Skip to content
Database tab absent on postgres-ssl:16 image service: supported credential rotation
jaoiad
OP

23 days ago

Hello Railway team,

We need the officially supported way to rotate credentials on one existing Preview PostgreSQL service after the October 21 incident. The entire Database tab is absent, rather than a disabled Regenerate button.

Target:

  • Project earnest-energy: ec8ac335-bd40-446b-a921-c586c286bbc6
  • Environment preview: 8b2d6c68-5e89-4a88-b8c5-e476062ac8a0
  • Service postgres-preview: 2298d086-913b-4b92-93da-34ed33acdbcc
  • Active deployment: 7723954e-18e8-41fb-954e-ddf3eb7ee0f6
  • Live source image: ghcr.io/railwayapp-templates/postgres-ssl:16
  • Existing persistent volume mounted at /var/lib/postgresql/data; PostgreSQL and API connectivity have been verified read-only.

The visible service tabs are Deployments, Backups, Variables, Metrics, Console and Settings. Console opens an SSH shell. Settings has no Database / Connection / Regenerate control. There are no staged changes. The only configured variable names are PGDATA, POSTGRES_DB, POSTGRES_USER and POSTGRES_PASSWORD. Service metadata reports templateId=null and templateServiceId=null, with createdAt=2026-10-01T12:23:14.325Z. We do not assume these facts establish the cause.

API and backup consumer credentials are reference variables pointing to this service's POSTGRES_PASSWORD. We have not edited any password variable, executed ALTER USER/ROLE, changed the image or volume, or restarted/redeployed anything to work around the missing UI. A short coordinated Preview maintenance is possible once the procedure is qualified; Production must remain untouched.

Could you confirm:

  1. Why this exact service is not recognized or eligible for Database View: service provenance/template detection, permissions, or a platform restriction?
  2. Whether the official Regenerate control can be restored for this existing instance without recreating the database/service or moving its volume?
  3. The officially supported synchronized credential-rotation procedure for this case, including consumer references and recovery if a step fails?

Please provide diagnosis/instructions only; do not modify resources, permissions, configuration, credentials, plan or data. No credentials or customer data are included here.

We checked the older missing-Data-tab discussion (44f122dc): that case had no source set, whereas our live image source is already set. The recent sealed-password discussion (321c091d) concerns a disabled button, while our whole Database tab is absent.

$10 Bounty

3 Replies

Status changed to Awaiting Railway Response Railway • 23 days ago


Railway
BOT

23 days ago

This thread has been opened as a bounty so the community can help solve it.

Status changed to Open Railway • 23 days ago


Try this:

  1. Create a new password.
  2. Go to the Console tab, run psql, then run Run ALTER USER postgres with password '<PASSWORD>'; where <PASSWORD> is the new password.
  3. Go to the Variables tab, and change POSTGRES_PASSWORD to the new password.
  4. Save changes and redeploy the database.

jaoiad
OP

23 days ago

Thank you. Before any action, could Railway confirm that this is an officially supported fallback for this exact service with no Database tab, and specify the synchronization and recovery steps if either side fails?

Our actual database role is media_os, not postgres. The two consumers are the Preview API and a scheduled backup producer, both using reference variables from this database. The current approved operation does not permit a manual SQL/password-variable workaround, so we have made no changes and have not accepted the solution.

We still need to establish why Database View is missing despite the live postgres-ssl:16 image source, and whether Regenerate can be enabled on the existing instance without recreating or moving the volume. The documented post-incident guidance warns about credential/variable mismatches; please clarify how the proposed fallback and dependent redeploys should be qualified for this case.

Diagnosis and instructions only, please; do not modify any resources or credentials. Production remains outside scope.


jaoiad

Thank you. Before any action, could Railway confirm that this is an officially supported fallback for this exact service with no Database tab, and specify the synchronization and recovery steps if either side fails? Our actual database role is media_os, not postgres. The two consumers are the Preview API and a scheduled backup producer, both using reference variables from this database. The current approved operation does not permit a manual SQL/password-variable workaround, so we have made no changes and have not accepted the solution. We still need to establish why Database View is missing despite the live postgres-ssl:16 image source, and whether Regenerate can be enabled on the existing instance without recreating or moving the volume. The documented post-incident guidance warns about credential/variable mismatches; please clarify how the proposed fallback and dependent redeploys should be qualified for this case. Diagnosis and instructions only, please; do not modify any resources or credentials. Production remains outside scope.

The Database tab is missing because you created the service by manually deploying the Postgres image instead of deploying a database natively by right-clicking the canvas.


Welcome!

Sign in to your Railway account to join the conversation.

Loading...