How to configure ORY Hydra to route read queries to a PostgreSQL Read Replica (AlloyDB) and also add the primary db

[vazidahmed]

I am currently running ORY Hydra backed by a primary GCP AlloyDB (PostgreSQL) instance. Under heavy load, I am experiencing high latency, which occasionally leads to database connection drops.

To mitigate this and improve stability, I want to introduce a read replica to handle all the read-heavy queries and offload the primary instance.

My questions for the community are:

• Does ORY Hydra natively support splitting read and write operations between a primary database and a read replica?
• If supported, what is the recommended way to configure the DSN (Data Source Name) or environment variables to achieve this routing?
• Are there any known caveats or best practices when scaling Hydra's database tier in this way?

Any guidance or examples would be highly appreciated. Thank you!

[Its-Nyein]

Hello @vazidahmed,

Ory Hydra doesn't natively support read/write splitting at the DSN level — you can only configure a single DSN connection string. However, there are a few proven approaches to achieve what you need:

Option 1: PgBouncer / PgPool-II as a proxy layer

Put a connection pooler in front of your databases that handles read/write routing transparently:

• PgPool-II can inspect queries and route SELECT statements to the replica and writes to the primary automatically.
• Hydra connects to the proxy as if it were a single database.

# Hydra config — points to proxy, not directly to DB
dsn: postgres://hydra:secret@pgpool-host:5432/hydra

This is the most common approach in production and requires zero changes to Hydra itself.

[vazidahmed]

Hi @Its-Nyein
Thank you for sharing this solution I will try to look into that if it works for me

[vinckr]

Hey @vazidahmed

If you're running Hydra for a commercial project at a scale where you're hitting connection limits and need read replicas, you might want to look at the Ory Enterprise License.
OEL gives you a commercially supported build of Hydra with "high-performance pooling" built in: https://www.ory.com/docs/self-hosted/oel/high-performance-pooling - as well as security SLAs, regular patched images, and direct access to the engineering team for deployment guidance, incident response, or troubleshooting
If that sounds relevant, feel free to reach out to our team at and we can walk through what that looks like for your setup.