Hard to see what `WaitFor` is really waiting on

[afscrome]

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.

With the WaitFor implementation in main, it's hard to see what a resource is actually waiting on. Logs are logged when a resource starts waiting on a resource, but nothing gets logged once that wait is complete.

In the following example, it looks like we're waiting for both sql and migrator, however we're actually only waiting on migrator as sql has started up.

Similarly with a longer chain of dependencies, You may see that you're waiting on B, but B is in turn waitign on C (which could in turn be waiting on D).

Describe the solution you'd like

A quick solution is to include a log entry when a wait is complete

2024-09-05T19:16:18 Waiting for resource 'sql' to enter the 'Running' state.
2024-09-05T19:16:19 Waiting for resource 'migrator' to complete.
2024-09-05T19:16:20 Finished waiting for resource 'sql' to enter the 'Running' state.

That would at least allow you to work out what is outstanding.

A slightly better solution would be if there was an easy way to see what was outstanding from a single place, without having to calculate across multiple log entries. Without adding additional UI, there's limited options with the append only logs - you could add a log along the lines of Waiting on 2/3 dependencies - 'migrator' and 'foo' remain log each time a dependency is removed. Even better is if this could include transitive dependencies, but that likely gets messy quickly.

The ideal would be some kind of UI on the resource that could present the dependencies as some kind of tree which you could follow to see how far the dependencies have gotten.

Additional context

No response

[davidfowl]

cc @mitchdenny

[mitchdenny]

We could add some extra log messages here to make this clearer. Also related to #5558

[mitchdenny]

I'll work on this after #5515 merges in.

[afscrome]

Related - when health checks fail, a failure message shows up in the terminal logs, but nothing shows up within the aspire console. Would be nice if these health check failures could be shown in the console logs for resources waiting on them.

[davidfowl]

We can't clutter the console logs for a resources with health check logs (those logs will go away in the next PR), there would be multiple things writing to the output stream. We need another log stream to show health checks logs.

[afscrome]

I was specifically thinking about including health check failures in the console logs during the Waiting phase to provide indication of the WaitFor progress. Although you probably don't need the full exception to indicate this. e.g. in the above example, the logs for the waiting resource could show something like

INFO: Resource test healthcheck reported unhealthy. Will try again in x seconds

I do agree that including them after the resource has fully started will be spammy & confusing.

[mitchdenny]

Should be fixed by #5842 which adds a WaitAnnotation and makes log entries clearer about what is blocking startup of a particular resource.