We have a FastAPI backend deployed on Railway and a Postgres database provisioned via Railway. Our backend uses the DATABASE_URL provided by Railway (e.g., postgresql://postgres:...@postgres.railway.internal:5432/railway). On startup, our backend runs code to create tables using both SQLAlchemy’s Base.metadata.create_all() and a raw SQL CREATE TABLE IF NOT EXISTS runs (...) statement.

Symptoms:

• The backend logs show that the table creation code runs successfully (e.g., “Runs table created (if it did not already exist).” and “Tables created on startup!”).

• However, when the backend tries to insert into the runs table, we get the error:

psycopg2.errors.UndefinedTable: relation "runs" does not exist

• When we check the Railway Postgres Data UI, it says “You have no tables.”

• We are certain the backend is using the correct DATABASE_URL (confirmed by printing it in logs).

• We do not see any errors during table creation in the logs.

What we’ve tried:

• Verified the DATABASE_URL in both the backend and Railway dashboard.

• Attempted to create the table both via SQLAlchemy and raw SQL on backend startup.

• Checked the Data UI for tables after deployment—still empty.

• Looked for a public connection string to connect via a local SQL client, but only see the internal hostname.

Request:

• Why are tables not being created or visible in the Railway Postgres Data UI, even though the backend logs indicate successful creation?

• Is there a network isolation or schema issue that could cause this?

• How can we reliably create and view tables in our Railway Postgres instance from our deployed backend?

Additional Info:

• Project: [your Railway project name or ID]

• Backend: FastAPI (Python)

• Database: Railway Postgres

• Example DATABASE_URL: postgresql://postgres:...@postgres.railway.internal:5432/railway