Bug Report

@dependabot made the following upgrades to packages (all synced from a common Directory.Build.props variable) :

• Bumps OTelExporterVersion from 1.0.0-rc9.4 to 1.0.0-rc9.5.
• Updates OpenTelemetry.Extensions.Hosting from 1.0.0-rc9.4 to 1.0.0-rc9.5
• Updates OpenTelemetry.Instrumentation.AspNetCore from 1.0.0-rc9.4 to 1.0.0-rc9.5
• Updates OpenTelemetry.Instrumentation.Http from 1.0.0-rc9.4 to 1.0.0-rc9.5
• Updates OpenTelemetry.Instrumentation.SqlClient from 1.0.0-rc9.4 to 1.0.0-rc9.5

We’re also using:

OpenTelemetry.Exporter.Jaeger 1.3.0 OpenTelemetry.Exporter.Prometheus 1.2.0-rc5

Runtime version: .NET Core & ASP.NET Core 3.1 / .NET & ASP.NET Core 6 .0

Symptom

.Configure(app =>
{
    app.UseRouting();
    app.UseOpenTelemetryPrometheusScrapingEndpoint(); // now throws a MethodAccessException under 1.0.0-rc9.5
    app.UseEndpoints(endpoints =>
    {
        endpoints.MapGet("/", async context => await context.Response.WriteAsync("hi!"));
        endpoints.MapControllers();
    });
})

Throws the following error:

MethodAccessException: Attempt by method 'Microsoft.AspNetCore.Builder.PrometheusExporterApplicationBuilderExtensions.UseOpenTelemetryPrometheusScrapingEndpoint(Microsoft.AspNetCore.Builder.IApplicationBuilder, OpenTelemetry.Metrics.MeterProvider)' to access method 'System.ServiceProviderExtensions.GetOptions<OpenTelemetry.Exporter.PrometheusExporterOptions>(System.IServiceProvider)' failed.

Rolling those other packages back to 1.0.0-rc9.4 fixes the issue.

What is the expected behavior?

The Prometheus metrics endpoint should start up without issue and /metrics served correctly.

What is the actual behavior?

A crash that causes the ASP.NET process to exit.

Reproduce

https://github.com/Aaronontheweb/OTelAspNetReproduction/blob/master/OTelRepro/Program.cs <-- running this bare minimum dotnet new webapp ASP.NET 6.0 sample will reproduce the error.