Deployment · CloudNativePG

Managed CloudNativePG, run on your Kubernetes or OpenShift

OSStend runs CloudNativePG, the open-source PostgreSQL operator for Kubernetes, on the Kubernetes or OpenShift cluster you own: failover, backups, upgrades and on-call, without your platform team carrying the pager.

45 minutes with an engineer: we review your workload and sketch the design.

  • Operated by AceMQ
  • Co-managed
  • Kubernetes and OpenShift
  • CloudNativePG

Managed CloudNativePG is a co-managed OSStend service: CloudNativePG, the open-source Kubernetes operator for PostgreSQL, run by OSStend engineers on your Kubernetes or OpenShift cluster. The operator automates failover, replicas and rolling updates; OSStend designs the cluster, owns the backups and upgrades, and is on call 24/7. You keep the platform.

160+AceMQ customers
10+Years running these systems
24/7Engineers on call
3Places it runs: our cloud, yours, your data center
Engines we run, on the platforms you already use
  • PostgreSQL
  • MySQL
  • MongoDB
  • Redis
  • Valkey
  • OpenSearch
  • Kafka
  • RabbitMQ
  • AWS
  • Google Cloud
  • Azure
  • Kubernetes
  • OpenShift
  • Prometheus
  • Grafana
  • Datadog
  • Dynatrace
  • Splunk

What is managed CloudNativePG?

CloudNativePG is an open-source operator that manages PostgreSQL on Kubernetes, and managed CloudNativePG means OSStend engineers run it for you on your own cluster. The operator handles the mechanics inside Kubernetes; OSStend handles the design, the recovery targets, the upgrades and the pager.

The CloudNativePG docs describe it as an operator for PostgreSQL workloads on any supported Kubernetes cluster, built by EDB, released under the Apache License 2.0 and now a CNCF incubating project (checked 17 Nov 2026). It adds a Cluster resource: one primary for writes, optional replicas for high availability and read scaling, reached through operator-managed services. The latest release on GitHub on 17 Nov 2026 was 1.30.2.

If you searched for a Postgres Kubernetes operator, this is the one OSStend runs. Which operator build, version and images a cluster uses is written into the design you sign off, on your Kubernetes or OpenShift version. The service is co-managed: the cluster and the network stay yours, and OSStend runs the database on top under a response-time SLA, as on the Kubernetes page.

Compare

What does the operator do, and what does it leave to you?

The operator automates failover, replica management and rolling updates inside the cluster. It does not decide your recovery targets, test your restores or answer the page at 3 a.m. The table splits the work three ways.

WorkCloudNativePG operatorOSStend engineersYour team
Cluster designReads the Cluster resource and builds itWrites the design: instances, synchronous replication, storage, backup target, versionsSigns it off
FailoverPromotes the replica with the most up-to-date dataSets quorum and synchronous settings, checks failover behaviour, responds to incidentsKeeps application retries and timeouts sane
Backups and recoveryRuns backups and WAL archiving through plugins or volume snapshotsChooses the method, sets the recovery window, checks the last good backup against your targetStates the recovery point and time you need
PostgreSQL upgradesRolling update for minor versions; in-place major upgrade when a new image is requestedPlans each upgrade, tests it, writes the rollback path, runs it on your calendarApproves the change window
Operator and monitoringExposes Prometheus metricsWatches replication lag, backup age and WAL archiving; forwards alerts to your toolsOwns the monitoring stack it plugs into
Kubernetes platformNeeds a supported Kubernetes versionChecks the design against your version and node maintenance plansRuns the cluster, nodes, storage and network
Your applicationn/an/aSchema, queries and connection handling

Operator column from the CloudNativePG docs and README, checked 17 Nov 2026. The OSStend and customer columns describe the co-managed service.

From our on-call engineers: what we verify on a CloudNativePG cluster

CloudNativePG automates failover and backups, but it cannot prove your data is recoverable or that the cluster has room to grow. We verify four things on every cluster: recovery, resource limits, table growth and the replica rebuild path.

  • Backups go to object storage through Barman Cloud or to volume snapshots, with WAL archiving enabling point-in-time recovery, and the documentation recommends testing recovery regularly, so we restore into a scratch cluster on a schedule.
  • Scheduled backups run from a standby by default, which keeps load off the primary, so we confirm a healthy standby exists before a backup window.
  • CloudNativePG recommends the Guaranteed QoS class for PostgreSQL pods, which means CPU and memory requests equal to limits, so we set both explicitly and size them for peak load, not the average.
  • In Postgres estates we were asked to stabilise on Kubernetes, one very large unpartitioned audit table was a recurring factor in slow backups and slow replica clones, so we check the largest tables before a go-live.
  • Changes are validated in pre-production before production, including operator and PostgreSQL minor upgrades.

How does CloudNativePG handle high availability and failover?

CloudNativePG uses the Kubernetes API for high availability instead of an external tool. When the primary fails, the operator promotes the replica with the most up-to-date data, and a per-cluster lease keeps two instances from promoting themselves at once.

Automated failover.
After the primary has been failing for longer than the failover delay (zero seconds by default), the operator shuts the failing primary down so its WAL receivers stop, runs an election and promotes the winner. The old primary comes back as a replica (failover docs, checked 17 Nov 2026).
Quorum-based and synchronous replication.
The docs list quorum-based and priority-based synchronous replication for stronger durability (features, checked 17 Nov 2026). PostgreSQL itself ships no software to detect a dead primary, which is the gap the operator fills.
Services that follow the primary.
A read-write service points at the primary and a read-only service spreads reads across replicas, and endpoints move during failover and rolling updates (architecture, checked 17 Nov 2026).
What OSStend adds.
The synchronous settings, replica count and zone spread are chosen from your recovery point and your tolerance for write latency, then tested before go-live. Applications still need retries that outlast a failover.

How do backups and point-in-time recovery work?

CloudNativePG supports physical backups, continuous WAL archiving and point-in-time recovery through plugins and volume snapshots, and OSStend sets the method and the recovery window for your workload. A backup that has not been restored is a guess, so restore tests are part of the run.

Per the backup docs and feature list (checked 17 Nov 2026), backup and recovery go through plugins, with hot and cold base backups, WAL archiving, full and point-in-time recovery, scheduled and on-demand backups, and backups taken from a standby to spare the primary. The community Barman Cloud plugin archives to object stores with retention based on a recovery window, and volume snapshots work where the storage class supports them. The older built-in Barman Cloud integration is deprecated, with 1.31 the last minor release to carry it. A pg_dump cannot restore to a moment; base backups plus archived WAL can.

OSStend picks the method for your storage and recovery targets, keeps the archive in a place you approve, and checks the newest backup and the last archived WAL segment against your recovery point. The target is written into the design before anything is built.

Compare

How are PostgreSQL and operator upgrades handled?

Minor PostgreSQL versions roll through the cluster one instance at a time, replicas first and the primary last. Major versions are a planned change with three documented methods, and OSStend runs it on your change calendar.

ChangeWhat the CloudNativePG docs sayHow OSStend runs it
Minor versionUpdate the image; the operator replaces instances one by one, starting with the replicas, then switches over or restarts the primary (upgrades, rolling updates, checked 17 Nov 2026).Tested on a copy first and applied in your change window; never forced.
Major versionThree methods: logical dump and restore, native logical replication (online) or an offline in-place upgrade with pg_upgrade. The docs recommend testing each in a controlled environment first (upgrades, checked 17 Nov 2026).The method is picked for your data size and downtime tolerance, with the rollback path written down before the window.
OperatorOperator updates trigger a rolling update of the instance manager unless in-place updates are enabled (rolling updates, checked 17 Nov 2026).Scheduled with Kubernetes upgrades so the two do not collide.
Kubernetes node workDraining a node moves database pods, which the docs cover under Kubernetes upgrade and maintenance (checked 17 Nov 2026).Planned with your platform team against pod disruption budgets and replica placement.

Each CloudNativePG minor release is supported until three months after the next one; the 1.30 line was released on 29 Jun 2026 and lists Kubernetes 1.34 to 1.36 and PostgreSQL 15 to 18 as supported (supported releases, checked 17 Nov 2026). The design names the operator and PostgreSQL versions for your cluster.

What gets monitored?

CloudNativePG exposes a Prometheus-compatible metrics port, and OSStend forwards alerts to the tools you already use. The checks are the ones that decide whether a failover or a restore will work.

From the operator.
A Prometheus-compatible exporter on port 9187, JSON logs on standard output and replica probes based on lag from the primary (features, monitoring, checked 17 Nov 2026).
What on-call engineers watch.
Replication lag, the age of the last base backup and of the last archived WAL segment, free space on the data and WAL volumes, the age of the oldest transaction ID, and pods that restart.
Where alerts go.
To your Datadog, Dynatrace, Splunk or Prometheus, with OSStend engineers paged first.

What changes on OpenShift?

The operator works the same way, and OpenShift adds stricter pod rules and its own operator catalog. The design step checks both before anything is built.

Installation.
The CloudNativePG docs list Helm and Operator Lifecycle Manager installs, and say EDB offers a certified version for Red Hat OpenShift through the Red Hat catalog; the operator images come in a Red Hat UBI 9 flavour intended for OLM (installation, overview, checked 17 Nov 2026). Which build your cluster runs is named in the design.
Security context constraints.
OpenShift limits what a pod may run as and which volumes it may use, so the operator and the PostgreSQL images are checked against them. The general OpenShift points are on the OpenShift page.
Disconnected clusters.
With no internet route, images and charts are mirrored in; see air-gapped.
Enterprise only

When should you run CloudNativePG yourself?

Run it yourself when your platform team already operates stateful workloads on Kubernetes and is happy to own backups, restores and upgrades. The software is free; the cost is the people and the on-call.

Contact us if

  • You want PostgreSQL on Kubernetes or OpenShift and your platform team would rather not be on call for it.
  • The data must stay in your data center or your cloud account, and you want one team accountable for failover, backups and upgrades.
  • You are choosing between CloudNativePG, Patroni and a managed service; read CloudNativePG vs Patroni vs a managed service first.
Talk to an engineer

Not a fit if

  • You have the people, the runbooks and the restore drills, and want to keep them. CloudNativePG is open source and has its own community.
  • You do not run Kubernetes: the project is Kubernetes-only and PostgreSQL-only (README, checked 17 Nov 2026). See managed PostgreSQL for clusters that do not need it.
  • You want the cluster itself run for you. See bring your own cloud.
Operated by AceMQCompany figures, as published on acemq.com
160+AceMQ customers
10+Years running these systems
24/7Engineers on call

What is CloudNativePG?

CloudNativePG is an open-source Kubernetes operator that manages PostgreSQL clusters: a primary, optional replicas, failover, backups, rolling updates and connection pooling. It was built by EDB, is released under the Apache License 2.0 and is a CNCF incubating project (project docs, checked 17 Nov 2026).

What is a Postgres Kubernetes operator?

A controller that watches a custom resource describing a database cluster and keeps the real cluster in that state. For PostgreSQL it creates the pods and volumes, promotes a replica when the primary fails and rolls out updates. Kubernetes documents the general operator pattern. OSStend runs CloudNativePG.

Does OSStend run CloudNativePG for me?

Yes. OSStend engineers run CloudNativePG on your Kubernetes or OpenShift cluster as a co-managed service: they design the cluster, run backups and upgrades and are on call 24/7, while you keep the platform. The response commitment is written into your contract.

Is this the same as managed PostgreSQL?

It is the same service delivered one way. Managed PostgreSQL covers the engine; this page covers running it on your Kubernetes or OpenShift through an operator. For the cluster run for you as well, see bring your own cloud.

CloudNativePG or Patroni?

CloudNativePG runs the whole PostgreSQL cluster on Kubernetes through the Kubernetes API. Patroni is a high-availability template that also runs on VMs and bare metal and keeps leader state in etcd, Consul, ZooKeeper or Kubernetes. The comparison sets out when each fits.

Can PostgreSQL on Kubernetes run in production?

Yes, with an operator that manages failover, replicas and backups, and with storage, zone spread and upgrade windows planned up front. See Kubernetes for what the design step settles.

How much does managed CloudNativePG cost?

OSStend prices each design after a workload review, from the cluster size, the service level and where it runs. The operator itself is free, open-source software. See pricing.

Talk to an engineer

Talk to an engineer about Postgres on your cluster

Bring your Kubernetes or OpenShift version, the size of the database and your recovery targets. An engineer will review them with you and sketch the design in 45 minutes.

  • A design sketch for your workload, not a plan menu
  • A zero-downtime cutover path for what you run today
  • A price for that design, after the review
See pricing →

Your request goes to the OSStend team. No newsletter sign-up.

Get a quoteTalk to an engineer
2Engines
3Where should it run?
4What do you need?

Your request goes to the OSStend team. No newsletter sign-up.