prole/etc/init_cloudnative_pg.sh
chrisfu f47c18fef7 feat(infrastructure): enhance k3s automation and OpenTofu integration
- Infrastructure:
    - Updated k3s Ansible role with mountpoint preflight checks and better permission management.
    - Automated deployment of prole configuration and port-forwarding scripts to cluster hosts.
    - Added systemd service for managing port forwards on k3s nodes.
    - Added prole-installer service account token automation.
- K8s Manifests:
    - Renamed and added Persistent Volumes in iscsi-pvs.yaml (including OpenBao support).
    - Updated StatefulSets for garage and openbao.
    - Migrated prole-db to CloudNativePG-based configuration.
    - Added comprehensive OpenTofu manifests for cluster deployment.
- Configuration:
    - Added cluster-specific configurations (k3d, k3s-hosts).
    - Added PostgreSQL configuration templates.
    - Updated .gitignore to track the conf/ directory.
- Tools:
    - Updated install.py and port-forwarding scripts.
    - Added render_manifest.py for manifest generation.
2026-02-07 22:53:18 -08:00

972 lines
31 KiB
Bash
Executable File
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

#!/usr/bin/env bash
set -euo pipefail
# init_cloudnative_pg.sh
# Purpose:
# - Distribute administrator ed25519 key pair to CloudNative‑PG as a Kubernetes Secret for cert auth
# - Configure Kerberos (GSSAPI) using external Kerberos KDC
# - Patch CNPG cluster to enable TLS and GSSAPI where possible
#
# Usage:
# ./init_cloudnative_pg.sh start|stop|status|restart
# ./init_cloudnative_pg.sh initialize # install CNPG, create cluster, configure secrets + Kerberos
# ./init_cloudnative_pg.sh update|reload # re-apply/patch
#
# Requirements:
# - init_openbao.sh has been run (OpenBao running as a local container)
# - $PROLE_HOME/env.sh or $HOME/.prole/env.sh defining PROLE_SERVICE
# - Optional: CNPG_MANIFEST_OVERRIDE to apply a recovery manifest instead of kustomize
# Initialize SCRIPT_DIR
SCRIPT_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
# Load environment and config via prole_cfg.sh
# shellcheck disable=SC1090
source "$SCRIPT_DIR/prole_cfg.sh"
if [[ "${1:-}" == "--mode" || "${1:-}" == "-m" ]]; then
prole_set_mode "${2:-}"
shift 2
elif [[ "${1:-}" == --mode=* || "${1:-}" == -m=* ]]; then
prole_set_mode "${1#*=}"
shift
fi
if [[ -z "${PROLE_SERVICE:-}" ]]; then
echo "ERROR: PROLE_SERVICE is not defined. Provide PROLE_HOME/env.sh or ~/.prole/env.sh" >&2
exit 1
fi
ACTION=${1:-}
CNPG_CLUSTER_NAME=${2:-${CNPG_CLUSTER_NAME:-prole-db}}
OPENBAO_NAME=${OPENBAO_NAME:-openbao}
REALM=${REALM:-PROLE.ORG}
DOMAIN=${DOMAIN:-prole.org}
KRB5_KDC=${KRB5_KDC:-}
KRB5_ADMIN=${KRB5_ADMIN:-}
CNPG_MANIFEST_OVERRIDE=${CNPG_MANIFEST_OVERRIDE:-}
PROLE_HOME=${PROLE_HOME:-$(cd "$SCRIPT_DIR/.." && pwd)}
BACKUP_DIR=${BACKUP_DIR:-$PROLE_HOME/prole/backup}
BACKUP_WAIT_TIMEOUT=${BACKUP_WAIT_TIMEOUT:-1800}
RECOVERY_TEMPLATE="$SCRIPT_DIR/../k8s/prole/prole-db-recovery.yaml.tpl"
CNPG_FORCE_ROLLOUT_SCRIPT="$SCRIPT_DIR/init_prole-db.sh"
BARMAN_PLUGIN_MANIFEST_URL=${BARMAN_PLUGIN_MANIFEST_URL:-}
BARMAN_PLUGIN_FALLBACK_VERSION=${BARMAN_PLUGIN_FALLBACK_VERSION:-0.9.0}
CERT_MANAGER_MANIFEST_URL=${CERT_MANAGER_MANIFEST_URL:-}
CERT_MANAGER_FALLBACK_VERSION=${CERT_MANAGER_FALLBACK_VERSION:-1.19.3}
# Support both PROLE_HOME/k8s and sibling k8s directory
if [[ -d "$SCRIPT_DIR/../k8s/prole" ]]; then
K8S_PROLE_DIR="$SCRIPT_DIR/../k8s/prole"
elif [[ -n "${PROLE_HOME:-}" && -d "$PROLE_HOME/k8s/prole" ]]; then
K8S_PROLE_DIR="$PROLE_HOME/k8s/prole"
else
K8S_PROLE_DIR="$SCRIPT_DIR/../k8s/prole"
fi
CNPG_MANIFEST="$K8S_PROLE_DIR/prole-db.yaml"
BARMAN_OBJECTSTORE_MANIFEST="$K8S_PROLE_DIR/prole-db-barman-objectstore.yaml"
SECRETS_DIR="$PROLE_SERVICE/secrets"
# Resolving CNPG admin keys.
# We prefer names without algorithm suffixes to be more generic, matching install.py fallback strategy.
ADMIN_PRIV_ED25519="$SECRETS_DIR/admin_ed25519.key"
ADMIN_PUB_ED25519="$SECRETS_DIR/admin_ed25519.pub"
ADMIN_PRIV_GENERIC="$SECRETS_DIR/admin.key"
ADMIN_PUB_GENERIC="$SECRETS_DIR/admin.pub"
OPENBAO_TOKEN_FILE="$SECRETS_DIR/openbao-root-token"
BAO_NAMESPACE="${NAMESPACE:-default}"
BAO_PATH_PREFIX="prole/${BAO_NAMESPACE}"
BAO_PATH_ADMIN="${BAO_PATH_PREFIX}/admin"
BAO_PATH_DB="${BAO_PATH_PREFIX}/db"
BAO_PATH_MONITORING="${BAO_PATH_PREFIX}/monitoring"
ensure_tools() {
for t in kubectl curl openssl base64 jq; do
command -v "$t" >/dev/null || { echo "Missing required tool: $t" >&2; exit 1; }
done
}
ensure_namespace() {
if [[ -z "${NAMESPACE:-}" ]]; then
echo "ERROR: NAMESPACE is empty. Check env.sh or conf/prole.cfg." >&2
exit 1
fi
if ! kubectl get namespace "$NAMESPACE" >/dev/null 2>&1; then
echo "Creating namespace '$NAMESPACE' ..."
kubectl create namespace "$NAMESPACE" >/dev/null 2>&1 || true
fi
}
openbao_url() {
if [[ -n "${PROLE_OPENBAO_URL:-}" ]]; then
echo "$PROLE_OPENBAO_URL"
elif curl -sS "http://127.0.0.1:18200/v1/sys/health" >/dev/null 2>&1; then
echo "http://127.0.0.1:18200"
else
echo "http://openbao.${NAMESPACE}.svc.cluster.local:8200"
fi
}
openbao_token() {
if [[ -f "$OPENBAO_TOKEN_FILE" ]]; then
cat "$OPENBAO_TOKEN_FILE"
else
echo "${OPENBAO_ROOT_TOKEN:-}"
fi
}
fetch_openbao_secret() {
local path="$1"
local key="$2"
local token url
token=$(openbao_token)
url=$(openbao_url)
if [[ -z "$token" ]]; then
echo ""
return 0
fi
curl -sS -H "X-Vault-Token: $token" "$url/v1/kv/data/$path" | jq -r ".data.data.\"$key\"" || echo ""
}
resolve_db_password() {
local db_pw="${DB_PASSWORD:-}"
if [[ -z "$db_pw" || "$db_pw" == '${OPENBAO:'* || "$db_pw" == '${PROLE_SECRET:'* ]]; then
local fetched_db
fetched_db=$(fetch_openbao_secret "$BAO_PATH_DB" "password")
if [[ -n "$fetched_db" && "$fetched_db" != "null" ]]; then
db_pw="$fetched_db"
fi
fi
printf '%s' "$db_pw"
}
ensure_grafana_admin_secret() {
local db_pw
db_pw=$(resolve_db_password)
if [[ -z "$db_pw" ]]; then
echo "ERROR: DB_PASSWORD unavailable; cannot create Grafana admin secret." >&2
return 1
fi
kubectl create secret generic grafana-admin \
-n "$NAMESPACE" \
--from-literal=admin-password="$db_pw" \
--dry-run=client -o yaml | kubectl apply -f - >/dev/null
local token url
token=$(openbao_token)
url=$(openbao_url)
if [[ -n "$token" ]]; then
curl -sS -H "X-Vault-Token: $token" -H 'Content-Type: application/json' \
-X POST "$url/v1/kv/data/$BAO_PATH_MONITORING" \
-d "{\"data\":{\"grafana_admin_password\":\"$db_pw\"}}" >/dev/null || true
fi
}
ensure_cnpg_operator() {
if kubectl get deployment -n cnpg-system cnpg-controller-manager >/dev/null 2>&1; then
kubectl -n cnpg-system rollout status deploy/cnpg-controller-manager --timeout=180s || true
wait_for_cnpg_webhook 180 || true
return 0
fi
local latest_version minor_version yaml_url
latest_version=$(get_latest_cnpg_version)
minor_version=$(echo "$latest_version" | cut -d. -f1,2)
yaml_url="https://raw.githubusercontent.com/cloudnative-pg/cloudnative-pg/release-${minor_version}/releases/cnpg-${latest_version}.yaml"
echo "Installing CloudNative-PG operator version ${latest_version} ..."
kubectl apply --server-side -f "$yaml_url"
if kubectl get deployment -n cnpg-system cnpg-controller-manager >/dev/null 2>&1; then
kubectl -n cnpg-system rollout status deploy/cnpg-controller-manager --timeout=180s || true
wait_for_cnpg_webhook 180 || true
fi
}
get_latest_barman_plugin_version() {
local version tag
tag=$(curl -s "https://api.github.com/repos/cloudnative-pg/plugin-barman-cloud/releases/latest" | jq -r '.tag_name' || echo "")
if [[ -z "$tag" || "$tag" == "null" ]]; then
echo "v${BARMAN_PLUGIN_FALLBACK_VERSION}"
return 0
fi
echo "$tag"
}
resolve_barman_plugin_manifest_url() {
if [[ -n "$BARMAN_PLUGIN_MANIFEST_URL" ]]; then
echo "$BARMAN_PLUGIN_MANIFEST_URL"
return 0
fi
local tag
tag=$(get_latest_barman_plugin_version)
echo "https://github.com/cloudnative-pg/plugin-barman-cloud/releases/download/${tag}/manifest.yaml"
}
cert_manager_ready() {
if ! kubectl get crd certificates.cert-manager.io >/dev/null 2>&1; then
return 1
fi
if ! kubectl -n cert-manager get deploy cert-manager >/dev/null 2>&1; then
return 1
fi
return 0
}
get_latest_cert_manager_version() {
local tag
tag=$(curl -s "https://api.github.com/repos/cert-manager/cert-manager/releases/latest" | jq -r '.tag_name' || echo "")
if [[ -z "$tag" || "$tag" == "null" ]]; then
echo "v${CERT_MANAGER_FALLBACK_VERSION}"
return 0
fi
echo "$tag"
}
resolve_cert_manager_manifest_url() {
if [[ -n "$CERT_MANAGER_MANIFEST_URL" ]]; then
echo "$CERT_MANAGER_MANIFEST_URL"
return 0
fi
local tag
tag=$(get_latest_cert_manager_version)
echo "https://github.com/cert-manager/cert-manager/releases/download/${tag}/cert-manager.yaml"
}
ensure_cert_manager() {
if cert_manager_ready; then
if kubectl -n cert-manager get deploy cert-manager >/dev/null 2>&1; then
kubectl -n cert-manager rollout status deploy/cert-manager --timeout=180s || true
kubectl -n cert-manager rollout status deploy/cert-manager-webhook --timeout=180s || true
kubectl -n cert-manager rollout status deploy/cert-manager-cainjector --timeout=180s || true
fi
return 0
fi
local cm_url
cm_url=$(resolve_cert_manager_manifest_url)
echo "Installing cert-manager from ${cm_url} ..."
kubectl apply -f "$cm_url"
if kubectl -n cert-manager get deploy cert-manager >/dev/null 2>&1; then
kubectl -n cert-manager rollout status deploy/cert-manager --timeout=180s || true
kubectl -n cert-manager rollout status deploy/cert-manager-webhook --timeout=180s || true
kubectl -n cert-manager rollout status deploy/cert-manager-cainjector --timeout=180s || true
fi
}
wait_for_barman_crd() {
local timeout=${1:-120}
local start_time
start_time=$(date +%s)
while true; do
if kubectl get crd objectstores.barmancloud.cnpg.io >/dev/null 2>&1; then
return 0
fi
if (( $(date +%s) - start_time > timeout )); then
return 1
fi
sleep 5
done
}
wait_for_barman_tls_secrets() {
local timeout=${1:-180}
local start_time
start_time=$(date +%s)
while true; do
if kubectl -n cnpg-system get secret barman-cloud-client-tls >/dev/null 2>&1 \
&& kubectl -n cnpg-system get secret barman-cloud-server-tls >/dev/null 2>&1; then
return 0
fi
if (( $(date +%s) - start_time > timeout )); then
return 1
fi
sleep 5
done
}
ensure_barman_plugin() {
ensure_cert_manager
local plugin_url
plugin_url=$(resolve_barman_plugin_manifest_url)
echo "Installing Barman Cloud plugin from ${plugin_url} ..."
kubectl apply -f "$plugin_url"
if ! wait_for_barman_crd 120; then
echo "WARN: Barman Cloud ObjectStore CRD not ready after install." >&2
fi
if ! wait_for_barman_tls_secrets 180; then
echo "WARN: Barman Cloud TLS secrets not ready after install." >&2
fi
if kubectl -n cnpg-system get deploy barman-cloud >/dev/null 2>&1; then
kubectl -n cnpg-system rollout status deploy/barman-cloud --timeout=180s || true
fi
}
pin_cnpg_controller() {
local selector="${CNPG_CONTROLLER_NODE_SELECTOR:-}"
if [[ -z "$selector" && "${PROLE_MODE:-}" == "k3s" ]]; then
selector="storage=primary"
fi
if [[ -z "$selector" ]]; then
return 0
fi
local key value
key="${selector%%=*}"
value="${selector#*=}"
if [[ -z "$key" || -z "$value" ]]; then
echo "WARN: CNPG_CONTROLLER_NODE_SELECTOR must be key=value (got '$selector'). Skipping pin." >&2
return 0
fi
if ! kubectl get nodes -l "${key}=${value}" >/dev/null 2>&1; then
echo "WARN: No nodes match CNPG_CONTROLLER_NODE_SELECTOR=${selector}. Skipping pin." >&2
return 0
fi
echo "Pinning cnpg-controller-manager to nodes with ${selector} ..."
kubectl -n cnpg-system patch deployment cnpg-controller-manager --type merge \
-p "{\"spec\":{\"template\":{\"spec\":{\"nodeSelector\":{\"${key}\":\"${value}\"}}}}}" >/dev/null 2>&1 || true
}
wait_for_cnpg_webhook() {
local timeout=${1:-120}
local start_time
start_time=$(date +%s)
echo "Waiting for CNPG webhook service endpoints to be ready..."
while true; do
local endpoints
endpoints=$(kubectl -n cnpg-system get endpoints cnpg-webhook-service -o jsonpath='{.subsets[*].addresses[*].ip}' 2>/dev/null || true)
if [[ -n "$endpoints" ]]; then
echo "CNPG webhook service has endpoints."
return 0
fi
if (( $(date +%s) - start_time > timeout )); then
echo "WARN: CNPG webhook endpoints not ready after ${timeout}s."
return 1
fi
sleep 5
done
}
apply_barman_objectstore_if_present() {
if [[ ! -f "$BARMAN_OBJECTSTORE_MANIFEST" ]]; then
return 0
fi
if kubectl get crd objectstores.barmancloud.cnpg.io >/dev/null 2>&1; then
prole_render_manifest "$BARMAN_OBJECTSTORE_MANIFEST" | kubectl apply -n "$NAMESPACE" -f -
else
echo "WARN: Barman Cloud ObjectStore CRD not found; skipping $BARMAN_OBJECTSTORE_MANIFEST."
fi
}
apply_prole_manifest_file() {
local file="$1"
local output=""
if output=$(prole_render_manifest "$file" | kubectl apply -n "$NAMESPACE" -f - 2>&1); then
printf '%s\n' "$output"
return 0
fi
if [[ "${PROLE_MODE:-}" == "k3d" && "$(basename "$file")" == "garage-statefulset.yaml" ]] \
&& echo "$output" | grep -q "updates to statefulset spec"; then
echo "WARN: Garage StatefulSet immutable in k3d; skipping apply."
return 0
fi
echo "$output" >&2
return 1
}
ensure_prole_stack_resources() {
echo "Applying CloudNative-PG cluster and related resources ..."
ensure_grafana_admin_secret
if [[ -n "$CNPG_MANIFEST_OVERRIDE" ]]; then
if [[ ! -f "$CNPG_MANIFEST_OVERRIDE" ]]; then
echo "ERROR: CNPG_MANIFEST_OVERRIDE not found: $CNPG_MANIFEST_OVERRIDE" >&2
return 1
fi
local dir file
dir="$K8S_PROLE_DIR"
for file in "$dir"/*.yaml; do
case "$(basename "$file")" in
prole-db.yaml|kustomization.yaml|supabase-*.yaml|prole-db-barman-objectstore.yaml)
continue
;;
esac
apply_prole_manifest_file "$file"
done
apply_barman_objectstore_if_present
apply_cnpg_cluster_manifest "$CNPG_MANIFEST_OVERRIDE"
else
local dir file
dir="$K8S_PROLE_DIR"
for file in "$dir"/*.yaml; do
case "$(basename "$file")" in
prole-db.yaml|kustomization.yaml|supabase-*.yaml|prole-db-barman-objectstore.yaml)
continue
;;
esac
apply_prole_manifest_file "$file"
done
apply_barman_objectstore_if_present
apply_cnpg_cluster_manifest "$CNPG_MANIFEST"
fi
# Ensure prole-index-html exists for prole deployment readiness probe
if ! kubectl get configmap prole-index-html -n "$NAMESPACE" >/dev/null 2>&1; then
echo "Creating prole-index-html configmap..."
printf "<html><body><h1>Prole</h1></body></html>" > /tmp/index.html
kubectl create configmap prole-index-html --from-file=/tmp/index.html -n "$NAMESPACE"
rm /tmp/index.html
fi
# Ensure prole-nginx-tls exists (self-signed for dev)
if ! kubectl get secret prole-nginx-tls -n "$NAMESPACE" >/dev/null 2>&1; then
echo "Generating self-signed prole-nginx-tls for development..."
openssl req -x509 -nodes -days 365 -newkey rsa:2048 \
-keyout /tmp/nginx-tls.key -out /tmp/nginx-tls.crt \
-subj "/CN=prole.org" >/dev/null 2>&1
kubectl create secret tls prole-nginx-tls --key /tmp/nginx-tls.key --cert /tmp/nginx-tls.crt -n "$NAMESPACE"
rm /tmp/nginx-tls.key /tmp/nginx-tls.crt
fi
}
apply_cnpg_cluster_manifest() {
local manifest="$1"
local attempts=${CNPG_APPLY_RETRIES:-6}
local i out
for ((i=1; i<=attempts; i++)); do
if out=$(prole_render_manifest "$manifest" | kubectl apply -n "$NAMESPACE" -f - 2>&1); then
printf '%s\n' "$out"
return 0
fi
if echo "$out" | grep -q "cnpg-webhook-service"; then
echo "CNPG webhook not ready yet (attempt $i/$attempts). Retrying..."
sleep 5
continue
fi
echo "$out" >&2
return 1
done
echo "ERROR: Failed to apply CNPG manifest after $attempts attempts." >&2
echo "$out" >&2
return 1
}
wait_for_cnpg_pods() {
local timeout=${1:-300}
local start_time
start_time=$(date +%s)
local target_pods="${CNPG_TARGET_PODS:-}"
if [[ -z "$target_pods" ]]; then
target_pods=$(kubectl -n "$NAMESPACE" get cluster "$CNPG_CLUSTER_NAME" -o jsonpath='{.spec.instances}' 2>/dev/null || true)
fi
if [[ -z "$target_pods" ]]; then
target_pods=3
fi
echo "Waiting for $target_pods CNPG pods for cluster '$CNPG_CLUSTER_NAME' in namespace '$NAMESPACE' to be Running..."
while true; do
local pods
pods=$(kubectl -n "$NAMESPACE" get pods -l "cnpg.io/cluster=$CNPG_CLUSTER_NAME" --no-headers 2>/dev/null || true)
if [[ -n "$pods" ]]; then
# Check for Error or CrashLoopBackOff
if echo "$pods" | grep -E "Error|CrashLoopBackOff" >/dev/null; then
echo "ERROR: Some CNPG pods are in Error or CrashLoopBackOff state:" >&2
echo "$pods" | grep -E "Error|CrashLoopBackOff" >&2
return 1
fi
# Count Running pods by name (exclude initdb)
local running_pods
running_pods=$(kubectl -n "$NAMESPACE" get pods --no-headers 2>/dev/null \
| grep "^${CNPG_CLUSTER_NAME}-" \
| grep -v "initdb" \
| grep "Running" \
| wc -l | xargs)
if [[ "$running_pods" -ge "$target_pods" ]]; then
echo "All $running_pods/$target_pods pods are Running."
return 0
fi
fi
if (( $(date +%s) - start_time > timeout )); then
echo "ERROR: Timed out waiting for $target_pods CNPG pods for cluster '$CNPG_CLUSTER_NAME' in namespace '$NAMESPACE'." >&2
return 1
fi
sleep 5
done
}
cluster_exists() {
kubectl -n "$NAMESPACE" get cluster "$CNPG_CLUSTER_NAME" >/dev/null 2>&1
}
cluster_has_pods() {
kubectl -n "$NAMESPACE" get pods -l "cnpg.io/cluster=$CNPG_CLUSTER_NAME" --no-headers 2>/dev/null | grep -q .
}
latest_backup_name() {
kubectl -n "$NAMESPACE" get backup \
--sort-by=.metadata.creationTimestamp \
-o jsonpath='{range .items[*]}{.metadata.name}{"\n"}{end}' 2>/dev/null | tail -n 1
}
latest_completed_backup_name() {
kubectl -n "$NAMESPACE" get backup \
--sort-by=.metadata.creationTimestamp \
-o jsonpath='{range .items[*]}{.metadata.name}{"|"}{.status.phase}{"\n"}{end}' 2>/dev/null | \
awk -F'|' '{p=tolower($2); if (p=="completed" || p=="succeeded") {name=$1}} END {print name}'
}
wait_for_backup() {
local backup_name="$1"
local start_time now phase phase_lc
start_time=$(date +%s)
while true; do
phase=$(kubectl -n "$NAMESPACE" get backup "$backup_name" -o jsonpath='{.status.phase}' 2>/dev/null || true)
phase_lc=$(printf '%s' "$phase" | tr '[:upper:]' '[:lower:]')
case "$phase_lc" in
completed|succeeded)
echo "Backup $backup_name completed."
return 0
;;
failed|error)
echo "Backup $backup_name failed (phase=$phase)." >&2
return 1
;;
esac
now=$(date +%s)
if (( now - start_time > BACKUP_WAIT_TIMEOUT )); then
echo "Timed out waiting for backup $backup_name." >&2
return 1
fi
echo "Waiting for backup $backup_name to complete (phase=${phase:-unknown}) ..."
sleep 10
done
}
run_garage_backup() {
if [[ ! -x "$SCRIPT_DIR/init_prole-db-backup.sh" ]]; then
echo "WARN: init_prole-db-backup.sh not found; skipping Garage backup." >&2
return 1
fi
echo "Running Garage backup via init_prole-db-backup.sh ..."
if ! "$SCRIPT_DIR/init_prole-db-backup.sh" start; then
echo "Garage backup script failed." >&2
return 1
fi
sleep 2
local backup_name
backup_name=$(latest_backup_name)
if [[ -z "$backup_name" ]]; then
echo "No backup resource detected after triggering backup." >&2
return 1
fi
wait_for_backup "$backup_name"
}
get_primary_pod() {
local primary
primary=$(kubectl -n "$NAMESPACE" get cluster "$CNPG_CLUSTER_NAME" -o jsonpath='{.status.currentPrimary}' 2>/dev/null || true)
if [[ -z "$primary" ]]; then
primary=$(kubectl -n "$NAMESPACE" get pods -l "cnpg.io/cluster=$CNPG_CLUSTER_NAME" -o jsonpath='{.items[0].metadata.name}' 2>/dev/null || true)
fi
printf '%s' "$primary"
}
decode_b64() {
local data="$1"
if [[ -z "$data" ]]; then
return 1
fi
printf '%s' "$data" | base64 -d 2>/dev/null
}
resolve_db_credentials() {
local user_b64 pass_b64 user pass
user_b64=$(kubectl -n "$NAMESPACE" get secret prole-db-user -o jsonpath='{.data.username}' 2>/dev/null || true)
pass_b64=$(kubectl -n "$NAMESPACE" get secret prole-db-user -o jsonpath='{.data.password}' 2>/dev/null || true)
user=$(decode_b64 "$user_b64" || true)
pass=$(decode_b64 "$pass_b64" || true)
if [[ -z "$user" || -z "$pass" ]]; then
user_b64=$(kubectl -n "$NAMESPACE" get secret prole-db-superuser -o jsonpath='{.data.username}' 2>/dev/null || true)
pass_b64=$(kubectl -n "$NAMESPACE" get secret prole-db-superuser -o jsonpath='{.data.password}' 2>/dev/null || true)
user=$(decode_b64 "$user_b64" || true)
pass=$(decode_b64 "$pass_b64" || true)
fi
if [[ -z "$user" || -z "$pass" ]]; then
return 1
fi
printf '%s\n%s' "$user" "$pass"
}
pgdump_local() {
local pod user pass db_name dump_dir dump_file timestamp
pod=$(get_primary_pod)
if [[ -z "$pod" ]]; then
echo "ERROR: No CNPG pod available for pg_dump." >&2
return 1
fi
local creds
if ! creds=$(resolve_db_credentials); then
echo "ERROR: Unable to resolve database credentials for pg_dump." >&2
return 1
fi
user=$(printf '%s' "$creds" | sed -n '1p')
pass=$(printf '%s' "$creds" | sed -n '2p')
db_name=${PROLE_DB_NAME:-prole-db}
dump_dir="$BACKUP_DIR"
mkdir -p "$dump_dir"
timestamp=$(date +%Y%m%d%H%M%S)
dump_file="$dump_dir/${CNPG_CLUSTER_NAME}-pgdump-${timestamp}.dump"
echo "Running pg_dump against pod $pod (db=$db_name) ..."
if kubectl -n "$NAMESPACE" exec "$pod" -c postgres -- env PGPASSWORD="$pass" \
pg_dump -U "$user" -d "$db_name" -Fc > "$dump_file"; then
echo "pg_dump saved to $dump_file"
return 0
fi
echo "pg_dump failed; removing partial file." >&2
rm -f "$dump_file"
return 1
}
attempt_backup_if_active() {
if ! cluster_exists; then
return 0
fi
if ! cluster_has_pods; then
echo "CNPG cluster '$CNPG_CLUSTER_NAME' exists but no pods detected; skipping backup." >&2
return 0
fi
echo "CNPG cluster '$CNPG_CLUSTER_NAME' detected; attempting Garage backup ..."
if run_garage_backup; then
return 0
fi
echo "Garage backup failed; attempting local pg_dump ..." >&2
if pgdump_local; then
return 0
fi
echo "WARN: Both Garage backup and local pg_dump failed." >&2
return 0
}
force_rollout() {
if [[ ! -x "$CNPG_FORCE_ROLLOUT_SCRIPT" ]]; then
echo "WARN: Force rollout script not found at $CNPG_FORCE_ROLLOUT_SCRIPT" >&2
return 1
fi
echo "Attempting force rollout of CNPG pods ..."
"$CNPG_FORCE_ROLLOUT_SCRIPT" rollout
}
reset_and_reinit() {
local backup_name tmp_manifest
backup_name=$(latest_completed_backup_name)
echo "Resetting CNPG cluster '$CNPG_CLUSTER_NAME' in namespace '$NAMESPACE' ..."
kubectl -n "$NAMESPACE" delete cluster "$CNPG_CLUSTER_NAME" --ignore-not-found
kubectl -n "$NAMESPACE" wait --for=delete pod -l "cnpg.io/cluster=$CNPG_CLUSTER_NAME" --timeout=180s >/dev/null 2>&1 || true
if [[ -n "$backup_name" && -f "$RECOVERY_TEMPLATE" ]]; then
tmp_manifest=$(mktemp)
sed "s/{{BACKUP_NAME}}/${backup_name}/g" "$RECOVERY_TEMPLATE" > "$tmp_manifest"
echo "Re-initializing from backup $backup_name ..."
CNPG_MANIFEST_OVERRIDE="$tmp_manifest" ensure_prole_stack_resources
rm -f "$tmp_manifest"
else
if [[ -n "$backup_name" && ! -f "$RECOVERY_TEMPLATE" ]]; then
echo "WARN: Recovery template not found: $RECOVERY_TEMPLATE" >&2
fi
if [[ -z "$backup_name" ]]; then
echo "No completed backups found; starting fresh initialization." >&2
fi
ensure_prole_stack_resources
fi
wait_for_cnpg_pods 300
}
# Resolve OpenBao URL: prefer explicit env, then localhost port-forward, then cluster DNS
bao_service_url() {
if [[ -n "${PROLE_OPENBAO_URL:-}" ]]; then
echo "$PROLE_OPENBAO_URL"
return 0
fi
# Prefer standard local port-forward managed by etc/init_port_forwards.sh
if curl -sS "http://127.0.0.1:18200/v1/sys/health" >/dev/null 2>&1; then
echo "http://127.0.0.1:18200"
return 0
fi
echo "http://$OPENBAO_NAME.$NAMESPACE.svc.cluster.local:8200"
}
fetch_admin_keys_and_db_pass_from_bao_or_local() {
local token url
if [[ -f "$OPENBAO_TOKEN_FILE" ]]; then
token=$(cat "$OPENBAO_TOKEN_FILE")
else
token=""
fi
url=$(bao_service_url)
if [[ -n "$token" ]]; then
echo "Attempting to read admin key pair from OpenBao kv/$BAO_PATH_ADMIN ..."
if curl -sS -H "X-Vault-Token: $token" "$url/v1/kv/data/$BAO_PATH_ADMIN" | jq -e '.data.data' >/dev/null 2>&1; then
local priv_b64 pub_b64
priv_b64=$(curl -sS -H "X-Vault-Token: $token" "$url/v1/kv/data/$BAO_PATH_ADMIN" | jq -r '.data.data.admin_private_key_b64')
pub_b64=$(curl -sS -H "X-Vault-Token: $token" "$url/v1/kv/data/$BAO_PATH_ADMIN" | jq -r '.data.data.admin_public_key_b64')
# Use a temporary file to determine where to save based on existing legacy or generic preference
local target_priv="$ADMIN_PRIV_GENERIC"
local target_pub="$ADMIN_PUB_GENERIC"
# If legacy keys exist, we might want to overwrite them too for compatibility
printf "%s" "$priv_b64" | base64 -d >"$target_priv"
printf "%s" "$pub_b64" | base64 -d >"$target_pub"
chmod 0600 "$target_priv"
# Mirror to legacy path if it was expected by other scripts
cp "$target_priv" "$ADMIN_PRIV_ED25519" 2>/dev/null || true
cp "$target_pub" "$ADMIN_PUB_ED25519" 2>/dev/null || true
fi
echo "Attempting to read database password from OpenBao kv/$BAO_PATH_DB ..."
if curl -sS -H "X-Vault-Token: $token" "$url/v1/kv/data/$BAO_PATH_DB" | jq -e '.data.data' >/dev/null 2>&1; then
local db_pass
db_pass=$(curl -sS -H "X-Vault-Token: $token" "$url/v1/kv/data/$BAO_PATH_DB" | jq -r '.data.data.password')
if [[ -n "$db_pass" ]]; then
echo "Updating database user secret 'prole-db-user' from OpenBao ..."
kubectl create secret generic prole-db-user -n "$NAMESPACE" \
--from-literal=username=prole \
--from-literal=password="$db_pass" \
--dry-run=client -o yaml | kubectl apply -f -
echo "Updating database superuser secret 'prole-db-superuser' from OpenBao ..."
kubectl create secret generic prole-db-superuser -n "$NAMESPACE" \
--from-literal=username=postgres \
--from-literal=password="$db_pass" \
--dry-run=client -o yaml | kubectl apply -f -
fi
fi
fi
if [[ -f "$ADMIN_PRIV_GENERIC" && -f "$ADMIN_PUB_GENERIC" ]]; then
echo "Using local admin key pair at $SECRETS_DIR"
return 0
fi
if [[ -f "$ADMIN_PRIV_ED25519" && -f "$ADMIN_PUB_ED25519" ]]; then
echo "Using local legacy admin key pair at $SECRETS_DIR"
# Link or copy to generic for consistent use below
cp "$ADMIN_PRIV_ED25519" "$ADMIN_PRIV_GENERIC"
cp "$ADMIN_PUB_ED25519" "$ADMIN_PUB_GENERIC"
return 0
fi
echo "ERROR: Could not obtain admin key pair from OpenBao and no local files found." >&2
exit 1
}
apply_cnpg_admin_secret() {
echo "Creating/updating Secret cnpg-admin-key ..."
kubectl create secret generic cnpg-admin-key -n "$NAMESPACE" \
--from-file=admin.key="$ADMIN_PRIV_GENERIC" \
--from-file=admin.pub="$ADMIN_PUB_GENERIC" \
--dry-run=client -o yaml | kubectl apply -f -
}
generate_tls_if_missing() {
local ca_secret_name="${CNPG_CLUSTER_NAME}-ca"
if kubectl -n "$NAMESPACE" get secret "$ca_secret_name" >/dev/null 2>&1; then
echo "CA secret $ca_secret_name already exists; skipping generation."
return 0
fi
echo "Generating self-signed CA (RSA 4096) for CNPG ..."
local TMPD
TMPD=$(mktemp -d)
openssl genrsa -out "$TMPD/ca.key" 4096
openssl req -x509 -new -key "$TMPD/ca.key" -out "$TMPD/ca.crt" -days 3650 -subj "/CN=Prole CNPG CA"
kubectl -n "$NAMESPACE" create secret generic "$ca_secret_name" \
--from-file=ca.crt="$TMPD/ca.crt" \
--from-file=ca.key="$TMPD/ca.key" \
--dry-run=client -o yaml | kubectl apply -f -
rm -rf "$TMPD"
}
# Resolve latest CNPG version from GitHub if possible, fallback to a sensible default.
get_latest_cnpg_version() {
local version
version=$(curl -s "https://api.github.com/repos/cloudnative-pg/cloudnative-pg/releases/latest" | jq -r '.tag_name' | sed 's/^v//' || echo "")
if [[ -z "$version" || "$version" == "null" ]]; then
echo "1.27.0"
else
echo "$version"
fi
}
initialize() {
ensure_tools
ensure_namespace
attempt_backup_if_active
ensure_cnpg_operator
pin_cnpg_controller
ensure_barman_plugin
echo "Using namespace: $NAMESPACE"
fetch_admin_keys_and_db_pass_from_bao_or_local
apply_cnpg_admin_secret
ensure_prole_stack_resources
if ! wait_for_cnpg_pods 300; then
echo "WARN: CNPG pods did not become ready after init; attempting force rollout ..." >&2
if force_rollout; then
if ! wait_for_cnpg_pods 300; then
echo "WARN: Force rollout did not recover CNPG; attempting reset and re-init ..." >&2
if ! reset_and_reinit; then
return 1
fi
fi
else
echo "WARN: Force rollout failed; attempting reset and re-init ..." >&2
if ! reset_and_reinit; then
return 1
fi
fi
fi
# Support both names for the toggle from prole.cfg
local kerberos_enabled="${KERBEROS_ENABLED:-${ENABLED:-false}}"
if [[ "$kerberos_enabled" == "true" || "$kerberos_enabled" == "True" || "$kerberos_enabled" == "1" ]]; then
if [[ -x "$SCRIPT_DIR/init_kerberos.sh" ]]; then
echo "Configuring Kerberos for CNPG pods ..."
if ! "$SCRIPT_DIR/init_kerberos.sh" initialize; then
echo "WARN: Kerberos initialization did not complete successfully."
fi
else
echo "WARN: init_kerberos.sh not found; skipping Kerberos configuration."
fi
else
echo "Kerberos is disabled; skipping Kerberos configuration."
fi
# Ensure port-forward is running for Postgres (local access)
echo "Ensuring port-forward for Postgres is active ..."
"$SCRIPT_DIR/init_port_forwards.sh" stop postgres || true
"$SCRIPT_DIR/init_port_forwards.sh" start postgres &
sleep 2
echo "Initialization complete for CNPG + Kerberos + cert artifacts."
}
update_reload() {
initialize
}
case "$ACTION" in
recreate)
ensure_tools
"$0" delete "$CNPG_CLUSTER_NAME"
"$0" create "$CNPG_CLUSTER_NAME"
;;
create)
ensure_tools
ensure_namespace
initialize
;;
delete)
ensure_tools
echo "Deleting all resources for '$CNPG_CLUSTER_NAME' ..."
kubectl delete -n "$NAMESPACE" -k "$SCRIPT_DIR/../k8s/prole" --ignore-not-found
;;
start)
ensure_tools
ensure_namespace
ensure_cnpg_operator
pin_cnpg_controller
ensure_barman_plugin
if [[ ! -f "$CNPG_MANIFEST" ]]; then
echo "ERROR: CNPG manifest not found at $CNPG_MANIFEST" >&2
exit 1
fi
echo "Starting CloudNative-PG cluster from $CNPG_MANIFEST in namespace $NAMESPACE..."
prole_render_manifest "$CNPG_MANIFEST" | kubectl apply -n "$NAMESPACE" -f -
;;
stop)
ensure_tools
if [[ ! -f "$CNPG_MANIFEST" ]]; then
echo "ERROR: CNPG manifest not found at $CNPG_MANIFEST" >&2
exit 1
fi
echo "Stopping CloudNative-PG cluster using $CNPG_MANIFEST ..."
kubectl delete -f "$CNPG_MANIFEST" --ignore-not-found
;;
status)
ensure_tools
echo "--- CloudNative-PG Cluster Status ($CNPG_CLUSTER_NAME) ---"
if kubectl -n "$NAMESPACE" get cluster "$CNPG_CLUSTER_NAME" >/dev/null 2>&1; then
kubectl -n "$NAMESPACE" get cluster "$CNPG_CLUSTER_NAME"
echo ""
echo "CNPG Plugin Status:"
if kubectl cnpg version >/dev/null 2>&1; then
kubectl cnpg status "$CNPG_CLUSTER_NAME" -n "$NAMESPACE"
else
echo "Note: 'kubectl cnpg' plugin not found; skipping detailed status."
fi
else
echo "Cluster '$CNPG_CLUSTER_NAME' not found in namespace '$NAMESPACE'."
fi
;;
restart)
ensure_tools
"$0" stop
"$0" start
;;
initialize)
if ! initialize; then
exit 1
fi
;;
update|reload)
if ! update_reload; then
exit 1
fi
;;
install-barman-plugin)
ensure_tools
ensure_cnpg_operator
pin_cnpg_controller
ensure_barman_plugin
;;
*)
echo "Usage: $0 {create|delete|recreate|start|stop|status|restart|initialize|update|reload|install-barman-plugin} [dbname]" >&2
exit 2
;;
esac