NAME Net::Prometheus - export monitoring metrics for prometheus SYNOPSIS use Net::Prometheus; my $client = Net::Prometheus->new; my $counter = $client->new_counter( name => "requests", help => "Number of received requests", ); sub handle_request { $counter->inc; ... } use Plack::Builder; builder { mount "/metrics" => $client->psgi_app; ... } DESCRIPTION This module provides the ability for a program to collect monitoring metrics and export them to the prometheus.io monitoring server. As prometheus will expect to collect the metrics by making an HTTP request, facilities are provided to yield a PSGI application that the containing program can embed in its own structure to provide the results, or the application can generate a plain-text result directly and serve them by its own means. Metrics::Any For more flexibility of metrics reporting, other modules may wish to use Metrics::Any as an abstraction interface instead of directly using this API. By using Metrics::Any instead, the module does not directly depend on Net::Prometheus, and in addition program ultimately using the module gets the flexibility to use Prometheus (via Metrics::Any::Adapter::Prometheus) or use another reporting system via a different adapter. CONSTRUCTOR new $prometheus = Net::Prometheus->new; Returns a new Net::Prometheus instance. Takes the following named arguments: disable_process_collector => BOOL If present and true, this instance will not load the default process collector from Net::Prometheus::ProcessCollector. If absent or false, such a collector will be loaded by default. disable_perl_collector => BOOL If present and true, this instance will not load perl-specific collector from Net::Prometheus::PerlCollector. If absent or false this collector is loaded by default. These two options are provided for testing purposes, or for specific use-cases where such features are not required. Usually it's best just to leave these enabled. METHODS register $collector = $prometheus->register( $collector ); Registers a new collector to be collected from by the render method. The collector instance itself is returned, for convenience. unregister $prometheus->unregister( $collector ); Removes a previously-registered collector. new_gauge $gauge = $prometheus->new_gauge( %args ); Constructs a new Net::Prometheus::Gauge using the arguments given and registers it with the exporter. The newly-constructed gauge is returned. new_counter $counter = $prometheus->new_counter( %args ); Constructs a new Net::Prometheus::Counter using the arguments given and registers it with the exporter. The newly-constructed counter is returned. new_summary $summary = $prometheus->new_summary( %args ); Constructs a new Net::Prometheus::Summary using the arguments given and registers it with the exporter. The newly-constructed summary is returned. new_histogram $histogram = $prometheus->new_histogram( %args ); Constructs a new Net::Prometheus::Histogram using the arguments given and registers it with the exporter. The newly-constructed histogram is returned. new_metricgroup $group = $prometheus->new_metricgroup( %args ); Returns a new Metric Group instance as a convenience for registering multiple metrics using the same namespace and subsystem arguments. Takes the following named arguments: namespace => STR subsystem => STR String values to pass by default into new metrics the group will construct. Once constructed, the group acts as a proxy to the other new_* methods, passing in these values as overrides. $gauge = $group->new_gauge( ... ); $counter = $group->new_counter( ... ); $summary = $group->new_summary( ... ); $histogram = $group->new_histogram( ... ); collect @metricsamples = $prometheus->collect( $opts ); Returns a list of "MetricSamples" in Net::Prometheus::Types obtained from all of the currently-registered collectors. render $str = $prometheus->render; Returns a string in the Prometheus text exposition format containing the current values of all the registered metrics. $str = $prometheus->render( { options => "for collectors" } ); An optional HASH reference may be provided; if so it will be passed into the collect method of every registered collector. Values that are set to undef will be absent from the output (this usually applies to gauges). Values set to NaN will be rendered as NaN. handle $response = $prometheus->handle( $request ); Given an HTTP request in an HTTP::Request instance, renders the metrics in response to it and returns an HTTP::Response instance. This application will respond to any GET request, and reject requests for any other method. If a query string is present on the URI it will be parsed for collector options to pass into the "render" method. This method is useful for integrating metrics into an existing HTTP server application which uses these objects. For example: my $prometheus = Net::Prometheus->new; sub serve_request { my ( $request ) = @_; if( $request->uri->path eq "/metrics" ) { return $prometheus->handle( $request ); } ... } psgi_app $app = $prometheus->psgi_app; Returns a new PSGI application as a CODE reference. This application will render the metrics in the Prometheus text exposition format, suitable for scraping by the Prometheus collector. This application will respond to any GET request, and reject requests for any other method. If a QUERY_STRING is present in the environment it will be parsed for collector options to pass into the "render" method. This method is useful for integrating metrics into an existing HTTP server application which is uses or is based on PSGI. For example: use Plack::Builder; my $prometheus = Net::Prometheus::->new; builder { mount "/metrics" => $prometheus->psgi_app; ... } export_to_Future_IO $f = $prometheus->export_to_Future_IO( %args ); Performs the necessary steps to create a minimal HTTP server for exporting metrics over HTTP, by using Future::IO directly. This requires Future::IO version 0.11 or above, and a containing process that has already loaded a non-default loop implementation that supports multiple filehandles. This new server will listen on its own port number for any incoming request, and will serve metrics regardless of path. This server is a very small, minimal implementation just sufficient to support prometheus itself, or simple tools like wget, curl or perhaps a web-browser for manual inspection. It is not intended to be a fully-featured HTTP server and certainly does not support many HTTP features at all. Takes the following named arguments: port => INT Port number on which to listen for incoming HTTP requests. The returned Future instance will remain pending for the entire lifetime of the process. If the containing program has nothing else to do it can call the await method on it, or else combine it with other toplevel event futures it is using for its own purposes. export_to_IO_Async $prometheus->export_to_IO_Async( $loop, %args ); Performs the necessary steps to create an HTTP server for exporting metrics over HTTP via IO::Async. This will involve creating a new Net::Async::HTTP::Server instance added to the loop. This new server will listen on its own port number for any incoming request, and will serve metrics regardless of path. Note this should only be used in applications that don't otherwise have an HTTP server, such as self-contained monitoring exporters or exporting metrics as a side-effect of other activity. For existing HTTP server applications it is better to integrate with the existing request/response processing of the application, such as by using the "handle" or "psgi_app" methods. Takes the following named arguments: port => INT Port number on which to listen for incoming HTTP requests. COLLECTORS The toplevel Net::Prometheus object stores a list of "collector" instances, which are used to generate the values that will be made visible via the "render" method. A collector can be any object instance that has a method called collect, which when invoked is passed no arguments and expected to return a list of "MetricSamples" in Net::Prometheus::Types structures. @metricsamples = $collector->collect( $opts ) The Net::Prometheus::Metric class is already a valid collector (and hence, so too are the individual metric type subclasses). This interface allows the creation of new custom collector objects, that more directly collect information to be exported. Collectors might choose to behave differently in the presence of some specifically-named option; typically to provide extra detail not normally provided (maybe at the expense of extra processing time to calculate it). Collectors must not complain about the presence of unrecognised options; the hash is shared among all potential collectors. TODO * Histogram/Summary 'start_timer' support * Add other export_to_* methods for other event systems and HTTP-serving frameworks, e.g. Mojo. AUTHOR Paul Evans