ddtrace 0.54.2 → 1.18.0

data/lib/ddtrace/profiling/pprof/pprof_pb.rb

@@ -1,82 +0,0 @@
-# typed: false
-# Generated by the protocol buffer compiler.  DO NOT EDIT!
-# source: lib/ddtrace/profiling/pprof/pprof.proto
-
-require 'google/protobuf'
-
-Google::Protobuf::DescriptorPool.generated_pool.build do
-  add_message "perftools.profiles.Profile" do
-    repeated :sample_type, :message, 1, "perftools.profiles.ValueType"
-    repeated :sample, :message, 2, "perftools.profiles.Sample"
-    repeated :mapping, :message, 3, "perftools.profiles.Mapping"
-    repeated :location, :message, 4, "perftools.profiles.Location"
-    repeated :function, :message, 5, "perftools.profiles.Function"
-    repeated :string_table, :string, 6
-    optional :drop_frames, :int64, 7
-    optional :keep_frames, :int64, 8
-    optional :time_nanos, :int64, 9
-    optional :duration_nanos, :int64, 10
-    optional :period_type, :message, 11, "perftools.profiles.ValueType"
-    optional :period, :int64, 12
-    repeated :comment, :int64, 13
-    optional :default_sample_type, :int64, 14
-  end
-  add_message "perftools.profiles.ValueType" do
-    optional :type, :int64, 1
-    optional :unit, :int64, 2
-  end
-  add_message "perftools.profiles.Sample" do
-    repeated :location_id, :uint64, 1
-    repeated :value, :int64, 2
-    repeated :label, :message, 3, "perftools.profiles.Label"
-  end
-  add_message "perftools.profiles.Label" do
-    optional :key, :int64, 1
-    optional :str, :int64, 2
-    optional :num, :int64, 3
-    optional :num_unit, :int64, 4
-  end
-  add_message "perftools.profiles.Mapping" do
-    optional :id, :uint64, 1
-    optional :memory_start, :uint64, 2
-    optional :memory_limit, :uint64, 3
-    optional :file_offset, :uint64, 4
-    optional :filename, :int64, 5
-    optional :build_id, :int64, 6
-    optional :has_functions, :bool, 7
-    optional :has_filenames, :bool, 8
-    optional :has_line_numbers, :bool, 9
-    optional :has_inline_frames, :bool, 10
-  end
-  add_message "perftools.profiles.Location" do
-    optional :id, :uint64, 1
-    optional :mapping_id, :uint64, 2
-    optional :address, :uint64, 3
-    repeated :line, :message, 4, "perftools.profiles.Line"
-    optional :is_folded, :bool, 5
-  end
-  add_message "perftools.profiles.Line" do
-    optional :function_id, :uint64, 1
-    optional :line, :int64, 2
-  end
-  add_message "perftools.profiles.Function" do
-    optional :id, :uint64, 1
-    optional :name, :int64, 2
-    optional :system_name, :int64, 3
-    optional :filename, :int64, 4
-    optional :start_line, :int64, 5
-  end
-end
-
-module Perftools
-  module Profiles
-    Profile = Google::Protobuf::DescriptorPool.generated_pool.lookup("perftools.profiles.Profile").msgclass
-    ValueType = Google::Protobuf::DescriptorPool.generated_pool.lookup("perftools.profiles.ValueType").msgclass
-    Sample = Google::Protobuf::DescriptorPool.generated_pool.lookup("perftools.profiles.Sample").msgclass
-    Label = Google::Protobuf::DescriptorPool.generated_pool.lookup("perftools.profiles.Label").msgclass
-    Mapping = Google::Protobuf::DescriptorPool.generated_pool.lookup("perftools.profiles.Mapping").msgclass
-    Location = Google::Protobuf::DescriptorPool.generated_pool.lookup("perftools.profiles.Location").msgclass
-    Line = Google::Protobuf::DescriptorPool.generated_pool.lookup("perftools.profiles.Line").msgclass
-    Function = Google::Protobuf::DescriptorPool.generated_pool.lookup("perftools.profiles.Function").msgclass
-  end
-end

data/lib/ddtrace/profiling/pprof/stack_sample.rb

@@ -1,140 +0,0 @@
-# typed: true
-require 'ddtrace/ext/profiling'
-require 'ddtrace/profiling/events/stack'
-require 'ddtrace/profiling/pprof/builder'
-require 'ddtrace/profiling/pprof/converter'
-
-module Datadog
-  module Profiling
-    module Pprof
-      # Builds a profile from a StackSample
-      #
-      # NOTE: This class may appear stateless but is in fact stateful; a new instance should be created for every
-      # encoded profile.
-      class StackSample < Converter
-        SAMPLE_TYPES = {
-          cpu_time_ns: [
-            Datadog::Ext::Profiling::Pprof::VALUE_TYPE_CPU,
-            Datadog::Ext::Profiling::Pprof::VALUE_UNIT_NANOSECONDS
-          ],
-          wall_time_ns: [
-            Datadog::Ext::Profiling::Pprof::VALUE_TYPE_WALL,
-            Datadog::Ext::Profiling::Pprof::VALUE_UNIT_NANOSECONDS
-          ]
-        }.freeze
-
-        def self.sample_value_types
-          SAMPLE_TYPES
-        end
-
-        def initialize(*_)
-          super
-
-          @most_recent_trace_samples = {}
-          @processed_unique_stacks = 0
-          @processed_with_trace = 0
-        end
-
-        def add_events!(stack_samples)
-          new_samples = build_samples(stack_samples)
-          builder.samples.concat(new_samples)
-        end
-
-        def stack_sample_group_key(stack_sample)
-          # We want to make sure we have the most recent sample for any trace.
-          # (This is done here to save an iteration over all samples.)
-          update_most_recent_trace_sample(stack_sample)
-
-          stack_sample.hash
-        end
-
-        # Track the most recent sample for each trace (identified by root span id)
-        def update_most_recent_trace_sample(stack_sample)
-          return unless stack_sample.root_span_id && stack_sample.trace_resource
-
-          # Update trace resource with most recent value
-          if (most_recent_trace_sample = @most_recent_trace_samples[stack_sample.root_span_id])
-            if most_recent_trace_sample.timestamp < stack_sample.timestamp
-              @most_recent_trace_samples[stack_sample.root_span_id] = stack_sample
-            end
-          else
-            # Add trace resource
-            @most_recent_trace_samples[stack_sample.root_span_id] = stack_sample
-          end
-        end
-
-        def build_samples(stack_samples)
-          groups = group_events(stack_samples, &method(:stack_sample_group_key))
-          groups.collect do |_group_key, group|
-            @processed_unique_stacks += 1
-            build_sample(group.sample, group.values)
-          end
-        end
-
-        def build_sample(stack_sample, values)
-          locations = builder.build_locations(
-            stack_sample.frames,
-            stack_sample.total_frame_count
-          )
-
-          Perftools::Profiles::Sample.new(
-            location_id: locations.collect { |location| location['id'.freeze] },
-            value: values,
-            label: build_sample_labels(stack_sample)
-          )
-        end
-
-        def build_event_values(stack_sample)
-          no_value = Datadog::Ext::Profiling::Pprof::SAMPLE_VALUE_NO_VALUE
-          values = super(stack_sample)
-          values[sample_value_index(:cpu_time_ns)] = stack_sample.cpu_time_interval_ns || no_value
-          values[sample_value_index(:wall_time_ns)] = stack_sample.wall_time_interval_ns || no_value
-          values
-        end
-
-        def build_sample_labels(stack_sample)
-          labels = [
-            Perftools::Profiles::Label.new(
-              key: builder.string_table.fetch(Datadog::Ext::Profiling::Pprof::LABEL_KEY_THREAD_ID),
-              str: builder.string_table.fetch(stack_sample.thread_id.to_s)
-            )
-          ]
-
-          root_span_id = stack_sample.root_span_id || 0
-          span_id = stack_sample.span_id || 0
-
-          if root_span_id != 0 && span_id != 0
-            @processed_with_trace += 1
-
-            labels << Perftools::Profiles::Label.new(
-              key: builder.string_table.fetch(Datadog::Ext::Profiling::Pprof::LABEL_KEY_LOCAL_ROOT_SPAN_ID),
-              str: builder.string_table.fetch(root_span_id.to_s)
-            )
-
-            labels << Perftools::Profiles::Label.new(
-              key: builder.string_table.fetch(Datadog::Ext::Profiling::Pprof::LABEL_KEY_SPAN_ID),
-              str: builder.string_table.fetch(span_id.to_s)
-            )
-
-            # Use most up-to-date trace resource, if available.
-            # Otherwise, use the trace resource provided.
-            trace_resource = @most_recent_trace_samples.fetch(stack_sample.root_span_id, stack_sample).trace_resource
-
-            if trace_resource && !trace_resource.empty?
-              labels << Perftools::Profiles::Label.new(
-                key: builder.string_table.fetch(Datadog::Ext::Profiling::Pprof::LABEL_KEY_TRACE_ENDPOINT),
-                str: builder.string_table.fetch(trace_resource)
-              )
-            end
-          end
-
-          labels
-        end
-
-        def debug_statistics
-          "unique stacks: #{@processed_unique_stacks}, of which had active traces: #{@processed_with_trace}"
-        end
-      end
-    end
-  end
-end

data/lib/ddtrace/profiling/pprof/string_table.rb

@@ -1,11 +0,0 @@
-# typed: strict
-require 'ddtrace/utils/string_table'
-
-module Datadog
-  module Profiling
-    module Pprof
-      # Tracks strings and returns IDs
-      class StringTable < Utils::StringTable; end
-    end
-  end
-end

data/lib/ddtrace/profiling/pprof/template.rb

@@ -1,119 +0,0 @@
-# typed: false
-require 'ddtrace/profiling/pprof/payload'
-require 'ddtrace/profiling/pprof/message_set'
-require 'ddtrace/profiling/pprof/builder'
-
-require 'ddtrace/profiling/events/stack'
-require 'ddtrace/profiling/pprof/stack_sample'
-
-module Datadog
-  module Profiling
-    module Pprof
-      # Converts a collection of profiling events into a Perftools::Profiles::Profile
-      class Template
-        DEFAULT_MAPPINGS = {
-          Events::StackSample => Pprof::StackSample
-        }.freeze
-
-        attr_reader \
-          :builder,
-          :converters,
-          :sample_type_mappings
-
-        def self.for_event_classes(event_classes)
-          # Build a map of event class --> converter class
-          mappings = event_classes.each_with_object({}) do |event_class, m|
-            converter_class = DEFAULT_MAPPINGS[event_class]
-            raise NoProfilingEventConversionError, event_class unless converter_class
-
-            m[event_class] = converter_class
-          end
-
-          new(mappings)
-        end
-
-        def initialize(mappings)
-          @builder = Builder.new
-          @converters = Hash.new { |_h, event_class| raise NoProfilingEventConversionError, event_class }
-          @sample_type_mappings = Hash.new { |_h, type| raise UnknownSampleTypeMappingError, type }
-
-          # Add default mapping
-          builder.mappings.fetch($PROGRAM_NAME, &builder.method(:build_mapping))
-
-          # Combine all sample types from each converter class
-          types = mappings.values.each_with_object({}) do |converter_class, t|
-            t.merge!(converter_class.sample_value_types)
-          end
-
-          # Build the sample types into sample type objects
-          types.each do |type_name, type_args|
-            index = nil
-
-            sample_type = builder.sample_types.fetch(*type_args) do |id, type, unit|
-              index = id
-              builder.build_value_type(type, unit)
-            end
-
-            # Create mapping between the type and index to which its assigned.
-            # Do this for faster lookup while building profile sample values.
-            sample_type_mappings[type_name] = index || builder.sample_types.messages.index(sample_type)
-          end
-
-          # Freeze them so they can't be modified.
-          # We don't want the number of sample types to vary between samples within the same profile.
-          builder.sample_types.freeze
-          sample_type_mappings.freeze
-
-          # Add converters
-          mappings.each do |event_class, converter_class|
-            converters[event_class] = converter_class.new(builder, sample_type_mappings)
-          end
-
-          converters.freeze
-        end
-
-        def add_events!(event_class, events)
-          converters[event_class].add_events!(events)
-        end
-
-        def debug_statistics
-          converters.values.map(&:debug_statistics).join(', ')
-        end
-
-        def to_pprof(start:, finish:)
-          profile = builder.build_profile(start: start, finish: finish)
-          data = builder.encode_profile(profile)
-          types = sample_type_mappings.keys
-
-          Payload.new(data, types)
-        end
-
-        # Error when an unknown event type is given to be converted
-        class NoProfilingEventConversionError < ArgumentError
-          attr_reader :type
-
-          def initialize(type)
-            @type = type
-          end
-
-          def message
-            "Profiling event type '#{type}' cannot be converted to pprof."
-          end
-        end
-
-        # Error when the mapping of a sample type to value index is unknown
-        class UnknownSampleTypeMappingError < ArgumentError
-          attr_reader :type
-
-          def initialize(type)
-            @type = type
-          end
-
-          def message
-            "Mapping for sample value type '#{type}' is unknown."
-          end
-        end
-      end
-    end
-  end
-end

data/lib/ddtrace/profiling/profiler.rb

@@ -1,31 +0,0 @@
-# typed: true
-module Datadog
-  # Profiling entry point, which coordinates collectors and a scheduler
-  class Profiler
-    attr_reader \
-      :collectors,
-      :scheduler
-
-    def initialize(collectors, scheduler)
-      @collectors = collectors
-      @scheduler = scheduler
-    end
-
-    def start
-      collectors.each(&:start)
-      scheduler.start
-    end
-
-    def shutdown!
-      Datadog.logger.debug('Shutting down profiler')
-
-      collectors.each do |collector|
-        collector.enabled = false
-        collector.stop(true)
-      end
-
-      scheduler.enabled = false
-      scheduler.stop(true)
-    end
-  end
-end

data/lib/ddtrace/profiling/recorder.rb

@@ -1,96 +0,0 @@
-# typed: true
-require 'ddtrace/profiling/buffer'
-require 'ddtrace/profiling/flush'
-
-module Datadog
-  module Profiling
-    # Stores profiling events gathered by `Collector`s
-    class Recorder
-      attr_reader :max_size
-
-      def initialize(event_classes, max_size, last_flush_time: Time.now.utc)
-        @buffers = {}
-        @last_flush_time = last_flush_time
-        @max_size = max_size
-
-        # Add a buffer for each class
-        event_classes.each do |event_class|
-          @buffers[event_class] = Profiling::Buffer.new(max_size)
-        end
-
-        # Event classes can only be added ahead of time
-        @buffers.freeze
-      end
-
-      def [](event_class)
-        @buffers[event_class]
-      end
-
-      def push(events)
-        if events.is_a?(Array)
-          # Push multiple events
-          event_class = events.first.class
-          raise UnknownEventError, event_class unless @buffers.key?(event_class)
-
-          @buffers[event_class].concat(events)
-        else
-          # Push single event
-          event_class = events.class
-          raise UnknownEventError, event_class unless @buffers.key?(event_class)
-
-          @buffers[event_class].push(events)
-        end
-      end
-
-      def flush
-        event_count = 0
-
-        event_groups, start, finish = update_time do
-          @buffers.collect do |event_class, buffer|
-            events = buffer.pop
-            next if events.empty?
-
-            event_count += events.length
-            EventGroup.new(event_class, events)
-          end.compact
-        end
-
-        Flush.new(
-          start,
-          finish,
-          event_groups,
-          event_count
-        )
-      end
-
-      # NOTE: Remember that if the recorder is being accessed by multiple threads, this is an inherently racy operation.
-      def empty?
-        @buffers.values.all?(&:empty?)
-      end
-
-      # Error when event of an unknown type is used with the Recorder
-      class UnknownEventError < StandardError
-        attr_reader :event_class
-
-        def initialize(event_class)
-          @event_class = event_class
-        end
-
-        def message
-          @message ||= "Unknown event class '#{event_class}' for profiling recorder."
-        end
-      end
-
-      private
-
-      def update_time
-        start = @last_flush_time
-        result = yield
-        @last_flush_time = Time.now.utc
-
-        # Return event groups, start time, finish time
-        [result, start, @last_flush_time]
-      end
-    end
-  end
-end

data/lib/ddtrace/profiling/scheduler.rb

@@ -1,150 +0,0 @@
-# typed: true
-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
-      MINIMUM_INTERVAL_SECONDS = 0
-
-      # Profiles with duration less than this will not be reported
-      PROFILE_DURATION_THRESHOLD_SECONDS = 1
-
-      # We sleep for at most this duration seconds before reporting data to avoid multi-process applications all
-      # reporting profiles at the exact same time
-      DEFAULT_FLUSH_JITTER_MAXIMUM_SECONDS = 3
-
-      private_constant :DEFAULT_INTERVAL_SECONDS, :MINIMUM_INTERVAL_SECONDS, :PROFILE_DURATION_THRESHOLD_SECONDS
-
-      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 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, MINIMUM_INTERVAL_SECONDS].max
-      end
-
-      def flush_events
-        # Get events from recorder
-        flush = recorder.flush
-
-        if duration_below_threshold?(flush)
-          Datadog.logger.debug do
-            "Skipped exporting profiling events as profile duration is below minimum (#{flush.event_count} events skipped)"
-          end
-
-          return flush
-        end
-
-        # Sleep for a bit to cause misalignment between profilers in multi-process applications
-        #
-        # When not being run in a loop, it means the scheduler has not been started or was stopped, and thus
-        # a) it's being shutting down (and is trying to report the last profile)
-        # b) it's being run as a one-shot, usually in a test
-        # ...so in those cases we don't sleep
-        #
-        # During PR review (https://github.com/DataDog/dd-trace-rb/pull/1807) we discussed the possible alternative of
-        # just sleeping before starting the scheduler loop. We ended up not going with that option to avoid the first
-        # profile containing up to DEFAULT_INTERVAL_SECONDS + DEFAULT_FLUSH_JITTER_MAXIMUM_SECONDS instead of the
-        # usual DEFAULT_INTERVAL_SECONDS size.
-        if run_loop?
-          jitter_seconds = rand * DEFAULT_FLUSH_JITTER_MAXIMUM_SECONDS # floating point number between (0.0...maximum)
-          sleep(jitter_seconds)
-        end
-
-        # 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: #{Array(e.backtrace).first}"
-              )
-            end
-          end
-        end
-
-        flush
-      end
-
-      def duration_below_threshold?(flush)
-        (flush.finish - flush.start) < PROFILE_DURATION_THRESHOLD_SECONDS
-      end
-    end
-  end
-end