Sort by label keys before generating labels key and value lists (#3698)

* sort by label keys

* changelog

* test

* review

* linting

* Update contrib SHA

---------

Co-authored-by: Leighton Chen <lechen@microsoft.com>
Co-authored-by: Diego Hurtado <ocelotl@users.noreply.github.com>

.github/workflows/test.yml

@@ -10,7 +10,7 @@ env:
   # Otherwise, set variable to the commit of your branch on
   # opentelemetry-python-contrib which is compatible with these Core repo
   # changes.
-  CONTRIB_REPO_SHA: 3c2788469834aa4f5976e1644d757f43d60bc219
+  CONTRIB_REPO_SHA: 9ce1c26d2732dfbdadbb492fc38c562dcd08ed2e
   # This is needed because we do not clone the core repo in contrib builds anymore.
   # When running contrib builds as part of core builds, we use actions/checkout@v2 which
   # does not set an environment variable (simply just runs tox), which is different when

CHANGELOG.md

@@ -19,6 +19,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   ([#3785](https://github.com/open-telemetry/opentelemetry-python/pull/3785))
 - Add capture the fully qualified type name for raised exceptions in spans
   ([#3837](https://github.com/open-telemetry/opentelemetry-python/pull/3837))
+- Prometheus exporter sort label keys to prevent duplicate metrics when user input changes order 
+  ([#3698](https://github.com/open-telemetry/opentelemetry-python/pull/3698))
 
 ## Version 1.24.0/0.45b0 (2024-03-28)
 

exporter/opentelemetry-exporter-prometheus/src/opentelemetry/exporter/prometheus/__init__.py

@@ -237,7 +237,7 @@ def _translate_to_prometheus(
                 label_keys = []
                 label_values = []
 
-                for key, value in number_data_point.attributes.items():
+                for key, value in sorted(number_data_point.attributes.items()):
                     label_keys.append(self._sanitize(key))
                     label_values.append(self._check_value(value))
 

exporter/opentelemetry-exporter-prometheus/tests/test_prometheus_exporter.py

@@ -436,3 +436,24 @@ def test_target_info_sanitize(self):
             prometheus_metric.samples[0].labels["ratio"],
             "0.1",
         )
+
+    def test_label_order_does_not_matter(self):
+        metric_reader = PrometheusMetricReader()
+        provider = MeterProvider(metric_readers=[metric_reader])
+        meter = provider.get_meter("getting-started", "0.1.2")
+        counter = meter.create_counter("counter")
+
+        counter.add(1, {"cause": "cause1", "reason": "reason1"})
+        counter.add(1, {"reason": "reason2", "cause": "cause2"})
+
+        prometheus_output = generate_latest().decode()
+
+        # All labels are mapped correctly
+        self.assertIn('cause="cause1"', prometheus_output)
+        self.assertIn('cause="cause2"', prometheus_output)
+        self.assertIn('reason="reason1"', prometheus_output)
+        self.assertIn('reason="reason2"', prometheus_output)
+
+        # Only one metric is generated
+        metric_count = prometheus_output.count("# HELP counter_total")
+        self.assertEqual(metric_count, 1)