feat: emit inferred aws.alb span for ALB-triggered Lambdas

datadog_lambda/tracing.py

@@ -837,6 +837,9 @@ def create_inferred_span(
         elif event_source.equals(EventTypes.LAMBDA_FUNCTION_URL):
             logger.debug("Function URL event detected. Inferring a span")
             return create_inferred_span_from_lambda_function_url_event(event, context)
+        elif event_source.equals(EventTypes.ALB, subtype=EventSubtypes.ALB):
+            logger.debug("ALB event detected. Inferring a span")
+            return create_inferred_span_from_alb_event(event, context)
         elif event_source.equals(
             EventTypes.API_GATEWAY, subtype=EventSubtypes.HTTP_API
         ):
@@ -952,6 +955,56 @@ def create_inferred_span_from_lambda_function_url_event(event, context):
     return span
 
 
+def create_inferred_span_from_alb_event(event, context):
+    request_context = event.get("requestContext") or {}
+    elb = request_context.get("elb") or {}
+    target_group_arn = elb.get("targetGroupArn")
+
+    headers = event.get("headers")
+    if not isinstance(headers, dict):
+        headers = {}
+    host = headers.get("host")
+    method = event.get("httpMethod")
+    path = event.get("path")
+    proto = headers.get("x-forwarded-proto", "http")
+
+    # ALB has no api id; key the service mapping off the load-balancer host and
+    # fall back to it when DD_TRACE_AWS_SERVICE_REPRESENTATION_ENABLED is on.
+    service_name = determine_service_name(service_mapping, host, "lambda_alb", host)
+
+    http_url = f"{proto}://{host}{path}" if host and path is not None else None
+    if method and path is not None:
+        resource = f"{method} {path}"
+    else:
+        resource = method or path
+
+    tags = {
+        "operation_name": "aws.alb",
+        "span.kind": "server",
+        "http.method": method,
+        "http.url": http_url,
+        "http.useragent": headers.get("user-agent"),
+        "endpoint": path,
+        "resource_names": resource,
+        "request_id": context.aws_request_id,
+        "target_group_arn": target_group_arn,
+    }
+    # Drop tags we couldn't derive so the span never carries malformed values.
+    tags = {key: value for key, value in tags.items() if value is not None}
+
+    InferredSpanInfo.set_tags(tags, tag_source="self", synchronicity="sync")
+    tracer.set_tags(_dd_origin)
+    # ALB events carry no request timestamp (unlike API GW requestTimeEpoch /
+    # Function URL timeEpoch), so the span starts at handler time.
+    span = tracer.trace(
+        "aws.alb", service=service_name, resource=resource, span_type="http"
+    )
+    if span:
+        span.set_tags(tags)
+        span.set_metric(InferredSpanInfo.METRIC, 1.0)
+    return span
+
+
 def is_api_gateway_invocation_async(event):
     hdrs = event.get("headers")
     if not hdrs:

datadog_lambda/trigger.py

@@ -324,6 +324,25 @@ def extract_http_tags(event):
             path = apigateway_v2_http.get("path")
             method = apigateway_v2_http.get("method")
 
+    elif request_context and request_context.get("elb"):
+        # ALB events have no requestContext.stage; derive the URL from the
+        # forwarded host/proto headers and the top-level path.
+        alb_headers = event.get("headers")
+        if not isinstance(alb_headers, dict):
+            alb_headers = {}
+        host = alb_headers.get("host")
+        if host:
+            proto = alb_headers.get("x-forwarded-proto", "http")
+            http_tags["http.url"] = f"{proto}://{host}"
+
+        user_agent = alb_headers.get("user-agent")
+        if user_agent:
+            http_tags["http.useragent"] = user_agent
+
+        # ALB carries no route template, so use the request path as the route.
+        if path:
+            http_tags["http.route"] = path
+
     if path:
         if http_tags.get("http.url"):
             http_tags["http.url"] += path

tests/test_tracing.py

@@ -112,6 +112,10 @@ def _wrap(*args, **kwargs):
         "api-gateway-websocket-disconnect",
         Context(trace_id=12345, span_id=67890, sampling_priority=2),
     ),
+    (
+        "application-load-balancer",
+        Context(trace_id=12345, span_id=67890, sampling_priority=2),
+    ),
     (
         "authorizer-request-api-gateway-v1",
         Context(
@@ -1953,6 +1957,133 @@ def test_remaps_specific_inferred_span_service_names_from_eventbridge_event(
         self.assertEqual(span2.get_tag("operation_name"), "aws.eventbridge")
         self.assertEqual(span2.service, "different.eventbridge.custom.event.sender")
 
+    def test_remaps_all_inferred_span_service_names_from_alb_event(self):
+        self.set_service_mapping({"lambda_alb": "new-name"})
+        with open(f"{event_samples}application-load-balancer.json") as event:
+            original_event = json.load(event)
+
+        ctx = get_mock_context()
+        ctx.aws_request_id = "123"
+
+        span1 = create_inferred_span(original_event, ctx)
+        self.assertEqual(span1.get_tag("operation_name"), "aws.alb")
+        self.assertEqual(span1.service, "new-name")
+
+        event2 = copy.deepcopy(original_event)
+        event2["headers"]["host"] = "different-alb.us-east-2.elb.amazonaws.com"
+        span2 = create_inferred_span(event2, ctx)
+        self.assertEqual(span2.get_tag("operation_name"), "aws.alb")
+        self.assertEqual(span2.service, "new-name")
+
+    def test_remaps_specific_inferred_span_service_names_from_alb_event(self):
+        host = "lambda-alb-123578498.us-east-2.elb.amazonaws.com"
+        self.set_service_mapping({host: "mapped-alb-service"})
+        with open(f"{event_samples}application-load-balancer.json") as event:
+            original_event = json.load(event)
+
+        ctx = get_mock_context()
+        ctx.aws_request_id = "123"
+
+        span1 = create_inferred_span(original_event, ctx)
+        self.assertEqual(span1.get_tag("operation_name"), "aws.alb")
+        self.assertEqual(span1.service, "mapped-alb-service")
+
+        event2 = copy.deepcopy(original_event)
+        event2["headers"]["host"] = "other-alb.us-east-2.elb.amazonaws.com"
+        span2 = create_inferred_span(event2, ctx)
+        self.assertEqual(span2.get_tag("operation_name"), "aws.alb")
+        self.assertEqual(span2.service, "other-alb.us-east-2.elb.amazonaws.com")
+
+
+class TestAlbInferredSpan(unittest.TestCase):
+    ALB_SAMPLE = "application-load-balancer"
+    ALB_MULTIVALUE = "application-load-balancer-multivalue-headers"
+    ALB_HOST = "lambda-alb-123578498.us-east-2.elb.amazonaws.com"
+    ALB_USER_AGENT = (
+        "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
+        "(KHTML, like Gecko) Chrome/71.0.3578.98 Safari/537.36"
+    )
+
+    def _load_event(self, sample_name):
+        with open(f"{event_samples}{sample_name}.json") as event_file:
+            return json.load(event_file)
+
+    def test_create_inferred_span_from_alb_event(self):
+        event = self._load_event(self.ALB_SAMPLE)
+        ctx = get_mock_context(aws_request_id="123")
+
+        span = create_inferred_span(event, ctx)
+
+        self.assertIsNotNone(span)
+        self.assertEqual(span.name, "aws.alb")
+        self.assertEqual(span.span_type, "http")
+        self.assertEqual(span.service, self.ALB_HOST)
+        self.assertEqual(span.resource, "GET /lambda")
+        self.assertEqual(span.get_tag("operation_name"), "aws.alb")
+        self.assertEqual(span.get_tag("span.kind"), "server")
+        self.assertEqual(span.get_tag("http.method"), "GET")
+        self.assertEqual(
+            span.get_tag("http.url"), f"http://{self.ALB_HOST}/lambda"
+        )
+        self.assertEqual(span.get_tag("http.useragent"), self.ALB_USER_AGENT)
+        self.assertEqual(span.get_tag("endpoint"), "/lambda")
+        self.assertEqual(span.get_tag("resource_names"), "GET /lambda")
+        self.assertEqual(span.get_tag("request_id"), "123")
+        self.assertEqual(span.get_tag("_inferred_span.synchronicity"), "sync")
+        self.assertEqual(span.get_tag("_inferred_span.tag_source"), "self")
+        self.assertEqual(span.get_metric("_dd._inferred_span"), 1.0)
+        self.assertEqual(
+            span.get_tag("target_group_arn"),
+            "arn:aws:elasticloadbalancing:us-east-2:123456789012:targetgroup/lambda-xyz/123abc",
+        )
+
+    def test_create_inferred_span_omits_tags_when_headers_missing(self):
+        event = self._load_event(self.ALB_SAMPLE)
+        del event["headers"]
+        event["httpMethod"] = None
+        event["path"] = None
+
+        span = create_inferred_span(event, get_mock_context())
+
+        self.assertIsNotNone(span)
+        self.assertNotIn("http.url", span.get_tags())
+        self.assertNotIn("http.method", span.get_tags())
+        self.assertNotIn("http.useragent", span.get_tags())
+
+    def test_multivalue_headers_subtype_returns_none(self):
+        event = self._load_event(self.ALB_MULTIVALUE)
+        span = create_inferred_span(event, get_mock_context())
+        self.assertIsNone(span)
+
+    @with_trace_propagation_style("datadog")
+    def test_inbound_datadog_context_parents_inferred_span(self):
+        event = self._load_event(self.ALB_SAMPLE)
+        ctx = get_mock_context()
+
+        parent_ctx, source, _ = extract_dd_trace_context(event, ctx)
+        set_dd_trace_py_root(source, merge_xray_traces=False)
+        span = create_inferred_span(event, ctx)
+
+        self.assertEqual(span.trace_id, parent_ctx.trace_id)
+        self.assertEqual(span.parent_id, parent_ctx.span_id)
+
+    def test_inbound_w3c_context_extracted_from_alb_event(self):
+        event = self._load_event(self.ALB_SAMPLE)
+        event["headers"] = {
+            "host": self.ALB_HOST,
+            "user-agent": self.ALB_USER_AGENT,
+            "x-forwarded-proto": "http",
+            "traceparent": "00-0000000000000000000000000000abcd-000000000000004d-01",
+            "tracestate": "dd=s:1",
+        }
+
+        ctx, source, _ = extract_dd_trace_context(event, get_mock_context())
+
+        self.assertIsNotNone(ctx)
+        self.assertEqual(source, TraceContextSource.EVENT)
+        self.assertEqual(ctx.trace_id, 0xABCD)
+        self.assertEqual(ctx.span_id, 0x4D)
+
 
 class _Span(object):
     def __init__(self, service, start, span_type, parent_name=None, tags=None):

tests/test_trigger.py

@@ -427,10 +427,28 @@ def test_extract_trigger_tags_application_load_balancer(self):
                 "function_trigger.event_source": "application-load-balancer",
                 "function_trigger.event_source_arn": "arn:aws:elasticloadbalancing:us-east-2:123456789012:targetgroup/lambda-xyz/123abc",
                 "http.method": "GET",
+                "http.url": "http://lambda-alb-123578498.us-east-2.elb.amazonaws.com/lambda",
+                "http.route": "/lambda",
+                "http.useragent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/71.0.3578.98 Safari/537.36",
                 "span.kind": "server",
             },
         )
 
+    def test_extract_trigger_tags_application_load_balancer_multivalue_headers(self):
+        event_sample_source = "application-load-balancer-multivalue-headers"
+        test_file = event_samples + event_sample_source + ".json"
+        ctx = get_mock_context()
+        with open(test_file, "r") as event:
+            event = json.load(event)
+        tags = extract_trigger_tags(event, ctx)
+
+        assert tags.get("function_trigger.event_source") == "application-load-balancer"
+        assert tags.get("http.method") == "GET"
+        assert tags.get("http.route") == "/lambda"
+        # multi-value subtype has no single-value ``headers`` map
+        assert "http.url" not in tags
+        assert "http.useragent" not in tags
+
     def test_extract_trigger_tags_cloudfront(self):
         event_sample_source = "cloudfront"
         test_file = event_samples + event_sample_source + ".json"

tests/test_wrapper.py

@@ -1001,3 +1001,109 @@ def test_profiling_import_errors_caught(monkeypatch):
     )  # force ModuleNotFoundError
     importlib.reload(wrapper)
     assert not hasattr(wrapper.datadog_lambda_wrapper, "prof")
+
+
+class TestAlbInferredSpanWrapper(unittest.TestCase):
+    """End-to-end wrapper tests for the inferred aws.alb span (FRSLES-851)."""
+
+    def setUp(self):
+        patch("ddtrace.internal.remoteconfig.worker.RemoteConfigPoller").start()
+        wrapper.datadog_lambda_wrapper._force_wrap = True
+
+        patcher = patch("datadog.threadstats.reporters.HttpReporter.flush_distributions")
+        self.mock_flush_distributions = patcher.start()
+        self.addCleanup(patcher.stop)
+
+        patcher = patch("datadog_lambda.xray.send_segment")
+        self.mock_send_segment = patcher.start()
+        self.addCleanup(patcher.stop)
+
+        patcher = patch("datadog_lambda.wrapper.create_dd_dummy_metadata_subsegment")
+        self.mock_create_dd_dummy_metadata_subsegment = patcher.start()
+        self.addCleanup(patcher.stop)
+
+        with open("tests/event_samples/application-load-balancer.json") as f:
+            self.alb_event = json.load(f)
+
+    def _alb_response(self, status_code=200):
+        return {
+            "statusCode": status_code,
+            "statusDescription": f"{status_code} OK",
+            "headers": {"Content-Type": "application/json"},
+            "body": "{}",
+            "isBase64Encoded": False,
+        }
+
+    @patch("datadog_lambda.config.Config.trace_enabled", True)
+    @patch("datadog_lambda.config.Config.make_inferred_span", True)
+    def test_wrapper_emits_inferred_alb_span_with_http_tags(self):
+        @wrapper.datadog_lambda_wrapper
+        def lambda_handler(event, context):
+            return self._alb_response(200)
+
+        lambda_handler(self.alb_event, get_mock_context())
+
+        inferred = lambda_handler.inferred_span
+        execution = lambda_handler.span
+
+        self.assertIsNotNone(inferred)
+        self.assertEqual(inferred.name, "aws.alb")
+        self.assertEqual(inferred.get_tag("operation_name"), "aws.alb")
+        self.assertEqual(inferred.get_tag("http.method"), "GET")
+        self.assertEqual(
+            inferred.get_tag("http.url"),
+            "http://lambda-alb-123578498.us-east-2.elb.amazonaws.com/lambda",
+        )
+        self.assertEqual(inferred.get_tag("http.status_code"), "200")
+        self.assertEqual(inferred.get_tag("http.route"), "/lambda")
+        self.assertEqual(execution.parent_id, inferred.span_id)
+        self.assertEqual(execution.get_tag("http.status_code"), "200")
+
+    @patch("datadog_lambda.config.Config.trace_enabled", True)
+    @patch("datadog_lambda.config.Config.make_inferred_span", True)
+    def test_wrapper_inferred_alb_span_joins_inbound_datadog_context(self):
+        @wrapper.datadog_lambda_wrapper
+        def lambda_handler(event, context):
+            return self._alb_response(200)
+
+        lambda_handler(self.alb_event, get_mock_context())
+
+        inferred = lambda_handler.inferred_span
+
+        self.assertIsNotNone(inferred)
+        # Fixture carries x-datadog-trace-id=12345, x-datadog-parent-id=67890
+        self.assertEqual(inferred.trace_id, 12345)
+        self.assertEqual(inferred.parent_id, 67890)
+
+    @patch("datadog_lambda.config.Config.trace_enabled", True)
+    @patch("datadog_lambda.config.Config.make_inferred_span", True)
+    def test_wrapper_sets_error_on_inferred_alb_span_for_5xx(self):
+        @wrapper.datadog_lambda_wrapper
+        def lambda_handler(event, context):
+            return self._alb_response(502)
+
+        lambda_handler(self.alb_event, get_mock_context())
+
+        inferred = lambda_handler.inferred_span
+        execution = lambda_handler.span
+
+        self.assertEqual(inferred.get_tag("http.status_code"), "502")
+        self.assertEqual(execution.get_tag("http.status_code"), "502")
+        self.assertEqual(execution.error, 1)
+
+    @patch("datadog_lambda.config.Config.trace_enabled", True)
+    @patch("datadog_lambda.config.Config.make_inferred_span", True)
+    def test_wrapper_multivalue_alb_event_has_no_inferred_span(self):
+        with open(
+            "tests/event_samples/application-load-balancer-multivalue-headers.json"
+        ) as f:
+            event = json.load(f)
+
+        @wrapper.datadog_lambda_wrapper
+        def lambda_handler(event, context):
+            return self._alb_response(200)
+
+        lambda_handler(event, get_mock_context())
+
+        self.assertIsNone(lambda_handler.inferred_span)
+        self.assertIsNotNone(lambda_handler.span)