Fix edge case in metrics collector (#3875)

* Fix edge case in metrics collector

For some reason, the default/fallback value returned by `get_collection_stats` is an empty list, not an empty object. This can trigger an edge case if Prometheus metrics are scraped after a collection has been created, but the stats for the new collection haven’t been computed yet.

To prevent unintended side effects, I haven’t changed the default/fallback value (as other parts of the code base might rely on it) and have instead explicitly handled empty values.

* Use Airplane instead of Person entities in tests

This shouldn’t make a difference, but unfortunately our test suite contains a bunch of tests that have side effects (e.g. they create entities that aren’t properly rolled back after the tests). In this particular case, other tests create Person entities, but not Airplane entities, so that’s a workaround for now.

Author: tillprochaska

aleph/metrics/collectors.py

@@ -283,6 +283,11 @@ def _entities(self):
             collection_stats = get_collection_stats(collection.id)
             schemata = collection_stats["schema"]["values"]
 
+            # For some reason, the default/fallback value returned by `get_collection_stats`
+            # is a list, not an empty object, so we need to handle this explicitly.
+            if not schemata:
+                continue
+
             for schema, count in schemata.items():
                 stats[schema] += count
 

aleph/tests/test_metrics.py

@@ -335,22 +335,29 @@ def test_entities(self):
         collection_1 = self.create_collection()
         entity_1 = self.create_entity(
             collection=collection_1,
-            data={"schema": "Person", "properties": {}},
+            data={"schema": "Airplane", "properties": {}},
         )
 
         collection_2 = self.create_collection()
         entity_2 = self.create_entity(
             collection=collection_2,
-            data={"schema": "Person", "properties": {}},
+            data={"schema": "Airplane", "properties": {}},
         )
 
         index_entity(entity_1)
         index_entity(entity_2)
 
-        # This is usually executed periodically by a worker
+        # Ensure that metric collection succeeds even if the stats for newly collections
+        # haven’t been computed yet
+        reg.collect()
+        generate_latest(reg)
+        planes = reg.get_sample_value("aleph_entities", {"schema": "Airplane"})
+        assert planes is None
+
+        # This is usually executed periodically by a worker or after collection contents are updated
         compute_collections()
 
         reg.collect()
         generate_latest(reg)
-        persons = reg.get_sample_value("aleph_entities", {"schema": "Person"})
-        assert persons == 2
+        planes = reg.get_sample_value("aleph_entities", {"schema": "Airplane"})
+        assert planes == 2