changelog

1. Install nuget package: App.Metrics.Reporting.ApplicationInsights
2. Obtain Application Insights instrumentation key.
3. Configure App.Metrics like so:

var instrumentationKey = "00000000-0000-0000-0000-000000000000";

var metrics = new MetricsBuilder()
    .Configuration.Configure(metricsOptions)
    .Report.ToApplicationInsights(instrumentationKey)
    .Build();

There are two ways how to deal with AppMetrics' items which would be referred to as dimension 1 in Application Insights.

1. Report them as part of the metric name. For example: if on the App.Metric side you would have a single metric "fruit_count" with two dimensions "apples" and "pears" than three metrics will be reported to Application Insights: "fruit_count", "fruit_count.apples" and "fruit_count.pears". This is default.
2. Report them under single name and distinguish dimension by using customDimensions; see section "AppMetrics item as Application Insights customDimension".

App.Metrics pre-aggregates metrics and reporters are responsible for publishing such aggregated data. Application Insights's type MetricTelemetry is used to describe pre-aggregated metrics and method Track(ITelemetry telemetry) of TelemetryClient publishes it.

It just boils down to translating MetricsDataValueSource into IEnumerable<MetricTelemetry> and publishing the collection using TelemetryClient.

(rather than part of the name, which is default behavior)

var metrics = new MetricsBuilder()
    .Configuration.Configure(metricsOptions)
    .Report.ToApplicationInsights(opts => {
        opts.InstrumentationKey = "00000000-0000-0000-0000-000000000000";
        opts.ItemsAsCustomDimensions = true;
        opts.DefaultCustomDimensionName = "item";
    })
    .Build();

If you would like the dimension to have more meaningful name than the default, you can add MetricTag with a name DimensionName to an AppMetrics metric and it will use it instead of a reporter-level default value.

• App.Metrics documentation
• Application Insights API for custom events and metrics

With regards to pre-aggregated data statistics (min, max, stddev) there are two types of App.Metric value sources:

1. Configurable cumulative - (counter), which are indefinately cummulative by default, but can be configured in their respective options using property ResetOnReporting to start new aggregation scope when the last one is reported.
2. Indefinately cumulative - (meter, histogram, apdex, timer), which means that they will start new aggregation scope only when IMetricsRoot is created (in practice this usually means when application restarts).

AI clearly favors the first approach.
It is what TelemetryClient does when non-pre-aggregation API (.GetMetric("mycounter").TrackValue(2)) is used and therefore aggregation strategy is under its control. I guess mostly because data statistics min, max, stddev are really only useful for describing smaller batches of uploaded data and not the whole "ever recorded" scope.

Bottom line is: when using App.Metric as a facade to Application Insights do not rely on metric properties min, max, stddev as they will contain something else than you would expect (compared to using TelemetryClient alone for example).

Reporter alone does not actively publishes the metric data. Something must periodically call its FlushAsync method.
In ASP.NET Core application this is done by MetricsReporterBackgroundService, which is a part of AppMetrics repository and nuget packages.
See Bootsrapping Startup.cs how it is registered.
Everywhere else it is up to you to implement this periodic "nudging" of the AppMetrics' reporters.

Simple example in this repository.