Merge pull request #138 from epimorphics/issue/135-task-prometheus-me…

…trics-update

Task: Prometheus Metrics - Update

CHANGELOG.md

@@ -5,7 +5,24 @@ including the SPARQL Qonsole
 
 ## 1.8.0 - 2024-09
 
-- (Bogdan) Updated both english and welsh copies of the accessibility statement [GH-136](https://github.com/epimorphics/lr-landing/issues/136)
+- (Jon) Create a `config/initializers/load_notification_subscribers.rb` file to
+  load all the notification subscribers in the application so that they are
+  registeredly correctly
+  [GH-135](https://github.com/epimorphics/lr-landing/issues/135)
+- (Jon) Updated the application exceptions controller to instrument the
+  `ActiveSupport::Notifications` for internal errors
+  [GH-135](https://github.com/epimorphics/lr-landing/issues/135)
+- (Jon) Updated `config/initializers/prometheus.rb` to include the `Middleware
+  instrumentation` fix for the 0 memory bug by notifying Action Dispatch
+  subscribers on Prometheus initialise
+  [GH-135](https://github.com/epimorphics/lr-landing/issues/135)
+- (Jon) Updated `config/puma.rb` to include metrics plugin and port information
+  for the metrics endpoint as environment variable, with default, to enable
+  running multiple sibling HMLR apps locally if needed without port conflicts
+  [GH-135](https://github.com/epimorphics/lr-landing/issues/135)
+- (Jon) Updated the `lr_common_styles` gem to the latest 1.9.9 patch release.
+- (Bogdan) Updated both english and welsh copies of the accessibility statement
+  [GH-136](https://github.com/epimorphics/lr-landing/issues/136)
 - (Jon) Moved all mirrored configuration settings from individual environments
   into the application configuration to reduce the need to manage multiple
   sources of truth

Gemfile.lock

@@ -265,7 +265,7 @@ GEM
       json
       lograge
       railties
-    lr_common_styles (1.9.8)
+    lr_common_styles (1.9.9)
       bootstrap-sass (~> 3.4.0)
       font-awesome-rails (~> 4.7.0.1)
       govuk_elements_rails (~> 2.0.0)

app/controllers/application_controller.rb

@@ -32,6 +32,7 @@ def change_default_caching_policy
   # @param [Exception] exp the exception that caused the error
   # @return [ActiveSupport::Notifications::Event] provides an object-oriented
   # interface to the event
+  #!IMPORTANT: This method is not used in the codebase and is only here for reference
   def instrument_internal_error(exception)
     ActiveSupport::Notifications.instrument('internal_error.application', exception: exception)
   end

app/subscribers/action_dispatch_prometheus_subscriber.rb

@@ -7,7 +7,7 @@
 class ActionDispatchPrometheusSubscriber < ActiveSupport::Subscriber
   attach_to :action_controller
 
-  # rubocop:disable Metrics/AbcSize
+  # rubocop:disable Metrics/AbcSize, Metrics/MethodLength
   def process_action(_event)
     mem = GetProcessMem.new
     Prometheus::Client.registry
@@ -59,5 +59,5 @@ def process_action(_event)
                         }
                       )
   end
-  # rubocop:enable Metrics/AbcSize
+  # rubocop:enable Metrics/AbcSize, Metrics/MethodLength
 end

config/initializers/load_notification_subscribers.rb

@@ -0,0 +1,6 @@
+  # frozen_string_literal: true
+
+  # Load all the notification subscribers in the application so that they are registered
+  Dir[Rails.root.join('app', 'subscribers', '**', '*_subscriber.rb')].sort.each do |source|
+    require source
+  end

config/initializers/prometheus.rb

@@ -53,3 +53,7 @@
   docstring: 'Histogram of back-end API response times',
   buckets: Prometheus::Client::Histogram.exponential_buckets(start: 0.0005, count: 16)
 )
+
+# Middleware instrumentation
+  # This fixes the 0 memory bug by notifying Action Dispatch subscribers on Prometheus initialise
+  ActiveSupport::Notifications.instrument('process_middleware.action_dispatch')

config/puma.rb

@@ -39,6 +39,13 @@
 # Allow puma to be restarted by `rails restart` command.
 plugin :tmp_restart
 
+# Uncomment the following line once ruby is updated to 2.7 or greater to allow
+# the use of the puma-metrics plugin as we're using puma 6.0.0 or greater
+# Additional metrics from the Puma server to be exposed in the /metrics endpoint
+# plugin :metrics
+# Bind the metric server to "url". "tcp://" is the only accepted protocol.
+# metrics_url 'tcp://0.0.0.0:9393'
+
 # Use a custom log formatter to emit Puma log messages in a JSON format
 log_formatter do |str|
   {