GEN-1493 - Fix paginate_es in opensearch (#17858)

* GEN-1493 - Fix opensearch pagination

* GEN-1494 - Add CI for py-tests with Postgres and Opensearch

* GEN-1494 - Add CI for py-tests with Postgres and Opensearch

.github/workflows/py-tests-postgres.yml

@@ -9,7 +9,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 
-name: py-tests
+name: py-tests-postgres
 on:
   workflow_dispatch:
   push:

docker/development/docker-compose.yml

@@ -63,7 +63,6 @@ services:
       retries: 10
     volumes:
       - es-data:/usr/share/elasticsearch/data
-    
   
   execute-migrate-all:
     build:

ingestion/src/metadata/ingestion/ometa/mixins/es_mixin.py

@@ -331,19 +331,9 @@ class ESMixin(Generic[T]):
                 else:
                     break
 
-            # Get the data
-            for hit in response.hits.hits:
-                try:
-                    yield self.get_by_name(
-                        entity=entity,
-                        fqn=hit.source["fullyQualifiedName"],
-                        fields=fields,
-                        nullable=False,  # Raise an error if we don't find the Entity
-                    )
-                except Exception as exc:
-                    logger.warning(
-                        f"Error while getting {hit.source['fullyQualifiedName']} - {exc}"
-                    )
+            yield from self._yield_hits_from_api(
+                response=response, entity=entity, fields=fields
+            )
 
             # Get next page
             last_hit = response.hits.hits[-1] if response.hits.hits else None
@@ -362,3 +352,20 @@ class ESMixin(Generic[T]):
             logger.debug(traceback.format_exc())
             logger.warning(f"Error while getting ES response: {exc}")
         return None
+
+    def _yield_hits_from_api(
+        self, response: ESResponse, entity: Type[T], fields: Optional[List[str]]
+    ) -> Iterator[T]:
+        """Get the data from the API based on ES responses"""
+        for hit in response.hits.hits:
+            try:
+                yield self.get_by_name(
+                    entity=entity,
+                    fqn=hit.source["fullyQualifiedName"],
+                    fields=fields,
+                    nullable=False,  # Raise an error if we don't find the Entity
+                )
+            except Exception as exc:
+                logger.warning(
+                    f"Error while getting {hit.source['fullyQualifiedName']} - {exc}"
+                )

ingestion/tests/integration/ometa/test_ometa_es_api.py

@@ -300,7 +300,6 @@ class OMetaESTest(TestCase):
         res = self.metadata.es_get_queries_with_lineage(self.service.name.root)
         self.assertIn(self.checksum, res)
 
-    @pytest.skip("This never finished with Opensearch", allow_module_level=True)
     def test_paginate_no_filter(self):
         """We can paginate all the data"""
         # Since the test can run in parallel with other tables being there, we just
@@ -309,7 +308,6 @@ class OMetaESTest(TestCase):
             assert asset
             break
 
-    @pytest.skip("This never finished with Opensearch", allow_module_level=True)
     def test_paginate_with_errors(self):
         """We don't want to stop the ES yields just because a single Entity has an error"""
         # 1. First, prepare some tables

openmetadata-service/src/main/java/org/openmetadata/service/search/opensearch/OpenSearchClient.java

@@ -385,6 +385,10 @@ public class OpenSearchClient implements SearchClient {
       }
     }
 
+    if (!nullOrEmpty(request.getSearchAfter())) {
+      searchSourceBuilder.searchAfter(request.getSearchAfter());
+    }
+
     /* For backward-compatibility we continue supporting the deleted argument, this should be removed in future versions */
     if (request
             .getIndex()