PgDog

PgDog is a PostgreSQL proxy that combines connection pooling, read/write load balancing, sharding, and failover-aware routing. It suits teams operating PostgreSQL at scale that want these controls without changing application code. The main appeal is unusually broad functionality in a free, open-source package; the main cautions are its PostgreSQL-only scope, AGPL-3.0 license, and unsigned macOS binary.

Overview

PgDog sits between applications and PostgreSQL servers, multiplexing client connections and deciding where queries should run. Its free Open source plan costs nothing and is released under AGPL-3.0. Documentation states that internal use does not require sharing source code, making it practical for many in-house deployments, while organizations distributing modified software should review the license carefully.

The project is available as self-hosted software and can be deployed through several infrastructure patterns. That flexibility favors engineering teams comfortable owning database proxy operations rather than buyers seeking a managed, database-agnostic service.

For a wider category view, see Database Proxy Software.

Key features

Pooling and session behavior

PgDog multiplexes thousands of application connections over a smaller number of PostgreSQL server connections. This can reduce connection pressure for busy services without discarding important PostgreSQL session features: SET, LISTEN/NOTIFY, and advisory locks are supported. It is a stronger fit than a narrowly focused pooler when applications depend on those behaviors.

Load balancing and failover

Read queries can be distributed across replicas while writes go to the primary. Health checks and failover detection inform that routing, giving teams a way to keep traffic aligned with changing database availability. The benefit is most relevant to read-heavy or highly available PostgreSQL installations; it does not broaden support beyond PostgreSQL.

Sharding and transactions

Queries containing a sharding key are routed to the appropriate database. Queries without a key can execute across databases in parallel, which helps reporting and other cross-shard work but may require careful workload design. Cross-shard transactions use PostgreSQL prepared transactions and two-phase commit, with automatic rollback when errors occur. That provides a consistency mechanism for distributed writes, alongside the operational complexity inherent in two-phase commit.

Online resharding

PgDog can copy data to another database and stream row changes while query traffic continues. Online movement avoids a planned pause during resharding and is valuable when changing shard placement on a live system. It still leaves the team responsible for planning and operating the migration.

Observability and security

Enterprise insights include Prometheus and OTEL exporters, with Incident.io integration for alerts. The enterprise control plane describes SSO for access protection, and installation guidance supports encrypting credentials at rest in environments such as Kubernetes or AWS Secrets Manager. These controls are aimed at production operations and are tied to the enterprise offering rather than the free core plan.

Pricing

The free plan is the logical choice for teams able to run PgDog themselves and accept community-led operations. Enterprise is for production users who need a control plane, integrated insights, a dedicated Slack channel, architecture reviews, and an SLA promising a response within 15 minutes. No seat, usage, or trial limits are stated, so buyers should obtain commercial terms before budgeting.

Platforms

PgDog runs on Linux, macOS, Windows, and self-hosted environments. Deployment options include Kubernetes Helm, Docker, AWS ECS Terraform, pre-built binaries, and source installation. Releases cover Linux arm64 and amd64, Apple Silicon Mac, and Debian packages. The macOS binary is unsigned, which can add trust and installation friction for Mac administrators.

Who it's for

Choose PgDog when PostgreSQL scaling requires pooling plus replica routing, sharding, cross-shard transaction handling, or online resharding in one layer. It is especially suitable for infrastructure teams that can operate Kubernetes, containers, cloud infrastructure, or source builds and want to avoid application rewrites.

Look elsewhere if you need a managed experience, a signed macOS application, or support for another database engine. The AGPL license also warrants review before embedding or redistributing modified software.

Pros and cons

• Pro: Combines pooling, load balancing, sharding, failover detection, and online resharding instead of limiting the deployment to one proxy function.
• Pro: Preserves SET, LISTEN/NOTIFY, and advisory locks while multiplexing large client populations.
• Pro: Supports multiple deployment paths, from Helm and Docker to ECS Terraform, binaries, and source.
• Con: PostgreSQL is the only supported database, so mixed-engine environments need another product.
• Con: The macOS binary is unsigned, creating an administrative hurdle on Mac systems.
• Con: Enterprise-only control-plane, insights, SSO, and rapid-response support are not part of the free core plan.

Alternatives

PgBouncer is the simpler choice when you need a free, self-hosted PostgreSQL connection pooler without PgDog’s broader sharding and routing scope. Choose pgagroal for a BSD-3-Clause self-hosted PostgreSQL pooler. PgDoorman fits buyers seeking a free pooler with no paid plans or usage limits stated.

ShardingSphere-Proxy is an alternative open-source proxy with Docker and Helm deployment. OceanBase Database suits teams wanting a distributed database with full kernel capabilities rather than a PostgreSQL proxy. MySQL is the relevant switch when the target engine is MySQL. ProxySQL offers an open-source community edition with its core features for supported platforms. TurbineProxy is a free Apache-2.0 option with no per-query fees.

Verdict

PgDog is a strong pick for PostgreSQL teams that want one open-source layer for pooling, replica routing, sharding, failover behavior, and online resharding without application rewrites. Its deployment flexibility and free core make it accessible to self-hosting engineers, while enterprise support adds a production path. Choose another product if you need a different database engine, a managed-first service, or a signed macOS binary.