HN Debrief

PgDog is funded and coming to a database near you

  • Infrastructure
  • Databases
  • Open Source
  • Developer Tools
  • Startups

The post is a funding announcement for PgDog, a Rust proxy that sits in front of PostgreSQL and handles connection pooling, read load balancing, and sharding so applications can keep talking standard Postgres while the data spreads across multiple servers. The company says it has already handled multi-terabyte deployments and millions of queries per second, and will make sharding features open source while charging for a control plane, query QoS, and support. People who had actually used PgDog were broadly positive about stability and day-to-day usefulness, especially as a smarter pooler that can smooth Aurora failovers and solve some rough edges people hit with PgBouncer.

If you are hitting Postgres limits, PgDog looks worth evaluating as an infrastructure layer, especially for pooling and sharding without changing application code. But you should read it as a proxy that helps orchestrate scale, not as magic Postgres throughput or built-in failover, and you should pressure-test its operational fit against Citus, Vitess-style systems, and simply switching databases.

Discussion mood

Cautiously positive. People liked that PgDog is a real open source project with users in production and a founder who knows Postgres internals, but they were quick to challenge inflated scale claims, fuzzy HA language, and the idea that a proxy can somehow sidestep the fundamental single-writer limits of Postgres.

Key insights

  1. 01

    HA still depends on external failover

    PgDog can steer traffic away from unhealthy nodes, but it does not promote replicas itself. That leaves the hard part of database high availability with tools like Patroni, CloudNativePG, AWS RDS, or Aurora. This sharpens the product boundary. PgDog improves the client-facing side of failover, not the control plane that decides who becomes primary.

    Treat PgDog as one layer in a Postgres HA stack, not the stack. If failover automation is your main pain, validate the orchestration story first and only then add a proxy.

      Attribution:
    • doctorpangloss #1
    • gchamonlive #1
    • parthdesai #1
    • nijave #1
  2. 02

    Sharding helps, but each shard stays single-writer

    The deepest technical pushback was that Dynamo and Postgres solve different problems. PgDog can shard writes across key ranges or tenants, but once a row lands on a shard, its writes still funnel through that shard's primary. That means the product improves aggregate scale and blast radius, yet it does not deliver the unbounded horizontal write scaling that a Dynamo-style system is built for.

    Use PgDog when your write load partitions cleanly by tenant, keyspace, or service boundary. If your hottest write path concentrates on shared records or global transactions, expect the bottleneck to move, not disappear.

      Attribution:
    • r7n #1 #2
    • zamalek #1
  3. 03

    Major upgrades are solved operationally, not architecturally

    The practical answer to Postgres major-version downtime was logical replication plus a controlled cutover, often with only seconds of paused writes if you rehearse it properly. The friction is not that the path is unknown. It is that PostgreSQL still lacks a polished upgrade-specific workflow that handles DDL, sequences, and switchover cleanly enough for this to feel routine.

    If version upgrades are your recurring outage source, invest in a rehearsed logical replication runbook or a managed blue-green workflow now. Do not wait for a sharding proxy to solve a problem that sits elsewhere in the stack.

      Attribution:
    • tempest_ #1
    • tux3 #1
    • tschellenbach #1
    • briffle #1
  4. 04

    The business model keeps core scaling open

    The licensing answer landed well because the company drew a clear line. Sharding and running Postgres at scale stay open source, while the paid layer is the control plane, QoS, and support. That is a more credible posture than the usual pattern where scale features gradually drift behind an enterprise wall after fundraising.

    If you care about adoption risk, watch whether future releases keep that boundary intact. For now, the commercial plan looks compatible with self-hosting the core data path and only paying for operational convenience.

      Attribution:
    • BowBun #1
    • levkk #1 #2
  5. 05

    Kubernetes and multi-tenant config are still rough

    The most concrete user report was not about throughput. It was about configuration churn in multi-tenant Kubernetes setups. Teams are already building controllers and secret-generation workflows around pgdog.toml and live reloads to make tenant creation atomic. The company acknowledged this is exactly where the enterprise control plane is headed, and also owned a known passthrough auth caching issue tied to SCRAM.

    If your environment creates and rotates many databases or credentials automatically, test configuration and auth workflows before betting on PgDog. Operational ergonomics may matter more than raw proxy performance.

      Attribution:
    • apt-get #1
    • levkk #1
    • maherbeg #1
    • aejm #1
  6. 06

    PgDog fills the Vitess-style gap for Postgres OLTP

    People kept mapping PgDog against the systems they already know. The most useful framing is not “better than Citus” in general. It is “closer to a Vitess-style proxy approach for PostgreSQL OLTP.” That means easier deployment because it stays outside the database, no need for upstream changes or extension support, and weaker distributed query capability than OLAP-oriented systems like Citus.

    Choose based on workload shape. If you need distributed SQL analytics, Citus still has the stronger query engine. If you need to keep existing Postgres apps running while splitting OLTP traffic across shards, PgDog is the more direct fit.

      Attribution:
    • parthdesai #1
    • levkk #1 #2
    • christoff12 #1

Against the grain

  1. 01

    Switching databases may be cleaner than proxy sharding

    For teams far enough along to need a sharding proxy, the harder problems may already have outgrown the appeal of preserving vanilla Postgres. Resharding very large datasets is still ugly, logical replication may not catch up on multi-terabyte systems, and application-side sharding can be more flexible than pushing routing logic into a proxy. In that view, PgDog is most attractive in the narrow middle where you are almost too big for one Postgres instance but not ready to change databases.

    Before adopting a proxy layer, compare it against a deliberate migration to a database built for your workload. The operational migration may be painful either way, so pick the end state you actually want.

      Attribution:
    • aeyes #1
  2. 02

    Mongo got the client experience right

    A sharp defense of Mongo argued that the biggest scaling advantage is not just the server architecture. It is the end-to-end client model. Drivers understand topology changes, server selection, and read preferences natively, so applications can choose consistency and routing behavior directly. A proxy in front of Postgres centralizes that logic, but it also hides those choices from application code and may offer less nuanced control.

    If per-query consistency and replica choice are important to your product, inspect how much control a proxy removes from the client. Operational simplicity can come at the cost of application-level intent.

      Attribution:
    • saghm #1
    • bourbonproof #1
    • dzonga #1
  3. 03

    Enterprise support may outgrow a tiny team fast

    Some skepticism was not technical at all. Selling and supporting infrastructure in the database path is brutal, and a three-person startup can feel thin when buyers expect long procurement cycles, P0 response, and deep QA. Funding buys time, but it does not instantly solve trust for conservative database teams.

    If you are a buyer, diligence the support model as hard as the code. If you are building infra software, assume go-to-market and customer assurance will consume as much energy as engineering.

      Attribution:
    • jeremyjh #1 #2
    • karolist #1

In plain english

Aurora
Amazon Aurora, a managed relational database service from Amazon Web Services with a custom storage layer.
Citus
A PostgreSQL extension for distributed queries and sharding across multiple nodes.
CloudNativePG
An open source Kubernetes operator for running and managing PostgreSQL clusters.
DDL
Data definition language, meaning SQL statements like CREATE, ALTER, and DROP that change database schema.
HA
High availability, meaning a system is designed to keep working during failures with minimal downtime.
Kubernetes
An orchestration platform for deploying and managing containerized applications across clusters of machines.
Multigres
A PostgreSQL scaling project mentioned alongside PgDog and Neki as part of the new proxy-based sharding wave.
Neki
A PlanetScale project discussed as a Vitess-style scaling approach for PostgreSQL.
OLAP
Online analytical processing, meaning databases or workloads focused on large aggregations, reporting, and analysis.
OLTP
Online transaction processing, meaning databases that handle many small operational reads and writes for live applications.
Patroni
An open source tool for PostgreSQL high availability that automates leader election, failover, and cluster management.
PgBouncer
A lightweight PostgreSQL connection pooler that reduces the overhead of many client connections.
QoS
Quality of service, meaning rules or controls that prioritize or limit workload behavior so one bad query or client does not degrade the whole system.
RDS
Relational Database Service, Amazon Web Services’ managed relational database offering.
SCRAM
Salted Challenge Response Authentication Mechanism, a password authentication protocol designed to avoid sending or storing plain passwords.
Vitess
An open source sharding and clustering system for MySQL that uses a proxy and control plane to scale databases.
Yugabyte
A distributed SQL database designed to provide PostgreSQL-compatible interfaces with built-in horizontal scaling.

Reference links

Postgres HA and upgrade tools

Comparable database scaling systems

PgDog docs and project references

Claims and due diligence references

Other media and references