Perform UnitTest with real application insights too long

[guenogu]

Hello,

I'm working on a Hapi server application.
I try to integrate the package applicationinsights to my application.

My application is cover by unit test (mocha), but when I use a real instrumentation key the unit test execution is quite long ( >60s) and when the instrumentation key is a fake ikey then the execution is more performant (<3s).

To help you to reproduce and understand my issue, I have push the code in this branch
https://github.com/guenogu/repro/tree/hapi-wreck-appinsights

Thanks for your help.

[OsvaldoRosado]

This SDK is optimized for use in long-running server scenarios, rather than quick termination. One such optimization is that it will wait for 15 seconds before sending telemetry to optimize network traffic by batching all of the telemetry into a single request. There are other optimizations in place similar to this - and I'm guessing that these are what impact its use within unit tests.

Some things to try, in no particular order (you may need all of them, or only one):

1. .setUseDiskRetryCaching(false)
2. appInsights.defaultClient.config.maxBatchIntervalMs = 0
3. .setAutoCollectPerformance(false)
4. appInsights.defaultClient.flush() (at the end of your test sequence)

If your tests don't actually need to emit telemetry, you can also use appInsights.defaultClient.config.disableAppInsights = true;

[markwolff]

@guenogu which version of mocha are you using?
Edit: looked at your repo, you are using 5.2

[markwolff]

@guenogu You can fix this for now by adding --exit to your mocha command, as per links below. We should investigate why our tests do not terminate.
mochajs/mocha#2879
https://boneskull.com/mocha-v4-nears-release/#mochawontforceexit