ddtrace 0.54.2 → 1.0.0

data/lib/datadog/tracing/sampling/priority_sampler.rb

@@ -0,0 +1,113 @@
+# typed: true
+
+require 'forwardable'
+
+require 'datadog/tracing/sampling/ext'
+require 'datadog/tracing/sampling/all_sampler'
+require 'datadog/tracing/sampling/rate_sampler'
+require 'datadog/tracing/sampling/rate_by_service_sampler'
+
+module Datadog
+  module Tracing
+    module Sampling
+      # {Datadog:::Tracing::Sampling::PrioritySampler}
+      # @public_api
+      class PrioritySampler
+        extend Forwardable
+
+        # NOTE: We do not advise using a pre-sampler. It can save resources,
+        # but pre-sampling at rates < 100% may result in partial traces, unless
+        # the pre-sampler knows exactly how to drop a span without dropping its ancestors.
+        #
+        # Additionally, as service metrics are calculated in the Datadog Agent,
+        # the service's throughput will be underestimated.
+        attr_reader :pre_sampler, :priority_sampler
+
+        def initialize(opts = {})
+          @pre_sampler = opts[:base_sampler] || AllSampler.new
+          @priority_sampler = opts[:post_sampler] || RateByServiceSampler.new
+        end
+
+        def sample?(trace)
+          @pre_sampler.sample?(trace)
+        end
+
+        def sample!(trace)
+          # If pre-sampling is configured, do it first. (By default, this will sample at 100%.)
+          # NOTE: Pre-sampling at rates < 100% may result in partial traces; not recommended.
+          trace.sampled = pre_sample?(trace) ? @pre_sampler.sample!(trace) : true
+
+          if trace.sampled?
+            # If priority sampling has already been applied upstream, use that value.
+            return true if priority_assigned?(trace)
+
+            # Check with post sampler how we set the priority.
+            sample = priority_sample!(trace)
+
+            # Check if post sampler has already assigned a priority.
+            return true if priority_assigned?(trace)
+
+            # If not, use agent priority values.
+            priority = sample ? Sampling::Ext::Priority::AUTO_KEEP : Sampling::Ext::Priority::AUTO_REJECT
+            assign_priority!(trace, priority)
+          else
+            # If discarded by pre-sampling, set "reject" priority, so other
+            # services for the same trace don't sample needlessly.
+            assign_priority!(trace, Sampling::Ext::Priority::AUTO_REJECT)
+          end
+
+          trace.sampled?
+        end
+
+        def_delegators :@priority_sampler, :update
+
+        private
+
+        def pre_sample?(trace)
+          case @pre_sampler
+          when RateSampler
+            @pre_sampler.sample_rate < 1.0
+          when RateByServiceSampler
+            @pre_sampler.sample_rate(trace) < 1.0
+          else
+            true
+          end
+        end
+
+        def priority_assigned?(trace)
+          !trace.sampling_priority.nil?
+        end
+
+        def priority_sample!(trace)
+          preserving_sampling(trace) do
+            @priority_sampler.sample!(trace)
+          end
+        end
+
+        # Ensures the trace is always propagated to the writer and that
+        # the sample rate metric represents the true client-side sampling.
+        def preserving_sampling(trace)
+          pre_sample_rate_metric = trace.sample_rate
+
+          yield.tap do
+            # NOTE: We'll want to leave `trace.sampled = true` here; all spans for priority sampling must
+            #       be sent to the agent. Otherwise metrics for traces will not be accurate, since the
+            #       agent will have an incomplete dataset.
+            #
+            #       We also ensure that the agent knows we that our `post_sampler` is not performing true sampling,
+            #       to avoid erroneous metric upscaling.
+            trace.sampled = true
+
+            # Restore true sampling metric, as only the @pre_sampler can reject traces.
+            # otherwise if @pre_sampler is not enabled, sending this metric would be misleading.
+            trace.sample_rate = pre_sample_rate_metric || nil
+          end
+        end
+
+        def assign_priority!(trace, priority)
+          trace.sampling_priority = priority
+        end
+      end
+    end
+  end
+end

data/lib/datadog/tracing/sampling/rate_by_key_sampler.rb

@@ -0,0 +1,97 @@
+# typed: true
+
+require 'datadog/tracing/sampling/sampler'
+require 'datadog/tracing/sampling/rate_sampler'
+
+module Datadog
+  module Tracing
+    module Sampling
+      # Samples at different rates by key.
+      # @public_api
+      class RateByKeySampler < Sampler
+        attr_reader \
+          :default_key
+
+        def initialize(default_key, default_rate = 1.0, &block)
+          super()
+
+          raise ArgumentError, 'No resolver given!' unless block
+
+          @default_key = default_key
+          @resolver = block
+          @mutex = Mutex.new
+          @samplers = {}
+
+          set_rate(default_key, default_rate)
+        end
+
+        def resolve(trace)
+          @resolver.call(trace)
+        end
+
+        def default_sampler
+          @samplers[default_key]
+        end
+
+        def sample?(trace)
+          key = resolve(trace)
+
+          @mutex.synchronize do
+            @samplers.fetch(key, default_sampler).sample?(trace)
+          end
+        end
+
+        def sample!(trace)
+          key = resolve(trace)
+
+          @mutex.synchronize do
+            @samplers.fetch(key, default_sampler).sample!(trace)
+          end
+        end
+
+        def sample_rate(trace)
+          key = resolve(trace)
+
+          @mutex.synchronize do
+            @samplers.fetch(key, default_sampler).sample_rate
+          end
+        end
+
+        def update(key, rate)
+          @mutex.synchronize do
+            set_rate(key, rate)
+          end
+        end
+
+        def update_all(rate_by_key)
+          @mutex.synchronize do
+            rate_by_key.each { |key, rate| set_rate(key, rate) }
+          end
+        end
+
+        def delete(key)
+          @mutex.synchronize do
+            @samplers.delete(key)
+          end
+        end
+
+        def delete_if(&block)
+          @mutex.synchronize do
+            @samplers.delete_if(&block)
+          end
+        end
+
+        def length
+          @samplers.length
+        end
+
+        private
+
+        def set_rate(key, rate)
+          @samplers[key] ||= RateSampler.new(rate)
+          @samplers[key].sample_rate = rate
+        end
+      end
+    end
+  end
+end

data/lib/datadog/tracing/sampling/rate_by_service_sampler.rb

@@ -0,0 +1,41 @@
+# typed: true
+
+require 'datadog/core'
+require 'datadog/tracing/sampling/rate_by_key_sampler'
+
+module Datadog
+  module Tracing
+    module Sampling
+      # {Datadog:::Tracing::Sampling::RateByServiceSampler} samples different services at different rates
+      # @public_api
+      class RateByServiceSampler < RateByKeySampler
+        DEFAULT_KEY = 'service:,env:'.freeze
+
+        def initialize(default_rate = 1.0, options = {})
+          super(DEFAULT_KEY, default_rate, &method(:key_for))
+          @env = options[:env]
+        end
+
+        def update(rate_by_service)
+          # Remove any old services
+          delete_if { |key, _| key != DEFAULT_KEY && !rate_by_service.key?(key) }
+
+          # Update each service rate
+          update_all(rate_by_service)
+
+          # Emit metric for service cache size
+          Datadog.health_metrics.sampling_service_cache_length(length)
+        end
+
+        private
+
+        def key_for(trace)
+          # Resolve env dynamically, if Proc is given.
+          env = @env.is_a?(Proc) ? @env.call : @env
+
+          "service:#{trace.service},env:#{env}"
+        end
+      end
+    end
+  end
+end

data/lib/datadog/tracing/sampling/rate_limiter.rb

@@ -0,0 +1,185 @@
+# typed: true
+
+require 'datadog/core/utils/time'
+
+module Datadog
+  module Tracing
+    module Sampling
+      # Checks for rate limiting on a resource.
+      # @public_api
+      class RateLimiter
+        # Checks if resource of specified size can be
+        # conforms with the current limit.
+        #
+        # Implementations of this method are not guaranteed
+        # to be side-effect free.
+        #
+        # @return [Boolean] whether a resource conforms with the current limit
+        def allow?(size); end
+
+        # The effective rate limiting ratio based on
+        # recent calls to `allow?`.
+        #
+        # @return [Float] recent allowance ratio
+        def effective_rate; end
+      end
+
+      # Implementation of the Token Bucket metering algorithm
+      # for rate limiting.
+      #
+      # @see https://en.wikipedia.org/wiki/Token_bucket Token bucket
+      # @public_api
+      class TokenBucket < RateLimiter
+        attr_reader :rate, :max_tokens
+
+        # @param rate [Numeric] Allowance rate, in units per second
+        #  if rate is negative, always allow
+        #  if rate is zero, never allow
+        # @param max_tokens [Numeric] Limit of available tokens
+        def initialize(rate, max_tokens = rate)
+          super()
+
+          @rate = rate
+          @max_tokens = max_tokens
+
+          @tokens = max_tokens
+          @total_messages = 0
+          @conforming_messages = 0
+          @prev_conforming_messages = nil
+          @prev_total_messages = nil
+          @current_window = nil
+
+          @last_refill = Core::Utils::Time.get_time
+        end
+
+        # Checks if a message of provided +size+
+        # conforms with the current bucket limit.
+        #
+        # If it does, return +true+ and remove +size+
+        # tokens from the bucket.
+        # If it does not, return +false+ without affecting
+        # the tokens from the bucket.
+        #
+        # @return [Boolean] +true+ if message conforms with current bucket limit
+        def allow?(size)
+          allowed = should_allow?(size)
+          update_rate_counts(allowed)
+          allowed
+        end
+
+        # Ratio of 'conformance' per 'total messages' checked
+        # averaged for the past 2 buckets
+        #
+        # Returns +1.0+ when no messages have been checked yet.
+        #
+        # @return [Float] Conformance ratio, between +[0,1]+
+        def effective_rate
+          return 0.0 if @rate.zero?
+          return 1.0 if @rate < 0 || @total_messages.zero?
+
+          return current_window_rate if @prev_conforming_messages.nil? || @prev_total_messages.nil?
+
+          (@conforming_messages.to_f + @prev_conforming_messages.to_f) / (@total_messages + @prev_total_messages)
+        end
+
+        # Ratio of 'conformance' per 'total messages' checked
+        # on this bucket
+        #
+        # Returns +1.0+ when no messages have been checked yet.
+        #
+        # @return [Float] Conformance ratio, between +[0,1]+
+        def current_window_rate
+          return 1.0 if @total_messages.zero?
+
+          @conforming_messages.to_f / @total_messages
+        end
+
+        # @return [Numeric] number of tokens currently available
+        def available_tokens
+          @tokens
+        end
+
+        private
+
+        def refill_since_last_message
+          now = Core::Utils::Time.get_time
+          elapsed = now - @last_refill
+
+          # Update the number of available tokens, but ensure we do not exceed the max
+          # we return the min of tokens + rate*elapsed, or max tokens
+          refill_tokens(@rate * elapsed)
+
+          @last_refill = now
+        end
+
+        def refill_tokens(size)
+          @tokens += size
+          @tokens = @max_tokens if @tokens > @max_tokens
+        end
+
+        def increment_total_count
+          @total_messages += 1
+        end
+
+        def increment_conforming_count
+          @conforming_messages += 1
+        end
+
+        def should_allow?(size)
+          # rate limit of 0 blocks everything
+          return false if @rate.zero?
+
+          # negative rate limit disables rate limiting
+          return true if @rate < 0
+
+          refill_since_last_message
+
+          # if tokens < 1 we don't allow?
+          return false if @tokens < size
+
+          @tokens -= size
+
+          true
+        end
+
+        # Sets and Updates the past two 1 second windows for which
+        # the rate limiter must compute it's rate over and updates
+        # the total count, and conforming message count if +allowed+
+        def update_rate_counts(allowed)
+          now = Core::Utils::Time.get_time
+
+          # No tokens have been seen yet, start a new window
+          if @current_window.nil?
+            @current_window = now
+          # If more than 1 second has past since last window, reset
+          elsif now - @current_window >= 1
+            @prev_conforming_messages = @conforming_messages
+            @prev_total_messages = @total_messages
+            @conforming_messages = 0
+            @total_messages = 0
+            @current_window = now
+          end
+
+          increment_conforming_count if allowed
+
+          increment_total_count
+        end
+      end
+
+      # {Datadog::Tracing::Sampling::RateLimiter} that accepts all resources,
+      # with no limits.
+      # @public_api
+      class UnlimitedLimiter < RateLimiter
+        # @return [Boolean] always +true+
+        def allow?(_)
+          true
+        end
+
+        # @return [Float] always 100%
+        def effective_rate
+          1.0
+        end
+      end
+    end
+  end
+end

data/lib/datadog/tracing/sampling/rate_sampler.rb

@@ -0,0 +1,55 @@
+# typed: true
+
+require 'datadog/core'
+
+require 'datadog/tracing/sampling/sampler'
+require 'datadog/tracing/span'
+
+module Datadog
+  module Tracing
+    module Sampling
+      # {Datadog:::Tracing::Sampling::RateSampler} is based on a sample rate.
+      # @public_api
+      class RateSampler < Sampler
+        KNUTH_FACTOR = 1111111111111111111
+
+        # Initialize a {Datadog:::Tracing::Sampling::RateSampler}.
+        # This sampler keeps a random subset of the traces. Its main purpose is to
+        # reduce the instrumentation footprint.
+        #
+        # * +sample_rate+: the sample rate as a {Float} between 0.0 and 1.0. 0.0
+        #   means that no trace will be sampled; 1.0 means that all traces will be
+        #   sampled.
+        def initialize(sample_rate = 1.0)
+          super()
+
+          unless sample_rate > 0.0 && sample_rate <= 1.0
+            Datadog.logger.error('sample rate is not between 0 and 1, disabling the sampler')
+            sample_rate = 1.0
+          end
+
+          self.sample_rate = sample_rate
+        end
+
+        def sample_rate(*_)
+          @sample_rate
+        end
+
+        def sample_rate=(sample_rate)
+          @sample_rate = sample_rate
+          @sampling_id_threshold = sample_rate * Span::EXTERNAL_MAX_ID
+        end
+
+        def sample?(trace)
+          ((trace.id * KNUTH_FACTOR) % Span::EXTERNAL_MAX_ID) <= @sampling_id_threshold
+        end
+
+        def sample!(trace)
+          sampled = trace.sampled = sample?(trace)
+          trace.sample_rate = @sample_rate if sampled
+          sampled
+        end
+      end
+    end
+  end
+end

data/lib/datadog/tracing/sampling/rule.rb

@@ -0,0 +1,70 @@
+# typed: true
+
+require 'forwardable'
+
+require 'datadog/core'
+
+require 'datadog/tracing/sampling/matcher'
+require 'datadog/tracing/sampling/rate_sampler'
+
+module Datadog
+  module Tracing
+    module Sampling
+      # Sampling rule that dictates if a trace matches
+      # a specific criteria and what sampling strategy to
+      # apply in case of a positive match.
+      # @public_api
+      class Rule
+        extend Forwardable
+
+        attr_reader :matcher, :sampler
+
+        # @param [Matcher] matcher A matcher to verify trace conformity against
+        # @param [Sampler] sampler A sampler to be consulted on a positive match
+        def initialize(matcher, sampler)
+          @matcher = matcher
+          @sampler = sampler
+        end
+
+        # Evaluates if the provided `trace` conforms to the `matcher`.
+        #
+        # @param [TraceOperation] trace
+        # @return [Boolean] whether this rules applies to the trace
+        # @return [NilClass] if the matcher fails errs during evaluation
+        def match?(trace)
+          @matcher.match?(trace)
+        rescue => e
+          Datadog.logger.error("Matcher failed. Cause: #{e.message} Source: #{Array(e.backtrace).first}")
+          nil
+        end
+
+        def_delegators :@sampler, :sample?, :sample_rate
+      end
+
+      # A {Datadog::Tracing::Sampling::Rule} that matches a trace based on
+      # trace name and/or service name and
+      # applies a fixed sampling to matching spans.
+      # @public_api
+      class SimpleRule < Rule
+        # @param name [String,Regexp,Proc] Matcher for case equality (===) with the trace name, defaults to always match
+        # @param service [String,Regexp,Proc] Matcher for case equality (===) with the service name,
+        #                defaults to always match
+        # @param sample_rate [Float] Sampling rate between +[0,1]+
+        def initialize(name: SimpleMatcher::MATCH_ALL, service: SimpleMatcher::MATCH_ALL, sample_rate: 1.0)
+          # We want to allow 0.0 to drop all traces, but {Datadog::Tracing::Sampling::RateSampler}
+          # considers 0.0 an invalid rate and falls back to 100% sampling.
+          #
+          # We address that here by not setting the rate in the constructor,
+          # but using the setter method.
+          #
+          # We don't want to make this change directly to {Datadog::Tracing::Sampling::RateSampler}
+          # because it breaks its current contract to existing users.
+          sampler = RateSampler.new
+          sampler.sample_rate = sample_rate
+
+          super(SimpleMatcher.new(name: name, service: service), sampler)
+        end
+      end
+    end
+  end
+end

data/lib/datadog/tracing/sampling/rule_sampler.rb

@@ -0,0 +1,133 @@
+# typed: true
+
+require 'forwardable'
+
+require 'datadog/core'
+
+require 'datadog/tracing/sampling/ext'
+require 'datadog/tracing/sampling/rate_limiter'
+require 'datadog/tracing/sampling/rule'
+
+module Datadog
+  module Tracing
+    module Sampling
+      # Span {Sampler} that applies a set of {Rule}s to decide
+      # on sampling outcome. Then, a rate limiter is applied.
+      #
+      # If a trace does not conform to any rules, a default
+      # sampling strategy is applied.
+      # @public_api
+      class RuleSampler
+        extend Forwardable
+
+        attr_reader :rules, :rate_limiter, :default_sampler
+
+        # @param rules [Array<Rule>] ordered list of rules to be applied to a trace
+        # @param rate_limit [Float] number of traces per second, defaults to +100+
+        # @param rate_limiter [RateLimiter] limiter applied after rule matching
+        # @param default_sample_rate [Float] fallback sample rate when no rules apply to a trace,
+        #   between +[0,1]+, defaults to +1+
+        # @param default_sampler [Sample] fallback strategy when no rules apply to a trace
+        def initialize(
+          rules = [],
+          rate_limit: Datadog.configuration.tracing.sampling.rate_limit,
+          rate_limiter: nil,
+          default_sample_rate: Datadog.configuration.tracing.sampling.default_rate,
+          default_sampler: nil
+        )
+          @rules = rules
+          @rate_limiter = if rate_limiter
+                            rate_limiter
+                          elsif rate_limit
+                            TokenBucket.new(rate_limit)
+                          else
+                            UnlimitedLimiter.new
+                          end
+
+          @default_sampler = if default_sampler
+                               default_sampler
+                             elsif default_sample_rate
+                               # Add to the end of the rule list a rule always matches any trace
+                               @rules << SimpleRule.new(sample_rate: default_sample_rate)
+                               nil
+                             else
+                               # TODO: Simplify .tags access, as `Tracer#tags` can't be arbitrarily changed anymore
+                               RateByServiceSampler.new(1.0, env: -> { Tracing.send(:tracer).tags[:env] })
+                             end
+        end
+
+        # /RuleSampler's components (it's rate limiter, for example) are
+        # not be guaranteed to be size-effect free.
+        # It is not possible to guarantee that a call to {#sample?} will
+        # return the same result as a successive call to {#sample!} with the same trace.
+        #
+        # Use {#sample!} instead
+        def sample?(_trace)
+          raise 'RuleSampler cannot be evaluated without side-effects'
+        end
+
+        def sample!(trace)
+          sampled = sample_trace(trace) do |t|
+            @default_sampler.sample!(t).tap do
+              # We want to make sure the trace is tagged with the agent-derived
+              # service rate. Retrieve this from the rate by service sampler.
+              # Only do this if it was set by a RateByServiceSampler.
+              trace.agent_sample_rate = @default_sampler.sample_rate(trace) if @default_sampler.is_a?(RateByServiceSampler)
+            end
+          end
+
+          trace.sampled = sampled
+        end
+
+        # @!visibility private
+        def update(*args)
+          return false unless @default_sampler.respond_to?(:update)
+
+          @default_sampler.update(*args)
+        end
+
+        private
+
+        def sample_trace(trace)
+          rule = @rules.find { |r| r.match?(trace) }
+
+          return yield(trace) if rule.nil?
+
+          sampled = rule.sample?(trace)
+          sample_rate = rule.sample_rate(trace)
+
+          set_priority(trace, sampled)
+          set_rule_metrics(trace, sample_rate)
+
+          return false unless sampled
+
+          rate_limiter.allow?(1).tap do |allowed|
+            set_priority(trace, allowed)
+            set_limiter_metrics(trace, rate_limiter.effective_rate)
+          end
+        rescue StandardError => e
+          Datadog.logger.error("Rule sampling failed. Cause: #{e.message} Source: #{Array(e.backtrace).first}")
+          yield(trace)
+        end
+
+        # Span priority should only be set when the {RuleSampler}
+        # was responsible for the sampling decision.
+        def set_priority(trace, sampled)
+          trace.sampling_priority = if sampled
+                                      Sampling::Ext::Priority::USER_KEEP
+                                    else
+                                      Sampling::Ext::Priority::USER_REJECT
+                                    end
+        end
+
+        def set_rule_metrics(trace, sample_rate)
+          trace.rule_sample_rate = sample_rate
+        end
+
+        def set_limiter_metrics(trace, limiter_rate)
+          trace.rate_limiter_rate = limiter_rate
+        end
+      end
+    end
+  end
+end