Starting a resource (WithExplicitStart) in Aspire tests

[m03chv13h]

Is there an existing issue for this?

• I have searched the existing issues

Is your feature request related to a problem? Please describe the problem.

Maybe there is already a solution for that but I cannot find it.

I want to start my resource via the code in the tests which I added with .WithExplicitStart() in my AppHost Program.cs file.

Program.cs

// ...
builder.AddExecutable("explicit-start-resource-name", ...)
    .WithExplicitStart();
// ...

In my tests I start the app host and want to start this resource manually via the code.

AppHostTests.cs

public class AppHostTests
{
    private static readonly TimeSpan BuildStopTimeout = TimeSpan.FromSeconds(15);
    private static readonly TimeSpan StartStopTimeout = TimeSpan.FromSeconds(120);

    [Fact]
    public async Task Test_ExplicitStartResource()
    {
        // Act
        var appHostBuilder = await DistributedApplicationTestingBuilder.CreateAsync<Projects.AppHost>();
        
        await using var appHost = await appHostBuilder.BuildAsync().WaitAsync(BuildStopTimeout);
        
        await appHost.StartAsync().WaitAsync(StartStopTimeout);
        
        // Some code here.

        // Arrange
        // Now I want to start the resource with a specific name.
        // For example: await appHost.StartResourceAsync(name: "explicit-start-resource-name");

        // Assert
    }
}

How can I do that? Is there already a possibility for example with the ResourceNotificationService?

Describe the solution you'd like

For example:

await appHost.StartResourceAsync(name: "explicit-start-resource-name");

Additional context

Aspire.Hosting.AppHost 9.1.0
Aspire.Hosting.Testing 9.1.0

[functario]

I have the same need. My application is a Kubernetes orchestrator that scale the number of pods or "Runner" services based on a request. I have implemented a Kubernetes client injected during the Aspire tests to simulate kube interactions. But I actually need to start 3 Runner services before to send the request and have multiple stages in my fake client to simulate scaling from 0 to multiple pods matching the request.

The fake client is a NSubstitute mock injected via ServiceDefaults. So I need to configure everything before the test run and I have no way to change the behavior during the test since the mock is in the context of the Orchestrator service and not the one of the test.

It makes the tests more complicated since I need to synchronize the number of requests by the Orchestrator to the fake Kubernetes client responses (simple polling).

I could replace NSubstitute by a WebService to control the fake client during the test, but it will still require synchronizing the calls.

Being able to control the start from the fake client will allow me to respond with the real current state of the Runner service.

[davidfowl]

Yep we need to expose an api. ResourceCommandService @JamesNK