Sy/om type override (#18054)

* Fix tests for openmetrics counter transformer

* Try dropping the '_total' restriction on metric samples

* add type override examples

* add test for type_override

* stricter test

* changelog

* remove f string

* remove f string

* spacing

* lint

* Update datadog_checks_base/changelog.d/18054.added

Co-authored-by: Ilia Kurenkov <ilia.kurenkov@datadoghq.com>

* Update datadog_checks_base/tests/base/checks/openmetrics/test_v2/test_transformers/test_type_override.py

Co-authored-by: Ilia Kurenkov <ilia.kurenkov@datadoghq.com>

* Update datadog_checks_base/tests/base/checks/openmetrics/test_v2/test_transformers/test_type_override.py

Co-authored-by: Ilia Kurenkov <ilia.kurenkov@datadoghq.com>

* Update datadog_checks_base/tests/base/checks/openmetrics/test_v2/test_transformers/test_type_override.py

Co-authored-by: Ilia Kurenkov <ilia.kurenkov@datadoghq.com>

* Update test_type_override.py

* lint

---------

Co-authored-by: Ilia Kurenkov <ilia.kurenkov@datadoghq.com>

datadog_checks_base/changelog.d/18054.added

@@ -0,0 +1 @@
+Allow untyped metrics that we coerce to `counter` to be collected regardless if they have `_total` or not.

datadog_checks_base/datadog_checks/base/checks/openmetrics/v2/transformers/counter.py

@@ -15,14 +15,13 @@ def counter(metric, sample_data, runtime_data):
         flush_first_value = runtime_data['flush_first_value']
 
         for sample, tags, hostname in sample_data:
-            if sample.name.endswith('_total'):
-                monotonic_count_method(
-                    metric_name,
-                    sample.value,
-                    tags=tags,
-                    hostname=hostname,
-                    flush_first_value=flush_first_value,
-                )
+            monotonic_count_method(
+                metric_name,
+                sample.value,
+                tags=tags,
+                hostname=hostname,
+                flush_first_value=flush_first_value,
+            )
 
     del check
     del modifiers

datadog_checks_base/tests/base/checks/openmetrics/test_v2/test_transformers/test_counter.py

@@ -11,107 +11,38 @@
 
 
 def test_basic(aggregator, dd_run_check, mock_http_response):
-    mock_http_response(
-        """
-        # HELP go_memstats_alloc_bytes_total Total number of bytes allocated, even if freed.
-        # TYPE go_memstats_alloc_bytes_total counter
-        go_memstats_alloc_bytes_total 9.339544592e+09
-        # HELP go_memstats_frees_total Total number of frees.
-        # TYPE go_memstats_frees_total counter
-        go_memstats_frees_total 1.28219257e+08
-        """
-    )
-    check = get_check({'metrics': ['.+']})
-    dd_run_check(check)
-
-    aggregator.assert_metric(
-        'test.go_memstats_alloc_bytes.count', 9339544592, metric_type=aggregator.MONOTONIC_COUNT, tags=['endpoint:test']
-    )
-    aggregator.assert_metric(
-        'test.go_memstats_frees.count', 128219257, metric_type=aggregator.MONOTONIC_COUNT, tags=['endpoint:test']
-    )
-
-    aggregator.assert_all_metrics_covered()
-
-
-def test_basic_incorrect_suffix(aggregator, dd_run_check, mock_http_response):
     """
-    We won't collect counter metrics unless they end in '_total'.
+    Metrics with type 'counter' in the raw payload get collected.
 
-    We commonly see metrics that end in '_count' or ones that have no meaningful suffix at all.
-    Here we make sure that by default these won't be collected.
+    If they end in '_total' that suffix is dropped.
     """
 
     mock_http_response(
         """
-        # HELP go_memstats_alloc_bytes_count Total number of bytes allocated, even if freed.
-        # TYPE go_memstats_alloc_bytes_count counter
-        go_memstats_alloc_bytes_total 9.339544592e+09
-        # HELP go_memstats_frees Total number of frees.
-        # TYPE go_memstats_frees counter
-        go_memstats_frees_total 1.28219257e+08
+        # HELP foo_total Example of '_total' getting dropped
+        # TYPE foo_total counter
+        foo_total 9.339544592e+09
+        # HELP bar_count Example of '_count' not getting dropped
+        # TYPE bar_count counter
+        bar_count 1.28219257e+08
+        # HELP baz Example that doesn't end in '_total' nor '_count'
+        # TYPE baz counter
+        baz 1.28219257e+08
         """
     )
     check = get_check({'metrics': ['.+']})
     dd_run_check(check)
 
-    assert not aggregator.metrics('test.go_memstats_alloc_bytes.count')
-    assert not aggregator.metrics('test.go_memstats_frees.count')
-    aggregator.assert_all_metrics_covered()
-
-
-def test_untyped_correct_suffix(aggregator, dd_run_check, mock_http_response):
-    """
-    We can force a metric that is 'untyped' in the raw payload to a counter as long as it ends in '_total'.
-    """
-
-    mock_http_response(
-        """
-        # HELP go_memstats_alloc_bytes_total Total number of bytes allocated, even if freed.
-        # TYPE go_memstats_alloc_bytes_total untyped
-        go_memstats_alloc_bytes_total 9.339544592e+09
-        """
-    )
-    check = get_check(
-        {'metrics': [{'go_memstats_alloc_bytes_total': {'name': 'go_memstats_alloc_bytes', 'type': 'counter'}}]}
-    )
-    dd_run_check(check)
-
     aggregator.assert_metric(
-        'test.go_memstats_alloc_bytes.count', 9339544592, metric_type=aggregator.MONOTONIC_COUNT, tags=['endpoint:test']
+        'test.foo.count', 9339544592, metric_type=aggregator.MONOTONIC_COUNT, tags=['endpoint:test']
     )
-    aggregator.assert_all_metrics_covered()
-
-
-def test_untyped_incorrect_suffix(aggregator, dd_run_check, mock_http_response):
-    """
-    Forcing an untyped metric as a counter won't work without it ending in '_total'.
-    """
-
-    mock_http_response(
-        """
-        # HELP go_memstats_alloc_bytes_count Total number of bytes allocated, even if freed.
-        # TYPE go_memstats_alloc_bytes_count counter
-        go_memstats_alloc_bytes_total 9.339544592e+09
-        # HELP go_memstats_frees Total number of frees.
-        # TYPE go_memstats_frees counter
-        go_memstats_frees_total 1.28219257e+08
-        """
+    aggregator.assert_metric(
+        'test.bar_count.count', 128219257, metric_type=aggregator.MONOTONIC_COUNT, tags=['endpoint:test']
     )
-    check = get_check(
-        {
-            'metrics': [
-                {
-                    'go_memstats_alloc_bytes': {'name': 'go_memstats_alloc_bytes', 'type': 'counter'},
-                    'go_memstats_frees': {'name': 'go_memstats_frees', 'type': 'counter'},
-                }
-            ]
-        }
+    aggregator.assert_metric(
+        'test.baz.count', 128219257, metric_type=aggregator.MONOTONIC_COUNT, tags=['endpoint:test']
     )
-    dd_run_check(check)
 
-    assert not aggregator.metrics('test.go_memstats_alloc_bytes.count')
-    assert not aggregator.metrics('test.go_memstats_frees.count')
     aggregator.assert_all_metrics_covered()
 
 

datadog_checks_base/tests/base/checks/openmetrics/test_v2/test_transformers/test_type_override.py

@@ -0,0 +1,84 @@
+# (C) Datadog, Inc. 2024-present
+# All rights reserved
+# Licensed under a 3-clause BSD style license (see LICENSE)
+import pytest
+
+from datadog_checks.dev.testing import requires_py3
+
+from ..utils import get_check
+
+pytestmark = [requires_py3]
+
+
+@pytest.mark.parametrize(
+    'metric_type',
+    [
+        'counter',
+        'gauge',
+    ],
+)
+def test_untyped_counter(aggregator, dd_run_check, mock_http_response, metric_type):
+    """
+    The Agent's OpenMetrics parser uses the # TYPE line to determine the type of a metric. However, when it encounters
+    an untyped metric, the Agent won't be able to derive the metric type, and the metric will be skipped. We can,
+    however, force the Agent to collect it as a specific type. In this scenario, the Agent will ignore the metadata
+    lines # HELP and # TYPE and will instead only look at the metric sample name.
+    Below are some of the scenarios we might encounter:
+
+    1. The metricset and metric sample does not end in '_total' and is untyped.
+    2. The metricset and metric sample does end in '_total' and is untyped.
+    3. The metricset does not end in '_total' and metric sample end in '_total' and is untyped.
+    4. The metricset and metric samples don't match
+    5. The metricset and metric samples are a non counter/gauge type
+
+    The test will show how these can be configured to be collected as gauge or counters. The metric sample just
+    has to match what the agent is looking for. The agent will then collect the metric as a we config
+    provided type. Only testing for counter and gauge here because these are the only 2 types I see we can accurately
+    collect with overrides.
+    """
+
+    mock_http_response(
+        """
+        # HELP foo The metricset and metric sample does not end in '_total' and is untyped.
+        # TYPE foo untyped
+        foo 0
+        # HELP bar_total The metricset and metric sample does end in '_total' and is untyped.
+        # TYPE bar_total untyped
+        bar_total 1
+        # HELP baz The metricset  does not end in '_total' and metric sample end in '_total' and is untyped.
+        # TYPE baz untyped
+        baz_total 2
+        # HELP qux The metricset and metric samples don't match
+        # TYPE qux untyped
+        fiz 3
+        # HELP bux The metricset and metric samples are a non counter type
+        # TYPE bux histogram
+        bux 4
+        """
+    )
+    check = get_check(
+        {
+            'metrics': [
+                {
+                    'foo': {'name': 'foo', 'type': metric_type},
+                    'bar_total': {'name': 'bar', 'type': metric_type},
+                    'baz_total': {'name': 'baz', 'type': metric_type},
+                    'fiz': {'name': 'fiz', 'type': metric_type},
+                    'bux': {'name': 'bux', 'type': metric_type},
+                }
+            ]
+        }
+    )
+    dd_run_check(check)
+
+    metrics = ["test.foo", "test.bar", "test.baz", "test.fiz", "test.bux"]
+
+    for metric in metrics:
+        aggregator.assert_metric(
+            '{}.count'.format(metric) if metric_type == 'counter' else metric,
+            metrics.index(metric),
+            metric_type=aggregator.MONOTONIC_COUNT if metric_type == 'counter' else aggregator.GAUGE,
+            tags=['endpoint:test'],
+        )
+
+    aggregator.assert_all_metrics_covered()