feat(otel): otel sampling support in distributed tracing

config.m4

@@ -331,6 +331,7 @@ if test "$PHP_DDTRACE" != "no" && test "$PHP_DDTRACE_PROFILING" = "no"; then
     tracer/live_debugger.c \
     tracer/limiter/limiter.c \
     tracer/memory_limit.c \
+    tracer/otel_sampling.c \
     tracer/tracer_otel_config.c \
     tracer/priority_sampling/priority_sampling.c \
     tracer/profiling.c \

config.w32

@@ -70,6 +70,7 @@ if (PHP_DDTRACE != 'no') {
 	DDTRACE_TRACER_SOURCES += " ip_extraction.c";
 	DDTRACE_TRACER_SOURCES += " live_debugger.c";
 	DDTRACE_TRACER_SOURCES += " memory_limit.c";
+	DDTRACE_TRACER_SOURCES += " otel_sampling.c";
 	DDTRACE_TRACER_SOURCES += " tracer_otel_config.c";
 	DDTRACE_TRACER_SOURCES += " profiling.c";
 	DDTRACE_TRACER_SOURCES += " random.c";

tests/OpenTelemetry/Integration/API/TracerTest.php

@@ -538,9 +538,9 @@ public function testGetSpanContextWithMultipleTraceStates()
                 '_dd.p.congo' => 't61rcWkgMzE',
                 '_dd.p.some_val' => 'tehehe'
             ]);
-            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.congo:t61rcWkgMzE;t.some_val:tehehe;t.dm:-0$/', (string)$span->getContext()->getTraceState());
+            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.congo:t61rcWkgMzE;t.some_val:tehehe;t.dm:-0,ot=rv:[0-9a-f]{14};th:0$/', (string)$span->getContext()->getTraceState());
             $span->end();
-            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.congo:t61rcWkgMzE;t.some_val:tehehe;t.dm:-0$/', (string)$span->getContext()->getTraceState());
+            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.congo:t61rcWkgMzE;t.some_val:tehehe;t.dm:-0,ot=rv:[0-9a-f]{14};th:0$/', (string)$span->getContext()->getTraceState());
         });
 
         $span = $traces[0][0];
@@ -659,11 +659,11 @@ public function getDescription(): string
             )))->getTracer('OpenTelemetry.TracerTest');
             $parent = $tracer->spanBuilder("parent")->startSpan(); // root sampler will be used
             $scope = $parent->activate();
-            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.dm:-0,root=yes,rojo=00f067aa0ba902b7,congo=t61rcWkgMzE$/', (string)$parent->getContext()->getTraceState());
+            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.dm:-0,ot=rv:[0-9a-f]{14};th:0,root=yes,rojo=00f067aa0ba902b7,congo=t61rcWkgMzE$/', (string)$parent->getContext()->getTraceState());
             $parent->setAttributes([
                 '_dd.p.some_val' => 'tehehe'
             ]);
-            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.dm:-0;t.some_val:tehehe,root=yes,rojo=00f067aa0ba902b7,congo=t61rcWkgMzE$/', (string)$parent->getContext()->getTraceState());
+            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.dm:-0;t.some_val:tehehe,ot=rv:[0-9a-f]{14};th:0,root=yes,rojo=00f067aa0ba902b7,congo=t61rcWkgMzE$/', (string)$parent->getContext()->getTraceState());
             try {
                 $child = $tracer->spanBuilder("child")->startSpan(); // local parent sampler will be used
 
@@ -673,7 +673,7 @@ public function getDescription(): string
                 $this->assertFalse($childContext->isRemote()); // "When creating children from remote spans, their IsRemote flag MUST be set to false."
                 $this->assertEquals(1, $childContext->getTraceFlags()); // RECORD_AND_SAMPLED ==> 01 (AlwaysOn sampler)
                 $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.dm:-0,localparent=yes,rojo=00f067aa0ba902b7,congo=t61rcWkgMzE$/', (string)$childContext->getTraceState());
-                $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.dm:-0;t.some_val:tehehe,root=yes,rojo=00f067aa0ba902b7,congo=t61rcWkgMzE$/', (string)$parent->getContext()->getTraceState());
+                $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.dm:-0;t.some_val:tehehe,ot=rv:[0-9a-f]{14};th:0,root=yes,rojo=00f067aa0ba902b7,congo=t61rcWkgMzE$/', (string)$parent->getContext()->getTraceState());
 
                 $grandChild = $tracer->spanBuilder("grandChild")
                     ->setParent(Context::getCurrent()->withContextValue($child))
@@ -692,7 +692,7 @@ public function getDescription(): string
 
                 $child->end();
             } finally {
-                $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.dm:-0;t.some_val:tehehe,root=yes,rojo=00f067aa0ba902b7,congo=t61rcWkgMzE$/', (string)$parent->getContext()->getTraceState());
+                $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.dm:-0;t.some_val:tehehe,ot=rv:[0-9a-f]{14};th:0,root=yes,rojo=00f067aa0ba902b7,congo=t61rcWkgMzE$/', (string)$parent->getContext()->getTraceState());
                 $scope->detach();
                 $parent->end();
             }
@@ -851,7 +851,7 @@ public function testAddItemToTracestate()
                 '_dd.p.congo' => 't61rcWkgMzE',
             ]);
 
-            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.congo:t61rcWkgMzE;t.dm:-0$/', (string)$span->getContext()->getTraceState());
+            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.congo:t61rcWkgMzE;t.dm:-0,ot=rv:[0-9a-f]{14};th:0$/', (string)$span->getContext()->getTraceState());
 
             $traceState = $span->getContext()->getTraceState()->with('rojo', '00f067aa0ba902b7');
             $context = SpanContext::create(
@@ -865,7 +865,7 @@ public function testAddItemToTracestate()
                 ->setParent(Context::getCurrent()->withContextValue(Span::wrap($context)))
                 ->startSpan();
 
-            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.congo:t61rcWkgMzE;t.dm:-0,rojo=00f067aa0ba902b7$/', (string)$child->getContext()->getTraceState());
+            $this->assertRegularExpression('/^dd=p:[0-9a-f]{16};t.congo:t61rcWkgMzE;t.dm:-0,rojo=00f067aa0ba902b7,ot=rv:[0-9a-f]{14};th:0$/', (string)$child->getContext()->getTraceState());
 
             $child->end();
             $span->end();

tests/ext/distributed_tracing/distributed_trace_span_link_from_headers.phpt

@@ -16,7 +16,7 @@ DDTrace\SpanLink Object
 (
     [traceId] => 0000000000000000000000000000002a
     [spanId] => %s
-    [traceState] => dd=o:datadog;t.custom_tag:inherited;t.second_tag:bar;t.dm:-0
+    [traceState] => dd=o:datadog;t.custom_tag:inherited;t.second_tag:bar;t.dm:-0,ot=rv:%x;th:0
     [attributes] => Array
         (
             [_dd.p.custom_tag] => inherited

tests/ext/distributed_tracing/distributed_trace_span_link_from_headers_long_origin.phpt

@@ -12,7 +12,7 @@ datadog.trace.propagation_style_extract=datadog
 --FILE--
 <?php
 
-$origin = str_repeat("=", 1000);
+$origin = str_repeat("=", 500);
 $link = DDTrace\SpanLink::fromHeaders([
     "x-datadog-trace-id" => "42",
     "x-datadog-parent-id" => "1",

tests/ext/distributed_tracing/otel_sampling_caps.phpt

@@ -0,0 +1,65 @@
+--TEST--
+OpenTelemetry tracestate sampling honors value, member, and byte caps
+--ENV--
+DD_TRACE_SAMPLE_RATE=0.5
+DD_TRACE_RATE_LIMIT=10000000
+--FILE--
+<?php
+
+function headersWithTracestate(string $tracestate): array
+{
+    $span = DDTrace\start_span();
+    $root = DDTrace\root_span();
+    $root->traceId = str_pad('1', 32, '0', STR_PAD_LEFT);
+    $root->tracestate = $tracestate;
+    $headers = DDTrace\generate_distributed_tracing_headers(['tracecontext']);
+    DDTrace\close_span();
+    return $headers;
+}
+
+$vendors = [];
+for ($i = 0; $i < 32; ++$i) {
+    $vendors[] = "vendor{$i}=value";
+}
+$tracestate = headersWithTracestate(implode(',', $vendors))['tracestate'];
+$members = explode(',', $tracestate);
+echo 'members=', count($members), ' leading=', implode(',', array_map(
+    function (string $member): string {
+        return strstr($member, '=', true);
+    },
+    array_slice($members, 0, 2)
+)), PHP_EOL;
+
+$largeVendors = [];
+for ($i = 0; $i < 32; ++$i) {
+    $largeVendors[] = "vendor{$i}=" . str_repeat('x', 30);
+}
+$tracestate = headersWithTracestate(implode(',', $largeVendors))['tracestate'];
+echo 'bytes=', strlen($tracestate) <= 512 ? 'within-cap' : 'over-cap',
+    ' complete=', substr($tracestate, -1) === 'x' ? 'yes' : 'no', PHP_EOL;
+
+$largeDatadog = 'dd=p:0000000000000001;t.large:' . str_repeat('x', 470);
+$tracestate = headersWithTracestate($largeDatadog)['tracestate'];
+$members = explode(',', $tracestate);
+echo 'owned-bytes=', strlen($tracestate) <= 512 ? 'within-cap' : 'over-cap',
+    ' leading=', implode(',', array_map(
+        function (string $member): string {
+            return strstr($member, '=', true);
+        },
+        array_slice($members, 0, 2)
+    )),
+    ' large=', strpos($tracestate, 't.large:') === false ? 'dropped' : 'kept', PHP_EOL;
+
+$oversizedUnknown = 'future:' . str_repeat('x', 230) . ';next:value';
+$tracestate = headersWithTracestate('ot=' . $oversizedUnknown)['tracestate'];
+preg_match('/(?:^|,)ot=([^,]+)/', $tracestate, $matches);
+echo 'ot-bytes=', strlen($matches[1]),
+    ' future=', strpos($matches[1], 'future:') === false ? 'dropped' : 'kept',
+    ' next=', strpos($matches[1], 'next:value') === false ? 'dropped' : 'kept', PHP_EOL;
+
+?>
+--EXPECTF--
+members=32 leading=dd,ot
+bytes=within-cap complete=yes
+owned-bytes=within-cap leading=dd,ot large=dropped
+ot-bytes=33 future=dropped next=kept

tests/ext/distributed_tracing/otel_sampling_inherited.phpt

@@ -0,0 +1,82 @@
+--TEST--
+Inherited OpenTelemetry tracestate sampling fields are validated and forwarded
+--ENV--
+DD_TRACE_SAMPLE_RATE=0.5
+DD_TRACE_RATE_LIMIT=10000000
+DD_TRACE_SAMPLING_RULES=[{"sample_rate":0.1,"service":"locally-decided","target_span":"any"}]
+--FILE--
+<?php
+
+function propagate(string $tracestate, bool $sampled = true, bool $manualKeep = false): string
+{
+    $span = DDTrace\start_span();
+    DDTrace\consume_distributed_tracing_headers([
+        'traceparent' => '00-0000000000000000fff972474538efff-0000000000000001-0' . ($sampled ? '1' : '0'),
+        'tracestate' => $tracestate,
+    ]);
+    if ($manualKeep) {
+        DDTrace\set_priority_sampling(DD_TRACE_PRIORITY_SAMPLING_USER_KEEP);
+    }
+
+    $headers = DDTrace\generate_distributed_tracing_headers(['tracecontext']);
+    DDTrace\close_span();
+    return $headers['tracestate'];
+}
+
+function ot(string $tracestate): string
+{
+    foreach (explode(',', $tracestate) as $member) {
+        $member = ltrim($member, " \t");
+        if (strncmp($member, 'ot=', 3) === 0) {
+            return substr($member, 3);
+        }
+    }
+    return '<absent>';
+}
+
+function locallyDecide(string $tracestate): string
+{
+    $span = DDTrace\start_span();
+    DDTrace\consume_distributed_tracing_headers([
+        'traceparent' => '00-0000000000000000fff972474538efff-0000000000000001-00',
+        'tracestate' => $tracestate,
+    ]);
+    $span->service = 'locally-decided';
+
+    $headers = DDTrace\generate_distributed_tracing_headers(['tracecontext']);
+    DDTrace\close_span();
+    return $headers['tracestate'];
+}
+
+echo ot(propagate('dd=s:2;t.dm:-3,ot=rv:ef284ace7a91e1;th:e6666666666668;foo:bar')), PHP_EOL;
+echo ot(propagate('ot=th:e6666666666668')), PHP_EOL;
+echo ot(propagate('dd=s:0,ot=rv:ef284ace7a91e1;th:e6666666666668', false)), PHP_EOL;
+echo ot(propagate('dd=s:1,ot=rv:not-hex;th:not-hex,congo=xyz123')), PHP_EOL;
+echo ot(propagate('ot=rv:1234567890abcd;th:not-hex')), PHP_EOL;
+echo ot(propagate('dd=s:1')), PHP_EOL;
+echo ot(propagate('ot=foo:bar')), PHP_EOL;
+echo ot(propagate('ot=rv:65cd67504a538e;th:e6666666666668', false, true)), PHP_EOL;
+echo ot(propagate('', false, true)), PHP_EOL;
+
+$ordered = propagate('dd=s:1,foo=bar,ot=rv:6e6d1a75832a2f,something=else');
+echo substr($ordered, strpos($ordered, ',') + 1), PHP_EOL;
+
+echo ot(locallyDecide('ot=rv:00000000000000;th:f0000000000000;foo:bar')), PHP_EOL;
+echo ot(locallyDecide('ot=rv:00000000000000;foo:bar')), PHP_EOL;
+echo ot(locallyDecide('ot=th:f0000000000000;foo:bar')), PHP_EOL;
+
+?>
+--EXPECT--
+rv:ef284ace7a91e1;th:e6666666666668;foo:bar
+th:e6666666666668
+rv:ef284ace7a91e1;th:e6666666666668
+<absent>
+rv:1234567890abcd
+<absent>
+foo:bar
+rv:65cd67504a538e
+<absent>
+foo=bar,ot=rv:6e6d1a75832a2f,something=else
+rv:ef284ace7a91e1;th:e6666666666668;foo:bar
+rv:ef284ace7a91e1;th:e6666666666668;foo:bar
+rv:ef284ace7a91e1;th:e6666666666668;foo:bar

tests/ext/distributed_tracing/otel_sampling_limiter.phpt

@@ -0,0 +1,30 @@
+--TEST--
+OpenTelemetry threshold is not emitted when the trace rate limiter makes the drop decision
+--SKIPIF--
+<?php if (getenv('USE_ZEND_ALLOC') === '0') die('skip timing sensitive test, does not make sense with valgrind'); ?>
+--ENV--
+DD_TRACE_AUTO_FLUSH_ENABLED=0
+DD_TRACE_GENERATE_ROOT_SPAN=0
+DD_TRACE_RATE_LIMIT=10
+DD_TRACE_SAMPLE_RATE=1
+--FILE--
+<?php
+
+for ($i = 0; $i < 1000; ++$i) {
+    DDTrace\start_span();
+    $headers = DDTrace\generate_distributed_tracing_headers(['tracecontext']);
+    DDTrace\close_span();
+    dd_trace_serialize_closed_spans();
+
+    $traceFlags = hexdec(substr($headers['traceparent'], -2));
+    if (($traceFlags & 1) === 0) {
+        echo strpos($headers['tracestate'], 'ot=') === false ? "OK\n" : "unexpected ot member\n";
+        return;
+    }
+}
+
+echo "rate limiter did not reject a trace\n";
+
+?>
+--EXPECT--
+OK

tests/ext/distributed_tracing/otel_sampling_precision.phpt

@@ -0,0 +1,21 @@
+--TEST--
+OpenTelemetry random value is reconciled with the 64-bit Datadog sampling decision
+--ENV--
+DD_TRACE_SAMPLE_RATE=0.1
+DD_TRACE_RATE_LIMIT=10000000
+--FILE--
+<?php
+
+$span = DDTrace\start_span();
+$root = DDTrace\root_span();
+$root->traceId = str_pad('03a93ee8b1999f00', 32, '0', STR_PAD_LEFT);
+
+$headers = DDTrace\generate_distributed_tracing_headers(['tracecontext']);
+preg_match('/(?:^|,)ot=([^,]+)/', $headers['tracestate'], $matches);
+echo $matches[1], ' sampled=', substr($headers['traceparent'], -1), PHP_EOL;
+
+DDTrace\close_span();
+
+?>
+--EXPECT--
+rv:e6666666666668;th:e6666666666668 sampled=1

tests/ext/distributed_tracing/otel_sampling_precision_drop.phpt

@@ -0,0 +1,21 @@
+--TEST--
+OpenTelemetry random value is reconciled for a 64-bit Datadog drop decision
+--ENV--
+DD_TRACE_SAMPLE_RATE=0.05
+DD_TRACE_RATE_LIMIT=10000000
+--FILE--
+<?php
+
+$span = DDTrace\start_span();
+$root = DDTrace\root_span();
+$root->traceId = str_pad(dechex(5401449561355763072), 32, '0', STR_PAD_LEFT);
+
+$headers = DDTrace\generate_distributed_tracing_headers(['tracecontext']);
+preg_match('/(?:^|,)ot=([^,]+)/', $headers['tracestate'], $matches);
+echo $matches[1], ' sampled=', substr($headers['traceparent'], -1), PHP_EOL;
+
+DDTrace\close_span();
+
+?>
+--EXPECT--
+rv:f333333333332f;th:f333333333333 sampled=0

tests/ext/distributed_tracing/otel_sampling_probability.phpt

@@ -0,0 +1,28 @@
+--TEST--
+OpenTelemetry tracestate fields are emitted for probability sampling decisions
+--ENV--
+DD_TRACE_SAMPLE_RATE=0.5
+DD_TRACE_RATE_LIMIT=10000000
+--FILE--
+<?php
+
+function sample(string $traceId)
+{
+    $span = DDTrace\start_span();
+    $root = DDTrace\root_span();
+    $root->traceId = str_pad($traceId, 32, '0', STR_PAD_LEFT);
+
+    $headers = DDTrace\generate_distributed_tracing_headers(['tracecontext']);
+    preg_match('/(?:^|,)ot=([^,]+)/', $headers['tracestate'], $matches);
+    echo $matches[1], ' sampled=', substr($headers['traceparent'], -1), PHP_EOL;
+
+    DDTrace\close_span();
+}
+
+sample('1');
+sample('a');
+
+?>
+--EXPECT--
+rv:f0948a54d43b8e;th:8 sampled=1
+rv:65cd67504a538e;th:8 sampled=0

tests/ext/distributed_tracing/otel_sampling_propagation_behavior_ignore.phpt

@@ -0,0 +1,34 @@
+--TEST--
+OpenTelemetry sampling state is regenerated when extraction ignores the context
+--ENV--
+DD_TRACE_GENERATE_ROOT_SPAN=0
+DD_TRACE_PROPAGATION_BEHAVIOR_EXTRACT=ignore
+DD_TRACE_PROPAGATION_STYLE_EXTRACT=tracecontext
+DD_TRACE_SAMPLE_RATE=0.5
+DD_TRACE_RATE_LIMIT=10000000
+--FILE--
+<?php
+
+$inboundTraceId = '0000000000000000fff972474538efff';
+$inboundRandomValue = '1234567890abcd';
+DDTrace\consume_distributed_tracing_headers([
+    'traceparent' => "00-$inboundTraceId-0000000000000001-01",
+    'tracestate' => "ot=rv:$inboundRandomValue;th:e6666666666668",
+]);
+
+DDTrace\start_span();
+$headers = DDTrace\generate_distributed_tracing_headers(['tracecontext']);
+preg_match('/(?:^|,)ot=([^,]+)/', $headers['tracestate'], $matches);
+$ot = $matches[1] ?? '';
+
+echo 'new trace: ', DDTrace\root_span()->traceId === $inboundTraceId ? 'no' : 'yes', PHP_EOL;
+echo 'new probability state: ', preg_match('/^rv:[0-9a-f]{14};th:8$/', $ot) ? 'yes' : 'no', PHP_EOL;
+echo 'inbound rv removed: ', strpos($ot, $inboundRandomValue) === false ? 'yes' : 'no', PHP_EOL;
+
+DDTrace\close_span();
+
+?>
+--EXPECT--
+new trace: yes
+new probability state: yes
+inbound rv removed: yes