SignalR in MVC skewing Application Insights

Question

We just started using SignalR in an MVC application and now we're getting a bunch of alerts due to high average response time. I suspect this to be misleading as the application isn't experiencing performance degradation. It appears that SignalR uses this URL to make a connection. This url not a controller/action of the project and just the built in SignalR code in the js file. jquery.signalR-2.2.1.js is the file. I suspect that it is just leaving the websocket connection open while they are on this page and it's skewing our numbers. Is this accurate? If so, is there a way to filter it out of the application insights?

Here is the counter. Is this the expected behavior?

Here is the signalR jquery code where it builds it's url:

// BUG #2953: The url needs to be same otherwise it will cause a memory leak
    getUrl: function (connection, transport, reconnecting, poll, ajaxPost) {
        /// <summary>Gets the url for making a GET based connect request</summary>
        var baseUrl = transport === "webSockets" ? "" : connection.baseUrl,
            url = baseUrl + connection.appRelativeUrl,
            qs = "transport=" + transport;

        if (!ajaxPost && connection.groupsToken) {
            qs += "&groupsToken=" + window.encodeURIComponent(connection.groupsToken);
        }

        if (!reconnecting) {
            url += "/connect";
        } else {
            if (poll) {
                // longPolling transport specific
                url += "/poll";
            } else {
                url += "/reconnect";
            }

            if (!ajaxPost && connection.messageId) {
                qs += "&messageId=" + window.encodeURIComponent(connection.messageId);
            }
        }
        url += "?" + qs;
        url = transportLogic.prepareQueryString(connection, url);

        if (!ajaxPost) {
            url += "&tid=" + Math.floor(Math.random() * 11);
        }

        return url;
    },

Answer 1

I fixed this by following the instructions on https://docs.microsoft.com/en-us/azure/application-insights/app-insights-api-filtering-sampling:

1. Update your ApplicationInsights Nuget package to 2.0.0 or later.
2. Create a class implementing ITelemetryProcessor:

    public class UnwantedTelemetryFilter : ITelemetryProcessor
    {
        private ITelemetryProcessor Next { get; set; }

        public UnwantedTelemetryFilter(ITelemetryProcessor next)
        {
            this.Next = next;
        }

        public void Process(ITelemetry item)
        {
            var request = item as RequestTelemetry;

            if (request != null && request.Name != null)
                if (request.Name.Contains("signalr"))
                    return;

            // Send everything else:
            this.Next.Process(item);
        }
    }

3. Add the processor to your Application_Start() in Global.asax.cs:

    var builder = TelemetryConfiguration.Active.TelemetryProcessorChainBuilder;
    builder.Use((next) => new UnwantedTelemetryFilter(next));
    builder.Build();

Answer 2

if the calls are coming from the C# part of the app, the easiest way is to write a custom telemetry processor: https://docs.microsoft.com/en-us/azure/application-insights/app-insights-api-filtering-sampling

public void Process(ITelemetry item)
{
    var request = item as RequestTelemetry;

    if (request != null && request.[some field here].Equals("[some signalr specific check here]", StringComparison.OrdinalIgnoreCase))
    {
        // To filter out an item, just terminate the chain:
        return;
    }
    // Send everything else:
    this.Next.Process(item);
}

and use that to explicitly filter out the signalr calls from being sent

or if the calls are coming from JS, then the telemetry initializer there does a similar thing to filter out telemetry if you return false in the initializer.