feat: make resource detector ordering deterministic (#5120)

* feat: make resource detector ordering deterministic

* fix lint error

* update CHANGELOG.md

* add more descriptive variable name

* make entry point iteration deterministic by sorting by name

---------

Co-authored-by: Riccardo Magliocchetti <riccardo.magliocchetti@gmail.com>

Author: herin049

CHANGELOG.md

@@ -28,6 +28,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   ([#5076](https://github.com/open-telemetry/opentelemetry-python/pull/5076))
 - `opentelemetry-semantic-conventions`: use `X | Y` union annotation
   ([#5096](https://github.com/open-telemetry/opentelemetry-python/pull/5096))
+- `opentelemetry-sdk`: make resource detector ordering deterministic
+  ([#5120](https://github.com/open-telemetry/opentelemetry-python/pull/5120))
 - Add WeaverLiveCheck test util
   ([#5088](https://github.com/open-telemetry/opentelemetry-python/pull/5088))
 

opentelemetry-sdk/src/opentelemetry/sdk/environment_variables/__init__.py

@@ -777,6 +777,13 @@ def channel_credential_provider() -> grpc.ChannelCredentials:
 entry points for the ```opentelemetry_resource_detector``` entry point. This is an
 experimental feature and the name of this variable and its behavior can change
 in a non-backwards compatible way.
+
+Detectors are run in the order they are listed and their attributes are merged
+in that order, with later detectors taking precedence over earlier ones on
+conflicting keys. The ``otel`` detector (which reads
+:envvar:`OTEL_RESOURCE_ATTRIBUTES` and :envvar:`OTEL_SERVICE_NAME`) is always
+appended last unless explicitly placed elsewhere in the list, ensuring
+environment variable attributes take highest priority among detectors.
 """
 
 OTEL_EXPORTER_PROMETHEUS_HOST = "OTEL_EXPORTER_PROMETHEUS_HOST"

opentelemetry-sdk/src/opentelemetry/sdk/resources/__init__.py

@@ -70,7 +70,7 @@
 from json import dumps
 from os import environ
 from types import ModuleType
-from typing import List, Optional, Set, cast
+from typing import List, Optional, cast
 from urllib import parse
 
 from opentelemetry.attributes import BoundedAttributes
@@ -195,22 +195,27 @@ def create(
         if not attributes:
             attributes = {}
 
-        otel_experimental_resource_detectors: Set[str] = {"otel"}.union(
-            {
-                otel_experimental_resource_detector.strip()
-                for otel_experimental_resource_detector in environ.get(
-                    OTEL_EXPERIMENTAL_RESOURCE_DETECTORS, ""
-                ).split(",")
-                if otel_experimental_resource_detector
-            }
-        )
+        otel_experimental_resource_detectors: list[str] = [
+            detector.strip()
+            for detector in environ.get(
+                OTEL_EXPERIMENTAL_RESOURCE_DETECTORS, ""
+            ).split(",")
+            if detector.strip()
+        ]
 
         resource_detectors: List[ResourceDetector] = []
 
         if "*" in otel_experimental_resource_detectors:
-            otel_experimental_resource_detectors = entry_points(
-                group="opentelemetry_resource_detector"
-            ).names
+            otel_experimental_resource_detectors = [
+                name
+                for name in sorted(
+                    entry_points(group="opentelemetry_resource_detector").names
+                )
+                if name != "otel"
+            ]
+            otel_experimental_resource_detectors.append("otel")
+        elif "otel" not in otel_experimental_resource_detectors:
+            otel_experimental_resource_detectors.append("otel")
 
         for resource_detector in otel_experimental_resource_detectors:
             try:

opentelemetry-sdk/tests/resources/test_resources.py

@@ -57,6 +57,9 @@
     _HostResourceDetector,
     get_aggregated_resources,
 )
+from opentelemetry.util._importlib_metadata import (
+    entry_points as real_entry_points,
+)
 
 try:
     import psutil
@@ -475,6 +478,16 @@ def test_service_name_env(self):
 
 
 # pylint: disable=too-many-public-methods
+def _make_detector_ep(resource):
+    return Mock(
+        **{
+            "load.return_value": Mock(
+                return_value=Mock(**{"detect.return_value": resource})
+            )
+        }
+    )
+
+
 class TestOTELResourceDetector(unittest.TestCase):
     def setUp(self) -> None:
         environ[OTEL_RESOURCE_ATTRIBUTES] = ""
@@ -774,6 +787,52 @@ def test_resource_detector_entry_points_otel(self):
             self.assertIn(PROCESS_RUNTIME_VERSION, resource.attributes.keys())
             self.assertEqual(resource.schema_url, "")
 
+    @patch.dict(
+        environ,
+        {OTEL_EXPERIMENTAL_RESOURCE_DETECTORS: "mock_a,mock_b"},
+        clear=True,
+    )
+    def test_resource_detector_ordering_last_wins(self):
+        """Last detector in OTEL_EXPERIMENTAL_RESOURCE_DETECTORS wins on conflict."""
+        ep_a = _make_detector_ep(Resource({"conflict_key": "from_a"}))
+        ep_b = _make_detector_ep(Resource({"conflict_key": "from_b"}))
+
+        def side_effect(*args, **kwargs):
+            return {"mock_a": [ep_a], "mock_b": [ep_b]}.get(
+                kwargs.get("name", ""), []
+            )
+
+        with patch(
+            "opentelemetry.sdk.resources.entry_points", side_effect=side_effect
+        ):
+            resource = Resource({}).create()
+
+        self.assertEqual(resource.attributes["conflict_key"], "from_b")
+
+    @patch.dict(
+        environ,
+        {
+            OTEL_EXPERIMENTAL_RESOURCE_DETECTORS: "mock",
+            OTEL_RESOURCE_ATTRIBUTES: "conflict_key=otel_value",
+        },
+        clear=True,
+    )
+    def test_otel_detector_appended_last(self):
+        """'otel' detector is always appended last, so its attributes win over earlier detectors."""
+        ep_mock = _make_detector_ep(Resource({"conflict_key": "mock_value"}))
+
+        def side_effect(*args, **kwargs):
+            if kwargs.get("name") == "mock":
+                return [ep_mock]
+            return real_entry_points(*args, **kwargs)
+
+        with patch(
+            "opentelemetry.sdk.resources.entry_points", side_effect=side_effect
+        ):
+            resource = Resource({}).create()
+
+        self.assertEqual(resource.attributes["conflict_key"], "otel_value")
+
     @patch("platform.system", lambda: "Linux")
     @patch("platform.release", lambda: "666.5.0-35-generic")
     def test_os_detector_linux(self):