you have your database in a different region than your other services, everything should be in the same region

I'm talking about this service:
b06a97c3-f560-4355-ad23-6aaeaf9b5e73
I moved it back to the same region as the database (I tested the metal to see if it did perform better on it)

did you delete the project?

no, project is still 9bd6efc6-3bb8-4c31-b1bf-885307cf2f9e 🙂

I modified the service to make sure I was on the same region as the database

please have everything in the same region

Here i'm just testing the cms / db relationship, I have added sentry on the payloadcms and I can see some things like this on requests

both the db and the cms service are on the same region

are you connecting for every db query?

After some investigation/updates/testing, I can reduce the issue database, I have setup a postgresql ha from the template provided by railway, but it seem that pgpool is always connecting to the US server even if the query come from the EU application.

I have tested to make 1 pg pool in the eu region and setting up PGPOOL_BACKEND_NODES with weight to try to make pgpool use the replica in priority

0:${{pg-primary.RAILWAY_PRIVATE_DOMAIN}}:5432:0,
1:${{pg-eu-west.RAILWAY_PRIVATE_DOMAIN}}:5432:1,
2:${{pg-us-west.RAILWAY_PRIVATE_DOMAIN}}:5432:0

But i still get an aweful latency.

Here is some stats:
Connected to pg-eu directly: 400ms average page load
Connected to pgpool both server: 2000ms average page load
Connect to pgpool with only the EU server: 400ms average page load

After enabling the pgpool logs:
I can see he is redirecting all my requests to the primary node :

2025-02-24 11:08:11.420: main pid 3: LOG:  find_primary_node: primary node is 0
2025-02-24 11:08:56.246: child pid 137: LOG:  new connection received
2025-02-24 11:08:56.098: [unknown] pid 148: LOG:  DB node id: 0 backend pid: 1253 statement: Bind: select distinct 

can you move pgpool and all the nodes into the same region as your main application?

They are, I could pin point the issue to pgpool.
Here is my testing setup:

• PG admin on europe-west4

• PG Pool on europe-west4

• 1 PG Replica on europe-west4, 1 PG Primary on us-west1

I connect on pg admin (EU) and run a basic select query

• query on pg-0.railway.internal (US) : Total query runtime: 1 secs 2 msec.

• query on pg-1.railway.internal (EU) : Total query runtime: 102 msec.

• query on pgpool-eu.railway.internal (EU): Total query runtime: 804 msec.

If i run show pool_nodes on pgpool, I get this result, and in the pgpool logs I can see that the queries are going to the pg-1 server but some unrelated queries are made to pg-0 which I guess, slow the whole process.

there will not be a lot of writing to the databasebut it is important is that I get fast read performance on all regions

Unless you have an other solution for a single app to serve request from all regions in a fast and efficient way, for now I enable multi region replica on my main app, used the RAILWAYREPLICAREGION env variable internally to pick a database url and pointing to a different pg pool in us or eu based on the region

And to be honest it seem to do the trick as it's fast if I bypass pgpool and connect directly to my read replica but unfortunatly I need to be able to make some write queries too 😦

unfortunately we cannot yet modify the laws of physics, so if your app or pgpool is connecting to a database in a region far away there will be increased latency

yes, I don't have issue with that, my issue is more why pgpool connect to the us db when it's not needed

i'm sorry but I don't have an answer for that, i think that would need to be asked in a forum of people who know pgpool