ddtrace 0.54.2 → 1.10.1

data/lib/datadog/profiling/stack_recorder.rb

@@ -0,0 +1,74 @@
+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(cpu_time_enabled:, alloc_samples_enabled:)
+        # 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
+
+        self.class._native_initialize(self, cpu_time_enabled, alloc_samples_enabled)
+      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
+
+      def reset_after_fork
+        self.class._native_reset_after_fork(self)
+      end
+    end
+  end
+end

data/lib/datadog/profiling/tag_builder.rb

@@ -0,0 +1,51 @@
+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

@@ -0,0 +1,48 @@
+module Datadog
+  module Profiling
+    module Tasks
+      # Wraps command with Datadog tracing
+      class Exec
+        attr_reader :args
+
+        def initialize(args)
+          @args = args
+        end
+
+        def run
+          set_rubyopt!
+          exec_with_error_handling(args)
+        end
+
+        def rubyopts
+          [
+            '-rdatadog/profiling/preload'
+          ]
+        end
+
+        private
+
+        def set_rubyopt!
+          existing_rubyopt = ENV['RUBYOPT']
+
+          ENV['RUBYOPT'] = existing_rubyopt ? "#{existing_rubyopt} #{rubyopts.join(' ')}" : rubyopts.join(' ')
+        end
+
+        # If there's an error here, rather than throwing a cryptic stack trace, let's instead have clearer messages, and
+        # follow the same status codes as the shell uses
+        # See also:
+        # * https://www.gnu.org/software/bash/manual/html_node/Exit-Status.html
+        # * https://github.com/rubygems/rubygems/blob/dd93966cac224532035deda533cba2685dfa30cc/bundler/lib/bundler/cli/exec.rb#L45
+        def exec_with_error_handling(args)
+          Kernel.exec(*args)
+        rescue Errno::ENOENT => e
+          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.class.name} #{e.message} (command was '#{args.join(' ')}')"
+          Kernel.exit 126
+        end
+      end
+    end
+  end
+end

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

@@ -0,0 +1,16 @@
+module Datadog
+  module Profiling
+    module Tasks
+      # Prints help message for usage of `ddtrace`
+      class Help
+        def run
+          puts %(
+  Usage: ddtracerb [command] [arguments]
+    exec [command]: Executes command with tracing & profiling preloaded.
+    help:           Prints this help message.
+          )
+        end
+      end
+    end
+  end
+end

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

@@ -0,0 +1,84 @@
+require_relative '../../core/utils/only_once'
+require_relative '../ext/forking'
+
+module Datadog
+  module Profiling
+    module Tasks
+      # Takes care of loading our extensions/monkey patches to handle fork() and validating if CPU-time profiling is usable
+      class Setup
+        ACTIVATE_EXTENSIONS_ONLY_ONCE = Core::Utils::OnlyOnce.new
+
+        def run
+          ACTIVATE_EXTENSIONS_ONLY_ONCE.run do
+            begin
+              check_if_cpu_time_profiling_is_supported
+              activate_forking_extensions
+              setup_at_fork_hooks
+            rescue StandardError, ScriptError => e
+              Datadog.logger.warn do
+                "Profiler extensions unavailable. Cause: #{e.class.name} #{e.message} " \
+                "Location: #{Array(e.backtrace).first}"
+              end
+            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.class.name} #{e.message} " \
+            "Location: #{Array(e.backtrace).first}"
+          end
+        end
+
+        def check_if_cpu_time_profiling_is_supported
+          unsupported = cpu_time_profiling_unsupported_reason
+
+          if unsupported
+            Datadog.logger.info do
+              'CPU time profiling skipped because native CPU time is not supported: ' \
+              "#{unsupported}. Profiles containing 'Wall time' data will still be reported."
+            end
+          end
+        end
+
+        def setup_at_fork_hooks
+          if Process.respond_to?(:at_fork)
+            Process.at_fork(:child) do
+              begin
+                # Restart profiler, if enabled
+                Profiling.start_if_enabled
+              rescue StandardError => e
+                Datadog.logger.warn do
+                  "Error during post-fork hooks. Cause: #{e.class.name} #{e.message} " \
+                  "Location: #{Array(e.backtrace).first}"
+                end
+              end
+            end
+          end
+        end
+
+        def cpu_time_profiling_unsupported_reason
+          # NOTE: Only the first matching reason is returned, so try to keep a nice order on reasons
+
+          if RUBY_ENGINE == 'jruby'
+            'JRuby is not supported'
+          elsif RUBY_PLATFORM.include?('darwin')
+            'Feature requires Linux; macOS is not supported'
+          elsif RUBY_PLATFORM =~ /(mswin|mingw)/
+            'Feature requires Linux; Windows is not supported'
+          elsif !RUBY_PLATFORM.include?('linux')
+            "Feature requires Linux; #{RUBY_PLATFORM} is not supported"
+          end
+        end
+      end
+    end
+  end
+end

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

@@ -0,0 +1,43 @@
+# frozen_string_literal: true
+
+require_relative '../../tracing'
+require_relative '../../tracing/metadata/ext'
+
+module Datadog
+  module Profiling
+    module TraceIdentifiers
+      # Used by Datadog::Profiling::TraceIdentifiers::Helper to get the trace identifiers (root span id and span id)
+      # for a given thread, if there is an active trace for that thread in the supplied tracer object.
+      class Ddtrace
+        def initialize(tracer:)
+          @tracer = (tracer if tracer.respond_to?(:active_trace))
+        end
+
+        def trace_identifiers_for(thread)
+          return unless @tracer
+
+          trace = @tracer.active_trace(thread)
+          return unless trace
+
+          root_span = trace.send(:root_span)
+          span = trace.active_span
+          return unless span && root_span
+
+          root_span_id = root_span.id || 0
+          span_id = span.id || 0
+
+          [root_span_id, span_id, maybe_extract_resource(trace, root_span)] if root_span_id != 0 && span_id != 0
+        end
+
+        private
+
+        # NOTE: Currently we're only interested in HTTP service endpoints. Over time, this list may be expanded.
+        # Resources MUST NOT include personal identifiable information (PII); this should not be the case with
+        # ddtrace integrations, but worth mentioning just in case :)
+        def maybe_extract_resource(trace, root_span)
+          trace.resource if root_span.span_type == Tracing::Metadata::Ext::HTTP::TYPE_INBOUND
+        end
+      end
+    end
+  end
+end

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

@@ -0,0 +1,45 @@
+# frozen_string_literal: true
+
+require_relative 'ddtrace'
+
+module Datadog
+  module Profiling
+    module TraceIdentifiers
+      # Helper used to retrieve the trace identifiers (root span id and span id) for a given thread,
+      # if there is an active trace for that thread for the supported tracing APIs.
+      #
+      # This data is used to connect profiles to the traces -- samples in a profile will be tagged with this data and
+      # the profile can be filtered down to look at only the samples for a given trace.
+      class Helper
+        DEFAULT_SUPPORTED_APIS = [
+          Profiling::TraceIdentifiers::Ddtrace
+        ].freeze
+        private_constant :DEFAULT_SUPPORTED_APIS
+
+        def initialize(
+          tracer:,
+          # If this is disabled, the helper will strip the optional trace_resource_container even if provided by the api
+          endpoint_collection_enabled:,
+          supported_apis: DEFAULT_SUPPORTED_APIS.map { |api| api.new(tracer: tracer) }
+        )
+          @endpoint_collection_enabled = endpoint_collection_enabled
+          @supported_apis = supported_apis
+        end
+
+        # Expected output of the #trace_identifiers_for
+        # duck type is [root_span_id, span_id, (optional trace_resource_container)]
+        def trace_identifiers_for(thread)
+          @supported_apis.each do |api|
+            trace_identifiers = api.trace_identifiers_for(thread)
+
+            if trace_identifiers
+              return @endpoint_collection_enabled ? trace_identifiers : trace_identifiers[0..1]
+            end
+          end
+
+          nil
+        end
+      end
+    end
+  end
+end

data/lib/datadog/profiling.rb

@@ -0,0 +1,210 @@
+require_relative 'core'
+require_relative 'core/environment/variable_helpers'
+require_relative 'core/utils/only_once'
+
+module Datadog
+  # Datadog Continuous Profiler implementation: https://docs.datadoghq.com/profiler/
+  module Profiling
+    GOOGLE_PROTOBUF_MINIMUM_VERSION = Gem::Version.new('3.0')
+    private_constant :GOOGLE_PROTOBUF_MINIMUM_VERSION
+
+    def self.supported?
+      unsupported_reason.nil?
+    end
+
+    def self.unsupported_reason
+      # 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
+
+      native_library_compilation_skipped? ||
+        native_library_failed_to_load? ||
+        protobuf_gem_unavailable? ||
+        protobuf_version_unsupported? ||
+        protobuf_failed_to_load?
+    end
+
+    # Starts the profiler, if the profiler is supported by in
+    # this runtime environment and if the profiler has been enabled
+    # in configuration.
+    #
+    # @return [Boolean] `true` if the profiler has successfully started, otherwise `false`.
+    # @public_api
+    def self.start_if_enabled
+      # If the profiler was not previously touched, getting the profiler instance triggers start as a side-effect
+      # otherwise we get nil
+      profiler = Datadog.send(:components).profiler
+      # ...but we still try to start it BECAUSE if the process forks, the profiler will exist but may
+      # not yet have been started in the fork
+      profiler.start if profiler
+      !!profiler
+    end
+
+    # Returns an ever-increasing counter of the number of allocations observed by the profiler in this thread.
+    #
+    # Note 1: This counter may not start from zero on new threads. It should only be used to measure how many
+    # allocations have happened between two calls to this API:
+    # ```
+    # allocations_before = Datadog::Profiling.allocation_count
+    # do_some_work()
+    # allocations_after = Datadog::Profiling.allocation_count
+    # puts "Allocations during do_some_work: #{allocations_after - allocations_before}"
+    # ```
+    # (This is similar to some OS-based time representations.)
+    #
+    # Note 2: All fibers in the same thread will share the same counter values.
+    #
+    # Only available when the profiler is running, the new CPU Profiling 2.0 profiler is in use, and allocation-related
+    # features are not disabled via configuration.
+    # For instructions on enabling CPU Profiling 2.0 see the ddtrace release notes.
+    #
+    # @return [Integer] number of allocations observed in the current thread.
+    # @return [nil] when not available.
+    # @public_api
+    def self.allocation_count
+      # This no-op implementation is used when profiling failed to load.
+      # It gets replaced inside #replace_noop_allocation_count.
+      nil
+    end
+
+    private_class_method def self.replace_noop_allocation_count
+      def self.allocation_count # rubocop:disable Lint/DuplicateMethods, Lint/NestedMethodDefinition (On purpose!)
+        Datadog::Profiling::Collectors::CpuAndWallTimeWorker._native_allocation_count
+      end
+    end
+
+    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?
+      # NOTE: On environments where protobuf is already loaded, we skip the check. This allows us to support environments
+      # where no Gem.loaded_version is NOT available but customers are able to load protobuf; see for instance
+      # https://github.com/teamcapybara/capybara/commit/caf3bcd7664f4f2691d0ca9ef3be9a2a954fecfb
+      if !protobuf_already_loaded? && Gem.loaded_specs['google-protobuf'].nil?
+        "Missing google-protobuf dependency; please add `gem 'google-protobuf', '~> 3.0'` to your Gemfile or gems.rb file"
+      end
+    end
+
+    private_class_method def self.protobuf_version_unsupported?
+      # See above for why we skip the check when protobuf is already loaded; note that when protobuf was already loaded
+      # we skip the version check to avoid the call to Gem.loaded_specs. Unfortunately, protobuf does not seem to
+      # expose the gem version constant elsewhere, so in that setup we are not able to check the version.
+      if !protobuf_already_loaded? && Gem.loaded_specs['google-protobuf'].version < GOOGLE_PROTOBUF_MINIMUM_VERSION
+        'Your google-protobuf is too old; ensure that you have google-protobuf >= 3.0 by ' \
+        "adding `gem 'google-protobuf', '~> 3.0'` to your Gemfile or gems.rb file"
+      end
+    end
+
+    private_class_method def self.protobuf_already_loaded?
+      defined?(::Google::Protobuf) && !defined?(::Protobuf)
+    end
+
+    private_class_method def self.protobuf_failed_to_load?
+      unless protobuf_loaded_successfully?
+        'There was an error loading the google-protobuf library; see previous warning message for details'
+      end
+    end
+
+    # The `google-protobuf` gem depends on a native component, and its creators helpfully tried to provide precompiled
+    # versions of this extension on rubygems.org.
+    #
+    # Unfortunately, for a long time, the supported Ruby versions metadata on these precompiled versions of the extension
+    # was not correctly set. (This is fixed in newer versions -- but not all Ruby versions we want to support can use
+    # these.)
+    #
+    # Thus, the gem can still be installed, but can be in a broken state. To avoid breaking customer applications, we
+    # use this helper to load it and gracefully handle failures.
+    private_class_method def self.protobuf_loaded_successfully?
+      return @protobuf_loaded if defined?(@protobuf_loaded)
+
+      begin
+        require 'google/protobuf'
+        @protobuf_loaded = true
+      rescue LoadError => e
+        # NOTE: We use Kernel#warn here because this code gets run BEFORE Datadog.logger is actually set up.
+        # 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.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 Datadog support at <https://docs.datadoghq.com/help/>.'
+        )
+        @protobuf_loaded = false
+      end
+    end
+
+    private_class_method def self.native_library_failed_to_load?
+      success, exception = try_loading_native_library
+
+      unless success
+        if exception
+          'There was an error loading the profiling native extension due to ' \
+          "'#{exception.class.name} #{exception.message}' at '#{exception.backtrace.first}'"
+        else
+          'The profiling native extension did not load correctly. ' \
+          '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
+      begin
+        require_relative 'profiling/load_native_extension'
+
+        success =
+          defined?(Profiling::NativeExtension) && Profiling::NativeExtension.send(:native_working?)
+        [success, nil]
+      rescue StandardError, LoadError => e
+        [false, e]
+      end
+    end
+
+    private_class_method def self.load_profiling
+      return false unless supported?
+
+      require_relative 'profiling/ext/forking'
+      require_relative 'profiling/collectors/code_provenance'
+      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/collectors/thread_context'
+      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'
+
+      replace_noop_allocation_count
+
+      true
+    end
+
+    load_profiling
+  end
+end

data/lib/datadog/tracing/analytics.rb

@@ -0,0 +1,23 @@
+require_relative 'metadata/ext'
+
+module Datadog
+  module Tracing
+    # Defines analytics behavior
+    module Analytics
+      class << self
+        def set_sample_rate(span_op, sample_rate)
+          return if span_op.nil? || !sample_rate.is_a?(Numeric)
+
+          span_op.set_metric(Metadata::Ext::Analytics::TAG_SAMPLE_RATE, sample_rate)
+        end
+
+        def set_measured(span_op, value = true)
+          return if span_op.nil?
+
+          value = value == true || value == 1 ? 1 : 0 # rubocop:disable Style/MultipleComparison
+          span_op.set_metric(Metadata::Ext::Analytics::TAG_MEASURED, value)
+        end
+      end
+    end
+  end
+end