Ensure Rotate Kubelet Server Certificate Argument Is Enabled

#!/usr/bin/env bash
#
# Automation: Enable RotateKubeletServerCertificate feature-gate on kube-controller-manager
# Scope: Run on every control plane node with direct access to /etc/kubernetes/manifests
# Usage: sudo ./enable-rotate-kubelet-server-cert.sh

set -euo pipefail

MANIFEST="/etc/kubernetes/manifests/kube-controller-manager.yaml"
BACKUP_DIR="/etc/kubernetes/manifests/backup-rotate-kubelet-server-cert"
TIMESTAMP="$(date +%Y%m%d%H%M%S)"

if [[ $EUID -ne 0 ]]; then
  echo "ERROR: This script must be run as root (or with sudo)." >&2
  exit 1
fi

if [[ ! -f "$MANIFEST" ]]; then
  echo "ERROR: Manifest not found at $MANIFEST. Are you on a control plane node?" >&2
  exit 1
fi

mkdir -p "$BACKUP_DIR"

backup_file="$BACKUP_DIR/kube-controller-manager.yaml.$TIMESTAMP"
cp -p "$MANIFEST" "$backup_file"
echo "Backup created at: $backup_file"

# Normalize feature-gates lines and ensure RotateKubeletServerCertificate=true is present
tmpfile="$(mktemp)"
trap 'rm -f "$tmpfile"' EXIT

# Idempotent edit:
# - If a --feature-gates arg exists:
#     * If it already includes RotateKubeletServerCertificate=true: leave as-is
#     * If it includes RotateKubeletServerCertificate=false: change to true
#     * Else append ,RotateKubeletServerCertificate=true
# - If no --feature-gates exists: add a new - --feature-gates=RotateKubeletServerCertificate=true
python3 - "$MANIFEST" > "$tmpfile" << 'PYEOF'
import sys, re

path = sys.argv[1]
with open(path, "r", encoding="utf-8") as f:
    lines = f.readlines()

out = []
feature_gates_key = "- --feature-gates="
found_fg = False

for i, line in enumerate(lines):
    stripped = line.lstrip()
    if stripped.startswith(feature_gates_key):
        found_fg = True
        indent = line[:len(line) - len(stripped)]
        value = stripped[len(feature_gates_key):].rstrip("\n")

        # Ensure RotateKubeletServerCertificate is set to true
        # Remove any existing RotateKubeletServerCertificate=<...> and append =true
        parts = [p for p in value.split(",") if not p.startswith("RotateKubeletServerCertificate=")]
        parts.append("RotateKubeletServerCertificate=true")
        # Deduplicate while preserving order
        seen = set()
        new_parts = []
        for p in parts:
            if p not in seen:
                seen.add(p)
                new_parts.append(p)
        new_value = ",".join(new_parts)
        out.append(f"{indent}{feature_gates_key}{new_value}\n")
    else:
        out.append(line)

if not found_fg:
    # Insert new feature-gates arg under the kube-controller-manager container args
    inserted = False
    for i, line in enumerate(out):
        if re.match(r'^\s*name:\s*kube-controller-manager\s*$', line):
            # Look for the args: block following this
            j = i + 1
            while j < len(out) and not re.match(r'^\s*args:\s*$', out[j]):
                j += 1
            if j < len(out):
                indent = re.match(r'^(\s*)args:\s*$', out[j]).group(1) + "  "
                out.insert(j + 1, f"{indent}- --feature-gates=RotateKubeletServerCertificate=true\n")
                inserted = True
                break
    if not inserted:
        # Fallback: append at end, minimally valid for static pod
        out.append("  - --feature-gates=RotateKubeletServerCertificate=true\n")

sys.stdout.writelines(out)
PYEOF

# Only overwrite if there is a change
if cmp -s "$MANIFEST" "$tmpfile"; then
  echo "No changes needed; RotateKubeletServerCertificate=true already configured."
else
  mv "$tmpfile" "$MANIFEST"
  sync
  echo "Updated $MANIFEST with RotateKubeletServerCertificate=true in --feature-gates."
fi

# Note: kubelet will automatically restart the static pod after manifest change.

echo "Waiting 30 seconds for kube-controller-manager to restart..."
sleep 30

echo "Verification: checking running kube-controller-manager arguments..."
if /bin/ps -ef | grep kube-controller-manager | grep -v grep | grep -q -- '--feature-gates=.*RotateKubeletServerCertificate=true'; then
  echo "SUCCESS: kube-controller-manager is running with RotateKubeletServerCertificate=true enabled."
  exit 0
else
  echo "WARNING: kube-controller-manager process not yet showing RotateKubeletServerCertificate=true."
  echo "Current kube-controller-manager processes:"
  /bin/ps -ef | grep kube-controller-manager | grep -v grep || true
  exit 1
fi