repeated failures deploying
Anonymous
HOBBYOP

a month ago

Hello Railway Support,

We're experiencing repeated failures deploying an update to our BridgeSpace staging application through the Railway CLI.

The application is currently running successfully on Railway, and our existing deployment remains healthy. However, two consecutive attempts to deploy a small, tested code change have failed before the build process begins.

Project: BridgeSpace

Environment: production (our established staging environment)

Service: bridgespace-staging

Service ID: 45ed61a4-7cab-44f2-a6f1-987f259930dc

Railway CLI version: 5.64.2

Source commit: 65f287169b3da98d05f8a625f4107c8ea9fc1ec2

Failed deployment IDs:

  • 72f7c5ef-d563-468b-8ea3-02f5ab47be37
  • 533aae95-b144-424a-a7e7-4c8141d23525

Both attempts completed local indexing and compression but timed out while uploading the archive to Railway's /up endpoint.

Railway subsequently marked the deployments as FAILED with the message:

"Failed to create code snapshot. Please review your last commit, or try again."

No build logs were generated, and the application never reached the build or startup stage.

We've verified that:

  • The source checkout is clean.
  • TypeScript checks, 83 focused tests, and the production build pass locally.
  • Railway authentication works.
  • The correct application service was explicitly selected.
  • The MySQL service was not targeted.
  • The existing successful deployment, 6103f31f-f292-4df7-a395-76f59eafd4aa, remains active and healthy.
  • The staging database is accessible.

We previously encountered a similar code-snapshot failure with this project, which was resolved after a later controlled CLI deployment attempt.

Could you please investigate why code snapshot creation is failing and advise whether there is a known issue with the CLI upload endpoint or an alternative supported deployment method?

We would prefer to preserve our current CLI-based deployment workflow rather than change repository connections or production configuration.

Thank you for your assistance.

$10 Bounty

3 Replies

Status changed to Awaiting Railway Response Railway • about 1 month ago


Railway
BOT

a month ago

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

Status changed to Open Railway • about 1 month ago


yuri-barreira-dev
FREE

a month ago

This looks like it’s getting stuck during the upload, before Railway even starts building the app. Finishing compression doesn’t necessarily mean the full archive reached their servers...

Do you have the full CLI error, the compressed upload size, and how long it took to time out? The timestamps would help us check their logs too. If you make another attempt, I’d add --verbose to your existing command and keep it to one upload at a time.

I’d keep your current setup for now. Trying the same command from another network could also help narrow things down, particularly if you’re using a VPN or proxy, while keeping your CLI workflow.


Anonymous
FREE

a month ago

AI-assisted source review of CLI 5.64.2; I have not accessed your project or run a deployment.

One useful detail beyond the existing network suggestion: a small Git diff does not imply a small CLI upload. In this version, the CLI builds a tar/gzip from the selected project files, rather than uploading only that commit's changes. Check the actual compressed byte count before treating this as a small upload.

The pinned source gives three concrete checks:

  • Archive selection: it reads .railwayignore and .gitignore, explicitly skips .git and node_modules, and includes other selected files. Review large generated assets or caches that are not excluded; keep files required for your build. Do not add --no-gitignore to troubleshoot this, because that disables Git ignore filtering. Archive and upload implementation
  • Failure stage: the upload code separates a request/send error from HTTP responses. A 413 is explicitly reported as an oversized upload, including its byte count; a 400 uses the server's message; other non-200 statuses are reported separately. Preserve the exact error chain. A timeout alone cannot establish whether the backend received the complete archive or why snapshot creation failed. Response handling
  • Flag limitation: --verbose prints the selected service, environment and compressed bytes before upload. --detach is checked only after upload has returned successfully and its response has been parsed, so it cannot bypass a failure at /up. Command flow

Keep the healthy deployment and current repository configuration in place. For Railway's investigation, attach the compressed byte count, exact UTC start/failure times, full redacted CLI error and the two deployment IDs you already supplied. Those distinguish a size rejection from a request timeout and let staff correlate snapshot processing. If you authorize one further upload, retain the same service/environment and collect that evidence with --verbose; railway up is a real deployment, not a diagnostic dry run. I cannot confirm a provider incident or promise that a retry fixes it from the public evidence alone.


ishusinghse
FREE

21 days ago

We’re interested in examining the CLI upload and code-snapshot failure path. If this is still open, could you provide the redacted command output and confirm whether a public reproduction is possible?


Welcome!

Sign in to your Railway account to join the conversation.

Loading...