openova/platform/cnpg

CNPG (CloudNative PostgreSQL)

Production-grade PostgreSQL operator. Application Blueprint (see docs/PLATFORM-TECH-STACK.md §4.1 — Data services). Used by Organizations that want managed Postgres; also the underlying engine for FerretDB (MongoDB-compatible) and Gitea metadata. Replication via WAL streaming to async standby (Application-tier choice).

Status: Accepted | Updated: 2026-04-27

---

Overview

CloudNative PostgreSQL (CNPG) provides production-grade PostgreSQL with:

• Kubernetes-native operator
• WAL streaming for multi-region DR
• Automated backups to SeaweedFS/S3
• High availability with automatic failover

---

Architecture

Single Region

flowchart TB
    subgraph Cluster["CNPG Cluster"]
        Primary[Primary]
        Replica1[Replica 1]
        Replica2[Replica 2]
    end

    subgraph Backup["Backup"]
        SeaweedFS[SeaweedFS]
    end

    Primary -->|"WAL Stream"| Replica1
    Primary -->|"WAL Stream"| Replica2
    Primary -->|"WAL Archive"| SeaweedFS

Multi-Region DR

flowchart TB
    subgraph Region1["Region 1 (Primary)"]
        PG1[CNPG Primary]
    end

    subgraph Region2["Region 2 (DR)"]
        PG2[CNPG Standby]
    end

    subgraph Backup["Backup"]
        SeaweedFS[SeaweedFS]
    end

    PG1 -->|"WAL Streaming"| PG2
    PG1 -->|"WAL Archive"| SeaweedFS
    PG2 -->|"WAL Restore"| SeaweedFS

---

Configuration

Cluster Definition

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: <org>-postgres
  namespace: databases
spec:
  instances: 3

  postgresql:
    parameters:
      max_connections: "200"
      shared_buffers: 256MB

  storage:
    size: 10Gi
    storageClass: <storage-class>

  backup:
    barmanObjectStore:
      destinationPath: s3://cnpg-backups/<org>
      endpointURL: http://seaweedfs.storage.svc:8333
      s3Credentials:
        accessKeyId:
          name: seaweedfs-credentials
          key: access-key
        secretAccessKey:
          name: seaweedfs-credentials
          key: secret-key
      wal:
        compression: gzip
    retentionPolicy: "30d"

  monitoring:
    enablePodMonitor: true

DR Replica (Region 2)

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: <org>-postgres-dr
  namespace: databases
spec:
  instances: 1

  replica:
    enabled: true
    source: <org>-postgres

  externalClusters:
    - name: <org>-postgres
      connectionParameters:
        host: postgres.<env>.<sovereign-domain>
        user: streaming_replica
      password:
        name: pg-replica-credentials
        key: password

---

Backup Strategy

Type	Schedule	Retention
WAL Archive	Continuous	7 days
Base Backup	Daily 2 AM	30 days
Point-in-Time	On-demand	Per backup

Scheduled Backup

apiVersion: postgresql.cnpg.io/v1
kind: ScheduledBackup
metadata:
  name: <org>-daily-backup
  namespace: databases
spec:
  schedule: "0 2 * * *"
  backupOwnerReference: self
  cluster:
    name: <org>-postgres

---

Failover

Automatic (Within Region)

CNPG automatically promotes replicas when primary fails.

Manual (Cross-Region)

# Promote DR cluster
kubectl cnpg promote <org>-postgres-dr -n databases

---

Monitoring

Metric	Description
cnpg_pg_replication_lag	Replication lag in seconds
cnpg_pg_database_size_bytes	Database size
cnpg_pg_stat_activity_count	Active connections

---

PgBouncer Integration

Connection pooling with PgBouncer:

apiVersion: postgresql.cnpg.io/v1
kind: Pooler
metadata:
  name: <org>-pooler
  namespace: databases
spec:
  cluster:
    name: <org>-postgres
  instances: 2
  type: rw
  pgbouncer:
    poolMode: transaction
    parameters:
      max_client_conn: "1000"
      default_pool_size: "20"

---

Part of OpenOva