NAME

Mojolicious::Plugin::Prometheus - Mojolicious Plugin

SYNOPSIS

# Mojolicious, no extra options
$self->plugin('Prometheus');

# Mojolicious::Lite, no extra options
plugin 'Prometheus';

# Mojolicious::Lite, with custom response buckets and metrics pr endpoint
plugin 'Prometheus' => {
  http_requests_total => {
    buckets => [qw/4 5 6/],
    labels  => [qw/worker method endpoint code/],
    cb      => sub {
      my $c = shift;
      my $endpoint = $c->match->endpoint ? $c->match->endpoint->to_string : undef;
      return ($$, $c->req->method, $endpoint || $c->req->url, $c->res->code);
    },
  },
};

# You can add your own route to do access control
my $under = app->routes->under('/secret' =>sub {
  my $c = shift;
  return 1 if $c->req->url->to_abs->userinfo eq 'Bender:rocks';
  $c->res->headers->www_authenticate('Basic');
  $c->render(text => 'Authentication required!', status => 401);
  return undef;
});
plugin Prometheus => {route => $under};

DESCRIPTION

Mojolicious::Plugin::Prometheus is a Mojolicious plugin that exports Prometheus metrics from Mojolicious.

Hooks are also installed to measure requests response time and count requests based on method and HTTP return code.

HELPERS

prometheus.instance

Create further instrumentation into your application by using this helper which gives access to the Net::Prometheus object. Please note that this represents a change from versions up to and including 1.4.x where prometheus was the method that returned the Prometheus instance.

See Net::Prometheus for further usage information.

prometheus.register($collector, $scope)

Register new / custom collectors. This method lets you differentiate those collectors that collect worker-specific metrics and those that collect more generic / global metrics. The difference between these two categories of collectors is that the results of calling collect on worker-specific collectors is cached to shared memory using IPC::ShareLite so that one worker can return collected results for all workers. Results of calling collect on generic / global collectors are not cached. This separation ensures that collectors that are global is not duplicated in the output, ie. does not appear once for each worker.

METHODS

Mojolicious::Plugin::Prometheus inherits all methods from Mojolicious::Plugin and implements the following new ones.

register

$plugin->register($app, \%config);

Register plugin in Mojolicious application. You may pass a hashref containing configuration overrides to register this will be merged with the default configuration values.

%config can have:

  • route

    Mojolicious::Routes::Route object to attach the metrics to, defaults to generating a new one for '/'.

    Default: /

  • path

    The path to mount the exporter.

    Default: /metrics

  • prometheus

    Override the Net::Prometheus object. The default is a new singleton instance of Net::Prometheus.

  • namespace, subsystem

    These will be prefixed to the metrics exported.

  • shm_key

    Key used for shared memory access between workers, see $key in IPC::ShareLite for details. Default is the process id read from $$.

  • http_request_duration_seconds

    Structure that overrides the configuration for the http_request_duration_seconds metric. See below.

  • http_request_size_bytes

    Structure that overrides the configuration for the http_request_size_bytes metric. See below.

  • http_response_size_bytes

    Structure that overrides the configuration for the http_response_size_bytes metric. See below.

  • http_requests_total

    Structure that overrides the configuration for the http_requests_total metric. See below.

  • perl_collector

    Structure that tells the plugin to enable or disable a Perl collector. Previously the Perl collector from Net::Prometheus was used, but that is no longer in use due to it not being possible to add dynamic label values. Now Mojolicious::Plugin::Prometheus::Collector::Perl is used. The configuration here need as follows:

    enabled

    Boolean-ish value indicating if this collector should be used.

    labels_cb

    A subref that the collector can call to dynamically resolve which labels and corresponding label values should be added to each metric. Default is:

    {
      enabled   => 1,
      labels_cb => sub { { worker => $$ } },
    }
  • process_collector

    Structure that tells the plugin to enable or disable a process collector. The process collector from Net::Prometheus is used for this. The configuration here need as follows:

    enabled

    Boolean-ish value indicating if this collector should be used.

    labels_cb

    A subref that the collector can call to dynamically resolve which labels and corresponding label values should be added to each metric. Default is:

    {
      enabled   => 1,
      labels_cb => sub { { worker => $$ } },
    }

METRICS

In addition to exposing the default process metrics that Net::Prometheus already expose this plugin will also expose

  • http_requests_total, request counter partitioned over HTTP method and HTTP response code

  • http_request_duration_seconds, request duration histogram partitioned over HTTP method

  • http_request_size_bytes, request size histogram partitioned over HTTP method

  • http_response_size_bytes, response size histogram partitioned over HTTP method

Custom configuration of the built in metrics is possible. An example structure can be seen in the synopsis. The four built in metrics from this plugin all have more or less the same structure. Metrics provided by the Perl- and Process-collectors from Net::Prometheus can be used and changed by providing a custom prometheus object instead of using the defaults as detailed previously.

Default configuration for the built in metrics are as follows:

http_request_duration_seconds => {
  buckets => [.005, .01, .025, .05, .075, .1, .25, .5, .75, 1.0, 2.5, 5.0, 7.5, 10],
  labels  => [qw/worker method/],
  cb      =>  sub($c) { $$, $c->req->method, tv_interval($c->stash('prometheus.start_time')) },
}

http_request_size_bytes => {
  buckets => [1, 50, 100, 1_000, 10_000, 50_000, 100_000, 500_000, 1_000_000],
  labels  => [qw/worker method/],
  cb      => sub($c) { $$, $c->req->method, $c->req->content->body_size },
}

http_response_size_bytes => {
  buckets => [5, 50, 100, 1_000, 10_000, 50_000, 100_000, 500_000, 1_000_000],
  labels  => [qw/worker method code/],
  cb      => sub($c) { $$, $c->req->method, $c->res->code, $c->res->content->body_size },
}

http_requests_total => {
  labels  => [qw/worker method code/],
  cb      => sub($c) { $$, $c->req->method, $c->res->code },
}

AUTHOR

  • Vidar Tyldum / TYLDUM - Author

  • Christopher Rasch-Olsen Raa / CRORAA - Co-maintainer

(the IPC::ShareLite parts of this code is shamelessly stolen from Mojolicious::Plugin::Status written by Sebastian Riedel and mangled into something that works for me)

COPYRIGHT AND LICENSE

Copyright (C) 2018, Vidar Tyldum

This program is free software, you can redistribute it and/or modify it under the terms of the Artistic License version 2.0.

SEE ALSO

Net::Prometheus
Mojolicious::Plugin::Status
Mojolicious
Mojolicious::Guides
http://mojolicious.org