Revert unintentional change for auth key generation API endpoint

[josecelano]

Tasks:

• Revert change.
• E2E test for the endpoint.

[josecelano]

hi @WarmBeer, I'm trying to add a test for the API endpoint. I'm doing the same as we did for the UDP tracker. I want to run the API and use an HTTP client to make a request to that endpoint.

For some reason I can't connect to the API:

running 1 test
2022-11-22T19:12:06.110254618+00:00 [torrust_tracker::logging][INFO] logging initialized.
2022-11-22T19:12:06.110283885+00:00 [torrust_tracker::jobs::tracker_api][INFO] Starting Torrust API server on: 127.0.0.1:55790
thread 'tracker_api::should_generate_a_new_auth_key' panicked at 'called `Result::unwrap()` on an `Err` value: reqwest::Error { kind: Request, url: Url { scheme: "http", cannot_be_a_base: false, username: "", password: None, host: Some(Ipv4(127.0.0.1)), port: Some(55790), path: "/api/key/60", query: Some("token=MyAccessToken"), fragment: None }, source: hyper::Error(Connect, ConnectError("tcp connect error", Os { code: 111, kind: ConnectionRefused, message: "Connection refused" })) }', tests/api.rs:103:14
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
test tracker_api::should_generate_a_new_auth_key ... FAILED

The client configuration is correct because it works when I make a request to a manually executed tracker in the standard 1212 port.

I think the problem could be the Tokyo library is not running a new thread for the API and when the client makes the request, the server is not responding. But I do not know if that makes sense. I suppose the server should be executed again when the client is waiting for the response, even is Tokio is using only one thread.

The other problem could be the setup. But It worked for the UDP tracker, and I do not see any error.

Do you have any idea why it's not working or how can I find out the issue? I've run out of ideas.

[mickvandijke]

hi @WarmBeer, I'm trying to add a test for the API endpoint. I'm doing the same as we did for the UDP tracker. I want to run the API and use an HTTP client to make a request to that endpoint.

For some reason I can't connect to the API:

running 1 test
2022-11-22T19:12:06.110254618+00:00 [torrust_tracker::logging][INFO] logging initialized.
2022-11-22T19:12:06.110283885+00:00 [torrust_tracker::jobs::tracker_api][INFO] Starting Torrust API server on: 127.0.0.1:55790
thread 'tracker_api::should_generate_a_new_auth_key' panicked at 'called `Result::unwrap()` on an `Err` value: reqwest::Error { kind: Request, url: Url { scheme: "http", cannot_be_a_base: false, username: "", password: None, host: Some(Ipv4(127.0.0.1)), port: Some(55790), path: "/api/key/60", query: Some("token=MyAccessToken"), fragment: None }, source: hyper::Error(Connect, ConnectError("tcp connect error", Os { code: 111, kind: ConnectionRefused, message: "Connection refused" })) }', tests/api.rs:103:14
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
test tracker_api::should_generate_a_new_auth_key ... FAILED

The client configuration is correct because it works when I make a request to a manually executed tracker in the standard 1212 port.

I think the problem could be the Tokyo library is not running a new thread for the API and when the client makes the request, the server is not responding. But I do not know if that makes sense. I suppose the server should be executed again when the client is waiting for the response, even is Tokio is using only one thread.

The other problem could be the setup. But It worked for the UDP tracker, and I do not see any error.

Do you have any idea why it's not working or how can I find out the issue? I've run out of ideas.

I will have a look.

[mickvandijke]

hi @WarmBeer, I'm trying to add a test for the API endpoint. I'm doing the same as we did for the UDP tracker. I want to run the API and use an HTTP client to make a request to that endpoint.

For some reason I can't connect to the API:

running 1 test
2022-11-22T19:12:06.110254618+00:00 [torrust_tracker::logging][INFO] logging initialized.
2022-11-22T19:12:06.110283885+00:00 [torrust_tracker::jobs::tracker_api][INFO] Starting Torrust API server on: 127.0.0.1:55790
thread 'tracker_api::should_generate_a_new_auth_key' panicked at 'called `Result::unwrap()` on an `Err` value: reqwest::Error { kind: Request, url: Url { scheme: "http", cannot_be_a_base: false, username: "", password: None, host: Some(Ipv4(127.0.0.1)), port: Some(55790), path: "/api/key/60", query: Some("token=MyAccessToken"), fragment: None }, source: hyper::Error(Connect, ConnectError("tcp connect error", Os { code: 111, kind: ConnectionRefused, message: "Connection refused" })) }', tests/api.rs:103:14
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
test tracker_api::should_generate_a_new_auth_key ... FAILED

The client configuration is correct because it works when I make a request to a manually executed tracker in the standard 1212 port.

I think the problem could be the Tokyo library is not running a new thread for the API and when the client makes the request, the server is not responding. But I do not know if that makes sense. I suppose the server should be executed again when the client is waiting for the response, even is Tokio is using only one thread.

The other problem could be the setup. But It worked for the UDP tracker, and I do not see any error.

Do you have any idea why it's not working or how can I find out the issue? I've run out of ideas.

Hey Jose,

Somehow the API server is not reachable during the test. I've tried connecting to the test API server from Postman, but I get the same error as Reqwest: "connection refused".

When running the API manually, it does work like you said. Strange..

I have to focus on an issue for the index frontend right now, but I will try again tomorrow if you haven't found a fix by then.

[josecelano]

hi @WarmBeer, I'm trying to add a test for the API endpoint. I'm doing the same as we did for the UDP tracker. I want to run the API and use an HTTP client to make a request to that endpoint.
For some reason I can't connect to the API:

running 1 test
2022-11-22T19:12:06.110254618+00:00 [torrust_tracker::logging][INFO] logging initialized.
2022-11-22T19:12:06.110283885+00:00 [torrust_tracker::jobs::tracker_api][INFO] Starting Torrust API server on: 127.0.0.1:55790
thread 'tracker_api::should_generate_a_new_auth_key' panicked at 'called `Result::unwrap()` on an `Err` value: reqwest::Error { kind: Request, url: Url { scheme: "http", cannot_be_a_base: false, username: "", password: None, host: Some(Ipv4(127.0.0.1)), port: Some(55790), path: "/api/key/60", query: Some("token=MyAccessToken"), fragment: None }, source: hyper::Error(Connect, ConnectError("tcp connect error", Os { code: 111, kind: ConnectionRefused, message: "Connection refused" })) }', tests/api.rs:103:14
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
test tracker_api::should_generate_a_new_auth_key ... FAILED

The client configuration is correct because it works when I make a request to a manually executed tracker in the standard 1212 port.
I think the problem could be the Tokyo library is not running a new thread for the API and when the client makes the request, the server is not responding. But I do not know if that makes sense. I suppose the server should be executed again when the client is waiting for the response, even is Tokio is using only one thread.
The other problem could be the setup. But It worked for the UDP tracker, and I do not see any error.
Do you have any idea why it's not working or how can I find out the issue? I've run out of ideas.

Hey Jose,

Somehow the API server is not reachable during the test. I've tried connecting to the test API server from Postman, but I get the same error as Reqwest: "connection refused".

When running the API manually, it does work like you said. Strange..

I have to focus on an issue for the index frontend right now, but I will try again tomorrow if you haven't found a fix by then.

Thank you very much @WarmBeer! I've just fixed it here. It seems we only needed to add a delay to give time to the server to be ready to accept requests. I saw @zupzup added a delay in this example here. Thanks @zupzup for sharing that repo. It's been very helpful for me.

Anyway, I do not like this solution very much because we wait for a random duration, and we make tests slower. Running the test should be a kind of callback executed when the server is ready.

[josecelano]

hi @WarmBeer @da2ce7, This is ready for review. I've added a simple e2e test. I was considering using warp test features:

https://docs.rs/warp/0.1.20/warp/test/index.html

but I do not know if it makes sense. It is similar to what other frameworks have. It allows you to test the "controller" without running the web server. In our case, we have a lot of bootstrapping, we have to setup the tracker anyway. I do not know if it is worth it since we only avoid creating the socket (and maybe the setup for the other endpoints).

On the other hand, the e2e test is less coupled with the implementation. I think @WarmBeer wanted to change the web framework. If we change it, we do not need to rewrite the tests.

[da2ce7]

ACK 92c50e6

[josecelano]

ACK 92c50e6

@da2ce7 I've rebased to include the latest changes in PR-111.

[da2ce7]

ACK 409f82a

[josecelano]

hi @WarmBeer @da2ce7, I've written an article about the problems I had testing the API endpoint:

https://github.com/Nautilus-Cyberneering/testing-in-rust/blob/main/docs/testing-apis-in-rust.md

While writing the article, I tried different options to avoid the nondeterministic "sleep". I managed to make it work with messages instead of using "sleep".