Managed PostgreSQL with Synchronous Replication — Without the Ops Headache

Setting up PostgreSQL with synchronous replication the hard way means Patroni configs, etcd clusters, pgBouncer, monitoring exporters, backup scripts, failover testing — easily a week of work before you even store a single row. And then you still need to maintain it. AWS RDS solves this but locks you into a cloud bill that grows faster than your data.

What if you could get managed PostgreSQL on your own hardware in two minutes?

The solution: Cozystack Managed PostgreSQL

Cozystack uses the CloudNativePG operator under the hood — one of the most mature Kubernetes-native Postgres operators available. You get automatic failover, streaming replication, and optional quorum-based synchronous replication, all managed by the platform.

Via Dashboard (the quick way)

  1. Open the Cozystack dashboard at https://dashboard.<your-domain>.
  2. Navigate to the Marketplace and find Postgres.
Cozystack dashboard Marketplace with the Postgres application
  1. Click Deploy and fill in the form:
    • Enter a name (e.g., app-postgres).
    • Set replicas to 2 (primary + standby with async replication) or 3 (for synchronous replication with quorum).
    • Choose your resourcesPreset (nano, micro, small, medium, large).
    • Set storage size (e.g., 10Gi).
    • Under PostgreSQL parameters, configure max_connections if needed.
Postgres deployment form with replicas and resources configured
  1. Click Deploy. Within two minutes, you have a primary + replica setup with automatic failover.
Deployed Postgres application reporting Ready state

Via kubectl (the GitOps way)

apiVersion: helm.toolkit.fluxcd.io/v2
kind: HelmRelease
metadata:
  name: postgres-myapp
  namespace: tenant-team1
spec:
  chart:
    spec:
      chart: postgres
      reconcileStrategy: Revision  # Reconcile on chart version change
      sourceRef:
        kind: HelmRepository
        name: cozystack-apps
        namespace: cozy-public
      version: 0.10.0
  interval: 0s  # Reconcile only on spec change, not periodically
  values:
    replicas: 3
    size: 10Gi
    resourcesPreset: small
    databases:
      production:
        roles:
          admin:
            - appuser
    users:
      appuser:
        password: "your-strong-password"  # or omit this line to auto-generate
    external: false

kubectl apply -f postgres.yaml

Getting connection credentials

# Primary (read-write)
kubectl get svc -n tenant-team1 | grep postgres-myapp-rw

# Replica (read-only)
kubectl get svc -n tenant-team1 | grep postgres-myapp-ro

# Password
kubectl get secret -n tenant-team1 postgres-myapp-app \
  -o jsonpath='{.data.password}' | base64 --decode

These service names are resolvable from any nested Kubernetes cluster in the same tenant — no external DNS or VPN needed.

Backups

Enable S3-compatible backup storage by setting backup.enabled: true in the values. Recovery is a one-line config change pointing to the source cluster name and an optional RFC 3339 timestamp.

Learn more

Join the community