostly idempotent. If the cluster is already
|
||||
# initialized, it will skip kubeadm init and preserve the existing cluster.
|
||||
###############################################################################
|
||||
|
||||
#######################################
|
||||
# User-tunable variables
|
||||
# User-configurable values
|
||||
#######################################
|
||||
|
||||
# Pod CIDR required by Flannel docs
|
||||
# Pod CIDR for Flannel
|
||||
POD_CIDR="${POD_CIDR:-10.244.0.0/16}"
|
||||
|
||||
# Flannel manifest URL from current flannel docs/releases
|
||||
FLANNEL_MANIFEST_URL="${FLANNEL_MANIFEST_URL:-https://github.com/flannel-io/flannel/releases/latest/download/kube-flannel.yml}"
|
||||
# Flannel manifest
|
||||
FLANNEL_URL="${FLANNEL_URL:-https://github.com/flannel-io/flannel/releases/latest/download/kube-flannel.yml}"
|
||||
|
||||
# Set to "true" if this is a single-node lab and you want workloads allowed
|
||||
# on the control-plane node after setup.
|
||||
ALLOW_PODS_ON_CONTROL_PLANE="${ALLOW_PODS_ON_CONTROL_PLANE:-true}"
|
||||
# Set to true if this is a single-node lab and you want normal workloads to run
|
||||
# on the control-plane node.
|
||||
ALLOW_SINGLE_NODE_SCHEDULING="${ALLOW_SINGLE_NODE_SCHEDULING:-true}"
|
||||
|
||||
# The user who should receive ~/.kube/config.
|
||||
# If script is run with sudo, prefer the original user.
|
||||
# Which user should receive ~/.kube/config
|
||||
TARGET_USER="${SUDO_USER:-root}"
|
||||
|
||||
# Figure out that user's home directory safely.
|
||||
if [[ "${TARGET_USER}" == "root" ]]; then
|
||||
TARGET_HOME="/root"
|
||||
else
|
||||
TARGET_HOME="$(getent passwd "${TARGET_USER}" | cut -d: -f6)"
|
||||
fi
|
||||
|
||||
#######################################
|
||||
# Helper functions
|
||||
#######################################
|
||||
|
|
@ -62,42 +64,58 @@ die() {
|
|||
exit 1
|
||||
}
|
||||
|
||||
require_root() {
|
||||
[[ "${EUID}" -eq 0 ]] || die "Run this script with sudo or as root."
|
||||
}
|
||||
|
||||
command_exists() {
|
||||
command -v "$1" >/dev/null 2>&1
|
||||
run_as_target_user() {
|
||||
if [[ "$TARGET_USER" == "root" ]]; then
|
||||
"$@"
|
||||
else
|
||||
sudo -u "$TARGET_USER" "$@"
|
||||
fi
|
||||
}
|
||||
|
||||
#######################################
|
||||
# Sanity checks
|
||||
# Root check
|
||||
#######################################
|
||||
|
||||
require_root
|
||||
|
||||
if [[ -z "${TARGET_HOME}" || ! -d "${TARGET_HOME}" ]]; then
|
||||
die "Could not determine home directory for target user: ${TARGET_USER}"
|
||||
if [[ "${EUID}" -ne 0 ]]; then
|
||||
die "Run this script with sudo or as root."
|
||||
fi
|
||||
|
||||
#######################################
|
||||
# Resolve target home directory
|
||||
#######################################
|
||||
|
||||
if [[ "$TARGET_USER" == "root" ]]; then
|
||||
TARGET_HOME="/root"
|
||||
else
|
||||
TARGET_HOME="$(getent passwd "$TARGET_USER" | cut -d: -f6)"
|
||||
fi
|
||||
|
||||
[[ -n "${TARGET_HOME:-}" && -d "$TARGET_HOME" ]] || die "Could not determine home directory for $TARGET_USER"
|
||||
|
||||
log "Starting Arch Kubernetes bootstrap"
|
||||
log "Target kubectl user: ${TARGET_USER}"
|
||||
log "Target home: ${TARGET_HOME}"
|
||||
log "Target kubectl user: $TARGET_USER"
|
||||
log "Target home: $TARGET_HOME"
|
||||
|
||||
#######################################
|
||||
# 1) Fully update Arch
|
||||
# 1) Update system
|
||||
#######################################
|
||||
|
||||
log "Updating package databases and upgrading system"
|
||||
pacman -Syu --noconfirm
|
||||
|
||||
#######################################
|
||||
# 2) Install required packages
|
||||
# 2) Install packages
|
||||
#
|
||||
# Important:
|
||||
# - iptables-nft must replace legacy iptables in the SAME transaction.
|
||||
# - We intentionally do not remove iptables first because that can break
|
||||
# dependency resolution temporarily for packages that need libxtables.
|
||||
# iptables-nft conflicts with legacy iptables, so we must allow pacman to
|
||||
# replace iptables during the SAME transaction.
|
||||
#
|
||||
# Why we use yes | pacman here:
|
||||
# pacman may ask:
|
||||
# "Remove iptables? [y/N]"
|
||||
# and then:
|
||||
# "Proceed with installation? [Y/n]"
|
||||
# Using yes feeds "y" to both prompts so the script does not stop there.
|
||||
#######################################
|
||||
|
||||
log "Installing container runtime, Kubernetes tools, networking tools, and iptables-nft"
|
||||
|
|
@ -125,45 +143,24 @@ log "Verifying iptables backend"
|
|||
iptables --version || true
|
||||
|
||||
#######################################
|
||||
# 4) Enable and configure containerd
|
||||
#
|
||||
# kubelet works best with containerd configured to use systemd cgroups.
|
||||
# 4) Persist required kernel modules
|
||||
#######################################
|
||||
|
||||
log "Enabling and starting containerd"
|
||||
systemctl enable --now containerd.service
|
||||
|
||||
log "Creating default containerd config if missing"
|
||||
mkdir -p /etc/containerd
|
||||
if [[ ! -f /etc/containerd/config.toml ]]; then
|
||||
containerd config default > /etc/containerd/config.toml
|
||||
fi
|
||||
|
||||
log "Setting SystemdCgroup = true in /etc/containerd/config.toml"
|
||||
sed -i 's/SystemdCgroup = false/SystemdCgroup = true/' /etc/containerd/config.toml
|
||||
|
||||
log "Restarting containerd to apply config"
|
||||
systemctl restart containerd.service
|
||||
|
||||
#######################################
|
||||
# 5) Load required kernel modules now and on boot
|
||||
#######################################
|
||||
|
||||
log "Persisting required kernel modules"
|
||||
log "Writing required kernel modules"
|
||||
cat > /etc/modules-load.d/k8s.conf <<'EOF'
|
||||
overlay
|
||||
br_netfilter
|
||||
EOF
|
||||
|
||||
log "Loading kernel modules immediately"
|
||||
log "Loading kernel modules now"
|
||||
modprobe overlay
|
||||
modprobe br_netfilter
|
||||
|
||||
#######################################
|
||||
# 6) Apply required sysctl settings now and on boot
|
||||
# 5) Persist required sysctl settings
|
||||
#######################################
|
||||
|
||||
log "Persisting required sysctl settings"
|
||||
log "Writing required sysctl settings"
|
||||
cat > /etc/sysctl.d/k8s.conf <<'EOF'
|
||||
net.bridge.bridge-nf-call-iptables = 1
|
||||
net.bridge.bridge-nf-call-ip6tables = 1
|
||||
|
|
@ -174,138 +171,156 @@ log "Applying sysctl settings"
|
|||
sysctl --system
|
||||
|
||||
#######################################
|
||||
# 7) Disable swap now and comment swap entries in /etc/fstab
|
||||
# 6) Disable swap now and on boot
|
||||
#######################################
|
||||
|
||||
log "Disabling swap immediately"
|
||||
swapoff -a || true
|
||||
|
||||
log "Commenting active swap entries in /etc/fstab"
|
||||
log "Commenting swap entries in /etc/fstab"
|
||||
if [[ -f /etc/fstab ]]; then
|
||||
cp /etc/fstab /etc/fstab.bak.$(date +%s)
|
||||
cp /etc/fstab "/etc/fstab.bak.$(date +%s)"
|
||||
sed -i '/^[^#].*\sswap\s/s/^/# /' /etc/fstab
|
||||
fi
|
||||
|
||||
#######################################
|
||||
# 8) Enable kubelet
|
||||
# 7) Configure containerd
|
||||
#
|
||||
# kubelet may show as active but waiting until kubeadm init finishes; that is
|
||||
# normal before the control plane exists.
|
||||
# Kubernetes works better with systemd cgroups.
|
||||
#######################################
|
||||
|
||||
log "Preparing /etc/containerd"
|
||||
mkdir -p /etc/containerd
|
||||
|
||||
log "Generating default containerd config if needed"
|
||||
if [[ ! -f /etc/containerd/config.toml ]]; then
|
||||
containerd config default > /etc/containerd/config.toml
|
||||
fi
|
||||
|
||||
log "Ensuring SystemdCgroup = true"
|
||||
sed -i 's/SystemdCgroup = false/SystemdCgroup = true/' /etc/containerd/config.toml
|
||||
|
||||
#######################################
|
||||
# 8) Enable and start services
|
||||
#######################################
|
||||
|
||||
log "Enabling and starting containerd"
|
||||
systemctl enable --now containerd.service
|
||||
|
||||
log "Enabling and starting kubelet"
|
||||
systemctl enable --now kubelet.service
|
||||
|
||||
#######################################
|
||||
# 9) Preflight visibility
|
||||
#######################################
|
||||
|
||||
log "Container runtime status"
|
||||
log "Showing containerd status"
|
||||
systemctl --no-pager --full status containerd.service || true
|
||||
|
||||
log "Kubelet status"
|
||||
log "Showing kubelet status"
|
||||
systemctl --no-pager --full status kubelet.service || true
|
||||
|
||||
#######################################
|
||||
# 10) Initialize cluster if not already initialized
|
||||
# 9) Initialize cluster if needed
|
||||
#######################################
|
||||
|
||||
if [[ ! -f /etc/kubernetes/admin.conf ]]; then
|
||||
log "Initializing Kubernetes control plane with kubeadm"
|
||||
kubeadm init --pod-network-cidr="${POD_CIDR}"
|
||||
log "Initializing Kubernetes control plane"
|
||||
kubeadm init --pod-network-cidr="$POD_CIDR"
|
||||
else
|
||||
log "Skipping kubeadm init because /etc/kubernetes/admin.conf already exists"
|
||||
log "Cluster already initialized; skipping kubeadm init"
|
||||
fi
|
||||
|
||||
#######################################
|
||||
# 11) Configure kubectl for target user
|
||||
# 10) Configure kubectl for target user
|
||||
#######################################
|
||||
|
||||
log "Configuring kubectl for ${TARGET_USER}"
|
||||
mkdir -p "${TARGET_HOME}/.kube"
|
||||
cp -f /etc/kubernetes/admin.conf "${TARGET_HOME}/.kube/config"
|
||||
chown -R "${TARGET_USER}:${TARGET_USER}" "${TARGET_HOME}/.kube"
|
||||
chmod 700 "${TARGET_HOME}/.kube"
|
||||
chmod 600 "${TARGET_HOME}/.kube/config"
|
||||
log "Configuring kubectl for $TARGET_USER"
|
||||
mkdir -p "$TARGET_HOME/.kube"
|
||||
cp -f /etc/kubernetes/admin.conf "$TARGET_HOME/.kube/config"
|
||||
chown -R "$TARGET_USER:$TARGET_USER" "$TARGET_HOME/.kube"
|
||||
chmod 700 "$TARGET_HOME/.kube"
|
||||
chmod 600 "$TARGET_HOME/.kube/config"
|
||||
|
||||
#######################################
|
||||
# 12) Wait briefly for API server to become responsive
|
||||
# 11) Wait for API responsiveness
|
||||
#######################################
|
||||
|
||||
log "Waiting for Kubernetes API to become reachable"
|
||||
for _ in $(seq 1 60); do
|
||||
if sudo -u "${TARGET_USER}" kubectl --kubeconfig="${TARGET_HOME}/.kube/config" get nodes >/dev/null 2>&1; then
|
||||
log "Waiting for Kubernetes API server"
|
||||
for _ in $(seq 1 90); do
|
||||
if run_as_target_user kubectl --kubeconfig="$TARGET_HOME/.kube/config" get nodes >/dev/null 2>&1; then
|
||||
break
|
||||
fi
|
||||
sleep 2
|
||||
done
|
||||
|
||||
#######################################
|
||||
# 13) Install Flannel if not already present
|
||||
# 12) Install Flannel if missing
|
||||
#######################################
|
||||
|
||||
log "Installing Flannel CNI"
|
||||
if ! sudo -u "${TARGET_USER}" kubectl --kubeconfig="${TARGET_HOME}/.kube/config" get namespace kube-flannel >/dev/null 2>&1; then
|
||||
sudo -u "${TARGET_USER}" kubectl --kubeconfig="${TARGET_HOME}/.kube/config" apply -f "${FLANNEL_MANIFEST_URL}"
|
||||
log "Installing Flannel CNI if needed"
|
||||
if ! run_as_target_user kubectl --kubeconfig="$TARGET_HOME/.kube/config" get namespace kube-flannel >/dev/null 2>&1; then
|
||||
run_as_target_user kubectl --kubeconfig="$TARGET_HOME/.kube/config" apply -f "$FLANNEL_URL"
|
||||
else
|
||||
log "kube-flannel namespace already exists; skipping Flannel install"
|
||||
log "Flannel already present; skipping apply"
|
||||
fi
|
||||
|
||||
#######################################
|
||||
# 14) Optionally allow scheduling on single-node control plane
|
||||
# 13) Allow single-node scheduling if requested
|
||||
#######################################
|
||||
|
||||
if [[ "${ALLOW_PODS_ON_CONTROL_PLANE}" == "true" ]]; then
|
||||
log "Allowing workloads on the control-plane node (single-node lab mode)"
|
||||
sudo -u "${TARGET_USER}" kubectl --kubeconfig="${TARGET_HOME}/.kube/config" taint nodes --all node-role.kubernetes.io/control-plane- || true
|
||||
if [[ "$ALLOW_SINGLE_NODE_SCHEDULING" == "true" ]]; then
|
||||
log "Removing control-plane taint for single-node scheduling"
|
||||
run_as_target_user kubectl --kubeconfig="$TARGET_HOME/.kube/config" taint nodes --all node-role.kubernetes.io/control-plane- || true
|
||||
else
|
||||
log "Leaving default control-plane taint in place"
|
||||
log "Leaving control-plane taint in place"
|
||||
fi
|
||||
|
||||
#######################################
|
||||
# 15) Final status / useful commands
|
||||
# 14) Final status
|
||||
#######################################
|
||||
|
||||
log "Final checks"
|
||||
log "Final cluster checks"
|
||||
|
||||
echo
|
||||
echo "==== iptables backend ===="
|
||||
iptables --version || true
|
||||
|
||||
echo
|
||||
echo "==== containerd info ===="
|
||||
crictl info >/dev/null 2>&1 && echo "crictl can talk to the runtime" || echo "crictl check did not succeed yet"
|
||||
echo "==== containerd status ===="
|
||||
systemctl --no-pager --full status containerd.service || true
|
||||
|
||||
echo
|
||||
echo "==== kubelet status ===="
|
||||
systemctl --no-pager --full status kubelet.service || true
|
||||
|
||||
echo
|
||||
echo "==== nodes ===="
|
||||
sudo -u "${TARGET_USER}" kubectl --kubeconfig="${TARGET_HOME}/.kube/config" get nodes -o wide || true
|
||||
run_as_target_user kubectl --kubeconfig="$TARGET_HOME/.kube/config" get nodes -o wide || true
|
||||
|
||||
echo
|
||||
echo "==== pods (all namespaces) ===="
|
||||
sudo -u "${TARGET_USER}" kubectl --kubeconfig="${TARGET_HOME}/.kube/config" get pods -A -o wide || true
|
||||
run_as_target_user kubectl --kubeconfig="$TARGET_HOME/.kube/config" get pods -A -o wide || true
|
||||
|
||||
cat <<EOF
|
||||
|
||||
Bootstrap complete.
|
||||
|
||||
kubectl is configured for:
|
||||
user: ${TARGET_USER}
|
||||
config: ${TARGET_HOME}/.kube/config
|
||||
kubectl config:
|
||||
User: $TARGET_USER
|
||||
Config: $TARGET_HOME/.kube/config
|
||||
|
||||
Common next commands:
|
||||
Useful commands:
|
||||
kubectl get nodes
|
||||
kubectl get pods -A
|
||||
kubectl cluster-info
|
||||
|
||||
If the node is not Ready yet, wait a minute and re-run:
|
||||
kubectl get nodes
|
||||
kubectl get pods -A
|
||||
|
||||
To test the cluster:
|
||||
Test workload:
|
||||
kubectl create deployment nginx --image=nginx
|
||||
kubectl expose deployment nginx --port=80 --type=NodePort
|
||||
kubectl get svc
|
||||
|
||||
To join worker nodes later, generate a join command with:
|
||||
Worker join command:
|
||||
kubeadm token create --print-join-command
|
||||
|
||||
If the node is still NotReady immediately after install, wait 30-60 seconds and re-check:
|
||||
kubectl get nodes
|
||||
kubectl get pods -A
|
||||
|
||||
EOF
|
||||
Loading…
Reference in New Issue