ddtrace 0.40.0 → 0.50.0

data/lib/ddtrace/profiling/scheduler.rb

@@ -0,0 +1,116 @@
+require 'ddtrace/utils/time'
+
+require 'ddtrace/worker'
+require 'ddtrace/workers/polling'
+
+module Datadog
+  module Profiling
+    # Periodically (every DEFAULT_INTERVAL_SECONDS) takes data from the `Recorder` and pushes them to all configured
+    # `Exporter`s. Runs on its own background thread.
+    class Scheduler < Worker
+      include Workers::Polling
+
+      DEFAULT_INTERVAL_SECONDS = 60
+      MIN_INTERVAL_SECONDS = 0
+
+      attr_reader \
+        :exporters,
+        :recorder
+
+      def initialize(
+        recorder,
+        exporters,
+        fork_policy: Workers::Async::Thread::FORK_POLICY_RESTART, # Restart in forks by default
+        interval: DEFAULT_INTERVAL_SECONDS,
+        enabled: true
+      )
+        @recorder = recorder
+        @exporters = [exporters].flatten
+
+        # Workers::Async::Thread settings
+        self.fork_policy = fork_policy
+
+        # Workers::IntervalLoop settings
+        self.loop_base_interval = interval
+
+        # Workers::Polling settings
+        self.enabled = enabled
+      end
+
+      def start
+        perform
+      end
+
+      def perform
+        # A profiling flush may be called while the VM is shutting down, to report the last profile. When we do so,
+        # we impose a strict timeout. This means this last profile may or may not be sent, depending on if the flush can
+        # successfully finish in the strict timeout.
+        # This can be somewhat confusing (why did it not get reported?), so let's at least log what happened.
+        interrupted = true
+
+        begin
+          flush_and_wait
+          interrupted = false
+        ensure
+          Datadog.logger.debug('#flush was interrupted or failed before it could complete') if interrupted
+        end
+      end
+
+      def loop_back_off?
+        false
+      end
+
+      def after_fork
+        # Clear recorder's buffers by flushing events.
+        # Objects from parent process will copy-on-write,
+        # and we don't want to send events for the wrong process.
+        recorder.flush
+      end
+
+      # Configure Workers::IntervalLoop to not report immediately when scheduler starts
+      #
+      # When a scheduler gets created (or reset), we don't want it to immediately try to flush; we want it to wait for
+      # the loop wait time first. This avoids an issue where the scheduler reported a mostly-empty profile if the
+      # application just started but this thread took a bit longer so there's already samples in the recorder.
+      def loop_wait_before_first_iteration?
+        true
+      end
+
+      def work_pending?
+        !recorder.empty?
+      end
+
+      private
+
+      def flush_and_wait
+        run_time = Datadog::Utils::Time.measure do
+          flush_events
+        end
+
+        # Update wait time to try to wake consistently on time.
+        # Don't drop below the minimum interval.
+        self.loop_wait_time = [loop_base_interval - run_time, MIN_INTERVAL_SECONDS].max
+      end
+
+      def flush_events
+        # Get events from recorder
+        flush = recorder.flush
+
+        # Send events to each exporter
+        if flush.event_count > 0
+          exporters.each do |exporter|
+            begin
+              exporter.export(flush)
+            rescue StandardError => e
+              Datadog.logger.error(
+                "Unable to export #{flush.event_count} profiling events. Cause: #{e} Location: #{e.backtrace.first}"
+              )
+            end
+          end
+        end
+
+        flush
+      end
+    end
+  end
+end

data/lib/ddtrace/profiling/tasks/setup.rb

@@ -0,0 +1,77 @@
+require 'ddtrace'
+require 'ddtrace/utils/only_once'
+require 'ddtrace/profiling'
+require 'ddtrace/profiling/ext/cpu'
+require 'ddtrace/profiling/ext/forking'
+
+module Datadog
+  module Profiling
+    module Tasks
+      # Takes care of loading our extensions/monkey patches to handle fork() and CPU profiling.
+      class Setup
+        ACTIVATE_EXTENSIONS_ONLY_ONCE = Datadog::Utils::OnlyOnce.new
+
+        def run
+          ACTIVATE_EXTENSIONS_ONLY_ONCE.run do
+            begin
+              activate_forking_extensions
+              activate_cpu_extensions
+              setup_at_fork_hooks
+            rescue StandardError, ScriptError => e
+              Datadog.logger.warn { "Profiler extensions unavailable. Cause: #{e.message} Location: #{e.backtrace.first}" }
+            end
+          end
+        end
+
+        private
+
+        def activate_forking_extensions
+          if Ext::Forking.supported?
+            Ext::Forking.apply!
+          elsif Datadog.configuration.profiling.enabled
+            Datadog.logger.debug('Profiler forking extensions skipped; forking not supported.')
+          end
+        rescue StandardError, ScriptError => e
+          Datadog.logger.warn do
+            "Profiler forking extensions unavailable. Cause: #{e.message} Location: #{e.backtrace.first}"
+          end
+        end
+
+        def activate_cpu_extensions
+          if Ext::CPU.supported?
+            Ext::CPU.apply!
+          elsif Datadog.configuration.profiling.enabled
+            Datadog.logger.info do
+              'CPU time profiling skipped because native CPU time is not supported: ' \
+              "#{Ext::CPU.unsupported_reason}. Profiles containing Wall time will still be reported."
+            end
+          end
+        rescue StandardError, ScriptError => e
+          Datadog.logger.warn do
+            "Profiler CPU profiling extensions unavailable. Cause: #{e.message} Location: #{e.backtrace.first}"
+          end
+        end
+
+        def setup_at_fork_hooks
+          if Process.respond_to?(:at_fork)
+            Process.at_fork(:child) do
+              begin
+                # When Ruby forks, clock IDs for each of the threads
+                # will change. We can only update these IDs from the
+                # execution context of the thread that owns it.
+                # This hook will update the IDs for the main thread
+                # after a fork occurs.
+                Thread.current.send(:update_native_ids) if Thread.current.respond_to?(:update_native_ids, true)
+
+                # Restart profiler, if enabled
+                Datadog.profiler.start if Datadog.profiler
+              rescue StandardError => e
+                Datadog.logger.warn { "Error during post-fork hooks. Cause: #{e.message} Location: #{e.backtrace.first}" }
+              end
+            end
+          end
+        end
+      end
+    end
+  end
+end

data/lib/ddtrace/profiling/transport/client.rb

@@ -0,0 +1,12 @@
+module Datadog
+  module Profiling
+    module Transport
+      # Generic interface for profiling transports
+      module Client
+        def send_profiling_flush(flush)
+          raise NotImplementedError
+        end
+      end
+    end
+  end
+end

data/lib/ddtrace/profiling/transport/http.rb

@@ -0,0 +1,118 @@
+require 'ddtrace/ext/runtime'
+require 'ddtrace/ext/transport'
+
+require 'ddtrace/runtime/container'
+
+require 'ddtrace/profiling/transport/http/builder'
+require 'ddtrace/profiling/transport/http/api'
+
+require 'ddtrace/transport/http/adapters/net'
+require 'ddtrace/transport/http/adapters/test'
+require 'ddtrace/transport/http/adapters/unix_socket'
+
+module Datadog
+  module Profiling
+    module Transport
+      # Namespace for HTTP transport components
+      module HTTP
+        module_function
+
+        # Builds a new Transport::HTTP::Client
+        def new(&block)
+          Builder.new(&block).to_transport
+        end
+
+        # Builds a new Transport::HTTP::Client with default settings
+        # Pass a block to override any settings.
+        def default(profiling_upload_timeout_seconds:, agent_settings: nil, site: nil, api_key: nil)
+          new do |transport|
+            transport.headers default_headers
+
+            # Configure adapter & API
+            if site && api_key
+              configure_for_agentless(
+                transport,
+                profiling_upload_timeout_seconds: profiling_upload_timeout_seconds,
+                site: site,
+                api_key: api_key
+              )
+            else
+              unless agent_settings
+                raise(
+                  ArgumentError,
+                  "Missing configuration for #{self}.default: All of `agent_settings`, `site` and `api_key` are nil"
+                )
+              end
+
+              configure_for_agent(
+                transport,
+                profiling_upload_timeout_seconds: profiling_upload_timeout_seconds,
+                agent_settings: agent_settings
+              )
+            end
+          end
+        end
+
+        def default_headers
+          {
+            Datadog::Ext::Transport::HTTP::HEADER_META_LANG => Datadog::Ext::Runtime::LANG,
+            Datadog::Ext::Transport::HTTP::HEADER_META_LANG_VERSION => Datadog::Ext::Runtime::LANG_VERSION,
+            Datadog::Ext::Transport::HTTP::HEADER_META_LANG_INTERPRETER => Datadog::Ext::Runtime::LANG_INTERPRETER,
+            Datadog::Ext::Transport::HTTP::HEADER_META_TRACER_VERSION => Datadog::Ext::Runtime::TRACER_VERSION
+          }.tap do |headers|
+            # Add container ID, if present.
+            container_id = Datadog::Runtime::Container.container_id
+            headers[Datadog::Ext::Transport::HTTP::HEADER_CONTAINER_ID] = container_id unless container_id.nil?
+          end
+        end
+
+        private_class_method def default_adapter
+          :net_http
+        end
+
+        private_class_method def configure_for_agent(transport, profiling_upload_timeout_seconds:, agent_settings:)
+          apis = API.agent_defaults
+
+          transport.adapter(
+            default_adapter,
+            agent_settings.hostname,
+            agent_settings.port,
+            # We explictly use profiling_upload_timeout_seconds instead of agent_settings.timeout because profile
+            # uploads are bigger and thus we employ a separate configuration.
+            timeout: profiling_upload_timeout_seconds,
+            ssl: agent_settings.ssl
+          )
+          transport.api(API::V1, apis[API::V1], default: true)
+
+          # NOTE: This proc, when it exists, usually overrides the transport specified above
+          if agent_settings.deprecated_for_removal_transport_configuration_proc
+            agent_settings.deprecated_for_removal_transport_configuration_proc.call(transport)
+          end
+        end
+
+        private_class_method def configure_for_agentless(transport, profiling_upload_timeout_seconds:, site:, api_key:)
+          apis = API.api_defaults
+
+          site_uri = URI(format(Datadog::Ext::Profiling::Transport::HTTP::URI_TEMPLATE_DD_API, site))
+          hostname = site_uri.host
+          port = site_uri.port
+
+          transport.adapter(
+            default_adapter,
+            hostname,
+            port,
+            timeout: profiling_upload_timeout_seconds,
+            ssl: site_uri.scheme == 'https'
+          )
+          transport.api(API::V1, apis[API::V1], default: true)
+          transport.headers(Datadog::Ext::Transport::HTTP::HEADER_DD_API_KEY => api_key)
+        end
+
+        # Add adapters to registry
+        Builder::REGISTRY.set(Datadog::Transport::HTTP::Adapters::Net, :net_http)
+        Builder::REGISTRY.set(Datadog::Transport::HTTP::Adapters::Test, :test)
+        Builder::REGISTRY.set(Datadog::Transport::HTTP::Adapters::UnixSocket, :unix)
+      end
+    end
+  end
+end

data/lib/ddtrace/profiling/transport/http/api.rb

@@ -0,0 +1,43 @@
+require 'ddtrace/transport/http/api/map'
+require 'ddtrace/profiling/encoding/profile'
+require 'ddtrace/profiling/transport/http/api/spec'
+require 'ddtrace/profiling/transport/http/api/instance'
+require 'ddtrace/profiling/transport/http/api/endpoint'
+
+module Datadog
+  module Profiling
+    module Transport
+      module HTTP
+        # Extensions for HTTP API Spec
+        module API
+          # Default API versions
+          V1 = 'v1'.freeze
+
+          module_function
+
+          def agent_defaults
+            @agent_defaults ||= Datadog::Transport::HTTP::API::Map[
+              V1 => Spec.new do |s|
+                s.profiles = Endpoint.new(
+                  '/profiling/v1/input'.freeze,
+                  Profiling::Encoding::Profile::Protobuf
+                )
+              end
+            ]
+          end
+
+          def api_defaults
+            @api_defaults ||= Datadog::Transport::HTTP::API::Map[
+              V1 => Spec.new do |s|
+                s.profiles = Endpoint.new(
+                  '/v1/input'.freeze,
+                  Profiling::Encoding::Profile::Protobuf
+                )
+              end
+            ]
+          end
+        end
+      end
+    end
+  end
+end

data/lib/ddtrace/profiling/transport/http/api/endpoint.rb

@@ -0,0 +1,100 @@
+require 'ddtrace/ext/profiling'
+require 'ddtrace/utils/compression'
+require 'ddtrace/vendor/multipart-post/multipart/post/composite_read_io'
+
+require 'ddtrace/transport/http/api/endpoint'
+require 'ddtrace/profiling/transport/http/response'
+
+module Datadog
+  module Profiling
+    module Transport
+      module HTTP
+        module API
+          # Datadog API endpoint for profiling
+          class Endpoint < Datadog::Transport::HTTP::API::Endpoint
+            include Datadog::Ext::Profiling::Transport::HTTP
+
+            # These tags are read from the flush object (see below) directly and so we ignore any extra copies that
+            # may come in the tags hash to avoid duplicates.
+            TAGS_TO_IGNORE_IN_TAGS_HASH = %w[service env version].freeze
+            private_constant :TAGS_TO_IGNORE_IN_TAGS_HASH
+
+            attr_reader \
+              :encoder
+
+            def initialize(path, encoder)
+              super(:post, path)
+              @encoder = encoder
+            end
+
+            def call(env, &block)
+              # Build request
+              env.form = build_form(env)
+
+              # Send request
+              http_response = super(env, &block)
+
+              # Build response
+              Profiling::Transport::HTTP::Response.new(http_response)
+            end
+
+            def build_form(env)
+              flush = env.request.parcel.data
+              pprof_file, types = build_pprof(flush)
+
+              form = {
+                # NOTE: Redundant w/ 'runtime-id' tag below; may want to remove this later.
+                FORM_FIELD_RUNTIME_ID => flush.runtime_id,
+                FORM_FIELD_RECORDING_START => flush.start.utc.iso8601,
+                FORM_FIELD_RECORDING_END => flush.finish.utc.iso8601,
+                FORM_FIELD_TAGS => [
+                  "#{FORM_FIELD_TAG_RUNTIME}:#{flush.language}",
+                  "#{FORM_FIELD_TAG_RUNTIME_ID}:#{flush.runtime_id}",
+                  "#{FORM_FIELD_TAG_RUNTIME_ENGINE}:#{flush.runtime_engine}",
+                  "#{FORM_FIELD_TAG_RUNTIME_PLATFORM}:#{flush.runtime_platform}",
+                  "#{FORM_FIELD_TAG_RUNTIME_VERSION}:#{flush.runtime_version}",
+                  "#{FORM_FIELD_TAG_PID}:#{Process.pid}",
+                  "#{FORM_FIELD_TAG_PROFILER_VERSION}:#{flush.profiler_version}",
+                  # NOTE: Redundant w/ 'runtime'; may want to remove this later.
+                  "#{FORM_FIELD_TAG_LANGUAGE}:#{flush.language}",
+                  "#{FORM_FIELD_TAG_HOST}:#{flush.host}",
+                  *flush
+                    .tags
+                    .reject { |tag_key| TAGS_TO_IGNORE_IN_TAGS_HASH.include?(tag_key) }
+                    .map { |tag_key, tag_value| "#{tag_key}:#{tag_value}" }
+                ],
+                FORM_FIELD_DATA => pprof_file,
+                FORM_FIELD_RUNTIME => flush.language,
+                FORM_FIELD_FORMAT => FORM_FIELD_FORMAT_PPROF
+              }
+
+              # Add types
+              form[FORM_FIELD_TYPES] = types.join(',')
+
+              # Optional fields
+              form[FORM_FIELD_TAGS] << "#{FORM_FIELD_TAG_SERVICE}:#{flush.service}" unless flush.service.nil?
+              form[FORM_FIELD_TAGS] << "#{FORM_FIELD_TAG_ENV}:#{flush.env}" unless flush.env.nil?
+              form[FORM_FIELD_TAGS] << "#{FORM_FIELD_TAG_VERSION}:#{flush.version}" unless flush.version.nil?
+
+              form
+            end
+
+            def build_pprof(flush)
+              pprof = encoder.encode(flush)
+
+              # Wrap pprof as a gzipped file
+              gzipped_data = Datadog::Utils::Compression.gzip(pprof.data)
+              pprof_file = Datadog::Vendor::Multipart::Post::UploadIO.new(
+                StringIO.new(gzipped_data),
+                HEADER_CONTENT_TYPE_OCTET_STREAM,
+                PPROF_DEFAULT_FILENAME
+              )
+
+              [pprof_file, [FORM_FIELD_TYPES_AUTO]]
+            end
+          end
+        end
+      end
+    end
+  end
+end