Get information from running Aspire Hosts

[vandenbergjp]

I want to use Aspire together with my integration and end-to-end tests. When I use the DistributedApplicationTestingBuilder, a new Aspire host is started each time I run the tests. This is useful because it ensures the tests run in a fresh environment. However, it also slows things down, since the Aspire host takes time to start—annoying when running tests frequently.

To speed things up, I currently start the Aspire environment manually, keep it running, and run my tests against that instance. The downside is that I then need to configure static endpoints and connection strings and hard code those in my tests. My goal, however, is to keep things “Aspirefied.”

So here’s my question: is it possible to detect if an Aspire host is already running and 'connect' to it in order to retrieve resource information (like connection strings and endpoints)?

Ideally, I’d like to continue using the DistributedApplicationTestingBuilder, but have the option to either connect to an existing instance or spin up a new one—something similar to how you can configure containers with .WithLifetime(ContainerLifetime.Persistent).

[engineering87]

Hello @vandenbergjp,
doesn’t look like there’s a built-in way to do that. Today, DistributedApplicationTestingBuilder always spins up its own Aspire host and you can’t attach to an already running AppHost to read endpoints/connection strings.
Try to start once per test run and wrap the builder in a test fixture and share it across tests.

[vandenbergjp]

Thanks! Sharing across tests is a good tip. In my case, I’m already doing that.

What I’d like to do instead is start Aspire once, keep it running, and then execute my tests against that instance. The part I’m struggling with is accessing the configuration that Aspire sets up within my tests—for example, database connection strings for data manipulation and the endpoints of the running web applications.

[davidfowl]

No there's nothing built in for this, and it would be a completely different model because the resource model you get lets you manipulate the resource state in process. What you are proposing cannot do that.