Skip to content

Problem

vanessaandersonbusiness-glitch
OP

5 days ago

Hello Railway Community,

I'm requesting technical assistance with a blocked DEVELOPMENT deployment in my Railway project.

Problem

I have an existing DEVELOPMENT service with a staged configuration patch. The deployment activation request returns INVALID_ARGUMENT before a deployment or workflow ID is created.

An earlier patch failed with this confirmed stored error (resource identifiers redacted):

Invalid patch payload: services.[REDACTED].deploy.restartPolicyMaxRetries: Too small: expected number to be >=1

That earlier patch specified:

  • restartPolicyType: NEVER
  • restartPolicyMaxRetries: 0

The current patch corrects the retry count to 1, retaining restartPolicyType: NEVER.

However, the corrected activation request still returns INVALID_ARGUMENT without a field-level explanation.

The current patch remains:

  • Status: STAGED
  • appliedAt: null
  • lastAppliedError: null
  • Zero deployments created

Railway's published configuration schema appears to permit NEVER with a retry count of 1. I haven't found evidence confirming that this combination causes the rejection.

Questions about deployment validation

  1. What could cause INVALID_ARGUMENT before a patch-application workflow is created?
  2. Is there a way to retrieve the exact API validation failure or rejected request trace?
  3. Does Railway's patch-commit validation impose additional constraints on restart policies beyond those in the public configuration schema?

Temporary volume and billing questions

The staged patch also includes a new temporary 500 MB volume attached to the DEVELOPMENT test service.

I need to confirm:

  1. When does storage billing stop after a volume-deletion request is accepted?
  2. What resource-specific recovery procedure is available if deletion fails or remains pending beyond 48 hours?
  3. Can deleting or unmounting the temporary volume trigger an unintended service redeployment, and how can that be prevented?

The test has a maximum incremental spending budget of $0.50, so reliable cleanup and cost control are important.

Current safety status

  • No deployment has been created.
  • The temporary volume has not been provisioned.
  • No incremental test charges have been observed.
  • The original production service remains unchanged.
  • The current DEVELOPMENT patch is preserved and staged.

I'm seeking diagnosis and guidance only. Please do not deploy, modify, or delete any resources.

I can provide exact project IDs, patch IDs, request traces, and timestamps privately to Railway staff if necessary.

Thank you for your help!

Awaiting User Response

5 Replies

Railway
BOT

5 days ago

The staged DEVELOPMENT patch sets a start command that base64-decodes and execs an inlined Python program several thousand characters long. Shipping code through the start command or variables isn't a supported setup, and that payload travels with every apply of the environment. Since the service already has a connected repo and branch in that same patch, the supported path is to commit the verification script to the repo and use a short start command that runs it (for example python path/to/script.py). Resubmitting the patch as it stands won't change the result.

On the restart policy, Hobby is a paid plan, so any policy and retry count can be set (restart policy docs). We haven't established that NEVER with a retry count of 1 is what's being rejected.

On the volume, a deleted volume is queued and permanently deleted within 48 hours, and it can be restored during that window using the link in the deletion email (volume docs). Since this volume hasn't been provisioned yet, removing it from the staged patch before you apply means it's never created.

We have not verified when storage metering stops after a deletion request, or whether unmounting a volume triggers a redeploy.


Status changed to Awaiting User Response Railway • 5 days ago


vanessaandersonbusiness-glitch
OP

5 days ago

Thank you. I’ll move the verification script into the existing GitHub repository and replace the encoded startup command with a short Python command.

Before activating the corrected patch, could you confirm whether storage billing stops when a volume deletion is requested or only when permanent deletion finishes?

Also, does detaching or deleting a volume automatically trigger a service redeployment?

I’m keeping the existing DEVELOPMENT patch staged until these details are confirmed.


Status changed to Awaiting Railway Response Railway • 5 days ago


Railway
BOT

5 days ago

Storage billing continues until the deletion actually completes, not from the moment you request it. A deleted volume that still holds data stays recoverable for up to 48 hours, and its stored data keeps counting as storage usage during that window. Detaching a volume from a service also does not stop its storage billing. Since this volume hasn't been provisioned yet, removing it from the staged patch before you apply means no volume is created and nothing is stored.

We haven't been able to confirm whether detaching or deleting a volume automatically starts a new deployment of the attached service, so we can't tell you how to prevent one.


Status changed to Awaiting User Response Railway • 5 days ago


Railway

Storage billing continues until the deletion actually completes, not from the moment you request it. A deleted volume that still holds data stays recoverable for up to 48 hours, and its stored data keeps counting as storage usage during that window. Detaching a volume from a service also does not stop its storage billing. Since this volume hasn't been provisioned yet, removing it from the staged patch before you apply means no volume is created and nothing is stored. We haven't been able to confirm whether detaching or deleting a volume automatically starts a new deployment of the attached service, so we can't tell you how to prevent one.

vanessaandersonbusiness-glitch
OP

5 days ago

Thank you for clarifying that storage is billed until deletion completes. I’ve moved the restart probe into the repository and staged the short Python startup command. The DEVELOPMENT patch is still staged and the temporary volume has not been provisioned.

Could a Railway support engineer confirm two things before I activate the test?

  1. What exact environment-scoped procedure removes the attached temporary volume after the test deployment is stopped, without triggering a fresh deployment? Does committing a staged volume removal with the dashboard’s no-redeploy option suppress the unmount-triggered redeploy?
  2. If deletion fails or is still pending after 48 hours, how can I stop metering for that specific volume and verify the billing cutoff? If this requires Railway intervention, what is the escalation path?

The test has a strict $0.50 incremental spending limit. I can provide the resource IDs privately. Please leave the patch and resources untouched while you review.


Status changed to Awaiting Railway Response Railway • 5 days ago


Railway
BOT

5 days ago

To remove the volume, stage the change in the DEVELOPMENT environment and review it in the staged changes banner. Clicking Deploy applies the change and redeploys any affected services. Holding Alt while you click Deploy commits the change without triggering a redeploy (staged changes docs). We haven't confirmed that the Alt commit also stops a service from redeploying when a volume is unmounted. Since this is a dev test, check that behavior on the test service before you rely on it.

After a volume deletion is applied, the volume stays recoverable for 48 hours and its stored data keeps billing during that time. There isn't a self-serve way to force-delete a single volume early or stop its metering sooner, and we don't purge individual volumes manually to shorten that window. You can see what was billed for the volume on your project's usage page.

The volume isn't billed until it exists. If you remove it from the staged patch before applying, no volume is created and there's nothing to clean up afterward.


Status changed to Awaiting User Response Railway • 5 days ago


Welcome!

Sign in to your Railway account to join the conversation.

Loading...