24 days ago
Hello Railway team,
I’m looking for official documentation or staff confirmation of the service Console’s general behavior before considering it for a bounded, read-only diagnostic.
This is a product-capability question, not a request to investigate a particular account or deployment. I do not wish to publish resource identifiers or configuration details in a public thread.
Could you please clarify the following?
- Authentication and prerequisites
Does the service Console use the existing authenticated Railway dashboard session without requiring a separately registered SSH key or local SSH host-key acceptance?
Please identify any plan, role or runtime prerequisites. Does opening the Console require a restart, rebuild, redeployment or installation?
- Deployment and instance selection
Can an operator select a specific running deployment and replica, or does the Console attach to the service’s current instance?
What UI information identifies the selected deployment and instance? If that instance is replaced or stops, does the session close, refuse further commands, or reconnect to a different instance?
- Execution context
Does a Console command use the selected container’s existing operating-system user and applied runtime environment, rather than newer configuration changes that have not been deployed?
Are there shell initialization steps, wrappers or privilege differences operators should account for?
I understand that shell access is not inherently read-only; the intended diagnostic would enforce its own limited database transaction and output rules.
- Results, retention and disconnection
How are command stdout, stderr and exit status delivered? Are they independent of the ordinary deployment-log ingestion system?
What command input, history or output is retained, and who can access it?
If the browser disconnects or the session expires, does the command continue? Is there a supported way to cancel only the command or session the operator started? Can reconnection automatically retry or replay any command?
Requested response
Please provide official documentation or Railway staff confirmation, identifying any unsupported behavior or limitations.
Guidance only: please do not open a shell, execute commands, inspect account resources, create credentials or change anything on my behalf.
If answering requires account-specific information, please identify an approved private channel first. I will not post resource identifiers, logs, connection strings, screenshots or credentials in this public thread.
Goal and desired outcome
We want to complete a planned credential refresh for an existing Staging application while protecting its database, keeping secrets private and avoiding unintended application or infrastructure changes.
Our immediate goal is narrower: identify an officially supported, authenticated method for performing one bounded, read-only database check from the existing running application container.
Why this check is needed
The application has startup routines that may perform database initialization or migration work when it is deployed again. Before approving that later deployment, we need to check a small set of database prerequisites rather than assume the current database still matches historical records.
The proposed check would confirm only whether the expected schema objects and migration records are present, whether required application-initialization records already exist, and whether a designated synthetic verification identity is absent.
It would return limited nonsecret identifiers, timestamps and Boolean results—not customer records, passwords, connection strings or a database export. The results would describe the database at that observation, not certify the entire database or guarantee future unchanged state.
What we need Railway’s guidance to enable
We need a supported procedure that lets an authorized operator:
- Identify the correct running deployment and instance, avoiding accidental execution in another environment or a replacement instance.
- Use that application’s existing database connection configuration and login role, keeping credentials inside the container rather than exporting them or substituting an administrator account.
- Run the reviewed check with fixed timeouts, stop if its prerequisite checks fail, and roll back and close the connection without modifying application data or schema.
- Obtain reliable completion information, including bounded output, exit status and session-closure behavior. Missing output or a disconnected browser must not be mistaken for successful completion.
Thank you.
3 Replies
Status changed to Awaiting Railway Response Railway • 24 days ago
24 days ago
This thread has been opened as a bounty so the community can help solve it.
Status changed to Open Railway • 24 days ago
24 days ago
-
Yes, the Console uses the existing authenticated session for the user.
-
The user can select the desired active replica for the Console session.
-
It uses the current running container, not staged changes.
-
Terminal output is separate from the deployment logs, and it combines
stdoutandstderr. Disconnection depends on runtime: Commands may continue even after the browser disconnects, and the output may be retained for replay. Closing the browser may not immediately cancel the command.
23 days ago
Thank you—the authentication, active-replica selection and running-container explanations are helpful.
One completion question remains: For a command run in dashboard Console, what supported read-only view or receipt identifies that same Console session and shows the command’s actual termination and exit status?
Can that information be viewed after browser disconnection without opening another remote session or running the command again? If no such mechanism is available, please confirm that limitation.
This is a general product question and requests guidance only. Please do not access or change customer resources.
23 days ago
IIRC no. If the Console session isn’t retained, or there isn’t a way to identify/read that session, nor check its exit status.