provisioning-catalog/components/buildkit_lite/cluster/install-buildkit_lite.sh

#!/bin/bash
set -euo pipefail

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
cd "$SCRIPT_DIR"
[ -f "${SCRIPT_DIR}/env-buildkit_lite" ]  && source "${SCRIPT_DIR}/env-buildkit_lite"
[ -f "${SCRIPT_DIR}/_credentials.env" ]   && source "${SCRIPT_DIR}/_credentials.env"

BUILDKIT_NAMESPACE="${BUILDKIT_NAMESPACE:-build-system}"
BUILDKIT_DEPLOYMENT="${BUILDKIT_DEPLOYMENT:-buildkitd}"
BUILDKIT_IMAGE="${BUILDKIT_IMAGE:-moby/buildkit:v0.29.0}"
BUILDKIT_PORT="${BUILDKIT_PORT:-1234}"
BUILDKIT_REGISTRY_SECRET="${BUILDKIT_REGISTRY_SECRET:-zot-credentials}"
CMD_TSK="${CMD_TSK:-${1:-install}}"

_resolve_kubeconfig() {
  if [ -n "${KUBECONFIG:-}" ] && [ -f "$KUBECONFIG" ]; then
    return
  fi
  for candidate in \
    /var/lib/k0s/pki/admin.conf \
    /etc/kubernetes/admin.conf \
    /root/.kube/config; do
    if [ -f "$candidate" ]; then
      export KUBECONFIG="$candidate"
      return
    fi
  done
  if command -v k0s >/dev/null 2>&1; then
    local tmp
    tmp="$(mktemp)"
    k0s kubeconfig admin > "$tmp" 2>/dev/null && export KUBECONFIG="$tmp" && return
    rm -f "$tmp"
  fi
  echo "ERROR: kubeconfig not found — set KUBECONFIG explicitly" >&2; exit 1
}
_resolve_kubeconfig

_kubectl() {
  if command -v kubectl >/dev/null 2>&1; then
    command kubectl "$@"
  elif command -v k0s >/dev/null 2>&1; then
    k0s kubectl "$@"
  else
    echo "ERROR: no kubectl or k0s binary found" >&2; exit 127
  fi
}

_wait_for_deploy() {
  local timeout="${1:-120}"
  _kubectl rollout status deployment/"$BUILDKIT_DEPLOYMENT" \
    --namespace "$BUILDKIT_NAMESPACE" \
    --timeout="${timeout}s"
}

_ensure_registry_secret() {
  if _kubectl get secret "$BUILDKIT_REGISTRY_SECRET" \
       --namespace "$BUILDKIT_NAMESPACE" >/dev/null 2>&1; then
    return
  fi
  if _kubectl get secret "$BUILDKIT_REGISTRY_SECRET" \
       --namespace registry >/dev/null 2>&1; then
    _kubectl get secret "$BUILDKIT_REGISTRY_SECRET" \
      --namespace registry -o yaml \
      | sed "s/namespace: registry/namespace: $BUILDKIT_NAMESPACE/" \
      | _kubectl apply -f -
    echo "Copied $BUILDKIT_REGISTRY_SECRET from registry namespace"
  else
    echo "WARNING: secret $BUILDKIT_REGISTRY_SECRET not found — buildkitd will have no registry auth" >&2
  fi
}

_apply_if_present() {
  local path="$1"
  [ -f "$path" ] || return 0
  if ! grep -q '[^[:space:]]' "$path"; then
    echo "  [skip] $(basename "$path") is empty (template inactive)"
    return 0
  fi
  _kubectl apply -f "$path"
}

_delete_if_present() {
  local path="$1"
  [ -f "$path" ] || return 0
  if ! grep -q '[^[:space:]]' "$path"; then
    return 0
  fi
  _kubectl delete -f "$path" --ignore-not-found
}

_wait_for_secret() {
  local secret="$1"
  local timeout="${2:-60}"
  local elapsed=0
  while [ "$elapsed" -lt "$timeout" ]; do
    if _kubectl get secret "$secret" --namespace "$BUILDKIT_NAMESPACE" >/dev/null 2>&1; then
      return 0
    fi
    sleep 2
    elapsed=$((elapsed + 2))
  done
  echo "WARNING: secret $secret not issued within ${timeout}s — buildkitd will fail until cert-manager catches up" >&2
  return 1
}

_apply_pki_if_enabled() {
  # Templates render empty when mtls.cert_manager.enabled is false → blank-skip.
  _apply_if_present "$SCRIPT_DIR/templates/mtls-clusterissuer.yaml"
  _apply_if_present "$SCRIPT_DIR/templates/mtls-ca-certificate.yaml"
  # CA Issuer depends on the CA secret being populated by cert-manager.
  if [ -f "$SCRIPT_DIR/templates/mtls-ca-certificate.yaml" ] \
       && grep -q '[^[:space:]]' "$SCRIPT_DIR/templates/mtls-ca-certificate.yaml"; then
    _wait_for_secret "${BUILDKIT_MTLS_CA_SECRET:-buildkit-mtls-ca}" 60 || true
  fi
  _apply_if_present "$SCRIPT_DIR/templates/mtls-ca-issuer.yaml"
  _apply_if_present "$SCRIPT_DIR/templates/mtls-server-certificate.yaml"
  if [ -f "$SCRIPT_DIR/templates/mtls-server-certificate.yaml" ] \
       && grep -q '[^[:space:]]' "$SCRIPT_DIR/templates/mtls-server-certificate.yaml"; then
    _wait_for_secret "${BUILDKIT_MTLS_SERVER_SECRET:-buildkit-mtls-server}" 60 || true
  fi
}

_do_install() {
  _kubectl apply -f "$SCRIPT_DIR/templates/namespace.yaml"
  _apply_pki_if_enabled
  _ensure_registry_secret
  # cache PVC (renders empty unless cache.enabled) — must exist before the Deployment mounts it
  _apply_if_present "$SCRIPT_DIR/templates/pvc.yaml"
  _kubectl apply -f "$SCRIPT_DIR/templates/buildkitd-config.yaml"
  _kubectl apply -f "$SCRIPT_DIR/templates/deployment.yaml"
  _kubectl apply -f "$SCRIPT_DIR/templates/service.yaml"
  _apply_if_present "$SCRIPT_DIR/templates/service-vpn.yaml"
  echo "Waiting for buildkitd to be ready..."
  _wait_for_deploy 120
  echo "buildkit_lite installed in namespace $BUILDKIT_NAMESPACE"
}

_do_update() {
  _apply_pki_if_enabled
  _apply_if_present "$SCRIPT_DIR/templates/pvc.yaml"
  _kubectl apply -f "$SCRIPT_DIR/templates/buildkitd-config.yaml"
  _kubectl apply -f "$SCRIPT_DIR/templates/deployment.yaml"
  _apply_if_present "$SCRIPT_DIR/templates/service-vpn.yaml"
  _kubectl rollout restart deployment/"$BUILDKIT_DEPLOYMENT" --namespace "$BUILDKIT_NAMESPACE"
  echo "Waiting for buildkitd to restart..."
  _wait_for_deploy 120
  echo "buildkit_lite updated"
}

_do_delete() {
  _kubectl delete -f "$SCRIPT_DIR/templates/deployment.yaml" --ignore-not-found
  _kubectl delete -f "$SCRIPT_DIR/templates/service.yaml"    --ignore-not-found
  _delete_if_present "$SCRIPT_DIR/templates/service-vpn.yaml"
  _delete_if_present "$SCRIPT_DIR/templates/mtls-server-certificate.yaml"
  _delete_if_present "$SCRIPT_DIR/templates/mtls-ca-issuer.yaml"
  _delete_if_present "$SCRIPT_DIR/templates/mtls-ca-certificate.yaml"
  # mtls-clusterissuer is cluster-scoped and shareable — never delete it from a component teardown.
  echo "buildkit_lite removed from namespace $BUILDKIT_NAMESPACE"
}

_do_health() {
  local ready
  ready=$(_kubectl get deployment "$BUILDKIT_DEPLOYMENT" \
    --namespace "$BUILDKIT_NAMESPACE" \
    -o jsonpath='{.status.readyReplicas}' 2>/dev/null || echo "0")
  if [ "${ready:-0}" -ge 1 ]; then
    echo "HEALTHY: $BUILDKIT_DEPLOYMENT $ready/$ready ready"
  else
    echo "UNHEALTHY: $BUILDKIT_DEPLOYMENT not ready in namespace $BUILDKIT_NAMESPACE" >&2
    exit 1
  fi
}

case "$CMD_TSK" in
  install) _do_install ;;
  update)  _do_update  ;;
  delete)  _do_delete  ;;
  health)  _do_health  ;;
  *)
    echo "ERROR: unknown CMD_TSK='$CMD_TSK'. Valid: install|update|delete|health" >&2
    exit 1
    ;;
esac