Bug: API returns HTTP 500 instead of 404 for non-existent deployment

[ygrishajev]

Labels: support

Describe the bug

When requesting deployment info for a non-existent deployment, the API returns HTTP 500, while the response body correctly indicates not found. This breaks expected HTTP semantics and client-side error handling.

• account address = akash1x5sp0raswdrum5en9l4htyetkcj6teuu8sucu
• deployment sequence = 24793434
• deploy.yaml = not available
• group sequence = not available / unknown

Endpoint:

GET /akash/deployment/v1beta4/deployments/info

Example:

https://consoleapi.akashnet.net/akash/deployment/v1beta4/deployments/info?d.owner=akash1x5sp0raswdrum5en9l4htyetkcj6teuu8sucu&d.dseq=24793434

To Reproduce

1. Call the deployments info endpoint with a non-existent deployment sequence.
2. Observe response body indicates “not found”.
3. Check HTTP status code.
4. Status code is 500.

Expected behavior

The API should return HTTP 404 (Not Found) when the deployment record does not exist.

Screenshots

See Slack thread showing 500 status with a “not found” response body.

Desktop

• OS: Unknown
• Browser: Unknown
• Version: Unknown

Smartphone

Not applicable.

Additional context

• Confirmed internally as a backend bug.
• Marked as a quick fix by maintainers.
• Client-side workaround is not recommended.

[vertex451]

The fix is to replace sdkerrors.Register() with sdkerrors.RegisterWithGRPCCode() for the proper error code propagation.

I updated the following repos:

• chain-sdk: fix!: replaced sdkerrors.Register with sdkerrors.RegisterWithGRPCCode for the proper error propagation. chain-sdk#173
• node: fix!: replaced sdkerrors.Register with sdkerrors.RegisterWithGRPCCode for the proper error propagation. node#2030

In the provider repo we already use the proper sdkerrors.RegisterWithGRPCCode().

In the rpc-proxy repo there is no entries of sdkerrors.Register().