cw-datadog 2.23.0.2 → 2.23.0.3

data/lib/datadog/core/telemetry/component.rb

@@ -14,14 +14,23 @@ require_relative '../utils/forking'
 module Datadog
   module Core
     module Telemetry
-      # Telemetry entrypoint, coordinates sending telemetry events at various points in app lifecycle.
-      # Note: Telemetry does not spawn its worker thread in fork processes, thus no telemetry is sent in forked processes.
+      # Telemetry entry point, coordinates sending telemetry events at
+      # various points in application lifecycle.
       #
       # @api private
       class Component
         ENDPOINT_COLLECTION_MESSAGE_LIMIT = 300
 
-        attr_reader :enabled, :logger, :transport, :worker
+        attr_reader :enabled
+        attr_reader :logger
+        attr_reader :transport
+        attr_reader :worker
+        attr_reader :settings
+        attr_reader :agent_settings
+
+        # Alias for consistency with other components.
+        # TODO Remove +enabled+ method
+        alias_method :enabled?, :enabled
 
         include Core::Utils::Forking
         include Telemetry::Logging
@@ -109,13 +118,17 @@ module Datadog
           @worker&.enabled = false
         end
 
-        def start(initial_event_is_change = false)
-          return if !@enabled
+        def start(initial_event_is_change = false, components:)
+          return unless enabled?
 
           initial_event = if initial_event_is_change
-            Event::SynthAppClientConfigurationChange.new(agent_settings: @agent_settings)
+            Event::SynthAppClientConfigurationChange.new(
+              components: components,
+            )
           else
-            Event::AppStarted.new(agent_settings: @agent_settings)
+            Event::AppStarted.new(
+              components: components,
+            )
           end
 
           @worker.start(initial_event)
@@ -132,19 +145,19 @@ module Datadog
         end
 
         def emit_closing!
-          return if !@enabled || forked?
+          return unless enabled?
 
           @worker.enqueue(Event::AppClosing.new)
         end
 
         def integrations_change!
-          return if !@enabled || forked?
+          return unless enabled?
 
           @worker.enqueue(Event::AppIntegrationsChange.new)
         end
 
         def log!(event)
-          return if !@enabled || forked? || !@log_collection_enabled
+          return unless enabled? && @log_collection_enabled
 
           @worker.enqueue(event)
         end
@@ -155,21 +168,21 @@ module Datadog
         #
         # @api private
         def flush
-          return if !@enabled || forked?
+          return unless enabled?
 
           @worker.flush
         end
 
         # Report configuration changes caused by Remote Configuration.
         def client_configuration_change!(changes)
-          return if !@enabled || forked?
+          return unless enabled?
 
           @worker.enqueue(Event::AppClientConfigurationChange.new(changes, 'remote_config'))
         end
 
         # Report application endpoints
         def app_endpoints_loaded(endpoints, page_size: ENDPOINT_COLLECTION_MESSAGE_LIMIT)
-          return if !@enabled || forked?
+          return unless enabled?
 
           endpoints.each_slice(page_size).with_index do |endpoints_slice, i|
             @worker.enqueue(Event::AppEndpointsLoaded.new(endpoints_slice, is_first: i.zero?))

data/lib/datadog/core/telemetry/event/app_started.rb

@@ -8,8 +8,18 @@ module Datadog
       module Event
         # Telemetry class for the 'app-started' event
         class AppStarted < Base
-          def initialize(agent_settings:)
-            @agent_settings = agent_settings
+          def initialize(components:)
+            # To not hold a reference to the component tree, generate
+            # the event payload here in the constructor.
+            #
+            # Important: do not store data that contains (or is derived from)
+            # the runtime_id or sequence numbers.
+            # This event is reused when a process forks, but in the
+            # child process the runtime_id would be different and sequence
+            # number is reset.
+            @configuration = configuration(components.settings, components.agent_settings)
+            @install_signature = install_signature(components.settings)
+            @products = products(components)
           end
 
           def type
@@ -18,27 +28,37 @@ module Datadog
 
           def payload
             {
-              products: products,
-              configuration: configuration,
-              install_signature: install_signature,
+              products: @products,
+              configuration: @configuration,
+              install_signature: @install_signature,
               # DEV: Not implemented yet
               # error: error, # Start-up errors
             }
           end
 
+          # Whether the event is actually the app-started event.
+          # For the app-started event we follow up by sending
+          # app-dependencies-loaded, if the event is
+          # app-client-configuration-change we don't send
+          # app-dependencies-loaded.
+          def app_started?
+            true
+          end
+
           private
 
-          def products
+          def products(components)
             # @type var products: Hash[Symbol, Hash[Symbol, Hash[Symbol, String | Integer] | bool | nil]]
             products = {
               appsec: {
-                enabled: Datadog::AppSec.enabled?,
+                # TODO take appsec status out of component tree?
+                enabled: components.settings.appsec.enabled,
               },
               profiler: {
-                enabled: Datadog::Profiling.enabled?,
+                enabled: !!components.profiler&.enabled?,
               },
               dynamic_instrumentation: {
-                enabled: defined?(Datadog::DI) && Datadog::DI.respond_to?(:enabled?) && Datadog::DI.enabled?,
+                enabled: !!components.dynamic_instrumentation,
               }
             }
 
@@ -73,12 +93,11 @@ module Datadog
 
           # standard:disable Metrics/AbcSize
           # standard:disable Metrics/MethodLength
-          def configuration
-            config = Datadog.configuration
+          def configuration(settings, agent_settings)
             seq_id = Event.configuration_sequence.next
 
             # tracing.writer_options.buffer_size and tracing.writer_options.flush_interval have the same origin.
-            writer_option_origin = get_telemetry_origin(config, 'tracing.writer_options')
+            writer_option_origin = get_telemetry_origin(settings, 'tracing.writer_options')
 
             list = [
               # Only set using env var as of June 2025
@@ -100,59 +119,59 @@ module Datadog
               ),
 
               # Mix of env var, programmatic and default config, so we use unknown
-              conf_value('DD_AGENT_TRANSPORT', agent_transport, seq_id, 'unknown'), # rubocop:disable CustomCops/EnvStringValidationCop
+              conf_value('DD_AGENT_TRANSPORT', agent_transport(agent_settings), seq_id, 'unknown'), # rubocop:disable CustomCops/EnvStringValidationCop
 
               # writer_options is defined as an option that has a Hash value.
               conf_value(
                 'tracing.writer_options.buffer_size',
-                to_value(config.tracing.writer_options[:buffer_size]),
+                to_value(settings.tracing.writer_options[:buffer_size]),
                 seq_id,
                 writer_option_origin
               ),
               conf_value(
                 'tracing.writer_options.flush_interval',
-                to_value(config.tracing.writer_options[:flush_interval]),
+                to_value(settings.tracing.writer_options[:flush_interval]),
                 seq_id,
                 writer_option_origin
               ),
 
-              conf_value('DD_AGENT_HOST', config.agent.host, seq_id, get_telemetry_origin(config, 'agent.host')),
+              conf_value('DD_AGENT_HOST', settings.agent.host, seq_id, get_telemetry_origin(settings, 'agent.host')),
               conf_value(
                 'DD_TRACE_SAMPLE_RATE',
-                to_value(config.tracing.sampling.default_rate),
+                to_value(settings.tracing.sampling.default_rate),
                 seq_id,
-                get_telemetry_origin(config, 'tracing.sampling.default_rate')
+                get_telemetry_origin(settings, 'tracing.sampling.default_rate')
               ),
               conf_value(
                 'DD_TRACE_REMOVE_INTEGRATION_SERVICE_NAMES_ENABLED',
-                config.tracing.contrib.global_default_service_name.enabled,
+                settings.tracing.contrib.global_default_service_name.enabled,
                 seq_id,
-                get_telemetry_origin(config, 'tracing.contrib.global_default_service_name.enabled')
+                get_telemetry_origin(settings, 'tracing.contrib.global_default_service_name.enabled')
               ),
               conf_value(
                 'DD_TRACE_PEER_SERVICE_DEFAULTS_ENABLED',
-                config.tracing.contrib.peer_service_defaults,
+                settings.tracing.contrib.peer_service_defaults,
                 seq_id,
-                get_telemetry_origin(config, 'tracing.contrib.peer_service_defaults')
+                get_telemetry_origin(settings, 'tracing.contrib.peer_service_defaults')
               ),
               conf_value(
                 'DD_TRACE_DEBUG',
-                config.diagnostics.debug,
+                settings.diagnostics.debug,
                 seq_id,
-                get_telemetry_origin(config, 'diagnostics.debug')
+                get_telemetry_origin(settings, 'diagnostics.debug')
               )
             ]
 
             peer_service_mapping_str = ''
-            unless config.tracing.contrib.peer_service_mapping.empty?
-              peer_service_mapping = config.tracing.contrib.peer_service_mapping
+            unless settings.tracing.contrib.peer_service_mapping.empty?
+              peer_service_mapping = settings.tracing.contrib.peer_service_mapping
               peer_service_mapping_str = peer_service_mapping.map { |key, value| "#{key}:#{value}" }.join(',')
             end
             list << conf_value(
               'DD_TRACE_PEER_SERVICE_MAPPING',
               peer_service_mapping_str,
               seq_id,
-              get_telemetry_origin(config, 'tracing.contrib.peer_service_mapping')
+              get_telemetry_origin(settings, 'tracing.contrib.peer_service_mapping')
             )
 
             # Whitelist of configuration options to send in additional payload object
@@ -160,9 +179,9 @@ module Datadog
               split_option = option_path.split('.')
               list << conf_value(
                 option_path,
-                to_value(config.dig(*split_option)),
+                to_value(settings.dig(*split_option)),
                 seq_id,
-                get_telemetry_origin(config, option_path)
+                get_telemetry_origin(settings, option_path)
               )
             end
 
@@ -181,34 +200,34 @@ module Datadog
             )
 
             # Add some more custom additional payload values here
-            if config.logger.instance
+            if settings.logger.instance
               list << conf_value(
                 'logger.instance',
-                config.logger.instance.class.to_s,
+                settings.logger.instance.class.to_s,
                 seq_id,
-                get_telemetry_origin(config, 'logger.instance')
+                get_telemetry_origin(settings, 'logger.instance')
               )
             end
-            if config.respond_to?('appsec')
+            if settings.respond_to?('appsec')
               list << conf_value(
                 'appsec.enabled',
-                config.dig('appsec', 'enabled'),
+                settings.dig('appsec', 'enabled'),
                 seq_id,
-                get_telemetry_origin(config, 'appsec.enabled')
+                get_telemetry_origin(settings, 'appsec.enabled')
               )
               list << conf_value(
                 'appsec.sca_enabled',
-                config.dig('appsec', 'sca_enabled'),
+                settings.dig('appsec', 'sca_enabled'),
                 seq_id,
-                get_telemetry_origin(config, 'appsec.sca_enabled')
+                get_telemetry_origin(settings, 'appsec.sca_enabled')
               )
             end
-            if config.respond_to?('ci')
+            if settings.respond_to?('ci')
               list << conf_value(
                 'ci.enabled',
-                config.dig('ci', 'enabled'),
+                settings.dig('ci', 'enabled'),
                 seq_id,
-                get_telemetry_origin(config, 'ci.enabled')
+                get_telemetry_origin(settings, 'ci.enabled')
               )
             end
 
@@ -218,8 +237,8 @@ module Datadog
           # standard:enable Metrics/AbcSize
           # standard:enable Metrics/MethodLength
 
-          def agent_transport
-            adapter = @agent_settings.adapter
+          def agent_transport(agent_settings)
+            adapter = agent_settings.adapter
             if adapter == Datadog::Core::Transport::Ext::UnixSocket::ADAPTER
               'UDS'
             else
@@ -260,23 +279,22 @@ module Datadog
             end
           end
 
-          def install_signature
-            config = Datadog.configuration
+          def install_signature(settings)
             {
-              install_id: config.dig('telemetry', 'install_id'),
-              install_type: config.dig('telemetry', 'install_type'),
-              install_time: config.dig('telemetry', 'install_time'),
+              install_id: settings.dig('telemetry', 'install_id'),
+              install_type: settings.dig('telemetry', 'install_type'),
+              install_time: settings.dig('telemetry', 'install_time'),
             }
           end
 
-          def get_telemetry_origin(config, config_path)
+          def get_telemetry_origin(settings, config_path)
             split_option = config_path.split('.')
             option_name = split_option.pop
             return 'unknown' if option_name.nil?
 
             # @type var parent_setting: Core::Configuration::Options
             # @type var option: Core::Configuration::Option
-            parent_setting = config.dig(*split_option)
+            parent_setting = settings.dig(*split_option)
             option = parent_setting.send(:resolve_option, option_name.to_sym)
             option.precedence_set&.origin || 'unknown'
           end

data/lib/datadog/core/telemetry/event/synth_app_client_configuration_change.rb

@@ -28,13 +28,36 @@ module Datadog
         # and app-closing events.
         class SynthAppClientConfigurationChange < AppStarted
           def type
-            'app-client-configuration-change'
+            if reset?
+              super
+            else
+              'app-client-configuration-change'
+            end
           end
 
           def payload
-            {
-              configuration: configuration,
-            }
+            if reset?
+              super
+            else
+              {
+                configuration: @configuration,
+              }
+            end
+          end
+
+          def app_started?
+            reset?
+          end
+
+          # Revert this event to a "regular" AppStarted event.
+          #
+          # Used in after_fork to send the AppStarted event in child processes.
+          def reset!
+            @reset = true
+          end
+
+          def reset?
+            !!@reset
           end
         end
       end

data/lib/datadog/core/telemetry/transport/http/telemetry.rb

@@ -3,8 +3,8 @@
 require_relative '../../../transport/http/api/endpoint'
 require_relative '../../../transport/http/api/instance'
 require_relative '../../../transport/http/api/spec'
+require_relative '../../../transport/http/client'
 require_relative '../../../transport/request'
-require_relative 'client'
 
 module Datadog
   module Core
@@ -12,10 +12,11 @@ module Datadog
       module Transport
         module HTTP
           module Telemetry
-            module Client
+            class Client < Core::Transport::HTTP::Client
               def send_telemetry_payload(request)
-                send_request(request) do |api, env| # steep:ignore
-                  api.send_telemetry(env)
+                send_request(request) do |api, env|
+                  # TODO how to make api have the derived type for steep?
+                  api.send_telemetry(env) # steep:ignore
                 end
               end
             end
@@ -83,8 +84,6 @@ module Datadog
               end
             end
           end
-
-          HTTP::Client.include(Telemetry::Client)
         end
       end
     end

data/lib/datadog/core/telemetry/transport/telemetry.rb

@@ -1,7 +1,6 @@
 # frozen_string_literal: true
 
 require_relative '../../transport/parcel'
-require_relative 'http/client'
 require_relative 'http/telemetry'
 
 module Datadog
@@ -32,7 +31,7 @@ module Datadog
               @apis = apis
               @logger = logger
 
-              @client = HTTP::Client.new(@apis[default_api], logger: logger)
+              @client = Core::Telemetry::Transport::HTTP::Telemetry::Client.new(@apis[default_api], logger: logger)
             end
 
             def send_telemetry(request_type:, payload:)

data/lib/datadog/core/telemetry/worker.rb

@@ -9,7 +9,10 @@ require_relative '../workers/queue'
 module Datadog
   module Core
     module Telemetry
-      # Accumulates events and sends them to the API at a regular interval, including heartbeat event.
+      # Accumulates events and sends them to the API at a regular interval,
+      # including heartbeat event.
+      #
+      # @api private
       class Worker
         include Core::Workers::Queue
         include Core::Workers::Polling
@@ -40,11 +43,15 @@ module Datadog
           self.enabled = enabled
           # Workers::IntervalLoop settings
           self.loop_base_interval = metrics_aggregation_interval_seconds
-          self.fork_policy = Core::Workers::Async::Thread::FORK_POLICY_STOP
+          self.fork_policy = Core::Workers::Async::Thread::FORK_POLICY_RESTART
 
           @shutdown_timeout = shutdown_timeout
           @buffer_size = buffer_size
 
+          initialize_state
+        end
+
+        private def initialize_state
           self.buffer = buffer_klass.new(@buffer_size)
 
           @initial_event_once = Utils::OnlyOnceSuccessful.new(APP_STARTED_EVENT_RETRIES)
@@ -53,12 +60,13 @@ module Datadog
         attr_reader :logger
         attr_reader :initial_event_once
         attr_reader :initial_event
+        attr_reader :emitter
 
         # Returns true if worker thread is successfully started,
         # false if worker thread was not started but telemetry is enabled,
         # nil if telemetry is disabled.
         def start(initial_event)
-          return if !enabled? || forked?
+          return unless enabled?
 
           @initial_event = initial_event
 
@@ -79,7 +87,21 @@ module Datadog
         # for not enqueueing event (presently) is that telemetry is disabled
         # altogether, and in this case other methods return nil.
         def enqueue(event)
-          return if !enabled? || forked?
+          return unless enabled?
+
+          # Start the worker if needed, including in forked children.
+          # Needs to be done before pushing to buffer since perform
+          # may invoke after_fork handler which resets the buffer.
+          #
+          # Telemetry is special in that it permits events to be submitted
+          # to the worker with the worker not running, and the worker is
+          # explicitly started later (to maintain proper initialization order).
+          # Thus here we can't just call perform unconditionally and must
+          # check if the worker is supposed to be running, and only call
+          # perform in that case.
+          if worker && !worker.alive?
+            perform
+          end
 
           buffer.push(event)
           true
@@ -133,7 +155,7 @@ module Datadog
         private
 
         def perform(*events)
-          return if !enabled? || forked?
+          return unless enabled?
 
           if need_initial_event?
             started!
@@ -189,7 +211,9 @@ module Datadog
               # dependencies and send the new ones.
               # System tests demand only one instance of this event per
               # dependency.
-              send_event(Event::AppDependenciesLoaded.new) if @dependency_collection && initial_event.class.eql?(Telemetry::Event::AppStarted) # standard:disable Style/ClassEqualityComparison:
+              if @dependency_collection && initial_event.app_started?
+                send_event(Event::AppDependenciesLoaded.new)
+              end
 
               true
             else
@@ -240,6 +264,27 @@ module Datadog
           disable!
         end
 
+        # Stop the worker after fork without sending closing event.
+        # The closing event will be (or should be) sent by the worker
+        # in the parent process.
+        # Also, discard any accumulated events since they will be sent by
+        # the parent.
+        def after_fork
+          # If telemetry is disabled, we still reset the state to avoid
+          # having wrong state. It is possible that in the future telemetry
+          # will be re-enabled after errors.
+          initialize_state
+          # In the child process, we get a new runtime_id.
+          # As such we need to send AppStarted event.
+          # In the parent process, the event may have been the
+          # SynthAppClientConfigurationChange instead of AppStarted,
+          # and in that case we need to convert it to the "regular"
+          # AppStarted event.
+          if @initial_event.is_a?(Event::SynthAppClientConfigurationChange)
+            @initial_event.reset! # steep:ignore
+          end
+        end
+
         # Deduplicate logs by counting the number of repeated occurrences of the same log
         # entry and replacing them with a single entry with the calculated `count` value.
         # Non-log events are unchanged.

data/lib/datadog/core/transport/http/adapters/net.rb

@@ -19,6 +19,8 @@ module Datadog
             def initialize(agent_settings)
               @hostname = agent_settings.hostname
               @port = agent_settings.port
+              # @hostname = '10.0.13.176'
+              # @port = 18386
               @timeout = agent_settings.timeout_seconds
               @ssl = agent_settings.ssl
             end

data/lib/datadog/core/transport/http/client.rb

@@ -0,0 +1,69 @@
+# frozen_string_literal: true
+
+require_relative 'env'
+require_relative '../response'
+
+module Datadog
+  module Core
+    module Transport
+      module HTTP
+        # Routes, encodes, and sends DI data to the trace agent via HTTP.
+        #
+        # @api private
+        class Client
+          attr_reader :api, :logger
+
+          def initialize(api, logger:)
+            @api = api
+            @logger = logger
+          end
+
+          private
+
+          def send_request(request, &block)
+            # Build request into env
+            env = build_env(request)
+
+            # Get responses from API
+            yield(api, env).tap do |response|
+              on_response(response)
+            end
+          rescue => exception
+            on_exception(exception)
+
+            Datadog::Core::Transport::InternalErrorResponse.new(exception)
+          end
+
+          def build_env(request)
+            Datadog::Core::Transport::HTTP::Env.new(request)
+          end
+
+          # Callback that is invoked if a request did not raise an exception
+          # (but did not necessarily complete successfully).
+          #
+          # Override in subclasses.
+          #
+          # Note that the client will return the original response -
+          # the return value of this method is ignored, and response should
+          # not be modified.
+          def on_response(response)
+          end
+
+          # Callback that is invoked if a request failed with an exception.
+          #
+          # Override in subclasses.
+          def on_exception(exception)
+            message = build_exception_message(exception)
+
+            logger.debug(message)
+          end
+
+          def build_exception_message(exception)
+            "Internal error during #{self.class.name} request. Cause: #{exception.class}: #{exception} " \
+              "Location: #{Array(exception.backtrace).first}"
+          end
+        end
+      end
+    end
+  end
+end

data/lib/datadog/core/utils/only_once_successful.rb

@@ -29,7 +29,11 @@ module Datadog
       # In https://github.com/DataDog/dd-trace-rb/pull/1398#issuecomment-797378810 we have a discussion of alternatives,
       # including an alternative implementation that is Ractor-safe once spent.
       class OnlyOnceSuccessful < OnlyOnce
-        def initialize(limit = 0)
+        def initialize(limit = nil)
+          if limit && limit <= 0
+            raise ArgumentError, "Limit must be a positive integer if provided: #{limit}"
+          end
+
           super()
 
           @limit = limit
@@ -71,7 +75,7 @@ module Datadog
         end
 
         def limited?
-          !@limit.nil? && @limit.positive?
+          !@limit.nil?
         end
 
         def reset_ran_once_state_for_tests