statsd-instrument 3.9.5 → 3.9.6

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 90a8111f77847ab11ea592d7b4a15c91a1d52b1f706b9296f29afa0c943658e8
-  data.tar.gz: 444ef926b623f13fa9cb453b4179969a3655071a20c39b0f429dee3f22272662
+  metadata.gz: 62a2601a6b425e15a8f3b616fcb182d48548179d95bc5b9bbdef74f5cdabcc1d
+  data.tar.gz: 64a4fd724bcee0e6fe8cac4001439c6a9e0b4b9e5234cd5b654b3df4045171cf
 SHA512:
-  metadata.gz: ead57ab89be89a51dabe4eb08c93fdd73cb5a439d5605f38305268b81e079a62c18555e1cf462ce1aa231b1928301dfe62376fa31f073c419408e478dd5cfef3
-  data.tar.gz: 372737145878f6b8eda3c1e2d498fd40dbb7f5aa38acf13e94b93b2ace670e2bf787162dcf54b1489fb9d057ea76511b07a83b03a3bbcc740edf1c64081b403f
+  metadata.gz: d74d19d5e29895763d423dd0daf67bedfeeb3e05ebff891f87eda90d4482648f5cd8bb3b163c000f2fe70473ece32f9f4febf36530bcfd16ab2d9d699cf1d49d
+  data.tar.gz: 18d287d1d24792d9e63d7d2f2db3bee74fd929d7e4fd5ea1dccb50447a7278de7cfe8f088184064298809e70ef82b3d719dc27c5d6dc07edbc01b5747dff7246

data/CHANGELOG.md

@@ -6,6 +6,11 @@ section below.
 
 ## Unreleased changes
 
+## Version 3.9.6
+
+- [#388](https://github.com/Shopify/statsd-instrument/pull/388) - Properly fixing the bug when using aggregation and sending sampled
+histograms, now the client will respect the sampling rate when sending the metrics and pass it down to the aggregator.
+
 ## Version 3.9.5
 
 - [#387](https://github.com/Shopify/statsd-instrument/pull/387) - Fixing bug when using aggregation and sending sampled

data/lib/statsd/instrument/aggregator.rb

@@ -138,7 +138,7 @@ module StatsD
       def aggregate_timing(name, value, tags: [], no_prefix: false, type: DISTRIBUTION, sample_rate: CONST_SAMPLE_RATE)
         unless thread_healthcheck
           @sink << datagram_builder(no_prefix: no_prefix).timing_value_packed(
-            name, type.to_s, [value], CONST_SAMPLE_RATE, tags
+            name, type.to_s, [value], sample_rate, tags
           )
           return
         end

data/lib/statsd/instrument/client.rb

@@ -318,25 +318,27 @@ module StatsD
       # @param tags (see #increment)
       # @return [void]
       def distribution(name, value = nil, sample_rate: nil, tags: nil, no_prefix: false, &block)
+        if block_given?
+          return latency(name, sample_rate: sample_rate, tags: tags, metric_type: :d, no_prefix: no_prefix, &block)
+        end
+
+        # For all timing metrics, we have to use the sampling logic.
+        # Not doing so would impact performance and CPU usage.
+        # See Datadog's documentation for more details: https://github.com/DataDog/datadog-go/blob/20af2dbfabbbe6bd0347780cd57ed931f903f223/statsd/aggregator.go#L281-L283
         sample_rate ||= @default_sample_rate
         if sample_rate && !sample?(sample_rate)
-          # For all timing metrics, we have to use the sampling logic.
-          # Not doing so would impact performance and CPU usage.
-          # See Datadog's documentation for more details: https://github.com/DataDog/datadog-go/blob/20af2dbfabbbe6bd0347780cd57ed931f903f223/statsd/aggregator.go#L281-L283
-
-          if block_given?
-            return yield
-          end
-
           return StatsD::Instrument::VOID
         end
 
-        if block_given?
-          return latency(name, sample_rate: sample_rate, tags: tags, metric_type: :d, no_prefix: no_prefix, &block)
-        end
-
         if @enable_aggregation
-          @aggregator.aggregate_timing(name, value, tags: tags, no_prefix: no_prefix, type: :d)
+          @aggregator.aggregate_timing(
+            name,
+            value,
+            tags: tags,
+            no_prefix: no_prefix,
+            type: :d,
+            sample_rate: sample_rate,
+          )
           return StatsD::Instrument::VOID
         end
 
@@ -392,13 +394,26 @@ module StatsD
         ensure
           stop = Process.clock_gettime(Process::CLOCK_MONOTONIC, :float_millisecond)
 
-          metric_type ||= datagram_builder(no_prefix: no_prefix).latency_metric_type
-          latency_in_ms = stop - start
-          if @enable_aggregation
-            @aggregator.aggregate_timing(name, latency_in_ms, tags: tags, no_prefix: no_prefix, type: metric_type)
-          else
-            sample_rate ||= @default_sample_rate
-            if sample_rate.nil? || sample?(sample_rate)
+          # For all timing metrics, we have to use the sampling logic.
+          # Not doing so would impact performance and CPU usage.
+          # See Datadog's documentation for more details:
+          # https://github.com/DataDog/datadog-go/blob/20af2dbfabbbe6bd0347780cd57ed931f903f223/statsd/aggregator.go#L281-L283
+          sample_rate ||= @default_sample_rate
+          if sample_rate.nil? || sample?(sample_rate)
+
+            metric_type ||= datagram_builder(no_prefix: no_prefix).latency_metric_type
+            latency_in_ms = stop - start
+
+            if @enable_aggregation
+              @aggregator.aggregate_timing(
+                name,
+                latency_in_ms,
+                tags: tags,
+                no_prefix: no_prefix,
+                type: metric_type,
+                sample_rate: sample_rate,
+              )
+            else
               emit(datagram_builder(no_prefix: no_prefix).send(metric_type, name, latency_in_ms, sample_rate, tags))
             end
           end

data/lib/statsd/instrument/version.rb

@@ -2,6 +2,6 @@
 
 module StatsD
   module Instrument
-    VERSION = "3.9.5"
+    VERSION = "3.9.6"
   end
 end

data/test/aggregator_test.rb

@@ -1,7 +1,6 @@
 # frozen_string_literal: true
 
 require "test_helper"
-require "ostruct"
 
 class AggregatorTest < Minitest::Test
   class CaptureLogger
@@ -70,6 +69,7 @@ class AggregatorTest < Minitest::Test
     sampled_datagram = @sink.datagrams.find { |d| d.name == "timing.sampled" }
     assert_equal([60.0, 80.0], sampled_datagram.value)
     assert_equal(0.01, sampled_datagram.sample_rate)
+    assert_equal("timing.sampled:60.0:80.0|d|@0.01", sampled_datagram.source)
 
     unsampled_datagram = @sink.datagrams.find { |d| d.name == "timing.unsampled" }
     assert_equal(60.0, unsampled_datagram.value)
@@ -208,7 +208,7 @@ class AggregatorTest < Minitest::Test
 
     # Additional metrics should also go through synchronously
     @subject.increment("foo", 1, tags: { foo: "bar" })
-    @subject.aggregate_timing("bar", 200, tags: { foo: "bar" })
+    @subject.aggregate_timing("bar", 200, tags: { foo: "bar" }, sample_rate: 0.5)
 
     # Verify new metrics were also sent immediately
     assert_equal(5, @sink.datagrams.size)
@@ -220,6 +220,7 @@ class AggregatorTest < Minitest::Test
     timing_datagram = @sink.datagrams.select { |d| d.name == "bar" }.last
     assert_equal([200.0], [timing_datagram.value])
     assert_equal(["foo:bar"], timing_datagram.tags)
+    assert_equal(0.5, timing_datagram.sample_rate)
 
     # undo the stubbing
     @subject.unstub(:thread_healthcheck)

data/test/client_test.rb

@@ -90,7 +90,7 @@ class ClientTest < Minitest::Test
     client.measure("block_duration_example") { 1 + 1 }
     client.force_flush
 
-    datagram = client.sink.datagrams.first
+    datagram = client.sink.datagrams.find { |d| d.name == "bar.foo" }
     assert_equal("bar.foo", datagram.name)
     assert_equal(2, datagram.value)
 
@@ -249,12 +249,17 @@ class ClientTest < Minitest::Test
     mock_sink = mock("sink")
     mock_sink.stubs(:sample?).returns(false, true, false, false, true)
     # Since we are aggregating, we only expect a single datagram.
-    mock_sink.expects(:<<).with("metric:60:60|d").once
+    mock_sink.expects(:<<).with("metric:60:60|d|@0.5").once
     mock_sink.expects(:flush).once
 
     client = StatsD::Instrument::Client.new(sink: mock_sink, default_sample_rate: 0.5, enable_aggregation: true)
     5.times { client.distribution("metric", 60) }
     client.force_flush
+
+    # undo mock
+    mock_sink.unstub(:sample?)
+    mock_sink.unstub(:<<)
+    mock_sink.unstub(:flush)
   end
 
   def test_clone_with_prefix_option

data/test/integration_test.rb

@@ -77,4 +77,30 @@ class IntegrationTest < Minitest::Test
     assert_equal("counter:20|c", @server.recvfrom(100).first)
     assert_operator(Time.now - before_flush, :<, 0.3, "Flush and ingest should have happened within 0.4s")
   end
+
+  def test_live_local_udp_socket_with_aggregation_sampled_scenario
+    client = StatsD::Instrument::Environment.new(
+      "STATSD_ADDR" => "#{@server.addr[2]}:#{@server.addr[1]}",
+      "STATSD_IMPLEMENTATION" => "dogstatsd",
+      "STATSD_ENV" => "production",
+      "STATSD_ENABLE_AGGREGATION" => "true",
+      "STATSD_AGGREGATION_INTERVAL" => "0.1",
+    ).client
+
+    100.times do
+      client.increment("counter", 2)
+      client.distribution("test_distribution", 3, sample_rate: 0.1)
+    end
+
+    sleep(0.2)
+
+    packets = []
+    while IO.select([@server], nil, nil, 0.1)
+      packets << @server.recvfrom(300).first
+    end
+    packets = packets.map { |packet| packet.split("\n") }.flatten
+
+    assert_match(/counter:\d+|c/, packets.find { |packet| packet.start_with?("counter:") })
+    assert_match(/test_distribution:\d+:3|d/, packets.find { |packet| packet.start_with?("test_distribution:") })
+  end
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: statsd-instrument
 version: !ruby/object:Gem::Version
-  version: 3.9.5
+  version: 3.9.6
 platform: ruby
 authors:
 - Jesse Storimer
@@ -10,7 +10,7 @@ authors:
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2024-10-30 00:00:00.000000000 Z
+date: 2024-10-31 00:00:00.000000000 Z
 dependencies: []
 description: A StatsD client for Ruby apps. Provides metaprogramming methods to inject
   StatsD instrumentation into your code.