Currently it is a known limitation of the system.

See this: microsoft/ApplicationInsights-JS#138

AI SDK should be able to read APPINSIGHTS_INSTRUMENTATIONKEY set as environment variable on Azure Web Apps to support additional scenarios for VS and decrease the probability of conflicts between SDK from VS and AI Extension. It should have less priority then key set in the code and/or key set in AI.config.

When using OWIN with the default ApplicationInsights.config all telemetry sent from inside the OWIN pipeline seems to be filtered out (and never makes it up to AI). I am guessing it is because of the below default configuration when adding application insights to a project:

<TelemetryProcessors>
  ...
  <Add Type="Microsoft.ApplicationInsights.Web.HandlerTelemetryProcessor, Microsoft.AI.Web">
    <Handlers>
      ...
      <Add Value="System.Web.Handlers.TransferRequestHandler"/>
      ...
    </Handlers>
  </Add>
</TelemetryProcessors>

For example, the out-of-the-box request tracking (RequestTelemetry) gets filtered out, and any custom telemetry sent, too.

If you e.g. upload a non-trivial blob in Azure (I think > 4mb), the Azure SDK chunks up the data into multiple requests of 4mb each.

In AI, the first "chunk" correctly shows as a dependency within the context of the request, but the remainder have no operation id (or parent operation id) so they appear as orphaned entries within AI. They should all have the correct operation id.

Copy of https://github.com/Microsoft/ApplicationInsights-Home/issues/38 from @isaacabraham

QuickPulse (aka ICU) feature enables real-time monitoring of critical application data. SDK will ping the QuickPulse service until it responds with a request to collect, at which point real-time data starts being collected and sent out to the service every second. As soon as the data is not being viewed any more, the service will signal SDK to stop collection and return to being idle.

503 are 404 is what I tried. Suggestion is to calculate success from status code in Src/DependencyCollector/Shared/Implementation/FrameworkHttpProcessing.cs

Customer claims SQL dependency calls are taking 70 seconds and yet AI is showing them as <100ms. More details in internal email.

Repros:

Scenario1: Create a new application, add 2.0.0 application insights, delete "AI.config" file, go to nuget manager and select "Update" that updates the ai version to 2.1.0, and the config file that is created is completely empty.
Scenario2: Create a new application, add 2.0.0 appilcatoin insights, uninstall the sdk and as a result the config file gets deleted, now install 2.1.0 and the same issue can be seen.

I'm not entirely sure what I think that the solution should be for this one.

There are API calls in the .NET Azure SDK that (I believe) convert e.g. HTTP 404s to true / false e.g. blob.Exists(). The problem is that our code calls these API calls in many places and we're ending up with lots of false failures showing in AI simply because the dependency tracking module sees this as a 404 ergo it must be a failure.

We could add a filter for this but I think that you're going to find many other people in this (or similar) situation.

When I use TelemetryClient.TrackDependency Method (String, String, DateTimeOffset, TimeSpan, Boolean) the CommandName parameter value is not shown in the Metrics Explorer in the Azure Portal.

A suggested workaround is to specify the DependencyTelemetry.DependencyTypeName property to be "SQL". This will get the CommandName value to be shown in the Dependency Property blade in the Azure Portal however it clutters the analytics if the type is not SQL.

Repro Steps:

1. telemetryClient.Track("MyCustomDependency", "MyCommandName", startTime, duration, success);

Expected Result:

• In the Dependency Property blade in the Azure Portal the Command value should be "MyCommandName"

Actual Result:

• Command value is "MyCustomDependency".
  
  

More info here:
https://social.msdn.microsoft.com/Forums/en-US/ApplicationInsights/thread/f8e069c1-d72e-4c89-ba28-9cebf8571aff/

@abaranch found this code that extracts exception, but never uses it in the callback two lines below:

https://github.com/Microsoft/ApplicationInsights-dotnet-server/blob/1f9afda651266bbebd09cb900d4393b7aa704817/Src/DependencyCollector/Shared/Implementation/ProfilerSqlProcessing.cs#L186

         public object OnEndForSqlAsync(object context, object returnValue, object thisObj, object asyncResult) 
         { 
             // See implementation of EndExecuteXXX - before calling EndExecuteXXX it will check exception  
             // set by BeginXXX and throw it if it had failed 
             Exception exc = null; 
             if (asyncResult is Task) 
             { 
                 exc = ((Task)asyncResult).Exception; 
             } 


             this.OnEnd(context, null, thisObj, true); 
             return returnValue; 
         } 

for example can not collect httpclient request

if I update Microsoft.AI.Agent.Intercept 2.0.1.0
I always can not be collect dependency data
I debug in a ITelemetryInitializer I receive a exception

[msg=RemoteDependencyModule failed];[msg=System.IO.FileLoadException: Could not load file or assembly 'Microsoft.AI.Agent.Intercept, Version=1.2.1.0, Culture=neutral, PublicKeyToken=31bf3856ad364e35' or one of its dependencies. The located assembly's manifest definition does not match the assembly reference.

The test ThrowsInvalidOperationExceptionWhenTransmissionIsAlreadySending has a race condition where the first call to SendAsync can return and run the finally Interlocked.Exchange to reset isSending before the second call to SendAsync gets CPU time to try the Interlocked.CompareExchange and throw the InvalidOperationException which is causing the test to fail sporadically.
I think this can be fixed by adding a little wrapper logic in the test to force the first thread to give up CPU time until the second has run into the error condition and thrown the expected exception.

We are having issues with the Microsoft.ApplicationInsights.PerfCounterCollector package causing exceptions on our dev boxes. The first one is:

System.InvalidOperationException occurred
Message: Exception thrown: 'System.InvalidOperationException' in System.dll
Additional information: Category does not exist.

Then if we click continue in the debugger we get:

System.InvalidOperationException occurred
Message: Exception thrown: 'System.InvalidOperationException' in System.dll
Additional information: Category does not exist. 

Both of these exceptions are very annoying as they are the type of exception we don't want to have to disable for obvious reasons.

I have reproduced the problem very easily:

System Info:

OS Name Microsoft Windows 10 Pro
Version 10.0.10586 Build 10586
Visual Studio:
Microsoft Visual Studio Professional 2015
Version 14.0.25425.01 Update 3
Microsoft .NET Framework
Version 4.6.01038
Installed Version: Professional

Repro Steps:

1. Create a Web MVC (including Web Api) project
2. Via nuget install "Microsoft.ApplicationInsights.Web v2.1.0"
3. Start debugging (F5)
4. Wait for a minute or two.
5. Results -> The two exceptions mentioned earlier

Hello,

Is there a command to ignore specific requests just like new relic's ignore method?

In our project on .NET 4.5 (ASP.NET Web-application for hosting WCF-service) used references to these NuGet-packages:

• Microsoft.ApplicationInsights 1.2.3
• Microsoft.ApplicationInsights.Agent.Intercept 1.2.1
• Microsoft.ApplicationInsights.DependencyCollector 1.2.3
• Microsoft.ApplicationInsights.JavaScript 0.15.0-build58334
• Microsoft.ApplicationInsights.PerfCounterCollector 1.2.3
• Microsoft.ApplicationInsights.Web 1.2.3
• Microsoft.ApplicationInsights.WindowsServer 1.2.3
• Microsoft.ApplicationInsights.WindowsServer.TelemetryChannel 1.2.3

We have rarely occurring issue, which lead to 100% CPU load.
When it happened again, we have made the process dump and used DebugDiag to find/diagnose the problem.

Here the result of the "Analysis report":

The following threads in w3wp.DMP are enumerating a System.Collections.Generic.Dictionary object:
Multiple threads enumerating through a collection is intrinsically not a thread-safe procedure. If the dictionary object accessed by these threads is declared as static then the threads can go in an infinite loop while trying to enumerate the dictionary if one of the threads writes to the dictionary while the other threads are reading\enumerating through the same dictionary. You may also experience High CPU during this stage. For more details refer to High CPU in .NET app using a static Generic.Dictionary

All threads have a same call stack:
System.Collections.Generic.Dictionary2[[System.__Canon, mscorlib],[System.__Canon, mscorlib]].FindEntry(System.__Canon)+1a6 System.Collections.Generic.Dictionary2[[System.__Canon, mscorlib],[System.__Canon, mscorlib]].TryGetValue(System.__Canon, System.__Canon ByRef)+14
Microsoft.ApplicationInsights.Extensibility.Implementation.LocationContext.get_Ip()+4f
Microsoft.ApplicationInsights.Web.ClientIpHeaderTelemetryInitializer.UpdateRequestTelemetry(System.Web.HttpContext, Microsoft.ApplicationInsights.Extensibility.Implementation.LocationContext)+199
Microsoft.ApplicationInsights.Web.ClientIpHeaderTelemetryInitializer.OnInitializeTelemetry(System.Web.HttpContext, Microsoft.ApplicationInsights.DataContracts.RequestTelemetry, Microsoft.ApplicationInsights.Channel.ITelemetry)+86
Microsoft.ApplicationInsights.Web.Implementation.WebTelemetryInitializerBase.Initialize(Microsoft.ApplicationInsights.Channel.ITelemetry)+86
Microsoft.ApplicationInsights.TelemetryClient.Initialize(Microsoft.ApplicationInsights.Channel.ITelemetry)+26e
Microsoft.ApplicationInsights.TelemetryClient.Track(Microsoft.ApplicationInsights.Channel.ITelemetry)+28
FIE.Common.Web.Code.ApplicationInsights.ApplicationInsightsLogExceptionAttribute.System.ServiceModel.Dispatcher.IErrorHandler.HandleError(System.Exception)+9b
System.ServiceModel.Dispatcher.ErrorBehavior.HandleErrorCommon(System.Exception, System.ServiceModel.Dispatcher.ErrorHandlerFaultInfo ByRef)+a7
System.ServiceModel.Dispatcher.ImmutableDispatchRuntime.ProcessMessageCleanup(System.ServiceModel.Dispatcher.MessageRpc ByRef)+307
System.ServiceModel.Dispatcher.ImmutableDispatchRuntime.ProcessMessage9(System.ServiceModel.Dispatcher.MessageRpc ByRef)+e2
System.ServiceModel.Dispatcher.ImmutableDispatchRuntime.ProcessMessage8(System.ServiceModel.Dispatcher.MessageRpc ByRef)+ff
System.ServiceModel.Dispatcher.MessageRpc.ProcessError(System.Exception)+2c8
System.ServiceModel.Dispatcher.MessageRpc.Process(Boolean)+169
[[HelperMethodFrame]]
System.ServiceModel.Dispatcher.ImmutableDispatchRuntime.ProcessMessage5(System.ServiceModel.Dispatcher.MessageRpc ByRef)+4e3
[[HelperMethodFrame]]
System.ServiceModel.Dispatcher.DispatchOperationRuntime.InvokeBegin(System.ServiceModel.Dispatcher.MessageRpc ByRef)+764
[[HelperMethodFrame]]

Attempt to trace events from WebEventsPublisher event source leads to the failure as message is null. This issue surfaces the worst on Azure Web Sites where it crashes an application with the verbose tracing enabled.

Need to trace those events in try{}catch so they wouldn't crash an application

Currently this message is verbose: Performance counters are unavailable when the application is running under IIS Express. Use EnableIISExpressPerformanceCounters element with a value of 'true' within the Performance Collector Module element to override this behavior.
So this message will never be sent to the portal.
Proposal is to change it to error (so it is sent to the portal) with diagnostics keyword (for it is not sampled out)

We currently support such user agent strings:
Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0; KTXN) which is used for the real browser Transaction Perspective monitors.

Also we need to consider the below keynote traffic as well.
•"KHTE" (for Application Perspective monitors),
•"Keynote" for Test Perspective load testing agents.

Why is the commandtext not logged for different kinds of commands?

Regards,
Matthijs ter Woord

Add telemetry initializer that prints to console:

    class ConsolePrintTelemetryInitializer : ITelemetryInitializer
    {
        public void Initialize(ITelemetry telemetry)
        {
            var bytes = JsonSerializer.Serialize(new ITelemetry[] { telemetry }, false);
            using (var stream = new MemoryStream(bytes))
            {
                using (var streamReader = new StreamReader(stream))
                {
                    Console.WriteLine(streamReader.ReadToEnd());
                }
            }
        }
    }

    class Program
    {
        static void Main(string[] args)
        {
            GetVersionFromRegistry();
            TelemetryConfiguration.Active.TelemetryInitializers.Add(new ConsolePrintTelemetryInitializer());
            new DependencyTrackingTelemetryModule().Initialize(TelemetryConfiguration.Active);
            Test().Wait();
            Console.ReadLine();
        }

        static async Task Test()
        {
            var result = await new HttpClient().GetAsync("https://www.bing.com/search?q=httpclient");
            var response = await WebRequest.Create("https://www.bing.com/search?q=webrequest").GetResponseAsync();                 
        }
}

You will see that for every RDD call you get 2 calls to telemetry initializer. On .NET 4.5.2 it is even worse - you get 3 calls to telemetry initializer for httpclient RDD and 4 calls to telemetry initializer for webrequest RDD.

Ref: microsoft/ApplicationInsights-dotnet#268

None of my http requests are correlated with it request. It seems that maybe the async nature of it breaks the correlation.

For example, Microsoft.ApplicationInsights.DependencyCollector.

this folder is obsolete: https://github.com/Microsoft/ApplicationInsights-server-dotnet/tree/aac44ef4cf526143d71683464e1c5a956cdeb5df/Src/PerformanceCollector/Net45/PerformanceCollector/Telemetry

Web SDK uses EventSource version 1.1.28
Dependency collector depends on Microsoft.ApplicationInsights.Agent.Intercept which uses EventSource version 1.1.24
We should update Intercept so there are no unnecessary binding redirects.

When an HTTP dependency is logged from the Web SDK, result code field is empty if a client error occurred, because .NET doesn't have a response to get status code from. We should somehow indicate that a client error occurred instead of just display "no name" in the UI for these dependencies.

All of our async HTTP requests are reporting as taking microseconds to execute when they are actually taking 1-30 seconds.

var address = new Uri("http://foo.bar");
var content = new StringContent("{'foo': 'bar'");
content.Headers.ContentType = new MediaTypeWithQualityHeaderValue("application/json");
var response = await new HttpClient().PostAsync(address, content);

Non-async HTTP requests have accurate timings. This is on a server running IIS with the status monitor configured.

Copy of https://github.com/Microsoft/ApplicationInsights-Home/issues/27 from @gregjhogan. See more comments there

HttpModule should use a singleton to generate Begin and End ETW events so we will not send duplicates when misconfigured

Try this example:

 static void Main(string[] args)
        {
            new DependencyTrackingTelemetryModule().Initialize(TelemetryConfiguration.Active);
            Test().Wait();
        }

        static async Task Test()
        {
            var result = await new HttpClient().GetAsync("https://www.bing.com/search?q=httpclient");
            var response = await WebRequest.Create("https://www.bing.com/search?q=webrequest").GetResponseAsync();                 
        }

These are dependencies that are getting logged on .NET 4.5.2. Note that we get nothing logged on 4.5 and 4.5.1 (probably expected) and on 4.6 everything is logged as expected

{"name":"Microsoft.ApplicationInsights.RemoteDependency","time":"2016-04-07T19:0
3:47.1438164Z","tags":{"ai.internal.sdkVersion":"rddf: 2.0.0.25000"},"data":{"ba
seType":"RemoteDependencyData","baseData":{"ver":2,"name":"https://www.bing.com/
search?q=httpclient","id":"R1YUAcsnqk8=","value":167.5677,"dependencyKind":1,"su
ccess":false}}}

{"name":"Microsoft.ApplicationInsights.RemoteDependency","time":"2016-04-07T19:0
3:47.7255489Z","tags":{"ai.internal.sdkVersion":"rddf: 2.0.0.25000"},"data":{"ba
seType":"RemoteDependencyData","baseData":{"ver":2,"name":"https://www.bing.com/
search?q=webrequest","id":"uGxqsnv1HTU=","value":487.8151,"dependencyKind":1,"su
ccess":false}}}

Currently worker role project references Web nuget package which is incorrect. It should reference WindowsServer nuget.
https://github.com/Microsoft/ApplicationInsights-Home/tree/master/Samples/AzureEmailService

For Application Insights Availability we set User.Id as test location:

https://github.com/Microsoft/ApplicationInsights-server-dotnet/blob/bdcec347a1b9476db070dae71062ce0d524536f8/Src/Web/Web.Shared.Net/WebTestTelemetryInitializer.cs#L62

https://github.com/Microsoft/ApplicationInsights-server-dotnet/blob/bdcec347a1b9476db070dae71062ce0d524536f8/Src/Web/Web.Nuget/Resources/ApplicationInsights.config.install.xdt#L6

So we will have ALL telemetry IN or OUT from a single location.

Proposal:

1. Do not use User/Session to correlate web tests - use RootID/ParentID, etc.
2. Use test run-run-id as a RootID for sampling so we will have better sampling distribution for web test results

CC: @abaranch @vgorbenko

We should make the code coverage script paths configurable through parameters as was done in the copy in the logging solution repository.

It would be useful to be able to embed some or all of the Live Metrics Stream charts into another page, like a dashboard, so you don't have to dig into the Azure portal each time to pull up the Live Metrics Stream information. New Relic has this feature.

Due to the second argument of XPath's contains expression not being quoted in Web NuGet transformation, it appears that XDT randomly removes the first TelemetryProcessor it encounters and the xdt:Locator filter is not being applied.

https://github.com/Microsoft/ApplicationInsights-server-dotnet/blob/aac44ef4cf526143d71683464e1c5a956cdeb5df/Src/Web/Web.Nuget/Resources/ApplicationInsights.config.install.xdt lines 50-52 currently in develop.

We are getting IndexOutOfRange exception from Microsoft.ApplicationInsights.Web.ApplicationInsightsHttpModule. This happened after the our service ran for over 8 hours.

Call Stack:
[IndexOutOfRangeException: Index was outside the bounds of the array.]
System.Collections.Generic.List`1.Add(T item) +34
Microsoft.ApplicationInsights.Extensibility.Implementation.Tracing.PortalDiagnosticsQueueSender.Send(TraceEvent eventData) +11
Microsoft.ApplicationInsights.Extensibility.Implementation.Tracing.DiagnosticsListener.WriteEvent(TraceEvent eventData) +90
Microsoft.ApplicationInsights.Extensibility.Implementation.Tracing.DiagnosticsListener.OnEventWritten(EventWrittenEventArgs eventSourceEvent) +134
Microsoft.Diagnostics.Tracing.EventSource.DispatchToAllListeners(Int32 eventId, Guid* childActivityID, EventWrittenEventArgs eventCallbackArgs) +91

[EventSourceException: An error occurred when writing to a listener. ]
Microsoft.Diagnostics.Tracing.EventSource.DispatchToAllListeners(Int32 eventId, Guid* childActivityID, EventWrittenEventArgs eventCallbackArgs) +201
Microsoft.Diagnostics.Tracing.EventSource.WriteToAllListeners(Int32 eventId, Guid* childActivityID, Object[] args) +206
Microsoft.Diagnostics.Tracing.EventSource.WriteToAllListeners(Int32 eventId, Guid* childActivityID, Int32 eventDataCount, EventData* data) +277
Microsoft.Diagnostics.Tracing.EventSource.WriteEventWithRelatedActivityIdCore(Int32 eventId, Guid* relatedActivityId, Int32 eventDataCount, EventData* data) +1245
Microsoft.Diagnostics.Tracing.EventSource.WriteEvent(Int32 eventId, String arg1, String arg2) +192
Microsoft.ApplicationInsights.Web.ApplicationInsightsHttpModule..ctor() +211

[TargetInvocationException: Exception has been thrown by the target of an invocation.]
System.RuntimeTypeHandle.CreateInstance(RuntimeType type, Boolean publicOnly, Boolean noCheck, Boolean& canBeCached, RuntimeMethodHandleInternal& ctor, Boolean& bNeedSecurityCheck) +0
System.RuntimeType.CreateInstanceSlow(Boolean publicOnly, Boolean skipCheckThis, Boolean fillCache, StackCrawlMark& stackMark) +113
System.RuntimeType.CreateInstanceDefaultCtor(Boolean publicOnly, Boolean skipCheckThis, Boolean fillCache, StackCrawlMark& stackMark) +232
System.Activator.CreateInstance(Type type, Boolean nonPublic) +83
System.RuntimeType.CreateInstanceImpl(BindingFlags bindingAttr, Binder binder, Object[] args, CultureInfo culture, Object[] activationAttributes, StackCrawlMark& stackMark) +1122
System.Activator.CreateInstance(Type type, BindingFlags bindingAttr, Binder binder, Object[] args, CultureInfo culture, Object[] activationAttributes) +128
System.Activator.CreateInstance(Type type, BindingFlags bindingAttr, Binder binder, Object[] args, CultureInfo culture) +18
System.Web.HttpRuntime.CreateNonPublicInstance(Type type, Object[] args) +60
System.Web.HttpApplication.BuildIntegratedModuleCollection(List`1 moduleList) +189
System.Web.HttpApplication.InitInternal(HttpContext context, HttpApplicationState state, MethodInfo[] handlers) +792
System.Web.HttpApplicationFactory.GetNormalApplicationInstance(HttpContext context) +336
System.Web.HttpApplicationFactory.GetApplicationInstance(HttpContext context) +107
System.Web.HttpRuntime.ProcessRequestNotificationPrivate(IIS7WorkerRequest wr, HttpContext context) +327

Our service details:
ASP.NET WebAPI and MVC application
.NET Framework: 4.5.1
Running as Web App in Azure
Application Insights version: 1.2.3

The Sql Dependency details are currently limited to a couple fields:

• success
• result code

Although the result code can help with troubleshooting issues, the full error message returned by the SQL server would be more useful, eg.

result code: 2601
vs.
error message: Cannot insert duplicate key row in object 'dbo.Products' with unique index 'UX_ProductSKU'. The duplicate key value is (Selenium SKU Min). The statement has been terminated.

Can the error message be added to the SQL dependency detail?

BeginExecuteNonQueryInternal expects 4 arguments. Actual signature:

// System.Data.SqlClient.SqlCommand
private IAsyncResult BeginExecuteNonQueryInternal(CommandBehavior behavior, AsyncCallback callback, object stateObject, int timeout, bool inRetry, bool asyncWrite = false)

Same for some other BeginFoo methods

When my Web API hits an exception and is about to return an HTTP 500, I'd like the response content to include the Session ID and Operation ID so users can report more actionable bugs. I read that AppInsights was setting that info in the OWIN environment, but I can't find it. Would you recommendations about how to retrieve those IDs?

Document DB has a REST API: https://msdn.microsoft.com/en-us/library/azure/mt622086.aspx that should be recognized and not sent as a regular HTTP

Currently all requests to QuickPulse service are blocking. Consider making requests to QuickPulse service async.

Issue reported by error telemetry:
weekly unique ikey count: 43
SDKs: 2.0.0.30671 | 2.1.0.1924 | dotnet: 2.1.0.14852 | w_2.0.0.30671

Error message:
AI (Internal): [msg=HanderFailure];[exception=System.NullReferenceException: Object reference not set to an instance of an object.
at Microsoft.ApplicationInsights.Web.RequestTrackingTelemetryModule.NeedProcessRequest(HttpContext httpContext)
at Microsoft.ApplicationInsights.Web.RequestTrackingTelemetryModule.OnEndRequest(EventWrittenEventArgs args)
at Microsoft.ApplicationInsights.Web.Implementation.WebEventsSubscriber.OnEventWritten(EventWrittenEventArgs eventData)];

If response or request is chunked duration reflects time only from beginning of the request till first chunk received/posted.
Related issue: #58

Repro:
Create web app that is doing:

protected void Page_Load(object sender, EventArgs e)
        {
            HttpContext.Current.Response.BufferOutput = false;
            HttpContext.Current.Response.Write("Chunck 1");
            HttpContext.Current.Response.Flush();

            Thread.Sleep(10000);

            HttpContext.Current.Response.Write("Chunck 2");
}

Create an app that does a request to that app like this:

var address = new Uri("http://localhost/Component2/Default);
var response = await new HttpClient().GetAsync(address);

Tracked dependency duration will be much less than 10 seconds.

It is not related to having StatusMonitor but rather to instrumenting GetResponse that fires when headers are sent. We need to instrument some other functions that read response and sum together all the durations.

For some people with few users sampling by operations is more appropriate.
Add a way to configure what is used by default.

Not sure what the logic is, or if async has anything to do with it, but (recently, I think) SQL dependencies have gone missing. They are indeed recorded as shown in the screenshot, but not linked to the request.

The ones marked with red arrows are missing.

I can't seem to find anything that correlates this data in the UI. Let me know if I can check somewhere else.

Edit: It seems in the 3 items marked, there is a bunch of missing info, eg Operation Id, USer Id, Session Id, etc.