What causes a 400 error with message Server not found for <url>?

[jtherrmann]

All of the API endpoints for the test deployment of our Flask app, which is deployed via Amazon API Gateway, are currently returning a response like the one below:

{
  "detail": "Server not found for https://vd2gh6uqw3.execute-api.us-west-2.amazonaws.com/api/user",
  "status": 400,
  "title": "Bad Request",
  "type": "about:blank"
}

We depend on the latest pre-release of openapi-core, https://github.com/python-openapi/openapi-core/releases/tag/0.19.0a1.

The error message leads me to believe that openapi-core is raising a ServerNotFound exception from BasePathFinder.find (here), which is called from BaseAPICallValidator._find_path (here).

I have a couple of questions:

• Is BaseAPICallValidator for validating requests, responses, or both?
• Why does the validator need to validate the server URL? What exactly does APICallPathFinder._get_servers_iter do? What is the fundamental meaning of a ServerNotFound exception?

[p1c2u]

HI @jtherrmann

the error "Server not found for https://vd2gh6uqw3.execute-api.us-west-2.amazonaws.com/api/user" means your OpenAPI specification doesn't define server that match url from your request.

Is BaseAPICallValidator for validating requests, responses, or both?

It's a base class for request and response validation

Why does the validator need to validate the server URL?

Server is part of OpneAPI specification

What exactly does APICallPathFinder._get_servers_iter do?

Finds valid server definition in your specification

What is the fundamental meaning of a ServerNotFound exception?

ServerNotFound exception is raised when your request server does not match any defined servers in your specification.

[jtherrmann]

@p1c2u This answer was extremely helpful, thank you! I had posted some comments with additional questions, but I deleted them because they were based on a partial misunderstanding of how OpenAPI works. But I understand how server URLs work now and we were able to resolve the issues with our app. Thanks again for your quick response!

[jtherrmann]

In case anyone reading this is having a similar issue, our problem was that we have an AWS CloudFront distribution that forwards API requests to our AWS ApiGateway app. The forwarded requests have a URL like <api-gateway-url>/api/<endpoint>, e.g. https://vd2gh6uqw3.execute-api.us-west-2.amazonaws.com/api/user, but our app does not expect the /api component. Fixing this issue is as simple as specifying the following relative server URLs in our OpenAPI spec:

servers:
  - url: /
  - url: /api

The / url is technically optional, but including it allows us to make queries via our Swagger UI (which would force us to send requests to /api if that were the only server url specified).

[p1c2u]

I'm glad you found the problem. Any suggestions about debugging improvements are welcome.

[jtherrmann]

@p1c2u I think the error message for the ServerNotFound exception could be improved. When I first saw the error:

Server not found for https://vd2gh6uqw3.execute-api.us-west-2.amazonaws.com/api/user

I assumed that the request had failed to reach the server. I did not even realize that the error was coming from the OpenAPI spec validator. At a minimum, I think the error message should make it clear that this is an OpenAPI spec validation problem. For example, it could say something like this:

Request is not valid for the given OpenAPI spec: Request URL https://vd2gh6uqw3.execute-api.us-west-2.amazonaws.com/api/user does not match any of the OpenAPI spec server URLs

In fact, I think that all request validation errors should begin with something like Request is not valid for the given OpenAPI spec: to make it clear that this is an OpenAPI spec error.