Library for collecting application metrics in .Net and exporting them to Prometheus
Nexogen.Libraries.Metrics API Reference
The default Prometheus metrics registration behavior in ASP.NET Core applications has changed. The request metrics collection is no longer enabled by default. This means that you need to add code to enable it. See the relevant section below for code samples. Explicitly defined metrics are not affected in any way.
dotnet add package Nexogen.Libraries.Metrics.Prometheus
dotnet add package Nexogen.Libraries.Metrics.Extensions
You can use an interface only nuget when writing libraries or when you want to use Metrics through dependency injection.
dotnet add package Nexogen.Libraries.Metrics
For exporting metrics you currently have to use ASP.NET Core.
dotnet add package Nexogen.Libraries.Metrics.Prometheus.AspCore
Or you can use a push gateway when measuring batch processes.
dotnet add package Nexogen.Libraries.Metrics.Prometheus.PushGateway
Counters can only increment, so they are most useful for counting things, like calls to API endpoints or backend services.
IMetrics metrics = new PrometheusMetrics();
ICounter counter = metrics.Counter()
.Name("nexogen_sort_calls_total")
.Help("Total calls to sort routine.")
.Register();
counter.Increment();
Gauges can take any value, so they are the most versatile metric type available. You can even measure durations or dates with them!
IGauge gauge = metrics.Gauge()
.Name("nexogen_sorted_items_count_last")
.Help("The last count of the sorted items.")
.Register();
gauge.Value = items.Length;
gauge.Increment();
gauge.Decrement(10.1);
Histograms are a trade off between measuring resolution and precision. With histograms you can avoid aliasing errors from Prometheus's scrape interval, but lose granularity. Histograms also need to have their buckets defined before use and we provide sevaral bucket generators to make it easy.
IHistogram histogram = metrics.Histogram()
.LinearBuckets(0.01, 0.01, 100)
.Name("nexogen_sort_time_seconds")
.Help("Time taken for sort in seconds.")
.Register();
var sw = Stopwatch.StartNew();
Array.Sort(items);
histogram.Observe(sw.Elapsed.TotalSeconds);
We provide an Extensions library for making common measurements easy.
using (histogram.Timer())
{
Array.Sort(items);
}
gauge.SetToCurrentTime();
gauge.TrackInProgress(() => Array.Sort(items));
There is a way to automatically collect metrics about the ASP.NET Core requests. This needs to be enabled when registering to the application.
public void Configure(IApplicationBuilder app, IHostingEnvironment env)
{
app.UsePrometheus(options => options.CollectHttpMetrics());
}
The options.CollectHttpMetrics()
registers automatic request metric collection, which was implicitly registered in versions prior to version 3.0.0.
The collected metrics can be exposed via ASP.NET Core middleware. To do this you need to register the feature during configuration.
public void Configure(IApplicationBuilder app, IHostingEnvironment env)
{
app.UsePrometheus();
}
There is a standalone server if you don't want to use ASP.NET Core just to expose your metrics.
dotnet add package Nexogen.Libraries.Metrics.Prometheus.Standalone
public void ConfigureServices(IServiceCollection services)
{
services.AddPrometheusStandalone(Configuration.GetSection("Prometheus"));
}
or
var metrics = new PrometheusMetrics();
await new PrometheusServer(new PrometheusServerOptions {Port = 9100}, metrics, loggerFactory).StartAsync();
We provide gRPC interceptors for capturing metrics.
dotnet add package Nexogen.Libraries.Metrics.Prometheus.Grpc
For gRPC servers:
services.AddSingleton<IGrpcServerMetrics, GrpcServerMetrics>()
.AddGrpc(options => options.Interceptors.Add<ServerMetricsInterceptor>());
For gRPC clients:
services.AddSingleton<IGrpcClientMetrics, GrpcClientMetrics>()
.AddGrpcClient<T>((provider, options) => options.Interceptors.Add(new ClientMetricsInterceptor(provider.GetRequiredService<IGrpcClientMetrics>())));