ddtrace 0.49.0 → 0.52.0

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

@@ -1,3 +1,4 @@
+# typed: false
 # Generated by the protocol buffer compiler.  DO NOT EDIT!
 # source: lib/ddtrace/profiling/pprof/pprof.proto
 

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

@@ -1,3 +1,4 @@
+# typed: true
 require 'ddtrace/ext/profiling'
 require 'ddtrace/profiling/events/stack'
 require 'ddtrace/profiling/pprof/builder'
@@ -7,6 +8,9 @@ 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: [
@@ -23,6 +27,13 @@ module Datadog
           SAMPLE_TYPES
         end
 
+        def initialize(*_)
+          super
+
+          @processed_unique_stacks = 0
+          @processed_with_trace_ids = 0
+        end
+
         def add_events!(stack_samples)
           new_samples = build_samples(stack_samples)
           builder.samples.concat(new_samples)
@@ -35,6 +46,7 @@ module Datadog
         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
@@ -46,7 +58,7 @@ module Datadog
           )
 
           Perftools::Profiles::Sample.new(
-            location_id: locations.collect(&:id),
+            location_id: locations.collect { |location| location['id'.freeze] },
             value: values,
             label: build_sample_labels(stack_sample)
           )
@@ -68,22 +80,37 @@ module Datadog
             )
           ]
 
-          unless stack_sample.trace_id.nil? || stack_sample.trace_id.zero?
+          trace_id = stack_sample.trace_id || 0
+          span_id = stack_sample.span_id || 0
+
+          if trace_id != 0 && span_id != 0
+            @processed_with_trace_ids += 1
+
             labels << Perftools::Profiles::Label.new(
               key: builder.string_table.fetch(Datadog::Ext::Profiling::Pprof::LABEL_KEY_TRACE_ID),
-              str: builder.string_table.fetch(stack_sample.trace_id.to_s)
+              str: builder.string_table.fetch(trace_id.to_s)
             )
-          end
 
-          unless stack_sample.span_id.nil? || stack_sample.span_id.zero?
             labels << Perftools::Profiles::Label.new(
               key: builder.string_table.fetch(Datadog::Ext::Profiling::Pprof::LABEL_KEY_SPAN_ID),
-              str: builder.string_table.fetch(stack_sample.span_id.to_s)
+              str: builder.string_table.fetch(span_id.to_s)
             )
+
+            trace_resource = stack_sample.trace_resource_container && stack_sample.trace_resource_container.latest
+            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_ids}"
+        end
       end
     end
   end

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

@@ -1,3 +1,4 @@
+# typed: strict
 require 'ddtrace/utils/string_table'
 
 module Datadog

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

@@ -1,3 +1,4 @@
+# typed: false
 require 'ddtrace/profiling/pprof/payload'
 require 'ddtrace/profiling/pprof/message_set'
 require 'ddtrace/profiling/pprof/builder'
@@ -75,6 +76,10 @@ module Datadog
           converters[event_class].add_events!(events)
         end
 
+        def debug_statistics
+          converters.values.map(&:debug_statistics).join(', ')
+        end
+
         def to_pprof
           profile = builder.build_profile
           data = builder.encode_profile(profile)

data/lib/ddtrace/profiling/preload.rb

@@ -1,3 +1,4 @@
+# typed: strict
 require 'ddtrace'
 
 Datadog.profiler.start if Datadog.profiler

data/lib/ddtrace/profiling/profiler.rb

@@ -1,3 +1,4 @@
+# typed: true
 module Datadog
   # Profiling entry point, which coordinates collectors and a scheduler
   class Profiler

data/lib/ddtrace/profiling/recorder.rb

@@ -1,3 +1,4 @@
+# typed: true
 require 'ddtrace/profiling/buffer'
 require 'ddtrace/profiling/flush'
 
@@ -7,9 +8,9 @@ module Datadog
     class Recorder
       attr_reader :max_size
 
-      def initialize(event_classes, max_size)
+      def initialize(event_classes, max_size, last_flush_time: Time.now.utc)
         @buffers = {}
-        @last_flush_time = Time.now.utc
+        @last_flush_time = last_flush_time
         @max_size = max_size
 
         # Add a buffer for each class

data/lib/ddtrace/profiling/scheduler.rb

@@ -1,3 +1,4 @@
+# typed: true
 require 'ddtrace/utils/time'
 
 require 'ddtrace/worker'
@@ -11,7 +12,12 @@ module Datadog
       include Workers::Polling
 
       DEFAULT_INTERVAL_SECONDS = 60
-      MIN_INTERVAL_SECONDS = 0
+      MINIMUM_INTERVAL_SECONDS = 0
+
+      # Profiles with duration less than this will not be reported
+      PROFILE_DURATION_THRESHOLD_SECONDS = 1
+
+      private_constant :DEFAULT_INTERVAL_SECONDS, :MINIMUM_INTERVAL_SECONDS, :PROFILE_DURATION_THRESHOLD_SECONDS
 
       attr_reader \
         :exporters,
@@ -42,7 +48,18 @@ module Datadog
       end
 
       def perform
-        flush_and_wait
+        # 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?
@@ -78,13 +95,21 @@ module Datadog
 
         # 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
+        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
+
         # Send events to each exporter
         if flush.event_count > 0
           exporters.each do |exporter|
@@ -92,7 +117,7 @@ module Datadog
               exporter.export(flush)
             rescue StandardError => e
               Datadog.logger.error(
-                "Unable to export #{flush.event_count} profiling events. Cause: #{e} Location: #{e.backtrace.first}"
+                "Unable to export #{flush.event_count} profiling events. Cause: #{e} Location: #{Array(e.backtrace).first}"
               )
             end
           end
@@ -100,6 +125,10 @@ module Datadog
 
         flush
       end
+
+      def duration_below_threshold?(flush)
+        (flush.finish - flush.start) < PROFILE_DURATION_THRESHOLD_SECONDS
+      end
     end
   end
 end

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

@@ -1,3 +1,4 @@
+# typed: false
 require 'ddtrace'
 require 'ddtrace/utils/only_once'
 require 'ddtrace/profiling'
@@ -18,7 +19,9 @@ module Datadog
               activate_cpu_extensions
               setup_at_fork_hooks
             rescue StandardError, ScriptError => e
-              log "[DDTRACE] Main extensions unavailable. Cause: #{e.message} Location: #{e.backtrace.first}"
+              Datadog.logger.warn do
+                "Profiler extensions unavailable. Cause: #{e.message} Location: #{Array(e.backtrace).first}"
+              end
             end
           end
         end
@@ -29,22 +32,27 @@ module Datadog
           if Ext::Forking.supported?
             Ext::Forking.apply!
           elsif Datadog.configuration.profiling.enabled
-            # Log warning if profiling was supposed to be activated.
-            log '[DDTRACE] Forking extensions skipped; forking not supported.'
+            Datadog.logger.debug('Profiler forking extensions skipped; forking not supported.')
           end
         rescue StandardError, ScriptError => e
-          log "[DDTRACE] Forking extensions unavailable. Cause: #{e.message} Location: #{e.backtrace.first}"
+          Datadog.logger.warn do
+            "Profiler forking extensions unavailable. Cause: #{e.message} Location: #{Array(e.backtrace).first}"
+          end
         end
 
         def activate_cpu_extensions
           if Ext::CPU.supported?
             Ext::CPU.apply!
           elsif Datadog.configuration.profiling.enabled
-            # Log warning if profiling was supposed to be activated.
-            log "[DDTRACE] CPU profiling skipped because native CPU time is not supported: #{Ext::CPU.unsupported_reason}."
+            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
-          log "[DDTRACE] CPU profiling unavailable. Cause: #{e.message} Location: #{e.backtrace.first}"
+          Datadog.logger.warn do
+            "Profiler CPU profiling extensions unavailable. Cause: #{e.message} Location: #{Array(e.backtrace).first}"
+          end
         end
 
         def setup_at_fork_hooks
@@ -61,16 +69,13 @@ module Datadog
                 # Restart profiler, if enabled
                 Datadog.profiler.start if Datadog.profiler
               rescue StandardError => e
-                log "[DDTRACE] Error during post-fork hooks. Cause: #{e.message} Location: #{e.backtrace.first}"
+                Datadog.logger.warn do
+                  "Error during post-fork hooks. Cause: #{e.message} Location: #{Array(e.backtrace).first}"
+                end
               end
             end
           end
         end
-
-        def log(message)
-          # Print to STDOUT for now because logging may not be setup yet...
-          puts message
-        end
       end
     end
   end

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

@@ -0,0 +1,41 @@
+# typed: true
+# frozen_string_literal: true
+
+require 'ddtrace/ext/http'
+
+module Datadog
+  module Profiling
+    module TraceIdentifiers
+      # Used by Datadog::Profiling::TraceIdentifiers::Helper to get the trace identifiers (trace id and span id) for a
+      # given thread, if there is an active trace for that thread in Datadog.tracer.
+      class Ddtrace
+        def initialize(tracer: nil)
+          @tracer = (tracer if tracer.respond_to?(:call_context))
+        end
+
+        def trace_identifiers_for(thread)
+          return unless @tracer
+
+          context = @tracer.call_context(thread)
+          return unless context
+
+          trace_id = context.trace_id || 0
+          span_id = context.span_id || 0
+
+          [trace_id, span_id, maybe_extract_resource(context.current_root_span)] if trace_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(root_span)
+          return unless root_span
+
+          root_span.resource_container if root_span.span_type == Datadog::Ext::HTTP::TYPE_INBOUND
+        end
+      end
+    end
+  end
+end

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

@@ -0,0 +1,46 @@
+# typed: true
+# frozen_string_literal: true
+
+require 'ddtrace/profiling/trace_identifiers/ddtrace'
+
+module Datadog
+  module Profiling
+    module TraceIdentifiers
+      # Helper used to retrieve the trace identifiers (trace 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 = [
+          ::Datadog::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
+          extract_trace_resource:,
+          supported_apis: DEFAULT_SUPPORTED_APIS.map { |api| api.new(tracer: tracer) }
+        )
+          @extract_trace_resource = extract_trace_resource
+          @supported_apis = supported_apis
+        end
+
+        # Expected output of the #trace_identifiers_for
+        # duck type is [trace_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 @extract_trace_resource ? trace_identifiers : trace_identifiers[0..1]
+            end
+          end
+
+          nil
+        end
+      end
+    end
+  end
+end

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

@@ -1,8 +1,11 @@
+# typed: true
 module Datadog
   module Profiling
     module Transport
       # Generic interface for profiling transports
       module Client
+        include Kernel # Ensure that kernel methods are always available (https://sorbet.org/docs/error-reference#7003)
+
         def send_profiling_flush(flush)
           raise NotImplementedError
         end

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

@@ -1,7 +1,9 @@
-require 'ddtrace/ext/runtime'
+# typed: true
+require 'datadog/core/environment/ext'
 require 'ddtrace/ext/transport'
 
-require 'ddtrace/runtime/container'
+require 'datadog/core/environment/container'
+require 'datadog/core/environment/variable_helpers'
 
 require 'ddtrace/profiling/transport/http/builder'
 require 'ddtrace/profiling/transport/http/api'
@@ -15,21 +17,24 @@ module Datadog
     module Transport
       # Namespace for HTTP transport components
       module HTTP
-        module_function
-
         # Builds a new Transport::HTTP::Client
-        def new(&block)
+        def self.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)
+        def self.default(
+          profiling_upload_timeout_seconds:,
+          agent_settings:,
+          site: nil,
+          api_key: nil,
+          agentless_allowed: agentless_allowed?
+        )
           new do |transport|
             transport.headers default_headers
 
             # Configure adapter & API
-            if site && api_key
+            if site && api_key && agentless_allowed
               configure_for_agentless(
                 transport,
                 profiling_upload_timeout_seconds: profiling_upload_timeout_seconds,
@@ -37,13 +42,6 @@ module Datadog
                 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,
@@ -53,24 +51,24 @@ module Datadog
           end
         end
 
-        def default_headers
+        def self.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
+            Datadog::Ext::Transport::HTTP::HEADER_META_LANG => Core::Environment::Ext::LANG,
+            Datadog::Ext::Transport::HTTP::HEADER_META_LANG_VERSION => Core::Environment::Ext::LANG_VERSION,
+            Datadog::Ext::Transport::HTTP::HEADER_META_LANG_INTERPRETER => Core::Environment::Ext::LANG_INTERPRETER,
+            Datadog::Ext::Transport::HTTP::HEADER_META_TRACER_VERSION => Core::Environment::Ext::TRACER_VERSION
           }.tap do |headers|
             # Add container ID, if present.
-            container_id = Datadog::Runtime::Container.container_id
+            container_id = Datadog::Core::Environment::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
+        private_class_method def self.default_adapter
           :net_http
         end
 
-        private_class_method def configure_for_agent(transport, profiling_upload_timeout_seconds:, agent_settings:)
+        private_class_method def self.configure_for_agent(transport, profiling_upload_timeout_seconds:, agent_settings:)
           apis = API.agent_defaults
 
           transport.adapter(
@@ -90,7 +88,7 @@ module Datadog
           end
         end
 
-        private_class_method def configure_for_agentless(transport, profiling_upload_timeout_seconds:, site:, api_key:)
+        private_class_method def self.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))
@@ -108,10 +106,14 @@ module Datadog
           transport.headers(Datadog::Ext::Transport::HTTP::HEADER_DD_API_KEY => api_key)
         end
 
+        private_class_method def self.agentless_allowed?
+          Datadog::Core::Environment::VariableHelpers.env_to_bool(Datadog::Ext::Profiling::ENV_AGENTLESS, false)
+        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)
+        Datadog::Transport::HTTP::Builder::REGISTRY.set(Datadog::Transport::HTTP::Adapters::Net, :net_http)
+        Datadog::Transport::HTTP::Builder::REGISTRY.set(Datadog::Transport::HTTP::Adapters::Test, :test)
+        Datadog::Transport::HTTP::Builder::REGISTRY.set(Datadog::Transport::HTTP::Adapters::UnixSocket, :unix)
       end
     end
   end