StackGres is a full PostgreSQL platform on Kubernetes with a web UI and distributed storage, but still lacks database branching, developer self-service, and org-wide RBAC. Compare with Vela BYOC.
Last updated: March 2026
StackGres (ongres.com/stackgres) takes a different approach from most Kubernetes PostgreSQL operators: it aims to be a complete PostgreSQL platform on Kubernetes with a built-in web UI, integrated connection pooling, distributed storage support (Rook/Ceph), and automated minor version upgrades. For teams wanting more than bare infrastructure primitives — but not wanting to manage 7+ Supabase services — StackGres offers an interesting middle ground. However, it still lacks the developer workflow capabilities that modern development teams need: no copy-on-write database branching, no developer self-service environments, and no organization-wide RBAC with SSO.
StackGres is an open-source AGPL-licensed PostgreSQL operator for Kubernetes developed by OnGres. It runs a 'PostgreSQL Stack' — PostgreSQL with PgBouncer, Envoy proxy, Fluentd for logging, and Prometheus for monitoring — all managed as a single unit. StackGres includes a web UI (SGWeb) for cluster management, supports distributed storage via Rook/Ceph or cloud PVs, and automates minor version upgrades. Its Patroni-based HA handles automatic failover. StackGres is opinionated — it makes choices about the full stack so you don't have to.
Best for: Teams who want a complete PostgreSQL platform on Kubernetes with a web UI, and are comfortable with StackGres's opinionated technology choices and AGPL licensing.
How StackGres compares to Vela BYOC across key dimensions
| Feature | StackGres | Vela BYOC |
|---|---|---|
| Deployment model | Kubernetes operator with full stack sidecar | BYOC — managed control plane in your cloud |
| Web UI | Yes — SGWeb for cluster management | Yes — built-in management interface |
| High availability | Patroni-based HA with automatic failover | Built-in HA with live migration support |
| Connection pooling | PgBouncer + Envoy sidecar (built in) | Built-in connection management |
| Monitoring | Prometheus + Grafana + Fluentd (built in) | Built-in observability dashboard |
| Distributed storage | Rook/Ceph or cloud PVs | Simplyblock high-performance distributed storage |
| Minor version upgrades | Automated | Managed upgrades by control plane |
| Instant database cloning | No — no copy-on-write branching | Yes — copy-on-write, any database size, seconds |
| Git-style DB branching | No — must implement yourself | Yes — branch per PR / pipeline / developer |
| Developer self-service | No — platform team manages clusters | Yes — developers spin up DB branches via UI/API |
| Org-wide RBAC | Kubernetes namespace isolation only | Organization-wide RBAC across all databases |
| SSO / SAML / LDAP | Not included | Built-in SSO/SAML/LDAP integration |
| License | AGPL (open source — commercial implications) | Commercial (BYOC — data stays in your cloud) |
StackGres is an AGPL-licensed open-source PostgreSQL operator for Kubernetes developed by OnGres. Unlike operators like CloudNativePG or Zalando that focus on HA clustering, StackGres provides a complete opinionated stack: PostgreSQL, PgBouncer, Envoy proxy, Fluentd logging, Prometheus monitoring, and a web UI — all deployed together as a single managed unit. This makes it easier to get started but means a heavier resource footprint than leaner operators.
No. StackGres does not support copy-on-write database cloning or Git-style branching. Like all Kubernetes PostgreSQL operators, creating an isolated copy of a database for testing or a feature branch requires provisioning a new cluster and restoring from backup. StackGres focuses on production cluster management, not developer workflow features like branching or self-service environments.
StackGres and CloudNativePG serve different philosophies. CloudNativePG is minimal — it manages HA PostgreSQL and lets you configure everything else (PgBouncer, monitoring, storage) separately. StackGres is opinionated — it deploys the full stack (PgBouncer, Envoy, Prometheus, Grafana, Fluentd) automatically with a web UI. CloudNativePG has much stronger community momentum (CNCF sandbox, EDB-backed) and a larger ecosystem. StackGres is better for teams wanting an all-in-one solution with less integration work.
AGPL requires that if you modify StackGres and expose it as a service (SaaS), you must release your modifications under AGPL. For internal use or self-hosting without exposing the software as a service, AGPL typically does not require releasing your application code. However, legal teams at many enterprises flag AGPL as a concern. CloudNativePG (Apache 2.0) and Zalando operator (MIT) have more permissive licenses for commercial use.
StackGres has a heavier resource footprint than CNPG or Zalando because it deploys PgBouncer, Envoy, Fluentd, and Prometheus as sidecars alongside each PostgreSQL cluster. Each StackGres cluster typically requires significantly more CPU and memory than a bare CNPG cluster with the same PostgreSQL workload. For resource-constrained environments or small clusters, CNPG or Zalando are more efficient. StackGres's overhead makes sense when you need the full monitoring and pooling stack anyway.
If you need a complete PostgreSQL platform on Kubernetes with better developer workflows: Vela BYOC provides instant copy-on-write database cloning, Git-style branching, org-wide RBAC, SSO, and a polished web interface — all in your own cloud. If you want an open-source K8s platform with a larger community, CloudNativePG with its ecosystem is the stronger choice over StackGres for new projects in 2026. If you need StackGres's web UI specifically, it remains unique in that respect among open-source operators.
Keep your data in your own cloud. Add instant database cloning, Git-style branching, and org-wide RBAC — without replacing your infrastructure.