prole/etc/init_cnpg_gke.sh

#!/usr/bin/env bash
# init_cnpg_gke.sh
# Provision CloudNativePG on GKE with GCS backup via Workload Identity.
#
# Usage:
#   ./etc/init_cnpg_gke.sh [--project PROJECT_ID] [--region REGION] [--cluster CLUSTER_NAME]
#
# Env vars (override flags):
#   GCP_PROJECT_ID, GCP_REGION, GKE_CLUSTER, GCS_BACKUP_BUCKET,
#   CNPG_NAMESPACE, ARTIFACT_REGISTRY, DB_PASSWORD, CLOUDSDK_AUTH_ACCESS_TOKEN

set -euo pipefail

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
REPO_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
GKE_MANIFEST_DIR="$REPO_ROOT/deploy/gcp/gke"

# ── Defaults ────────────────────────────────────────────────────────────────
GCP_PROJECT_ID="${GCP_PROJECT_ID:-}"
GCP_REGION="${GCP_REGION:-us-central1}"
GKE_CLUSTER="${GKE_CLUSTER:-knoe-dev-0}"
GCS_BACKUP_BUCKET="${GCS_BACKUP_BUCKET:-knoe-0-backups}"
GCS_WAL_BUCKET="${GCS_WAL_BUCKET:-knoe-0-wal}"
CNPG_NAMESPACE="${CNPG_NAMESPACE:-knoe-db-0}"
CNPG_CLUSTER_NAME="${CNPG_CLUSTER_NAME:-knoe-db}"
CNPG_BACKUP_SA="${CNPG_BACKUP_SA:-cnpg-backup}"
ARTIFACT_REGISTRY="${ARTIFACT_REGISTRY:-}"
KNOE_DB_IMAGE_TAG="${KNOE_DB_IMAGE_TAG:-}"
DB_PASSWORD="${DB_PASSWORD:-}"
CNPG_DB_OWNER="${CNPG_DB_OWNER:-knoe}"
CNPG_STORAGE_CLASS_ORDER="${CNPG_STORAGE_CLASS_ORDER:-premium-rwo,ssd,premium,standard-rwo,garage-hdd,standard,dynamic-rwo}"
CNPG_STORAGE_WAIT_TIMEOUT="${CNPG_STORAGE_WAIT_TIMEOUT:-240}"
# v1.29.0+ is required: spec.serviceAccountName lets the cluster pods run as
# cnpg-backup-sa (annotated with iam.gke.io/gcp-service-account for WI), which
# is how the GCS-backed barman ObjectStore authenticates without a static key.
# Older operators (we used 1.24.0 historically) silently drop the field.
CNPG_OPERATOR_VERSION="${CNPG_OPERATOR_VERSION:-1.29.0}"

# ── Argument parsing ─────────────────────────────────────────────────────────
while [[ $# -gt 0 ]]; do
  case "$1" in
    --project)   GCP_PROJECT_ID="$2"; shift 2 ;;
    --region)    GCP_REGION="$2";     shift 2 ;;
    --cluster)   GKE_CLUSTER="$2";   shift 2 ;;
    --bucket)    GCS_BACKUP_BUCKET="$2"; shift 2 ;;
    --namespace) CNPG_NAMESPACE="$2"; shift 2 ;;
    --registry)  ARTIFACT_REGISTRY="$2"; shift 2 ;;
    *) echo "Unknown argument: $1" >&2; exit 1 ;;
  esac
done

[[ -z "$GCP_PROJECT_ID" ]] && {
  # Try to read from gcp.cfg
  GCP_CFG="$REPO_ROOT/conf/prod/gcp.cfg"
  if [[ -f "$GCP_CFG" ]]; then
    GCP_PROJECT_ID=$(grep '^project_id' "$GCP_CFG" | sed 's/.*=\s*"\?\([^"]*\)"\?.*/\1/' | tr -d '[:space:]')
  fi
}
[[ -z "$GCP_PROJECT_ID" ]] && { echo "ERROR: GCP_PROJECT_ID not set and conf/prod/gcp.cfg not found." >&2; exit 1; }

GCP_SA_EMAIL="${CNPG_BACKUP_SA}@${GCP_PROJECT_ID}.iam.gserviceaccount.com"

log() { echo "[init_cnpg_gke] $*"; }

resolve_knoe_db_image_tag() {
  [[ -n "$KNOE_DB_IMAGE_TAG" ]] && return

  local mode_key="k8s"
  local knoe_home
  knoe_home="${KNOE_HOME:-$REPO_ROOT}"

  local pg_version_file="$knoe_home/modes/$mode_key/conf/postgresql/.version"
  local release_file="$knoe_home/modes/$mode_key/knoe-db/.version"

  [[ -f "$pg_version_file" ]] || pg_version_file="$knoe_home/conf/postgresql/.version"
  [[ -f "$release_file" ]] || release_file="$knoe_home/knoe-db/.version"

  local pg_version release
  if [[ -f "$pg_version_file" ]]; then
    pg_version="$(tr -d '[:space:]' < "$pg_version_file")"
  else
    pg_version="17.7"
  fi

  if [[ -f "$release_file" ]]; then
    release="$(tr -d '[:space:]' < "$release_file")"
  else
    release="43"
  fi

  if [[ "$release" =~ ^[0-9]+$ ]]; then
    release="$(printf '%03d' "$release")"
  fi

  KNOE_DB_IMAGE_TAG="${pg_version}-${release}"
}

storage_class_available() {
  local candidate="$1"
  [[ -n "$candidate" ]] || return 1
  kubectl get storageclass "$candidate" >/dev/null 2>&1
}

append_storage_candidate() {
  local candidate="$1"
  [[ -n "$candidate" ]] || return
  case ",${CNPG_STORAGE_CLASS_CANDIDATES_RESOLVED}," in
    *",${candidate},"*) return ;;
  esac
  if [[ -n "$CNPG_STORAGE_CLASS_CANDIDATES_RESOLVED" ]]; then
    CNPG_STORAGE_CLASS_CANDIDATES_RESOLVED+="${IFS_COMMA}${candidate}"
  else
    CNPG_STORAGE_CLASS_CANDIDATES_RESOLVED="$candidate"
  fi
}

storage_class_is_compatible() {
  local candidate="$1"
  case "$candidate" in
    *rwx*|*RWX*|gcsfuse*|parallelstore-*|enterprise-rwx|enterprise-multishare-rwx|regional-rwx|premium-rwx|standard-rwx|synology-*|*iscsi*)
      return 1
      ;;
    *)
      return 0
      ;;
  esac
}

append_by_name_pattern() {
  local pattern="$1"
  local sc
  while IFS= read -r sc; do
    [[ -n "$sc" ]] || continue
    storage_class_is_compatible "$sc" || continue
    case "$sc" in
      *"$pattern"*) append_storage_candidate "$sc" ;;
    esac
  done <<< "$AVAILABLE_STORAGE_CLASSES"
}

resolve_storage_class_candidates() {
  AVAILABLE_STORAGE_CLASSES="$(kubectl get storageclass -o jsonpath='{range .items[*]}{.metadata.name}{"\n"}{end}' 2>/dev/null || true)"
  [[ -n "$AVAILABLE_STORAGE_CLASSES" ]] || {
    echo "ERROR: unable to enumerate Kubernetes storage classes." >&2
    exit 1
  }

  CNPG_STORAGE_CLASS_CANDIDATES_RESOLVED=""
  IFS=',' read -r -a requested_classes <<< "$CNPG_STORAGE_CLASS_ORDER"
  local requested
  for requested in "${requested_classes[@]}"; do
    requested="${requested//[[:space:]]/}"
    [[ -n "$requested" ]] || continue
    case "$requested" in
      ssd)
        append_by_name_pattern "ssd"
        ;;
      premium)
        append_by_name_pattern "premium"
        ;;
      *)
        if storage_class_available "$requested" && storage_class_is_compatible "$requested"; then
          append_storage_candidate "$requested"
        fi
        ;;
    esac
  done

  local sc
  while IFS= read -r sc; do
    [[ -n "$sc" ]] || continue
    storage_class_is_compatible "$sc" || continue
    append_storage_candidate "$sc"
  done <<< "$AVAILABLE_STORAGE_CLASSES"

  [[ -n "$CNPG_STORAGE_CLASS_CANDIDATES_RESOLVED" ]] || {
    echo "ERROR: no compatible storage classes available for CNPG provisioning." >&2
    exit 1
  }
}

render_cnpg_manifest() {
  local storage_class="$1"
  local rendered_manifest="$2"
  ARTIFACT_REGISTRY="${ARTIFACT_REGISTRY}" KNOE_DB_IMAGE_TAG="${KNOE_DB_IMAGE_TAG}" \
    envsubst '${ARTIFACT_REGISTRY} ${KNOE_DB_IMAGE_TAG}' < "$GKE_MANIFEST_DIR/knoe-db.yaml" > "$rendered_manifest"
  sed -E "s#^([[:space:]]*storageClassName:).*#\\1 ${storage_class}#" "$rendered_manifest" > "${rendered_manifest}.tmp"
  mv "${rendered_manifest}.tmp" "$rendered_manifest"
}

cnpg_attempt_ready() {
  local expected_storage_class="$1"
  local pfx current_storage_class
  pfx="${CNPG_CLUSTER_NAME}-"
  current_storage_class="$(kubectl -n "$CNPG_NAMESPACE" get cluster "$CNPG_CLUSTER_NAME" -o jsonpath='{.spec.storage.pvcTemplate.storageClassName}' 2>/dev/null || true)"
  [[ "$current_storage_class" == "$expected_storage_class" ]] || return 1

  kubectl -n "$CNPG_NAMESPACE" get pvc -o jsonpath='{range .items[*]}{.metadata.name}{"|"}{.spec.storageClassName}{"|"}{.status.phase}{"\n"}{end}' 2>/dev/null \
    | awk -F'|' -v expected="$expected_storage_class" -v prefix="$pfx" '$1 ~ ("^" prefix) && $2 == expected && $3 == "Bound" { found=1 } END { exit(found ? 0 : 1) }'
}

cnpg_attempt_has_pvc_for_class() {
  local expected_storage_class="$1"
  local pfx
  pfx="${CNPG_CLUSTER_NAME}-"
  kubectl -n "$CNPG_NAMESPACE" get pvc -o jsonpath='{range .items[*]}{.metadata.name}{"|"}{.spec.storageClassName}{"\n"}{end}' 2>/dev/null \
    | awk -F'|' -v expected="$expected_storage_class" -v prefix="$pfx" '$1 ~ ("^" prefix) && $2 == expected { found=1 } END { exit(found ? 0 : 1) }'
}

cnpg_recent_events() {
  kubectl -n "$CNPG_NAMESPACE" get events --sort-by=.lastTimestamp 2>/dev/null | tail -n 80 || true
}

wait_for_storage_outcome() {
  local expected_storage_class="$1"
  local started_at current_ts events
  started_at="$(date +%s)"

  while true; do
    if cnpg_attempt_ready "$expected_storage_class"; then
      return 0
    fi

    if cnpg_attempt_has_pvc_for_class "$expected_storage_class"; then
      events="$(cnpg_recent_events)"
      if printf '%s\n' "$events" | grep -q "SSD_TOTAL_GB"; then
        return 10
      fi
      if printf '%s\n' "$events" | grep -Eqi "storageclass.*not found|failed to provision volume"; then
        return 11
      fi
    fi

    current_ts="$(date +%s)"
    if (( current_ts - started_at >= CNPG_STORAGE_WAIT_TIMEOUT )); then
      return 1
    fi
    sleep 10
  done
}

reset_cnpg_cluster_attempt() {
  kubectl -n "$CNPG_NAMESPACE" delete cluster "$CNPG_CLUSTER_NAME" --ignore-not-found=true --wait=true --timeout=180s >/dev/null 2>&1 || true
  kubectl -n "$CNPG_NAMESPACE" delete job -l cnpg.io/cluster="$CNPG_CLUSTER_NAME" --ignore-not-found=true >/dev/null 2>&1 || true
  kubectl -n "$CNPG_NAMESPACE" delete pvc -l cnpg.io/cluster="$CNPG_CLUSTER_NAME" --ignore-not-found=true >/dev/null 2>&1 || true
  kubectl -n "$CNPG_NAMESPACE" delete pod -l cnpg.io/cluster="$CNPG_CLUSTER_NAME" --ignore-not-found=true >/dev/null 2>&1 || true
}

ensure_cnpg_tls_secrets() {
  local cluster_name="knoe-db"
  if kubectl -n "$CNPG_NAMESPACE" get secret "${cluster_name}-ca" >/dev/null 2>&1 \
    && kubectl -n "$CNPG_NAMESPACE" get secret "${cluster_name}-tls" >/dev/null 2>&1; then
    local ca_key_b64
    local ca_key_pem
    ca_key_b64="$(kubectl -n "$CNPG_NAMESPACE" get secret "${cluster_name}-ca" -o jsonpath='{.data.ca\.key}' 2>/dev/null || true)"
    ca_key_pem="$(printf '%s' "$ca_key_b64" | openssl base64 -d -A 2>/dev/null || true)"
    if [[ -n "$ca_key_b64" && "$ca_key_pem" == *"BEGIN EC PRIVATE KEY"* ]]; then
      log "CNPG TLS secrets already present in ${CNPG_NAMESPACE}."
      return
    fi
    log "CNPG CA secret is missing/invalid for CNPG in ${CNPG_NAMESPACE}; regenerating CNPG TLS secrets ..."
  fi

  log "Bootstrapping CNPG TLS secrets in ${CNPG_NAMESPACE} ..."
  local tmp_dir
  tmp_dir="$(mktemp -d)"
  local fqdn cn
  fqdn="${cluster_name}-rw.${CNPG_NAMESPACE}.svc.cluster.local"
  cn="$fqdn"
  if [[ ${#cn} -gt 64 ]]; then
    cn="${cluster_name}-rw"
  fi

  openssl ecparam -name prime256v1 -genkey -noout -out "$tmp_dir/ca.key" >/dev/null 2>&1
  openssl req -x509 -new -nodes -key "$tmp_dir/ca.key" -sha256 -days 3650 \
    -subj "/CN=${cluster_name}-ca" -out "$tmp_dir/ca.crt" >/dev/null 2>&1

  openssl ecparam -name prime256v1 -genkey -noout -out "$tmp_dir/tls.key" >/dev/null 2>&1
  openssl req -new -key "$tmp_dir/tls.key" -subj "/CN=${cn}" -out "$tmp_dir/tls.csr" >/dev/null 2>&1

  cat > "$tmp_dir/ext.cnf" <<EOF
subjectAltName=DNS:${fqdn},DNS:${cluster_name},DNS:localhost,IP:127.0.0.1
extendedKeyUsage=serverAuth
EOF

  openssl x509 -req -in "$tmp_dir/tls.csr" -CA "$tmp_dir/ca.crt" -CAkey "$tmp_dir/ca.key" \
    -CAcreateserial -out "$tmp_dir/tls.crt" -days 3650 -sha256 -extfile "$tmp_dir/ext.cnf" >/dev/null 2>&1

  kubectl -n "$CNPG_NAMESPACE" create secret generic "${cluster_name}-ca" \
    --from-file=ca.crt="$tmp_dir/ca.crt" \
    --from-file=ca.key="$tmp_dir/ca.key" \
    --dry-run=client -o yaml | kubectl apply -f -

  kubectl -n "$CNPG_NAMESPACE" create secret tls "${cluster_name}-tls" \
    --cert="$tmp_dir/tls.crt" \
    --key="$tmp_dir/tls.key" \
    --dry-run=client -o yaml | kubectl apply -f -

  rm -rf "$tmp_dir"
}

ensure_db_user_secret() {
  if kubectl -n "$CNPG_NAMESPACE" get secret knoe-db-user >/dev/null 2>&1; then
    log "knoe-db-user secret already present in ${CNPG_NAMESPACE}."
  else
    [[ -n "$DB_PASSWORD" && "$DB_PASSWORD" != \$\{* ]] || {
      echo "ERROR: secret 'knoe-db-user' missing in '${CNPG_NAMESPACE}' and DB_PASSWORD is not set." >&2
      exit 1
    }
    log "Creating missing knoe-db-user secret in ${CNPG_NAMESPACE} ..."
    kubectl -n "$CNPG_NAMESPACE" create secret generic knoe-db-user \
      --from-literal=username="$CNPG_DB_OWNER" \
      --from-literal=password="$DB_PASSWORD" \
      --dry-run=client -o yaml | kubectl apply -f -
  fi

  if kubectl -n "$CNPG_NAMESPACE" get secret knoe-db-superuser >/dev/null 2>&1; then
    return
  fi
  [[ -n "$DB_PASSWORD" && "$DB_PASSWORD" != \$\{* ]] || return
  log "Creating missing knoe-db-superuser secret in ${CNPG_NAMESPACE} ..."
  kubectl -n "$CNPG_NAMESPACE" create secret generic knoe-db-superuser \
    --from-literal=username=postgres \
    --from-literal=password="$DB_PASSWORD" \
    --dry-run=client -o yaml | kubectl apply -f -
}

# ── Tool checks ──────────────────────────────────────────────────────────────
for tool in gcloud kubectl gsutil openssl; do
  command -v "$tool" >/dev/null || { echo "ERROR: $tool not found in PATH." >&2; exit 1; }
done

# ── 1. Acquire kubeconfig ────────────────────────────────────────────────────
log "Fetching kubeconfig for $GKE_CLUSTER in $GCP_REGION ..."
gcloud container clusters get-credentials "$GKE_CLUSTER" \
  --project "$GCP_PROJECT_ID" \
  --region  "$GCP_REGION"

# ── 2. Create GCS buckets ────────────────────────────────────────────────────
for bucket in "$GCS_BACKUP_BUCKET" "$GCS_WAL_BUCKET"; do
  if gsutil ls -b "gs://$bucket" >/dev/null 2>&1; then
    log "Bucket gs://$bucket already exists."
  else
    log "Creating bucket gs://$bucket ..."
    gsutil mb -p "$GCP_PROJECT_ID" -l "$GCP_REGION" "gs://$bucket"
    gsutil versioning set on "gs://$bucket"
    gsutil lifecycle set /dev/stdin "gs://$bucket" <<EOF
{"rule":[{"action":{"type":"Delete"},"condition":{"age":30}}]}
EOF
  fi
done

# ── 3. Create GCP service account ────────────────────────────────────────────
if gcloud iam service-accounts describe "$GCP_SA_EMAIL" --project "$GCP_PROJECT_ID" >/dev/null 2>&1; then
  log "Service account $GCP_SA_EMAIL already exists."
else
  log "Creating service account $GCP_SA_EMAIL ..."
  gcloud iam service-accounts create "$CNPG_BACKUP_SA" \
    --display-name="CNPG GCS Backup" \
    --project="$GCP_PROJECT_ID"
fi

# IAM propagation is eventually consistent right after service-account creation.
# Wait briefly so downstream gsutil IAM grants don't fail with transient 400.
for i in $(seq 1 12); do
  if gcloud iam service-accounts describe "$GCP_SA_EMAIL" --project "$GCP_PROJECT_ID" >/dev/null 2>&1; then
    break
  fi
  log "Waiting for service account propagation ($i/12) ..."
  sleep 5
done

# ── 4. Grant storage access ───────────────────────────────────────────────────
log "Granting objectAdmin on backup buckets to $GCP_SA_EMAIL ..."
for bucket in "$GCS_BACKUP_BUCKET" "$GCS_WAL_BUCKET"; do
  ok=0
  for i in $(seq 1 6); do
    if gsutil iam ch \
      "serviceAccount:${GCP_SA_EMAIL}:objectAdmin" \
      "gs://$bucket"; then
      ok=1
      break
    fi
    log "WARN: IAM grant failed for gs://$bucket (attempt $i/6); retrying in 5s ..."
    sleep 5
  done
  [[ $ok -eq 1 ]] || {
    echo "ERROR: failed to grant objectAdmin on gs://$bucket to $GCP_SA_EMAIL" >&2
    exit 1
  }
done

# ── 5. Bind Workload Identity ─────────────────────────────────────────────────
log "Binding Workload Identity ..."
gcloud iam service-accounts add-iam-policy-binding "$GCP_SA_EMAIL" \
  --role=roles/iam.workloadIdentityUser \
  --member="serviceAccount:${GCP_PROJECT_ID}.svc.id.goog[${CNPG_NAMESPACE}/cnpg-backup-sa]" \
  --project="$GCP_PROJECT_ID"

# ── 6. Apply namespace ────────────────────────────────────────────────────────
log "Applying namespace $CNPG_NAMESPACE ..."
kubectl apply -f "$GKE_MANIFEST_DIR/namespace.yaml"

# ── 7. Install CNPG operator (if not present) ─────────────────────────────────
if ! kubectl get crd clusters.postgresql.cnpg.io >/dev/null 2>&1; then
  log "Installing CloudNativePG operator v${CNPG_OPERATOR_VERSION} ..."
  kubectl apply --server-side -f \
    "https://raw.githubusercontent.com/cloudnative-pg/cloudnative-pg/main/releases/cnpg-${CNPG_OPERATOR_VERSION}.yaml"
  log "Waiting for CNPG operator to be ready ..."
  kubectl rollout status deployment/cnpg-controller-manager -n cnpg-system --timeout=120s
else
  log "CNPG operator already installed."
fi

# ── 8. Apply ServiceAccount + annotate with WI ───────────────────────────────
log "Applying CNPG backup ServiceAccount with Workload Identity annotation ..."
GCP_PROJECT_ID="$GCP_PROJECT_ID" envsubst '${GCP_PROJECT_ID}' < "$GKE_MANIFEST_DIR/knoe-db-backup-gcs.yaml" \
  | kubectl apply -f -

kubectl annotate serviceaccount cnpg-backup-sa \
  -n "$CNPG_NAMESPACE" \
  "iam.gke.io/gcp-service-account=${GCP_SA_EMAIL}" \
  --overwrite

# ── 9. Bootstrap CNPG secrets ───────────────────────────────────────────────
ensure_cnpg_tls_secrets
ensure_db_user_secret

# ── 10. Apply CNPG cluster ───────────────────────────────────────────────────
IFS_COMMA=","
resolve_knoe_db_image_tag
log "Using knoe-db image tag: ${KNOE_DB_IMAGE_TAG}"
resolve_storage_class_candidates
IFS=',' read -r -a storage_classes <<< "$CNPG_STORAGE_CLASS_CANDIDATES_RESOLVED"
selected_storage_class=""
attempt_total="${#storage_classes[@]}"
attempt_index=0

if kubectl -n "$CNPG_NAMESPACE" get cluster "$CNPG_CLUSTER_NAME" >/dev/null 2>&1; then
  existing_phase="$(kubectl -n "$CNPG_NAMESPACE" get cluster "$CNPG_CLUSTER_NAME" -o jsonpath='{.status.phase}' 2>/dev/null || true)"
  if [[ "$existing_phase" != "Cluster in healthy state" ]]; then
    log "Existing cluster phase ${existing_phase:-unknown} is not healthy; resetting stale resources before storage-class evaluation."
    reset_cnpg_cluster_attempt
  fi
fi

for storage_class in "${storage_classes[@]}"; do
  storage_class="${storage_class//[[:space:]]/}"
  [[ -n "$storage_class" ]] || continue
  attempt_index=$((attempt_index + 1))

  log "Applying knoe-db CNPG cluster with storageClass=${storage_class} (${attempt_index}/${attempt_total}) ..."
  tmp_manifest="$(mktemp)"
  render_cnpg_manifest "$storage_class" "$tmp_manifest"
  kubectl apply -f "$tmp_manifest"
  rm -f "$tmp_manifest"

  if wait_for_storage_outcome "$storage_class"; then
    selected_storage_class="$storage_class"
    break
  fi

  outcome_rc=$?
  case "$outcome_rc" in
    10)
      log "WARN: storageClass=${storage_class} failed due to SSD quota pressure; stepping down."
      ;;
    11)
      log "WARN: storageClass=${storage_class} failed provisioning; stepping down."
      ;;
    *)
      log "WARN: storageClass=${storage_class} did not become ready within ${CNPG_STORAGE_WAIT_TIMEOUT}s; stepping down."
      ;;
  esac

  reset_cnpg_cluster_attempt
done

[[ -n "$selected_storage_class" ]] || {
  echo "ERROR: CNPG failed to provision across storage class candidates: ${CNPG_STORAGE_CLASS_CANDIDATES_RESOLVED}" >&2
  exit 1
}

log "CNPG provisioning selected storageClass=${selected_storage_class}."

# ── 11. Ensure RoleBindings reference cnpg-backup-sa ─────────────────────────
# CNPG auto-creates knoe-db and knoe-db-barman-cloud RoleBindings at cluster
# creation. With spec.serviceAccountName set from the start the operator may
# use cnpg-backup-sa as the sole subject, but if it still uses the default SA
# name (cluster name) we patch both bindings to add cnpg-backup-sa — matching
# the live state after the 2026-04-29 stabilization session.
log "Ensuring CNPG-managed RoleBindings include cnpg-backup-sa ..."
for rb in "${CNPG_CLUSTER_NAME}" "${CNPG_CLUSTER_NAME}-barman-cloud"; do
  for _ in 1 2 3 4 5; do
    kubectl -n "${CNPG_NAMESPACE}" get rolebinding "${rb}" >/dev/null 2>&1 && break
    sleep 2
  done
  if kubectl -n "${CNPG_NAMESPACE}" get rolebinding "${rb}" >/dev/null 2>&1; then
    if ! kubectl -n "${CNPG_NAMESPACE}" get rolebinding "${rb}" \
         -o jsonpath='{.subjects[*].name}' | grep -qw cnpg-backup-sa; then
      patch_file="$(mktemp)"
      cat > "${patch_file}" <<EOF
[{"op":"add","path":"/subjects/-","value":{"kind":"ServiceAccount","name":"cnpg-backup-sa","namespace":"${CNPG_NAMESPACE}"}}]
EOF
      kubectl -n "${CNPG_NAMESPACE}" patch rolebinding "${rb}" \
        --type='json' --patch-file="${patch_file}" || true
      rm -f "${patch_file}"
      log "  Patched RoleBinding ${rb} to add cnpg-backup-sa."
    else
      log "  RoleBinding ${rb} already includes cnpg-backup-sa."
    fi
  else
    log "  WARNING: RoleBinding ${rb} not found — skipping patch."
  fi
done

log ""
log "Done. Monitor with:"
log "  kubectl -n $CNPG_NAMESPACE get cluster knoe-db"
log "  kubectl -n $CNPG_NAMESPACE get pods -l cnpg.io/cluster=knoe-db"
log "  kubectl -n $CNPG_NAMESPACE get scheduledbackup"