Your staging Postgres service has its volume attached and in a ready state, with plenty of free space, so the database and its data are still in place. We could not find recent log output from the database that explains the terminated connections, so the cause is not established yet.

The supported way to rotate the password is the Regenerate button under the Postgres service's Database > Config tab, which changes the password inside Postgres, updates the service variables, and restarts the database without touching the volume. After it succeeds, redeploy every service that references the Postgres variables, and update any connection string you pasted by hand. Please do not edit POSTGRES_PASSWORD or PGPASSWORD directly or run ALTER USER, since that leaves the database and its variables out of sync.

Because Regenerate needs a working connection to the database, it will keep failing while the management connection is terminating. To help pin that down, please share the exact UTC times of your most recent failed attempts and any error lines from the Postgres service's deploy logs around those times, with secrets removed.

Hello Railway Support,

Thank you for your response.

I have attempted to regenerate the PostgreSQL password multiple times, but the operation continues to fail with the following message:

"Failed to regenerate password. Couldn't connect to the database to change its password: Connection terminated unexpectedly."

I have documented failed attempts around:

• October 23, 2026, approximately 03:10 UTC
• November 1, 2026, approximately 20:59 UTC

Additionally:

• The database viewer reports "Connection terminated unexpectedly."
• The PostgreSQL extensions page fails with the same error.
• Railway CLI SSH connections close unexpectedly.
• The database tunnel reports "channel 1: open failed: unknown channel type: unsupported."
• The PostgreSQL deployment shows Online, but the Deploy Logs page displays "No logs in this time range."

I have not manually modified any PostgreSQL passwords, database variables, or database configuration.

Because the administrator password was previously exposed, I need assistance rotating it securely.

Could your team investigate why Railway's management connection cannot reach this PostgreSQL service, and help resolve the issue without affecting the existing database or persistent volume?

Please let me know if you need additional diagnostics.

Thank you.

Thanks for the timestamps. Your staging Postgres has been running on the same deployment since October 15, its volume is attached and ready with most of its space free, and that deployment has produced no log output we can see, which matches the empty Deploy Logs page. That does not explain why the database viewer, extensions page, Regenerate, and SSH tunnel all get terminated connections, so the cause is not established yet.

Because Regenerate has to connect to the database to change the password inside Postgres, it cannot complete while those connections are being terminated, and editing POSTGRES_PASSWORD or PGPASSWORD by hand would leave the database and its variables out of sync. Once Regenerate succeeds, it changes the password, updates the service variables, and restarts the database without touching the volume. After that, redeploy every service that references the Postgres variables and update any connection string you entered by hand.

In the meantime, you can reduce exposure of the current credential by removing the database's TCP proxy (public access), which removes its public endpoint. Before you do, check that nothing outside Railway still connects through it, because those connections will stop. Removing the proxy does not rotate the password.

Thank you for the response. Unfortunately, this has not resolved my issue.

My PostgreSQL 16 database remains Online, but the database viewer, extensions page, password regeneration, and Railway SSH tunneling continue to fail.

I have already attempted the recommended password regeneration several times without success.

Because my PostgreSQL administrator password was exposed, I need to rotate it urgently. However, I cannot safely perform the rotation while Railway's management connection is failing.

Could a Railway engineer please manually review this issue and advise how to restore connectivity and safely rotate the credential without losing the existing database or volume?

I am currently on the Railway Trial plan. Please let me know whether this issue can be escalated for technical investigation on my current plan.

Thank you.

Try selecting the service, press Cmd/Ctrl + K, and deploy source image.