Conditionally initialize the global opentracing tracer

This change will instantiate an OpenTracing tracer and configure it
as the global tracer when the GITLAB_TRACING environment variable is
configured. GITLAB_TRACING takes a "connection string"-like value,
encapsulating the driver (eg jaeger, etc) and options for the driver.

Since each service, whether it's written in Ruby or Golang, uses the
same connection-string, it should be very easy to configure all
services in a cluster, or even a single development machine to be
setup to use tracing.

Note that this change does not include instrumentation or propagation
changes as this is a way of breaking a previous larger change into
components. The instrumentation and propagation changes will follow
in separate changes.

2 files changed, 158 insertions, 0 deletions

diff --git a/lib/gitlab/tracing/factory.rb b/lib/gitlab/tracing/factory.rb
new file mode 100644
index 00000000000..fc714164353
--- /dev/null
+++ b/lib/gitlab/tracing/factory.rb
@@ -0,0 +1,61 @@
+# frozen_string_literal: true
+
+require "cgi"
+
+module Gitlab
+  module Tracing
+    class Factory
+      OPENTRACING_SCHEME = "opentracing"
+
+      def self.create_tracer(service_name, connection_string)
+        return unless connection_string.present?
+
+        begin
+          opentracing_details = parse_connection_string(connection_string)
+          driver_name = opentracing_details[:driver_name]
+
+          case driver_name
+          when "jaeger"
+            JaegerFactory.create_tracer(service_name, opentracing_details[:options])
+          else
+            raise "Unknown driver: #{driver_name}"
+          end
+        rescue => e
+          # Can't create the tracer? Warn and continue sans tracer
+          warn "Unable to instantiate tracer: #{e}"
+          nil
+        end
+      end
+
+      def self.parse_connection_string(connection_string)
+        parsed = URI.parse(connection_string)
+
+        unless valid_uri?(parsed)
+          raise "Invalid tracing connection string"
+        end
+
+        {
+          driver_name: parsed.host,
+          options: parse_query(parsed.query)
+        }
+      end
+      private_class_method :parse_connection_string
+
+      def self.parse_query(query)
+        return {} unless query
+
+        CGI.parse(query).symbolize_keys.transform_values(&:first)
+      end
+      private_class_method :parse_query
+
+      def self.valid_uri?(uri)
+        return false unless uri
+
+        uri.scheme == OPENTRACING_SCHEME &&
+          uri.host.to_s =~ /^[a-z0-9_]+$/ &&
+          uri.path.empty?
+      end
+      private_class_method :valid_uri?
+    end
+  end
+end
diff --git a/lib/gitlab/tracing/jaeger_factory.rb b/lib/gitlab/tracing/jaeger_factory.rb
new file mode 100644
index 00000000000..0726f6b67f4
--- /dev/null
+++ b/lib/gitlab/tracing/jaeger_factory.rb
@@ -0,0 +1,97 @@
+# frozen_string_literal: true
+
+require 'jaeger/client'
+
+module Gitlab
+  module Tracing
+    class JaegerFactory
+      # When the probabilistic sampler is used, by default 0.1% of requests will be traced
+      DEFAULT_PROBABILISTIC_RATE = 0.001
+
+      # The default port for the Jaeger agent UDP listener
+      DEFAULT_UDP_PORT = 6831
+
+      # Reduce this from default of 10 seconds as the Ruby jaeger
+      # client doesn't have overflow control, leading to very large
+      # messages which fail to send over UDP (max packet = 64k)
+      # Flush more often, with smaller packets
+      FLUSH_INTERVAL = 5
+
+      def self.create_tracer(service_name, options)
+        kwargs = {
+          service_name: service_name,
+          sampler: get_sampler(options[:sampler], options[:sampler_param]),
+          reporter: get_reporter(service_name, options[:http_endpoint], options[:udp_endpoint])
+        }
+
+        extra_params = options.except(:sampler, :sampler_param, :http_endpoint, :udp_endpoint, :strict_parsing, :debug) # rubocop: disable CodeReuse/ActiveRecord
+        if extra_params.present?
+          message = "jaeger tracer: invalid option: #{extra_params.keys.join(", ")}"
+
+          if options[:strict_parsing]
+            raise message
+          else
+            warn message
+          end
+        end
+
+        Jaeger::Client.build(kwargs)
+      end
+
+      def self.get_sampler(sampler_type, sampler_param)
+        case sampler_type
+        when "probabilistic"
+          sampler_rate = sampler_param ? sampler_param.to_f : DEFAULT_PROBABILISTIC_RATE
+          Jaeger::Samplers::Probabilistic.new(rate: sampler_rate)
+        when "const"
+          const_value = sampler_param == "1"
+          Jaeger::Samplers::Const.new(const_value)
+        else
+          nil
+        end
+      end
+      private_class_method :get_sampler
+
+      def self.get_reporter(service_name, http_endpoint, udp_endpoint)
+        encoder = Jaeger::Encoders::ThriftEncoder.new(service_name: service_name)
+
+        if http_endpoint.present?
+          sender = get_http_sender(encoder, http_endpoint)
+        elsif udp_endpoint.present?
+          sender = get_udp_sender(encoder, udp_endpoint)
+        else
+          return nil
+        end
+
+        Jaeger::Reporters::RemoteReporter.new(
+          sender: sender,
+          flush_interval: FLUSH_INTERVAL
+        )
+      end
+      private_class_method :get_reporter
+
+      def self.get_http_sender(encoder, address)
+        Jaeger::HttpSender.new(
+          url: address,
+          encoder: encoder,
+          logger: Logger.new(STDOUT)
+        )
+      end
+      private_class_method :get_http_sender
+
+      def self.get_udp_sender(encoder, address)
+        pair = address.split(":", 2)
+        host = pair[0]
+        port = pair[1] ? pair[1].to_i : DEFAULT_UDP_PORT
+
+        Jaeger::UdpSender.new(
+          host: host,
+          port: port,
+          encoder: encoder,
+          logger: Logger.new(STDOUT)
+        )
+      end
+      private_class_method :get_udp_sender
+    end
+  end
+end