Hi guys,

I've been struggling this for a while, and I can almost completely confirm it, as I've no managed to reproduce the issue without Application Insights - it seems that over time, AI (most likely it's the profiler) locks up out web application's sockets/threads over time, causing non-sense like this to appear:

StackExchange.Redis.RedisConnectionException
No connection is available to service this operation: GET chapters:228901; It was not possible to connect to the redis server(s); ConnectTimeout; IOCP: (Busy=3,Free=997,Min=4,Max=1000), WORKER: (Busy=75,Free=948,Min=4,Max=1023), Local-CPU: n/a

Along with Entity Framework connection timeouts, which usually cause the request itself to time out.

Restarting the application usually mitigates this, but then it comes back. I've managed to reproduce this both in ASP.NET Core and ASP.NET MVC 5.

Reproducing this artificially is very hard, since we have a moderately large website with a few millions of pageviews per day, and the error surfaces anywhere from 24 hours to a week.

Based on my experiments, the NuGet package isn't actually required for reproduction, only the Azure Extension.

Version Info

SDK Version : v2.3.0
.NET Version : Core 2.1.1.0
How Application was onboarded with SDK(VisualStudio/StatusMonitor/Azure Extension) :Azure Extension, NuGet
OS : Windows Server
Hosting Info (IIS/Azure WebApps/ etc) : Azure WebApp