Merge remote-tracking branch 'parent/main' into upstream-20240128

config/initializers/prometheus_exporter.rb

@@ -0,0 +1,24 @@
+# frozen_string_literal: true
+
+if ENV['MASTODON_PROMETHEUS_EXPORTER_ENABLED'] == 'true'
+  if ENV['MASTODON_PROMETHEUS_EXPORTER_LOCAL'] == 'true'
+    require 'prometheus_exporter/server'
+    require 'prometheus_exporter/client'
+
+    # bind is the address, on which the webserver will listen
+    # port is the port that will provide the /metrics route
+    server = PrometheusExporter::Server::WebServer.new bind: ENV.fetch('MASTODON_PROMETHEUS_EXPORTER_HOST', 'localhost'), port: ENV.fetch('MASTODON_PROMETHEUS_EXPORTER_PORT', '9394').to_i
+    server.start
+
+    # wire up a default local client
+    PrometheusExporter::Client.default = PrometheusExporter::LocalClient.new(collector: server.collector)
+  end
+
+  if ENV['MASTODON_PROMETHEUS_EXPORTER_WEB_DETAILED_METRICS'] == 'true'
+    # Optional, as those metrics might generate extra overhead and be redundant with what OTEL provides
+    require 'prometheus_exporter/middleware'
+
+    # Per-action/controller request stats like HTTP status and timings
+    Rails.application.middleware.unshift PrometheusExporter::Middleware
+  end
+end

config/initializers/sidekiq.rb

@@ -22,6 +22,48 @@ Sidekiq.configure_server do |config|
     end
   end
 
+  if ENV['MASTODON_PROMETHEUS_EXPORTER_ENABLED'] == 'true'
+    require 'prometheus_exporter'
+    require 'prometheus_exporter/instrumentation'
+
+    config.on :startup do
+      # Ruby process metrics (memory, GC, etc)
+      PrometheusExporter::Instrumentation::Process.start type: 'sidekiq'
+
+      # Sidekiq process metrics (concurrency, busy, etc)
+      PrometheusExporter::Instrumentation::SidekiqProcess.start
+
+      # ActiveRecord metrics (connection pool usage)
+      PrometheusExporter::Instrumentation::ActiveRecord.start(
+        custom_labels: { type: 'sidekiq' },
+        config_labels: [:database, :host]
+      )
+
+      if ENV['MASTODON_PROMETHEUS_EXPORTER_SIDEKIQ_DETAILED_METRICS'] == 'true'
+        # Optional, as those metrics might generate extra overhead and be redundant with what OTEL provides
+
+        # Per-job metrics
+        config.server_middleware do |chain|
+          chain.add PrometheusExporter::Instrumentation::Sidekiq
+        end
+        config.death_handlers << PrometheusExporter::Instrumentation::Sidekiq.death_handler
+
+        # Per-queue metrics for queues handled by this process (size, latency, etc)
+        # They will be reported by every process handling those queues, so do not sum them up
+        PrometheusExporter::Instrumentation::SidekiqQueue.start
+
+        # Global Sidekiq metrics (size of the global queues, number of jobs, etc)
+        # Will be the same for every Sidekiq process
+        PrometheusExporter::Instrumentation::SidekiqStats.start
+      end
+    end
+
+    at_exit do
+      # Wait for the latest metrics to be reported before shutting down
+      PrometheusExporter::Client.default.stop(wait_timeout_seconds: 10)
+    end
+  end
+
   config.server_middleware do |chain|
     chain.add Mastodon::SidekiqMiddleware
   end