[BUG] No auto dependency tracking when application insights is initialized and running on web worker

[riteshbhoi]

• SDK Version [e.g. 22]: 2.5.7
• How you initialized the SDK: In the web worker role

So have offloaded the app insights initialization and logging to a web worker. And we invoke the required methods from web worker to log to application insights. But it seems as the dependency calls are taking place on main thread and worker is unaware of this, dependency calls are not getting tracker. Is there any resolution to this issue?

[MSNev]

Correct, a web worker is a self contained javascript instance and do not share the same XMLHttpRequest / fetch references.

If you want to track dependency calls from both the main thread and any worker(s) then all instances would need to be instrumented.

In your case as you are offloading the initialization into a worker, you would need to wrap and track the dependency calls (same as the dependency plugin) and then delegate the request (trackDependencyData) call to your worker.

[riteshbhoi]

If i just delegate the XMLHttpRequest/Fetch calls to my worker and my worker makes the call (not sure if that is the right pattern), won't the dependency calls be tracked automatically, as worker has application insights initialized? Do i need to wrap and track the dependency calls separately as such?

[MSNev]

If i just delegate the XMLHttpRequest/Fetch calls to my worker and my worker makes the call (not sure if that is the right pattern)

You could, but I wouldn't as you would incur additional marshalling costs for every request.

won't the dependency calls be tracked automatically

Yes, they should as long as you have also enabled fetch tracking (currently disabled by default), but as above I don't believe this would be the best approach -- it does of course depend on your application and performance requirements etc.

Do i need to wrap and track the dependency calls separately as such?

There are really 2 general approaches (or 3 based on your delegation pattern), so for the 2

1. Wrapping: Yes, you would need to wrap the request and once you have a response (same as the dependency code) you could then delegate the "track" call to your worker. And you would have to wrap the XHR and fetch yourself -- potentially taking the dependency extension as a guide on how you would do this
2. Initialize on page: As mentioned above, just initialize an instance of the AI SDK in the main (page/document) of your application and this will automatically wrap etc. But this of course depends on why you have created the delegation pattern in the first place and whether this would work, it may also cause some event duplication depending on how your measuring the automatic events such as page load / performance -- so you would probably need to have a telemetry initializer so you can drop the events on the page or worker that you don't want to have reported twice.

[MSNev]

Closing as this doesn't seem to an issue that we can address directly in the SDK (as a bug/enhancement)

[github-actions]

This issue has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.