ddtrace 1.0.0 → 1.9.0

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

@@ -1,6 +1,6 @@
 # typed: strict
 
-require 'datadog/core/utils/string_table'
+require_relative '../../core/utils/string_table'
 
 module Datadog
   module Profiling

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

@@ -1,11 +1,11 @@
 # typed: false
 
-require 'datadog/profiling/pprof/payload'
-require 'datadog/profiling/pprof/message_set'
-require 'datadog/profiling/pprof/builder'
+require_relative 'payload'
+require_relative 'message_set'
+require_relative 'builder'
 
-require 'datadog/profiling/events/stack'
-require 'datadog/profiling/pprof/stack_sample'
+require_relative '../events/stack'
+require_relative 'stack_sample'
 
 module Datadog
   module Profiling

data/lib/datadog/profiling/preload.rb

@@ -1,5 +1,5 @@
 # typed: strict
 
-require 'ddtrace'
+require_relative '../../ddtrace'
 
 Datadog::Profiling.start_if_enabled

data/lib/datadog/profiling/profiler.rb

@@ -4,6 +4,8 @@ module Datadog
   module Profiling
     # Profiling entry point, which coordinates collectors and a scheduler
     class Profiler
+      include Datadog::Core::Utils::Forking
+
       attr_reader \
         :collectors,
         :scheduler
@@ -14,6 +16,11 @@ module Datadog
       end
 
       def start
+        after_fork! do
+          collectors.each(&:reset_after_fork)
+          scheduler.reset_after_fork
+        end
+
         collectors.each(&:start)
         scheduler.start
       end

data/lib/datadog/profiling/scheduler.rb

@@ -1,42 +1,41 @@
 # typed: true
 
-require 'datadog/core/utils/time'
+require_relative '../core/utils/time'
 
-require 'datadog/core/worker'
-require 'datadog/core/workers/polling'
+require_relative '../core/worker'
+require_relative '../core/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.
+    # Periodically (every DEFAULT_INTERVAL_SECONDS) takes a profile from the `Exporter` and reports it using the
+    # configured transport. Runs on its own background thread.
     class Scheduler < Core::Worker
       include Core::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
+      private
 
       attr_reader \
-        :exporters,
-        :recorder
+        :exporter,
+        :transport
+
+      public
 
       def initialize(
-        recorder,
-        exporters,
+        exporter:,
+        transport:,
         fork_policy: Core::Workers::Async::Thread::FORK_POLICY_RESTART, # Restart in forks by default
         interval: DEFAULT_INTERVAL_SECONDS,
         enabled: true
       )
-        @recorder = recorder
-        @exporters = [exporters].flatten
+        @exporter = exporter
+        @transport = transport
 
         # Workers::Async::Thread settings
         self.fork_policy = fork_policy
@@ -67,24 +66,21 @@ module Datadog
         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.
+      # application just started but this thread took a bit longer so there's already profiling data in the exporter.
       def loop_wait_before_first_iteration?
         true
       end
 
       def work_pending?
-        !recorder.empty?
+        exporter.can_flush?
+      end
+
+      def reset_after_fork
+        exporter.reset_after_fork
       end
 
       private
@@ -100,21 +96,15 @@ module Datadog
       end
 
       def flush_events
-        # Get events from recorder
-        flush = recorder.flush
+        # Collect data to be exported
+        flush = exporter.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
+        return false unless flush
 
         # 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)
+        # a) it's being shut 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
         #
@@ -127,24 +117,13 @@ module Datadog
           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
+        begin
+          transport.export(flush)
+        rescue StandardError => e
+          Datadog.logger.error("Unable to report profile. Cause: #{e} Location: #{Array(e.backtrace).first}")
         end
 
-        flush
-      end
-
-      def duration_below_threshold?(flush)
-        (flush.finish - flush.start) < PROFILE_DURATION_THRESHOLD_SECONDS
+        true
       end
     end
   end

data/lib/datadog/profiling/stack_recorder.rb

@@ -0,0 +1,79 @@
+# typed: false
+
+module Datadog
+  module Profiling
+    # Stores stack samples in a native libdatadog data structure and expose Ruby-level serialization APIs
+    # Note that `record_sample` is only accessible from native code.
+    # Methods prefixed with _native_ are implemented in `stack_recorder.c`
+    class StackRecorder
+      def initialize
+        # This mutex works in addition to the fancy C-level mutexes we have in the native side (see the docs there).
+        # It prevents multiple Ruby threads calling serialize at the same time -- something like
+        # `10.times { Thread.new { stack_recorder.serialize } }`.
+        # This isn't something we expect to happen normally, but because it would break the assumptions of the
+        # C-level mutexes (that there is a single serializer thread), we add it here as an extra safeguard against it
+        # accidentally happening.
+        @no_concurrent_synchronize_mutex = Mutex.new
+      end
+
+      def serialize
+        status, result = @no_concurrent_synchronize_mutex.synchronize { self.class._native_serialize(self) }
+
+        if status == :ok
+          start, finish, encoded_pprof = result
+
+          Datadog.logger.debug { "Encoded profile covering #{start.iso8601} to #{finish.iso8601}" }
+
+          [start, finish, encoded_pprof]
+        else
+          error_message = result
+
+          Datadog.logger.error("Failed to serialize profiling data: #{error_message}")
+
+          nil
+        end
+      end
+
+      def serialize!
+        status, result = @no_concurrent_synchronize_mutex.synchronize { self.class._native_serialize(self) }
+
+        if status == :ok
+          _start, _finish, encoded_pprof = result
+
+          encoded_pprof
+        else
+          error_message = result
+
+          raise("Failed to serialize profiling data: #{error_message}")
+        end
+      end
+
+      def clear
+        status, result = @no_concurrent_synchronize_mutex.synchronize { self.class._native_clear(self) }
+
+        if status == :ok
+          finish_timestamp = result
+
+          Datadog.logger.debug { "Cleared profile at #{finish_timestamp}" }
+
+          finish_timestamp
+        else
+          error_message = result
+
+          Datadog.logger.error("Failed to clear profiling data: #{error_message}")
+
+          nil
+        end
+      end
+
+      # Used only for Ruby 2.2 which doesn't have the native `rb_time_timespec_new` API; called from native code
+      def self.ruby_time_from(timespec_seconds, timespec_nanoseconds)
+        Time.at(0).utc + timespec_seconds + (timespec_nanoseconds.to_r / 1_000_000_000)
+      end
+
+      def reset_after_fork
+        self.class._native_reset_after_fork(self)
+      end
+    end
+  end
+end

data/lib/datadog/profiling/tag_builder.rb

@@ -0,0 +1,53 @@
+# typed: true
+
+require_relative '../core/utils'
+
+module Datadog
+  module Profiling
+    # Builds a hash of default plus user tags to be included in a profile
+    module TagBuilder
+      include Datadog::Profiling::Ext::Transport::HTTP # Tag name constants
+
+      def self.call(
+        settings:,
+        # Unified service tagging
+        env: settings.env,
+        service: settings.service,
+        version: settings.version,
+        # Other metadata
+        host: Core::Environment::Socket.hostname,
+        language: Core::Environment::Identity.lang,
+        pid: Process.pid.to_s,
+        profiler_version: Core::Environment::Identity.tracer_version,
+        runtime_engine: Core::Environment::Identity.lang_engine,
+        runtime_id: Core::Environment::Identity.id,
+        runtime_platform: Core::Environment::Identity.lang_platform,
+        runtime_version: Core::Environment::Identity.lang_version,
+        # User-provided tags
+        user_tags: settings.tags
+      )
+        tags = {
+          # When changing or adding these, make sure they are kept in sync with
+          # https://docs.google.com/spreadsheets/d/1LOGMf4c4Avbtn36uZ2SWvhIGKRPLM1BoWkUP4JYj7hA/ (Datadog internal link)
+          FORM_FIELD_TAG_HOST => host,
+          FORM_FIELD_TAG_LANGUAGE => language,
+          FORM_FIELD_TAG_PID => pid,
+          FORM_FIELD_TAG_PROFILER_VERSION => profiler_version,
+          FORM_FIELD_TAG_RUNTIME => language, # This is known to be repeated from language, above
+          FORM_FIELD_TAG_RUNTIME_ENGINE => runtime_engine,
+          FORM_FIELD_TAG_RUNTIME_ID => runtime_id,
+          FORM_FIELD_TAG_RUNTIME_PLATFORM => runtime_platform,
+          FORM_FIELD_TAG_RUNTIME_VERSION => runtime_version,
+        }
+        tags[FORM_FIELD_TAG_ENV] = env if env
+        tags[FORM_FIELD_TAG_SERVICE] = service if service
+        tags[FORM_FIELD_TAG_VERSION] = version if version
+
+        # Make sure everything is an utf-8 string, to avoid encoding issues in native code/libddprof/further downstream
+        user_tags.merge(tags).map do |key, value|
+          [Datadog::Core::Utils.utf8_encode(key), Datadog::Core::Utils.utf8_encode(value)]
+        end.to_h
+      end
+    end
+  end
+end

data/lib/datadog/profiling/tasks/exec.rb

@@ -38,10 +38,10 @@ module Datadog
         def exec_with_error_handling(args)
           Kernel.exec(*args)
         rescue Errno::ENOENT => e
-          Kernel.warn "ddtracerb exec failed: #{e.message} (command was '#{args.join(' ')}')"
+          Kernel.warn "ddtracerb exec failed: #{e.class.name} #{e.message} (command was '#{args.join(' ')}')"
           Kernel.exit 127
         rescue Errno::EACCES, Errno::ENOEXEC => e
-          Kernel.warn "ddtracerb exec failed: #{e.message} (command was '#{args.join(' ')}')"
+          Kernel.warn "ddtracerb exec failed: #{e.class.name} #{e.message} (command was '#{args.join(' ')}')"
           Kernel.exit 126
         end
       end

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

@@ -1,8 +1,7 @@
 # typed: false
 
-require 'datadog/core/utils/only_once'
-require 'datadog/profiling'
-require 'datadog/profiling/ext/forking'
+require_relative '../../core/utils/only_once'
+require_relative '../ext/forking'
 
 module Datadog
   module Profiling
@@ -19,7 +18,8 @@ module Datadog
               setup_at_fork_hooks
             rescue StandardError, ScriptError => e
               Datadog.logger.warn do
-                "Profiler extensions unavailable. Cause: #{e.message} Location: #{Array(e.backtrace).first}"
+                "Profiler extensions unavailable. Cause: #{e.class.name} #{e.message} " \
+                "Location: #{Array(e.backtrace).first}"
               end
             end
           end
@@ -35,7 +35,8 @@ module Datadog
           end
         rescue StandardError, ScriptError => e
           Datadog.logger.warn do
-            "Profiler forking extensions unavailable. Cause: #{e.message} Location: #{Array(e.backtrace).first}"
+            "Profiler forking extensions unavailable. Cause: #{e.class.name} #{e.message} " \
+            "Location: #{Array(e.backtrace).first}"
           end
         end
 
@@ -54,18 +55,12 @@ module Datadog
           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
                 Profiling.start_if_enabled
               rescue StandardError => e
                 Datadog.logger.warn do
-                  "Error during post-fork hooks. Cause: #{e.message} Location: #{Array(e.backtrace).first}"
+                  "Error during post-fork hooks. Cause: #{e.class.name} #{e.message} " \
+                  "Location: #{Array(e.backtrace).first}"
                 end
               end
             end

data/lib/datadog/profiling/trace_identifiers/ddtrace.rb

@@ -2,8 +2,8 @@
 
 # typed: true
 
-require 'datadog/tracing'
-require 'datadog/tracing/metadata/ext'
+require_relative '../../tracing'
+require_relative '../../tracing/metadata/ext'
 
 module Datadog
   module Profiling

data/lib/datadog/profiling/trace_identifiers/helper.rb

@@ -2,7 +2,7 @@
 
 # typed: true
 
-require 'datadog/profiling/trace_identifiers/ddtrace'
+require_relative 'ddtrace'
 
 module Datadog
   module Profiling

data/lib/datadog/profiling.rb

@@ -1,18 +1,15 @@
 # typed: true
 
-require 'datadog/core'
-require 'datadog/core/environment/variable_helpers'
-require 'datadog/core/utils/only_once'
+require_relative 'core'
+require_relative 'core/environment/variable_helpers'
+require_relative 'core/utils/only_once'
 
 module Datadog
   # Contains profiler for generating stack profiles, etc.
-  module Profiling # rubocop:disable Metrics/ModuleLength
+  module Profiling
     GOOGLE_PROTOBUF_MINIMUM_VERSION = Gem::Version.new('3.0')
     private_constant :GOOGLE_PROTOBUF_MINIMUM_VERSION
 
-    SKIPPED_NATIVE_EXTENSION_ONLY_ONCE = Core::Utils::OnlyOnce.new
-    private_constant :SKIPPED_NATIVE_EXTENSION_ONLY_ONCE
-
     def self.supported?
       unsupported_reason.nil?
     end
@@ -21,7 +18,7 @@ module Datadog
       # NOTE: Only the first matching reason is returned, so try to keep a nice order on reasons -- e.g. tell users
       # first that they can't use this on JRuby before telling them that they are missing protobuf
 
-      ruby_engine_unsupported? ||
+      native_library_compilation_skipped? ||
         native_library_failed_to_load? ||
         protobuf_gem_unavailable? ||
         protobuf_version_unsupported? ||
@@ -44,8 +41,24 @@ module Datadog
       !!profiler
     end
 
-    private_class_method def self.ruby_engine_unsupported?
-      'JRuby is not supported' if RUBY_ENGINE == 'jruby'
+    private_class_method def self.native_library_compilation_skipped?
+      skipped_reason = try_reading_skipped_reason_file
+
+      "Your ddtrace installation is missing support for the Continuous Profiler because #{skipped_reason}" if skipped_reason
+    end
+
+    private_class_method def self.try_reading_skipped_reason_file(file_api = File)
+      # This file, if it exists, is recorded by extconf.rb during compilation of the native extension
+      skipped_reason_file = "#{__dir__}/../../ext/ddtrace_profiling_native_extension/skipped_reason.txt"
+
+      begin
+        return unless file_api.exist?(skipped_reason_file)
+
+        contents = file_api.read(skipped_reason_file).strip
+        contents unless contents.empty?
+      rescue StandardError
+        # Do nothing
+      end
     end
 
     private_class_method def self.protobuf_gem_unavailable?
@@ -93,12 +106,11 @@ module Datadog
         # In the future it'd be nice to shuffle the logger startup to happen first to avoid this special case.
         Kernel.warn(
           '[DDTRACE] Error while loading google-protobuf gem. ' \
-          "Cause: '#{e.message}' Location: '#{Array(e.backtrace).first}'. " \
+          "Cause: '#{e.class.name} #{e.message}' Location: '#{Array(e.backtrace).first}'. " \
           'This can happen when google-protobuf is missing its native components. ' \
           'To fix this, try removing and reinstalling the gem, forcing it to recompile the components: ' \
           '`gem uninstall google-protobuf -a; BUNDLE_FORCE_RUBY_PLATFORM=true bundle install`. ' \
-          'If the error persists, please contact support via <https://docs.datadoghq.com/help/> or ' \
-          'file a bug at <https://github.com/DataDog/dd-trace-rb/blob/master/CONTRIBUTING.md#found-a-bug>.'
+          'If the error persists, please contact Datadog support at <https://docs.datadoghq.com/help/>.'
         )
         @protobuf_loaded = false
       end
@@ -110,31 +122,18 @@ module Datadog
       unless success
         if exception
           'There was an error loading the profiling native extension due to ' \
-          "'#{exception.message}' at '#{exception.backtrace.first}'"
+          "'#{exception.class.name} #{exception.message}' at '#{exception.backtrace.first}'"
         else
           'The profiling native extension did not load correctly. ' \
-          'If the error persists, please contact support via <https://docs.datadoghq.com/help/> or ' \
-          'file a bug at <https://github.com/DataDog/dd-trace-rb/blob/master/CONTRIBUTING.md#found-a-bug>.'
+          'For help solving this issue, please contact Datadog support at <https://docs.datadoghq.com/help/>.' \
         end
       end
     end
 
     private_class_method def self.try_loading_native_library
-      if Core::Environment::VariableHelpers.env_to_bool('DD_PROFILING_NO_EXTENSION', false)
-        SKIPPED_NATIVE_EXTENSION_ONLY_ONCE.run do
-          Kernel.warn(
-            '[DDTRACE] Skipped loading of profiling native extension due to DD_PROFILING_NO_EXTENSION environment ' \
-            'variable being set. ' \
-            'This option is experimental and will lead to the profiler not working in future releases. ' \
-            'If you needed to use this, please tell us why on <https://github.com/DataDog/dd-trace-rb/issues/new>.'
-          )
-        end
-
-        return [true, nil]
-      end
-
       begin
-        require "ddtrace_profiling_native_extension.#{RUBY_VERSION}_#{RUBY_PLATFORM}"
+        require_relative 'profiling/load_native_extension'
+
         success =
           defined?(Profiling::NativeExtension) && Profiling::NativeExtension.send(:native_working?)
         [success, nil]
@@ -146,23 +145,29 @@ module Datadog
     private_class_method def self.load_profiling
       return false unless supported?
 
-      require 'datadog/profiling/ext/forking'
-      require 'datadog/profiling/collectors/code_provenance'
-      require 'datadog/profiling/collectors/stack'
-      require 'datadog/profiling/exporter'
-      require 'datadog/profiling/recorder'
-      require 'datadog/profiling/scheduler'
-      require 'datadog/profiling/tasks/setup'
-      require 'datadog/profiling/transport/io'
-      require 'datadog/profiling/transport/http'
-      require 'datadog/profiling/profiler'
-      require 'datadog/profiling/native_extension'
-      require 'datadog/profiling/trace_identifiers/helper'
-      require 'datadog/profiling/pprof/pprof_pb'
+      require_relative 'profiling/ext/forking'
+      require_relative 'profiling/collectors/code_provenance'
+      require_relative 'profiling/collectors/cpu_and_wall_time'
+      require_relative 'profiling/collectors/cpu_and_wall_time_worker'
+      require_relative 'profiling/collectors/dynamic_sampling_rate'
+      require_relative 'profiling/collectors/idle_sampling_helper'
+      require_relative 'profiling/collectors/old_stack'
+      require_relative 'profiling/collectors/stack'
+      require_relative 'profiling/stack_recorder'
+      require_relative 'profiling/old_recorder'
+      require_relative 'profiling/exporter'
+      require_relative 'profiling/scheduler'
+      require_relative 'profiling/tasks/setup'
+      require_relative 'profiling/profiler'
+      require_relative 'profiling/native_extension'
+      require_relative 'profiling/trace_identifiers/helper'
+      require_relative 'profiling/pprof/pprof_pb'
+      require_relative 'profiling/tag_builder'
+      require_relative 'profiling/http_transport'
 
       true
     end
 
-    load_profiling if supported?
+    load_profiling
   end
 end

data/lib/datadog/tracing/analytics.rb

@@ -1,6 +1,6 @@
 # typed: true
 
-require 'datadog/tracing/metadata/ext'
+require_relative 'metadata/ext'
 
 module Datadog
   module Tracing

data/lib/datadog/tracing/buffer.rb

@@ -1,10 +1,10 @@
 # typed: true
 
-require 'datadog/core'
-require 'datadog/core/environment/ext'
-require 'datadog/core/buffer/thread_safe'
-require 'datadog/core/buffer/cruby'
-require 'datadog/core/diagnostics/health'
+require_relative '../core'
+require_relative '../core/environment/ext'
+require_relative '../core/buffer/thread_safe'
+require_relative '../core/buffer/cruby'
+require_relative '../core/diagnostics/health'
 
 module Datadog
   module Tracing
@@ -58,7 +58,9 @@ module Datadog
 
         @buffer_spans += trace.length
       rescue StandardError => e
-        Datadog.logger.debug("Failed to measure queue accept. Cause: #{e.message} Source: #{Array(e.backtrace).first}")
+        Datadog.logger.debug(
+          "Failed to measure queue accept. Cause: #{e.class.name} #{e.message} Source: #{Array(e.backtrace).first}"
+        )
       end
 
       def measure_drop(trace)
@@ -66,7 +68,9 @@ module Datadog
 
         @buffer_spans -= trace.length
       rescue StandardError => e
-        Datadog.logger.debug("Failed to measure queue drop. Cause: #{e.message} Source: #{Array(e.backtrace).first}")
+        Datadog.logger.debug(
+          "Failed to measure queue drop. Cause: #{e.class.name} #{e.message} Source: #{Array(e.backtrace).first}"
+        )
       end
 
       def measure_pop(traces)
@@ -89,7 +93,9 @@ module Datadog
         @buffer_dropped = 0
         @buffer_spans = 0
       rescue StandardError => e
-        Datadog.logger.debug("Failed to measure queue. Cause: #{e.message} Source: #{Array(e.backtrace).first}")
+        Datadog.logger.debug(
+          "Failed to measure queue. Cause: #{e.class.name} #{e.message} Source: #{Array(e.backtrace).first}"
+        )
       end
     end