Metrics

DotVVM is instrumented with System.Diagnostics.Metrics, which should be consumable by most observability platforms (i.e. OpenTelemetry, Prometheus exporter). The metrics which we consider important and generally useful:

• dotvvm_viewmodel_size_bytes histogram - Size of the returned viewmodel JSON in bytes.
  • Labels route=RouteName and request_type=Navigate/SpaNavigate/Command/StaticCommand
• dotvvm_request_duration_seconds - Time it took to stringify the resulting JSON view model.
  • Labels route, request_type and dothtml_file
  • You can use ASP.NET Core statistics about request duration, this metric makes it possible to split the measurement by DotVVM route name.
• dotvvm_viewmodel_serialization_seconds histogram - Time it took to serialize view model to JSON objects.
  • Labels route and request_type
• dotvvm_control_lifecycle_seconds histogram - Time it took to process a request on the specific route
  • Labels route and lifecycle_type=PreInit/Init/Load/PreRender/PreRenderComplete
• dotvvm_command_invocation_seconds - Time it took to invoke a specific command method. Compared to request_duration_seconds this only includes the time spent in the command method, and is labeled by the executed binding (command)
  • Labels command and result=Ok/Exception/UnhandledException
• dotvvm_staticcommand_invocation_seconds - Similar to command_invocation_seconds, but for staticCommand invocations
  • Labels command and result=Ok/Exception/UnhandledException
• dotvvm_viewmodel_validation_errors_total histogram - Number view model validation errors returned to the client.
  • Labels route and request_type
• dotvvm_uploaded_file_bytes - Total size of user-uploaded files
• dotvvm_returned_file_bytes - Total size of returned files. Measured when the file is returned, not when downloaded by the client

If you are using Server-side viewmodel cache you might be also interested in the related metrics in order to measure if/how is the cache helping:

• dotvvm_viewmodel_cache_hit_total counter - number of cache hits
• dotvvm_viewmodel_cache_miss_total counter - number of cache misses - on a cache miss, the request fails and client has to retry uploading the entire view model.
• viewmodel_cache_loaded_bytes_total counter - number of bytes loaded from the cache, you may interpret it as an upper bound estimate on the bandwidth saved

The more niche / internal metrics are:

• dotvvm_view_compiled_total - number of views compiled (should stay constant at runtime).
• dotvvm_view_compilation_seconds_total - time spent in view compiler.
• dotvvm_property_initialized_total - number of DotVVM properties initialized (should be constant at runtime).
• dotvvm_binding_initialized_total - number of bindings instantiated.
• dotvvm_binding_compiled_total - number of binding expressions compiled (if it continues growing, it signifies problem with a missing cache).
• dotvvm_request_rejected_total - number of requests rejected (for security reasons) on the specific route.
• dotvvm_resource_serve_seconds - time it took to lookup and serve a resource.
• dotvvm_lazy_csrf_token_created_total - number of lazy CSRF tokens created.

All metrics are listed in the DotvvmMetrics.cs source code file.

Prometheus configuration

If you want to use the prometheus-net library to expose the metrics in prometheus format, we recommend calling the Prometheus.MeterAdapter.StartListening method as soon as possible (at the start of Startup/DotvvmStartup is a good place), and configuring the buckets for histograms.

MeterAdapter.StartListening(new MeterAdapterOptions {
    ResolveHistogramBuckets = instrument => {
        // prometheus-net does not know which buckets will make sense for each histogram and System.Diagnostics.Metrics API
        // does not provide a way to specify it. The ResolveHistogramBuckets function will be called for each exported histogram in to define the buckets.
        return DotvvmMetrics.TryGetRecommendedBuckets(instrument) ?? MeterAdapterOptions.DefaultHistogramBuckets;
    }
});