datadog 2.21.0 → 2.38.0

data/lib/datadog/profiling/profiler.rb

@@ -8,56 +8,73 @@ module Datadog
 
       private
 
-      attr_reader :worker, :scheduler
+      attr_reader :worker #: Datadog::Profiling::Collectors::CpuAndWallTimeWorker
+      attr_reader :scheduler #: Datadog::Profiling::Scheduler
 
       public
 
+      # @rbs worker: Datadog::Profiling::Collectors::CpuAndWallTimeWorker
+      # @rbs scheduler: Datadog::Profiling::Scheduler
+      # @rbs return: void
       def initialize(worker:, scheduler:)
         @worker = worker
         @scheduler = scheduler
       end
 
+      #: () -> void
       def start
         after_fork! do
           worker.reset_after_fork
           scheduler.reset_after_fork
         end
 
-        worker.start(on_failure_proc: proc { component_failed(:worker) })
+        worker.start(
+          on_failure_proc: ->(log_failure: true) do
+            # @type var log_failure: bool
+            component_failed(:worker, log_failure: log_failure)
+          end
+        )
         scheduler.start(on_failure_proc: proc { component_failed(:scheduler) })
       end
 
-      def shutdown!
+      # @rbs report_last_profile: bool
+      # @rbs return: void
+      def shutdown!(report_last_profile: true)
         Datadog.logger.debug("Shutting down profiler")
 
         stop_worker
+        scheduler.disable_reporting unless report_last_profile
         stop_scheduler
       end
 
       private
 
+      #: () -> void
       def stop_worker
         worker.stop
       end
 
+      #: () -> void
       def stop_scheduler
         scheduler.enabled = false
         scheduler.stop(true)
       end
 
-      def component_failed(failed_component)
-        Datadog.logger.warn(
-          "Detected issue with profiler (#{failed_component} component), stopping profiling. " \
-          "See previous log messages for details."
-        )
-        Datadog::Core::Telemetry::Logger
-          .error("Detected issue with profiler (#{failed_component} component), stopping profiling")
-
-        # We explicitly not stop the crash tracker in this situation, under the assumption that, if a component failed,
-        # we're operating in a degraded state and crash tracking may still be helpful.
+      # @rbs failed_component: :worker | :scheduler | ::Symbol
+      # @rbs log_failure: bool
+      # @rbs return: void
+      def component_failed(failed_component, log_failure: true)
+        if log_failure
+          Datadog.logger.warn(
+            "Detected issue with profiler (#{failed_component} component), stopping profiling. " \
+            "See previous log messages for details."
+          )
+          Datadog::Core::Telemetry::Logger
+            .error("Detected issue with profiler (#{failed_component} component), stopping profiling")
+        end
 
         if failed_component == :worker
-          scheduler.mark_profiler_failed
+          scheduler.disable_reporting
           stop_scheduler
         elsif failed_component == :scheduler
           stop_worker

data/lib/datadog/profiling/scheduler.rb

@@ -24,7 +24,7 @@ module Datadog
       attr_reader \
         :exporter,
         :transport,
-        :profiler_failed
+        :reporting_disabled
 
       public
 
@@ -36,7 +36,7 @@ module Datadog
       )
         @exporter = exporter
         @transport = transport
-        @profiler_failed = false
+        @reporting_disabled = false
         @stop_requested = false
 
         # Workers::Async::Thread settings
@@ -65,7 +65,7 @@ module Datadog
       rescue Exception => e # rubocop:disable Lint/RescueException
         Datadog.logger.warn(
           "Profiling::Scheduler thread error. " \
-          "Cause: #{e.class.name} #{e.message} Location: #{Array(e.backtrace).first}"
+          "Cause: #{e.class}: #{e.message} Location: #{Array(e.backtrace).first}"
         )
         on_failure_proc&.call
         Datadog::Core::Telemetry::Logger.report(e, description: "Profiling::Scheduler thread error")
@@ -83,14 +83,15 @@ module Datadog
         true
       end
 
-      # This is called by the Profiler class whenever an issue happened in the profiler. This makes sure that even
-      # if there is data to be flushed, we don't try to flush it.
-      def mark_profiler_failed
-        @profiler_failed = true
+      # Called by the Profiler when it wants to prevent any further reporting,
+      # even if there is data to be flushed.
+      # Used when the profiler failed or we're otherwise in a hurry to shut down.
+      def disable_reporting
+        @reporting_disabled = true
       end
 
       def work_pending?
-        !profiler_failed && exporter.can_flush? && (run_loop? || !stop_requested?)
+        !reporting_disabled && exporter.can_flush? && (run_loop? || !stop_requested?)
       end
 
       def reset_after_fork
@@ -135,7 +136,7 @@ module Datadog
           transport.export(flush)
         rescue => e
           Datadog.logger.warn(
-            "Unable to report profile. Cause: #{e.class.name} #{e.message} Location: #{Array(e.backtrace).first}"
+            "Unable to report profile. Cause: #{e.class}: #{e.message} Location: #{Array(e.backtrace).first}"
           )
           Datadog::Core::Telemetry::Logger.report(e, description: "Unable to report profile")
         end

data/lib/datadog/profiling/stack_recorder.rb

@@ -9,12 +9,10 @@ module Datadog
     # Methods prefixed with _native_ are implemented in `stack_recorder.c`
     class StackRecorder
       def initialize(
-        cpu_time_enabled:,
         alloc_samples_enabled:,
         heap_samples_enabled:,
         heap_size_enabled:,
         heap_sample_every:,
-        timeline_enabled:,
         heap_clean_after_gc_enabled:
       )
         # This mutex works in addition to the fancy C-level mutexes we have in the native side (see the docs there).
@@ -27,33 +25,27 @@ module Datadog
 
         self.class._native_initialize(
           self_instance: self,
-          cpu_time_enabled: cpu_time_enabled,
           alloc_samples_enabled: alloc_samples_enabled,
           heap_samples_enabled: heap_samples_enabled,
           heap_size_enabled: heap_size_enabled,
           heap_sample_every: heap_sample_every,
-          timeline_enabled: timeline_enabled,
           heap_clean_after_gc_enabled: heap_clean_after_gc_enabled,
         )
       end
 
       def self.for_testing(
-        cpu_time_enabled: true,
         alloc_samples_enabled: false,
         heap_samples_enabled: false,
         heap_size_enabled: false,
         heap_sample_every: 1,
-        timeline_enabled: false,
         heap_clean_after_gc_enabled: true,
         **options
       )
         new(
-          cpu_time_enabled: cpu_time_enabled,
           alloc_samples_enabled: alloc_samples_enabled,
           heap_samples_enabled: heap_samples_enabled,
           heap_size_enabled: heap_size_enabled,
           heap_sample_every: heap_sample_every,
-          timeline_enabled: timeline_enabled,
           heap_clean_after_gc_enabled: heap_clean_after_gc_enabled,
           **options,
         )

data/lib/datadog/profiling/tag_builder.rb

@@ -1,17 +1,48 @@
 # frozen_string_literal: true
 
-require_relative "../core/tag_builder"
-require_relative "../core/utils"
+require_relative '../core/tag_builder'
+require_relative '../core/utils'
+
+require 'set'
 
 module Datadog
   module Profiling
     # Builds a hash of default plus user tags to be included in a profile
+    #
+    # @api private
     module TagBuilder
       include Datadog::Profiling::Ext::Transport::HTTP # Tag name constants
 
+      # When changing or adding profiling-related tags, make sure they are
+      # kept in sync with
+      # https://docs.google.com/spreadsheets/d/1LOGMf4c4Avbtn36uZ2SWvhIGKRPLM1BoWkUP4JYj7hA/
+      # (Datadog internal link).
+      #
+      # For consistency between the different profilers, every tag should be
+      # vetted before it gets reported with a profile, as otherwise it's too
+      # easy to end up with different tags in different languages.
+      ALLOWED_TAGS = Set.new(
+        [
+          'env',
+          'service',
+          'version',
+          'git.commit.sha',
+          'git.repository_url',
+          'host',
+          'language',
+          'runtime',
+          'runtime_engine',
+          'runtime_platform',
+          'runtime_version',
+          'runtime-id',
+          'process_id',
+          'profiler_version',
+          'profile_seq',
+        ]
+      ).freeze
+
       def self.call(
         settings:,
-        # Other metadata
         profile_seq:,
         profiler_version: Core::Environment::Identity.gem_datadog_version
       )
@@ -19,6 +50,9 @@ module Datadog
           FORM_FIELD_TAG_PROFILER_VERSION => profiler_version,
           'profile_seq' => profile_seq.to_s,
         )
+
+        user_tag_keys = settings.tags.keys
+        hash.keep_if { |tag| user_tag_keys.include?(tag) || ALLOWED_TAGS.include?(tag) }
         Core::Utils.encode_tags(hash)
       end
     end

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

@@ -5,17 +5,20 @@ module Datadog
     module Tasks
       # Wraps command with Datadog profiling
       class Exec
-        attr_reader :args
+        attr_reader :args #: untyped
 
+        #: (untyped) -> void
         def initialize(args)
           @args = args
         end
 
+        #: () -> untyped
         def run
           set_rubyopt!
           exec_with_error_handling(args)
         end
 
+        #: () -> ::Array["-rdatadog/profiling/preload"]
         def rubyopts
           [
             "-rdatadog/profiling/preload"
@@ -24,6 +27,7 @@ module Datadog
 
         private
 
+        #: () -> untyped
         def set_rubyopt!
           existing_rubyopt = ENV["RUBYOPT"] # rubocop:disable CustomCops/EnvUsageCop
 
@@ -35,13 +39,14 @@ module Datadog
         # 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
+        #: (untyped) -> untyped
         def exec_with_error_handling(args)
           Kernel.exec(*args)
         rescue Errno::ENOENT => e
-          Kernel.warn "ddprofrb exec failed: #{e.class.name} #{e.message} (command was '#{args.join(" ")}')"
+          Kernel.warn "ddprofrb exec failed: #{e.class}: #{e.message} (command was '#{args.join(" ")}')"
           Kernel.exit 127
         rescue Errno::EACCES, Errno::ENOEXEC => e
-          Kernel.warn "ddprofrb exec failed: #{e.class.name} #{e.message} (command was '#{args.join(" ")}')"
+          Kernel.warn "ddprofrb exec failed: #{e.class}: #{e.message} (command was '#{args.join(" ")}')"
           Kernel.exit 126
         end
       end

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

@@ -5,6 +5,7 @@ module Datadog
     module Tasks
       # Prints help message for usage of `ddprofrb`
       class Help
+        #: () -> untyped
         def run
           puts %(
   Usage: ddprofrb [command] [arguments]

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

@@ -16,7 +16,7 @@ module Datadog
             setup_at_fork_hooks
           rescue StandardError, ScriptError => e
             Datadog.logger.warn do
-              "Profiler extensions unavailable. Cause: #{e.class.name} #{e.message} " \
+              "Profiler extensions unavailable. Cause: #{e.class}: #{e.message} " \
               "Location: #{Array(e.backtrace).first}"
             end
             Datadog::Core::Telemetry::Logger.report(e, description: "Profiler extensions unavailable")
@@ -31,7 +31,7 @@ module Datadog
             Profiling.start_if_enabled
           rescue => e
             Datadog.logger.warn do
-              "Error during post-fork hooks. Cause: #{e.class.name} #{e.message} " \
+              "Error during post-fork hooks. Cause: #{e.class}: #{e.message} " \
               "Location: #{Array(e.backtrace).first}"
             end
             Datadog::Core::Telemetry::Logger.report(e, description: "Error during post-fork hooks")

data/lib/datadog/profiling.rb

@@ -61,9 +61,7 @@ module Datadog
     end
 
     def self.enabled?
-      profiler = Datadog.send(:components).profiler
-      # Use .send(...) to avoid exposing the attr_reader as an API to the outside
-      !!profiler&.send(:scheduler)&.running?
+      !!Datadog.send(:components).profiler
     end
 
     def self.wait_until_running(timeout_seconds: 5)

data/lib/datadog/ruby_version.rb

@@ -0,0 +1,25 @@
+# frozen_string_literal: true
+
+module Datadog
+  # Compares the running Ruby against version requirements.
+  #
+  # Lexical comparisons against `RUBY_VERSION` are subtly wrong: `RUBY_VERSION < "3.2.3"` is also
+  # `true` on 3.2.10 and 3.2.11, since those sort *before* "3.2.3" as strings.
+  #
+  # @example
+  #   RubyVersion.is?(">= 3.2", "< 3.2.3") # => true on Ruby 3.2.0, 3.2.1, 3.2.2 (NOT 3.2.3+, NOT 3.2.10+)
+  module RubyVersion
+    extend RubyVersion # steep currently needs this (instead of extend self or def self.is?) for the inline rbs to work
+
+    CURRENT_RUBY_VERSION = Gem::Version.new(RUBY_VERSION) #: ::Gem::Version
+    private_constant :CURRENT_RUBY_VERSION
+
+    # Returns `true` when the running Ruby satisfies ALL of the given requirements. Each uses the
+    # same syntax as a gem dependency (e.g. `">= 3.1"`, `">= 3.2"`, `"< 3.2.3"`).
+    #
+    # @rbs (*String requirements, ?ruby_version: ::Gem::Version) -> bool
+    def is?(*requirements, ruby_version: CURRENT_RUBY_VERSION)
+      Gem::Requirement.new(*requirements).satisfied_by?(ruby_version)
+    end
+  end
+end

data/lib/datadog/single_step_instrument.rb

@@ -15,7 +15,7 @@ end
 
 begin
   require_relative 'auto_instrument'
-  Datadog::SingleStepInstrument::LOADED = true
+  Datadog::SingleStepInstrument.const_set(:LOADED, true)
 rescue StandardError, LoadError => e
-  warn "Single step instrumentation failed: #{e.class}:#{e.message}\n\tSource:\n\t#{Array(e.backtrace).join("\n\t")}"
+  warn "Single step instrumentation failed: #{e.class}: #{e.message}\n\tSource:\n\t#{Array(e.backtrace).join("\n\t")}"
 end