Skip to content
Verifying the actual digest of a private GHCR deployment on Railway Pro
strev
OP

14 days ago

Hi Railway team,

We’re evaluating Pro for an isolated acceptance worker. We would build and review a Linux amd64 image outside Railway, store it in private GHCR, and configure Railway to pull it by an exact SHA-256 digest. A Railway employee mentioned that deployment.meta.imageDigest records the digest the runtime pulled. Before we rely on that field, could you clarify two things?

  1. For a private image pinned to a single-platform OCI manifest, does meta.imageDigest report that exact manifest digest as observed by the runtime? What read-only API query or CLI command returns it alongside the deployment ID?
  2. When does that field become available? Can we verify it for a deployment whose worker starts in a disabled mode and exits, then check that same deployment’s record after it has zero running replicas? If enabling the job requires a new deployment, is there a supported way to verify the new deployment’s digest before the job starts?

A documentation link or a sanitized example response would help. This is a capability question only. Thanks!

Solved

3 Replies

Railway
BOT

14 days ago

The read-only way to get this is our public GraphQL API: the deployment(id) query returns id, status, createdAt and meta for a single deployment, and the deployments query with a projectId and serviceId lists a service's deployment IDs so you can fetch each one's meta. You can try both in the GraphiQL playground with a token from your account.

An image digest is reported in deployment metadata, but it is a record of the deployment, not a runtime attestation that the files inside the running container were unchanged. Deployment records stay queryable by ID after the process exits, and changing a service's configuration produces a new deployment with its own ID and its own meta.

We could not confirm for this answer that meta.imageDigest always equals the exact single-platform manifest digest you pin, or at which lifecycle step it is populated, so we cannot say whether a new deployment's digest is readable before its start command runs. The reliable check is to query a test deployment of your own private image and compare the reported value with the digest you pinned, both while it runs and after it exits.


Status changed to Awaiting User Response Railway • 14 days ago


Railway

The read-only way to get this is our public GraphQL API: the [`deployment(id)` query](https://docs.railway.com/integrations/api/manage-deployments) returns `id`, `status`, `createdAt` and `meta` for a single deployment, and the `deployments` query with a `projectId` and `serviceId` lists a service's deployment IDs so you can fetch each one's `meta`. You can try both in the [GraphiQL playground](https://railway.com/graphiql) with a token from your account. An image digest is reported in deployment metadata, but it is a record of the deployment, not a runtime attestation that the files inside the running container were unchanged. Deployment records stay queryable by ID after the process exits, and changing a service's configuration produces a new deployment with its own ID and its own `meta`. We could not confirm for this answer that `meta.imageDigest` always equals the exact single-platform manifest digest you pin, or at which lifecycle step it is populated, so we cannot say whether a new deployment's digest is readable before its start command runs. The reliable check is to query a test deployment of your own private image and compare the reported value with the digest you pinned, both while it runs and after it exits.

strev
OP

14 days ago

Thank you. We still need clarification from a Railway employee before relying on this mechanism.

  1. For a Pro service using a private GHCR single-platform image pinned by SHA-256, does deployment.meta.imageDigest equal the exact OCI manifest digest supplied in the image reference, or can it represent a different Railway/provider digest?

  2. At what lifecycle event is meta.imageDigest populated? Is there any supported read-only way to observe it before the deployment’s start command executes?

  3. If pre-start observation is unavailable, please confirm that the supported verification point is only while or after a test deployment has executed.

We only need confirmation of the platform-selected image identity. We are not asking for runtime filesystem attestation or application debugging. Thanks!


Status changed to Awaiting Railway Response Railway • 14 days ago


Railway
BOT

14 days ago

Every deployment has its own ID, and its record, including meta, can be read by that ID with the deployment(id) query in our public GraphQL API. The record stays readable after the process exits. Enabling the job through a config change creates a new deployment with its own ID and its own meta.

We can't yet confirm two things: whether meta.imageDigest always equals the exact single-platform manifest digest in your image reference (rather than a digest resolved during the pull), and at which deployment step it is written. That means we can't say it is readable before the start command runs.

Until those are confirmed, the supported way to verify the image identity is with a test deployment: compare its reported meta.imageDigest with the digest you pinned, both while it runs and after it exits.


Status changed to Awaiting User Response Railway • 14 days ago


Railway
BOT

7 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 • 7 days ago


Welcome!

Sign in to your Railway account to join the conversation.

Loading...