Skip to main content

Managed PostgreSQL (CloudNativePG)

Okustera Managed PostgreSQL runs upstream PostgreSQL 16 orchestrated by the cloud-native CloudNativePG operator.

It combines the battle-tested durability of PostgreSQL with Kubernetes-native automated self-healing, rolling minor upgrades, synchronous replication, and Point-in-Time Recovery (PITR).

Architectural Highlights​

  • Instantaneous Failover: Quorum election and promotion take $< 5$ seconds using native PostgreSQL replication.
  • Separated Read/Write Endpoints: Built-in Kubernetes services routing write traffic to the primary (<name>-rw) and read queries to standbys (<name>-ro).
  • Continuous S3 Archiving (Barman): Every write-ahead log (WAL) is immediately streamed to S3-compatible object storage for Point-in-Time Recovery (PITR) up to the exact minute.
  • Zero-Downtime Rolling Minor Upgrades: Operator applies minor version patches sequentially with graceful switchovers.

Provisioning with Terraform​

The official Okustera Terraform provider delivers full declarative control over your PostgreSQL instances:

resource "okustera_database_postgresql" "app_database" {
  name            = "production-crm-db"
  instances       = 3
  storage_size_gb = 100
  storage_class   = "ceph-nvme"
  database_name   = "crm"
  enable_backups  = true

  parameters = {
    "max_connections"       = "300"
    "shared_buffers"        = "2GB"
    "effective_cache_size"  = "6GB"
    "maintenance_work_mem"  = "512MB"
  }
}

output "db_connection" {
  value     = okustera_database_postgresql.app_database.connection_string
  sensitive = true
}

Refer to the Terraform Resource Reference for full argument documentation.

Connecting to PostgreSQL​

1. Connection String Format​

postgresql://<user>:<password>@<name>-rw.omc-system.svc.cluster.local:5432/<database_name>?sslmode=require

2. Python (psycopg2 / asyncpg) Example​

import asyncpg
import asyncio

async def main():
    conn = await asyncpg.connect(
        user='app_user',
        password='<PASSWORD>',
        database='crm',
        host='production-crm-db-rw.omc-system.svc.cluster.local',
        port=5432,
        ssl='require'
    )
    version = await conn.fetchval('SELECT version()')
    print("Connected to:", version)
    await conn.close()

asyncio.run(main())

Point-in-Time Recovery (PITR)​

To restore a database cluster to a specific timestamp in the past:

apiVersion: postgresql.cnpg.io/v1
kind: Cluster
metadata:
  name: crm-db-restored
spec:
  instances: 3
  storage:
    size: 100Gi
    storageClass: ceph-nvme
  bootstrap:
    recovery:
      source: production-crm-db
      recoveryTarget:
        targetTime: "2026-09-10T14:30:00Z"

For complete architecture details on Barman S3 WAL streaming, RPO/RTO SLAs, and disaster recovery runbooks, see the Replication, Backups & Disaster Recovery Guide.