Merge pull request #2543 from DataDog/tonycthsu/128bits-trace-id-prop…

…agation

128 bits trace id

docs/GettingStarted.md

@@ -2140,6 +2140,7 @@ end
 | `tracing.sampling.default_rate`                         | `DD_TRACE_SAMPLE_RATE`         | `nil`                                                             | Sets the trace sampling rate between `0.0` (0%) and `1.0` (100%). See [Application-side sampling](#application-side-sampling) for details.                                                                                                  |
 | `tracing.sampling.rate_limit`                           | `DD_TRACE_RATE_LIMIT`          | `100` (per second)                                                | Sets a maximum number of traces per second to sample. Set a rate limit to avoid the ingestion volume overages in the case of traffic spikes.                                                                    |
 | `tracing.sampling.span_rules`                           | `DD_SPAN_SAMPLING_RULES`,`ENV_SPAN_SAMPLING_RULES_FILE` | `nil`                                    | Sets [Single Span Sampling](#single-span-sampling) rules. These rules allow you to keep spans even when their respective traces are dropped.                                                                                              |
+| `tracing.trace_id_128_bit_generation_enabled` | `DD_TRACE_128_BIT_TRACEID_GENERATION_ENABLED` | `false` | `true` to generate 128 bits trace ID and `false` to generate 64 bits trace ID  |
 | `tracing.report_hostname`                               | `DD_TRACE_REPORT_HOSTNAME`     | `false`                                                           | Adds hostname tag to traces.                                                                                                                                                                                                              |
 | `tracing.test_mode.enabled`                             | `DD_TRACE_TEST_MODE_ENABLED`   | `false`                                                           | Enables or disables test mode, for use of tracing in test suites.                                                                                                                                                                         |
 | `tracing.test_mode.trace_flush`                         |                                | `nil`                                                             | Object that determines trace flushing behavior.                                                                                                                                                                                           |

lib/datadog/tracing/configuration/ext.rb

@@ -5,6 +5,7 @@ module Configuration
       # e.g. Env vars, default values, enums, etc...
       module Ext
         ENV_ENABLED = 'DD_TRACE_ENABLED'.freeze
+        ENV_TRACE_ID_128_BIT_GENERATION_ENABLED = 'DD_TRACE_128_BIT_TRACEID_GENERATION_ENABLED'.freeze
 
         # @public_api
         module Analytics
@@ -14,6 +15,7 @@ module Analytics
         # @public_api
         module Correlation
           ENV_LOGS_INJECTION_ENABLED = 'DD_LOGS_INJECTION'.freeze
+          ENV_TRACE_ID_128_BIT_LOGGING_ENABLED = 'DD_TRACE_128_BIT_TRACEID_LOGGING_ENABLED'.freeze
         end
 
         # @public_api

lib/datadog/tracing/configuration/settings.rb

@@ -173,6 +173,26 @@ def self.extended(base)
                 o.lazy
               end
 
+              # Enable 128 bit trace id generation.
+              #
+              # @default `DD_TRACE_128_BIT_TRACEID_GENERATION_ENABLED` environment variable, otherwise `false`
+              # @return [Boolean]
+              option :trace_id_128_bit_generation_enabled do |o|
+                o.default { env_to_bool(Tracing::Configuration::Ext::ENV_TRACE_ID_128_BIT_GENERATION_ENABLED, false) }
+                o.lazy
+              end
+
+              # Enable 128 bit trace id injected for logging.
+              #
+              # @default `DD_TRACE_128_BIT_TRACEID_LOGGING_ENABLED` environment variable, otherwise `false`
+              # @return [Boolean]
+              #
+              # It is not supported by our backend yet. Do not enable it.
+              option :trace_id_128_bit_logging_enabled do |o|
+                o.default { env_to_bool(Tracing::Configuration::Ext::Correlation::ENV_TRACE_ID_128_BIT_LOGGING_ENABLED, false) }
+                o.lazy
+              end
+
               # A custom tracer instance.
               #
               # It must respect the contract of {Datadog::Tracing::Tracer}.

lib/datadog/tracing/correlation.rb

@@ -1,4 +1,6 @@
 require_relative '../core'
+require_relative 'utils'
+require_relative 'metadata/ext'
 
 module Datadog
   module Tracing
@@ -64,11 +66,23 @@ def to_log_format
             attributes << "#{LOG_ATTR_ENV}=#{env}" unless env.nil?
             attributes << "#{LOG_ATTR_SERVICE}=#{service}"
             attributes << "#{LOG_ATTR_VERSION}=#{version}" unless version.nil?
-            attributes << "#{LOG_ATTR_TRACE_ID}=#{trace_id}"
+            attributes << "#{LOG_ATTR_TRACE_ID}=#{logging_trace_id}"
             attributes << "#{LOG_ATTR_SPAN_ID}=#{span_id}"
             attributes.join(' ')
           end
         end
+
+        private
+
+        def logging_trace_id
+          @logging_trace_id ||=
+            if Datadog.configuration.tracing.trace_id_128_bit_logging_enabled &&
+                !Tracing::Utils::TraceId.to_high_order(@trace_id).zero?
+              Kernel.format('%032x', trace_id)
+            else
+              Tracing::Utils::TraceId.to_low_order(@trace_id)
+            end
+        end
       end
 
       module_function

lib/datadog/tracing/distributed/b3_multi.rb

@@ -2,6 +2,7 @@
 
 require_relative 'helpers'
 require_relative '../trace_digest'
+require_relative '../utils'
 
 module Datadog
   module Tracing
@@ -45,13 +46,19 @@ def inject!(digest, data = {})
         def extract(data)
           # DEV: B3 doesn't have "origin"
           fetcher = @fetcher.new(data)
-          trace_id = fetcher.id(@trace_id_key, base: 16)
-          span_id = fetcher.id(@span_id_key, base: 16)
-          # We don't need to try and convert sampled since B3 supports 0/1 (AUTO_REJECT/AUTO_KEEP)
-          sampling_priority = fetcher.number(@sampled_key)
+
+          trace_id = Helpers.parse_hex_id(fetcher[@trace_id_key])
 
           # Return early if this propagation is not valid
-          return unless trace_id && span_id
+          return if trace_id.nil? || trace_id <= 0 || trace_id > Tracing::Utils::TraceId::MAX
+
+          span_id = Helpers.parse_hex_id(fetcher[@span_id_key])
+
+          # Return early if this propagation is not valid
+          return if span_id.nil? || span_id <= 0 || span_id >= Tracing::Utils::EXTERNAL_MAX_ID
+
+          # We don't need to try and convert sampled since B3 supports 0/1 (AUTO_REJECT/AUTO_KEEP)
+          sampling_priority = Helpers.parse_decimal_id(fetcher[@sampled_key])
 
           TraceDigest.new(
             trace_id: trace_id,

lib/datadog/tracing/distributed/b3_single.rb

@@ -46,12 +46,15 @@ def extract(env)
           return unless value
 
           parts = value.split('-')
-          trace_id = Helpers.value_to_id(parts[0], base: 16) unless parts.empty?
-          span_id = Helpers.value_to_id(parts[1], base: 16) if parts.length > 1
-          sampling_priority = Helpers.value_to_number(parts[2]) if parts.length > 2
+          trace_id = Helpers.parse_hex_id(parts[0]) unless parts.empty?
+          # Return early if this propagation is not valid
+          return if trace_id.nil? || trace_id <= 0 || trace_id > Tracing::Utils::TraceId::MAX
 
-          # Return if this propagation is not valid
-          return unless trace_id && span_id
+          span_id = Helpers.parse_hex_id(parts[1]) if parts.length > 1
+          # Return early if this propagation is not valid
+          return if span_id.nil? || span_id <= 0 || span_id >= Tracing::Utils::EXTERNAL_MAX_ID
+
+          sampling_priority = Helpers.parse_decimal_id(parts[2]) if parts.length > 2
 
           TraceDigest.new(
             span_id: span_id,

lib/datadog/tracing/distributed/datadog.rb

@@ -3,6 +3,8 @@
 require_relative '../metadata/ext'
 require_relative '../trace_digest'
 require_relative 'datadog_tags_codec'
+require_relative '../utils'
+require_relative 'helpers'
 
 module Datadog
   module Tracing
@@ -38,21 +40,26 @@ def initialize(
         def inject!(digest, data)
           return if digest.nil?
 
-          data[@trace_id_key] = digest.trace_id.to_s
+          data[@trace_id_key] = Tracing::Utils::TraceId.to_low_order(digest.trace_id).to_s
+
           data[@parent_id_key] = digest.span_id.to_s
           data[@sampling_priority_key] = digest.trace_sampling_priority.to_s if digest.trace_sampling_priority
           data[@origin_key] = digest.trace_origin.to_s if digest.trace_origin
 
-          inject_tags(digest, data)
+          build_tags(digest).tap do |tags|
+            inject_tags!(tags, data) unless tags.empty?
+          end
 
           data
         end
 
         def extract(data)
           fetcher = @fetcher.new(data)
-          trace_id = fetcher.id(@trace_id_key)
-          parent_id = fetcher.id(@parent_id_key)
-          sampling_priority = fetcher.number(@sampling_priority_key)
+
+          trace_id  = parse_trace_id(fetcher)
+          parent_id = parse_parent_id(fetcher)
+
+          sampling_priority = Helpers.parse_decimal_id(fetcher[@sampling_priority_key])
           origin = fetcher[@origin_key]
 
           # Return early if this propagation is not valid
@@ -63,6 +70,10 @@ def extract(data)
 
           trace_distributed_tags = extract_tags(fetcher)
 
+          # If trace id is 128 bits long,
+          # Concatentated high order 64 bit hex-encoded `tid` tag
+          trace_id = extract_trace_id!(trace_id, trace_distributed_tags)
+
           TraceDigest.new(
             span_id: parent_id,
             trace_id: trace_id,
@@ -74,20 +85,57 @@ def extract(data)
 
         private
 
+        def parse_trace_id(fetcher_object)
+          trace_id = Helpers.parse_decimal_id(fetcher_object[@trace_id_key])
+
+          return unless trace_id
+          return if trace_id <= 0 || trace_id >= Tracing::Utils::EXTERNAL_MAX_ID
+
+          trace_id
+        end
+
+        def parse_parent_id(fetcher_object)
+          parent_id = Helpers.parse_decimal_id(fetcher_object[@parent_id_key])
+
+          return unless parent_id
+          return if parent_id <= 0 || parent_id >= Tracing::Utils::EXTERNAL_MAX_ID
+
+          parent_id
+        end
+
+        def build_tags(digest)
+          high_order = Tracing::Utils::TraceId.to_high_order(digest.trace_id)
+          tags = digest.trace_distributed_tags || {}
+
+          return tags if high_order == 0
+
+          tags.merge(Tracing::Metadata::Ext::Distributed::TAG_TID => high_order.to_s(16))
+        end
+
+        # Side effect: Remove high order 64 bit hex-encoded `tid` tag from distributed tags
+        def extract_trace_id!(trace_id, tags)
+          return trace_id unless tags
+          return trace_id unless (high_order = tags.delete(Tracing::Metadata::Ext::Distributed::TAG_TID))
+
+          Tracing::Utils::TraceId.concatenate(high_order.to_i(16), trace_id)
+        end
+
         # Export trace distributed tags through the `x-datadog-tags` key.
         #
         # DEV: This method accesses global state (the active trace) to record its error state as a trace tag.
         # DEV: This means errors cannot be reported if there's not active span.
         # DEV: Ideally, we'd have a dedicated error reporting stream for all of ddtrace.
-        def inject_tags(digest, data)
-          return if digest.trace_distributed_tags.nil? || digest.trace_distributed_tags.empty?
+        def inject_tags!(tags, data)
           return set_tags_propagation_error(reason: 'disabled') if tags_disabled?
 
-          tags = DatadogTagsCodec.encode(digest.trace_distributed_tags)
+          encoded_tags = DatadogTagsCodec.encode(tags)
 
-          return set_tags_propagation_error(reason: 'inject_max_size') if tags_too_large?(tags.size, scenario: 'inject')
+          return set_tags_propagation_error(reason: 'inject_max_size') if tags_too_large?(
+            encoded_tags.size,
+            scenario: 'inject'
+          )
 
-          data[@tags_key] = tags
+          data[@tags_key] = encoded_tags
         rescue => e
           set_tags_propagation_error(reason: 'encoding_error')
           ::Datadog.logger.warn(

lib/datadog/tracing/distributed/fetcher.rb

@@ -15,14 +15,6 @@ def initialize(data)
         def [](key)
           @data[key]
         end
-
-        def id(key, base: 10)
-          Helpers.value_to_id(self[key], base: base)
-        end
-
-        def number(key, base: 10)
-          Helpers.value_to_number(self[key], base: base)
-        end
       end
     end
   end

lib/datadog/tracing/distributed/helpers.rb

@@ -20,54 +20,42 @@ def self.clamp_sampling_priority(sampling_priority)
           sampling_priority
         end
 
-        def self.truncate_base16_number(value)
+        def self.parse_decimal_id(value)
+          return unless value
+
+          value = value.to_s
+
+          num   = value.to_i
+
+          # Ensure the parsed number is the same as the original string value
+          # e.g. We want to make sure to throw away `'nan'.to_i == 0`
+          return unless num.to_s(10) == value
+
+          num
+        end
+
+        def self.parse_hex_id(value)
+          return unless value
+
           # Lowercase if we want to parse base16 e.g. 3E8 => 3e8
           # DEV: Ruby will parse `3E8` just fine, but to test
           #      `num.to_s(base) == value` we need to lowercase
-          value = value.downcase
-
-          # Truncate to trailing 16 characters if length is greater than 16
-          # https://github.com/apache/incubator-zipkin/blob/21fe362899fef5c593370466bc5707d3837070c2/zipkin/src/main/java/zipkin2/storage/StorageComponent.java#L49-L53
-          # DEV: This ensures we truncate B3 128-bit trace and span ids to 64-bit
-          value = value[value.length - 16, 16] if value.length > 16
+          value = value.to_s.downcase
 
           # Remove any leading zeros
           # DEV: When we call `num.to_s(16)` later Ruby will not add leading zeros
           #      for us so we want to make sure the comparision will work as expected
           # DEV: regex, remove all leading zeros up until we find the last 0 in the string
           #      or we find the first non-zero, this allows `'0000' -> '0'` and `'00001' -> '1'`
-          value.sub(/^0*(?=(0$)|[^0])/, '')
-        end
-
-        def self.value_to_id(value, base: 10)
-          id = value_to_number(value, base: base)
-
-          # Return early if we could not parse a number
-          return if id.nil?
-
-          # Zero or greater than max allowed value of 2**64
-          return if id.zero? || id > Tracing::Utils::EXTERNAL_MAX_ID
-
-          id < 0 ? id + (2**64) : id
-        end
-
-        def self.value_to_number(value, base: 10)
-          # It's important to make a difference between no data and zero.
-          return if value.nil?
-
-          # Be sure we have a string
-          value = value.to_s
-
-          # If we are parsing base16 number then truncate to 64-bit
-          value = Helpers.truncate_base16_number(value) if base == 16
+          value = value.sub(/^0*(?=(0$)|[^0])/, '')
 
           # Convert value to an integer
           # DEV: Ruby `.to_i` will return `0` if a number could not be parsed
-          num = value.to_i(base)
+          num = value.to_i(16)
 
           # Ensure the parsed number is the same as the original string value
           # e.g. We want to make sure to throw away `'nan'.to_i == 0`
-          return unless num.to_s(base) == value
+          return unless num.to_s(16) == value
 
           num
         end