3 years ago and this is still open? It is a default with other providers, e.g. Render

Watching

We have a few HA templates that users deploy while on the platform. Melissa, our Solutions Engineer, have a few of them being flighted and then we plan to take the learnings from that project over to the native offering on the platform.

Just +1'ing this as this would be great to just add this with a click

Then I'm happy to announce we now have a Highly available Postgres template - https://railway.app/template/ha-postgres

Try it out and let us know what you think please!

watching. I would like to be able to deploy read replicas across regions too

We are looking to split this work up- what would you say is “Minimally Lovable Managed Postgres”?

I am thinking one button version upgrades. I know some others want smart replication.

+1 for multi region replication to have multi region reads.

native integration with AWS or Supabase for db would be great

What is the status of this task?

I think there is a new template for this, you should be able to find it in the template page

PostgreSQL 18 HA Cluster (AI & GIS Ready)
https://railway.com/deploy/postgresql-18-ha-cluster-ai-and-gis-read
TimescaleDB HA (Primary + Replica)
https://railway.com/deploy/timescaledb-replica-postgresql
MongoDB 8.2.3 Replica
https://railway.com/deploy/mongodb-823-replica

Hi folks, the High Availability PostgreSQL project has been moved to planning !!

TLDR: we're making it easy to upgrade a standalone PostgreSQL database to a highly available cluster with automatic failover, and just as easy to roll back if you need to.

What's your ideal HA database experience look like?

This is really great news. Railway needed a feature like this. Hopefully it will be available soon.

will this be a managed db? or is it still an unmanaged db? will we be able to have replica in different regions?

Watching

This link is no longer working

Can Railway impliment this? Please?

Something like this: https://render.com/docs/postgresql-high-availability
So when something goes down on Railway's end (like today) our apps don't go down.

This sounds really cool and I'm glad to see its now 'Completed', but can't seem to see any documentation on this, or option to upgrade to the HA setup. How is this done please?

thanks to ship this and to have shipped pg 18 as well, I have just one question, I know you can't answer with a real number, but more or less, how much would cost to run postgres in HA instead of a single node? for sure the cost of the replicas cpu/memory/storage, but what about etcd Nodes & HAProxy how much these will consume in terms of resources?

Also, what are the exact steps to upgrade an existing postgres to high availability?

in theory db->settings->enable HA

https://railway.com/changelog/2026-03-13-highly-available-postgres

You have to enable it in Priority Boarding then select the db mode > config > convert to HA

I turned on Priority Boarding and and Postgres High Availablity, but still not seeing anything in my db->settings to enable HA

You should see the option to convert under Database > Config

Ah, thank you!

I'm using a pgvector template for one of my databases, so for that one, looks like it is unavailable. But I see it on other normal Postgres installs.

Does anyone have a rough estimate on how much more HA Postgres costs vs normal Postgres?

I want to be able to control these options, seems not possible right now. It could be part of interface (Config)

ALTER SYSTEM SET max_connections = 500;
ALTER SYSTEM SET shared_buffers = '8GB';
ALTER SYSTEM SET effective_cache_size = '24GB';
ALTER SYSTEM SET maintenance_work_mem = '1GB';
ALTER SYSTEM SET work_mem = '16MB';
ALTER SYSTEM SET max_worker_processes = '24';
ALTER SYSTEM SET max_parallel_workers_per_gather = '6';
ALTER SYSTEM SET max_parallel_workers = '24';

SELECT pg_reload_conf();

RAILWAY_SHM_SIZE_BYTES=524288000

I'm relatively new to this architecture - what criteria should I look at to determine if I should upgrade to a HA cluster instead of my standard single DB?

do we have any expected downtime and/or connection url changes when upgrading to HA?