23 days ago
Hello,
On our Hobby account, the web and PostgreSQL services in Production are shown as Online, but the PostgreSQL Console displays "WebSocket connection failed" and remains Disconnected. The last attempt was on 28 October 2026 at approximately 13:35-13:36 Israel time; Reconnect did not resolve the issue either. The Database view had previously displayed "We are unable to connect to the database via SSH". We have not established that the database itself is unavailable.
We are preparing a deployment but have not carried it out. We would appreciate focused guidance:
- How can we diagnose and resolve the Console failure without a restart or redeployment?
- What is the supported way, using existing permissions, to perform read-only queries of the schema, migration history, and the number of records eligible for backfill? Startup runs DDL checks and existing migration and backfill mechanisms; we have not yet verified whether these would actually change data or schema. We are not asking you to execute SQL for us.
- What costs could these reads incur, and is it possible to enforce a per-operation cost cap without putting Production at risk? A general cap that could shut down services is not suitable.
- How can we create a fresh logical backup on the current plan, and what are the costs and ways to limit them? Creating a Volume backup in the current dashboard requires Pro, and we have not upgraded.
- Is it possible to build and validate a Docker image of a Rollback candidate, including startup and health, using existing infrastructure, without a live deployment or creating a new service?
This request is for diagnosis and guidance only. Please do not change services, the database, permissions, configuration, or the plan, and do not restart, deploy, run a migration, or perform a paid operation. We will review the proposed minimum action and its cost before authorizing execution.
1 Reply
Status changed to Awaiting Railway Response Railway • 23 days ago
23 days ago
Your Production Postgres has a DATABASE_PUBLIC_URL variable. We couldn't establish why the dashboard Console fails, but you don't need the Console for any of the reads you listed.
For read-only queries with no restart, connect with psql, pgAdmin or DBeaver using that connection string, or run railway connect postgres --tunnel-only to get a local tunnel. To guarantee nothing gets written, use a role with read-only grants or run SET default_transaction_read_only = on at the start of the session. Schema, migration-history and backfill counts are all plain SELECTs.
For a logical backup on Hobby, run pg_dump from your own machine against the same connection string. This doesn't need the Pro volume backup feature.
Traffic through the public TCP proxy is billed as network egress. Small queries move very little data, while a full dump costs roughly in line with the database size. To keep it down, dump only the tables you need and compress the output with pg_dump -Fc. Usage limits apply to the whole workspace, not to individual operations, so we can't offer a per-query cost cap.
You can build and test a rollback image locally with Docker and point it at a copy restored from that dump. Its startup DDL, migration and backfill steps will run against whatever database it connects to, so keep it away from Production.
Status changed to Awaiting User Response brody • 23 days ago
16 days ago
This thread has been marked as solved automatically due to a lack of recent activity. Please re-open this thread or create a new one if you require further assistance. Thank you!
Status changed to Solved Railway • 16 days ago