statefulset.yml

apiVersion: v1
kind: Secret
metadata:
  name: pg-postgres-secret
  namespace: vacation-planner-postgres-in-cluster
type: Opaque
stringData:
  POSTGRES_PASSWORD: "SuperStrongPass123"
  REPL_PASSWORD: "ReplStrongPass123"
---
apiVersion: v1
kind: ConfigMap
metadata:
  name: pg-postgres-scripts
  namespace: vacation-planner-postgres-in-cluster
data:
  # Runs only on the primary (pod-0): the official entrypoint executes
  # everything in /docker-entrypoint-initdb.d only when it runs initdb,
  # which happens solely on a fresh primary. Replicas get their data via
  # pg_basebackup, so their PGDATA is already populated and these are skipped.
  primary-init.sh: |
    #!/bin/bash
    set -euo pipefail
    ORD="${HOSTNAME##*-}"
    if [ "$ORD" != "0" ]; then
      echo "not primary, skipping primary-init"
      exit 0
    fi
    echo "configuring primary for streaming replication"
    psql -v ON_ERROR_STOP=1 --username "$POSTGRES_USER" --dbname postgres <<EOSQL
      CREATE ROLE replicator WITH REPLICATION LOGIN PASSWORD '$REPL_PASSWORD';
      ALTER SYSTEM SET wal_level = replica;
      ALTER SYSTEM SET max_wal_senders = 10;
      ALTER SYSTEM SET max_replication_slots = 10;
      ALTER SYSTEM SET hot_standby = on;
      ALTER SYSTEM SET wal_keep_size = '512MB';
    EOSQL
    echo "host replication replicator 0.0.0.0/0 scram-sha-256" >> "$PGDATA/pg_hba.conf"
    echo "primary-init done (wal_level applied on the real server start)"

  # Runs as an initContainer on every pod. No-op on the primary and on any
  # pod whose data dir is already initialized. On a fresh standby it clones
  # the primary with pg_basebackup and wires up primary_conninfo.
  replica-init.sh: |
    #!/bin/bash
    set -euo pipefail
    ORD="${HOSTNAME##*-}"
    if [ "$ORD" = "0" ]; then
      echo "primary, no basebackup needed"
      exit 0
    fi
    if [ -s "$PGDATA/PG_VERSION" ]; then
      echo "data dir already initialized, skipping basebackup"
      exit 0
    fi
    echo "waiting for primary pg-postgres-0 to accept connections..."
    until pg_isready -h pg-postgres-0.pg-postgres-hl -p 5432 -U "$POSTGRES_USER"; do
      sleep 2
    done
    echo "cloning primary via pg_basebackup..."
    export PGPASSWORD="$REPL_PASSWORD"
    pg_basebackup -h pg-postgres-0.pg-postgres-hl -p 5432 -U replicator \
      -D "$PGDATA" -Fp -Xs -R -P
    # -R already wrote standby.signal + a password-less primary_conninfo.
    # Append a complete primary_conninfo (last value wins) so the walreceiver
    # can authenticate.
    cat >> "$PGDATA/postgresql.auto.conf" <<EOF
    primary_conninfo = 'host=pg-postgres-0.pg-postgres-hl port=5432 user=replicator password=$REPL_PASSWORD application_name=$HOSTNAME'
    EOF
    echo "standby $HOSTNAME initialized"
---
# Headless service: stable per-pod DNS (pg-postgres-0.pg-postgres-hl, ...)
apiVersion: v1
kind: Service
metadata:
  name: pg-postgres-hl
  namespace: vacation-planner-postgres-in-cluster
  labels:
    app: pg-postgres
spec:
  clusterIP: None
  selector:
    app: pg-postgres
  ports:
    - name: postgres
      port: 5432
      targetPort: 5432
---
# Write endpoint: targets ONLY pod-0 (the primary) via the per-pod label.
apiVersion: v1
kind: Service
metadata:
  name: pg-postgres-primary
  namespace: vacation-planner-postgres-in-cluster
  labels:
    app: pg-postgres
    role: primary
spec:
  type: ClusterIP
  selector:
    app: pg-postgres
    statefulset.kubernetes.io/pod-name: pg-postgres-0
  ports:
    - name: postgres
      port: 5432
      targetPort: 5432
---
# Read endpoint: round-robins across all pods (primary + standbys).
apiVersion: v1
kind: Service
metadata:
  name: pg-postgres-read
  namespace: vacation-planner-postgres-in-cluster
  labels:
    app: pg-postgres
    role: read
spec:
  type: ClusterIP
  selector:
    app: pg-postgres
  ports:
    - name: postgres
      port: 5432
      targetPort: 5432
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: pg-postgres
  namespace: vacation-planner-postgres-in-cluster
spec:
  serviceName: pg-postgres-hl
  replicas: 3
  podManagementPolicy: OrderedReady   # primary (pod-0) becomes Ready before standbys start
  selector:
    matchLabels:
      app: pg-postgres
  template:
    metadata:
      labels:
        app: pg-postgres
    spec:
      securityContext:
        runAsUser: 999     # postgres uid in the official image
        runAsGroup: 999
        fsGroup: 999       # makes the Azure disk PVC writable by postgres
      initContainers:
        - name: init-replica
          image: postgres:16
          command: ["/bin/bash", "/scripts/replica-init.sh"]
          env:
            - name: PGDATA
              value: /var/lib/postgresql/data/pgdata
            - name: POSTGRES_USER
              value: postgres
            - name: REPL_PASSWORD
              valueFrom:
                secretKeyRef:
                  name: pg-postgres-secret
                  key: REPL_PASSWORD
          volumeMounts:
            - name: data
              mountPath: /var/lib/postgresql/data
            - name: scripts
              mountPath: /scripts
      containers:
        - name: postgres
          image: postgres:16
          ports:
            - name: postgres
              containerPort: 5432
          env:
            - name: POSTGRES_USER
              value: postgres
            - name: POSTGRES_PASSWORD
              valueFrom:
                secretKeyRef:
                  name: pg-postgres-secret
                  key: POSTGRES_PASSWORD
            - name: REPL_PASSWORD
              valueFrom:
                secretKeyRef:
                  name: pg-postgres-secret
                  key: REPL_PASSWORD
            - name: PGDATA
              value: /var/lib/postgresql/data/pgdata
          volumeMounts:
            - name: data
              mountPath: /var/lib/postgresql/data
            - name: scripts
              mountPath: /docker-entrypoint-initdb.d
          readinessProbe:
            exec:
              command: ["bash", "-c", "pg_isready -U postgres -h 127.0.0.1"]
            initialDelaySeconds: 15
            periodSeconds: 10
            timeoutSeconds: 5
          livenessProbe:
            exec:
              command: ["bash", "-c", "pg_isready -U postgres -h 127.0.0.1"]
            initialDelaySeconds: 30
            periodSeconds: 15
            timeoutSeconds: 5
          resources:
            requests:
              cpu: "250m"
              memory: "256Mi"
            limits:
              cpu: "1"
              memory: "1Gi"
      volumes:
        - name: scripts
          configMap:
            name: pg-postgres-scripts
            defaultMode: 0755
  volumeClaimTemplates:
    - metadata:
        name: data
      spec:
        accessModes: ["ReadWriteOnce"]
        storageClassName: managed-csi-premium
        resources:
          requests:
            storage: 20Gi