Vertica comments, dbs, profiler and docs (#9845)

* Vertica comments, dbs, profiler and docs

* Revert metabase changes

* Format

* Fix median

ingestion/src/metadata/ingestion/source/database/vertica/metadata.py

@@ -12,7 +12,9 @@
 Vertica source implementation.
 """
 import re
+import traceback
 from textwrap import dedent
+from typing import Iterable
 
 from sqlalchemy import sql, util
 from sqlalchemy.engine import reflection
@@ -20,6 +22,7 @@ from sqlalchemy.sql import sqltypes
 from sqlalchemy.sql.sqltypes import String
 from sqlalchemy_vertica.base import VerticaDialect
 
+from metadata.generated.schema.entity.data.database import Database
 from metadata.generated.schema.entity.services.connections.database.verticaConnection import (
     VerticaConnection,
 )
@@ -34,8 +37,19 @@ from metadata.ingestion.source.database.common_db_source import CommonDbSourceSe
 from metadata.ingestion.source.database.vertica.queries import (
     VERTICA_GET_COLUMNS,
     VERTICA_GET_PRIMARY_KEYS,
+    VERTICA_LIST_DATABASES,
+    VERTICA_TABLE_COMMENTS,
     VERTICA_VIEW_DEFINITION,
 )
+from metadata.utils import fqn
+from metadata.utils.filters import filter_by_database
+from metadata.utils.logger import ingestion_logger
+from metadata.utils.sqlalchemy_utils import (
+    get_all_table_comments,
+    get_table_comment_wrapper,
+)
+
+logger = ingestion_logger()
 
 
 class UUID(String):
@@ -219,9 +233,24 @@ def get_view_definition(
     return None
 
 
+@reflection.cache
+def get_table_comment(
+    self, connection, table_name, schema=None, **kw  # pylint: disable=unused-argument
+):
+    return get_table_comment_wrapper(
+        self,
+        connection,
+        table_name=table_name,
+        schema=schema,
+        query=VERTICA_TABLE_COMMENTS,
+    )
+
+
 VerticaDialect.get_columns = get_columns
 VerticaDialect._get_column_info = _get_column_info  # pylint: disable=protected-access
 VerticaDialect.get_view_definition = get_view_definition
+VerticaDialect.get_all_table_comments = get_all_table_comments
+VerticaDialect.get_table_comment = get_table_comment
 
 
 class VerticaSource(CommonDbSourceService):
@@ -239,3 +268,38 @@ class VerticaSource(CommonDbSourceService):
                 f"Expected VerticaConnection, but got {connection}"
             )
         return cls(config, metadata_config)
+
+    def get_database_names(self) -> Iterable[str]:
+        configured_db = self.config.serviceConnection.__root__.config.database
+        if configured_db:
+            self.set_inspector(database_name=configured_db)
+            yield configured_db
+        else:
+            results = self.connection.execute(VERTICA_LIST_DATABASES)
+            for res in results:
+                row = list(res)
+                new_database = row[0]
+                database_fqn = fqn.build(
+                    self.metadata,
+                    entity_type=Database,
+                    service_name=self.context.database_service.name.__root__,
+                    database_name=new_database,
+                )
+
+                if filter_by_database(
+                    self.source_config.databaseFilterPattern,
+                    database_fqn
+                    if self.source_config.useFqnForFiltering
+                    else new_database,
+                ):
+                    self.status.filter(database_fqn, "Database Filtered Out")
+                    continue
+
+                try:
+                    self.set_inspector(database_name=new_database)
+                    yield new_database
+                except Exception as exc:
+                    logger.debug(traceback.format_exc())
+                    logger.error(
+                        f"Error trying to connect to database {new_database}: {exc}"
+                    )

ingestion/src/metadata/ingestion/source/database/vertica/queries.py

@@ -14,15 +14,37 @@ SQL Queries used during ingestion
 
 import textwrap
 
+# Column comments in Vertica can only happen on Projections
+#   https://forum.vertica.com/discussion/238945/vertica-try-to-create-comment
+# And Vertica projections follow this naming:
+#   https://www.vertica.com/docs/9.2.x/HTML/Content/Authoring/AdministratorsGuide/Projections/WorkingWithProjections.htm
+# So to fetch column comments we need to concat the table_name + projection infix + column name.
+# Example: querying `v_catalog.comments` we find an object_name for a column in the table vendor_dimension as
+# `vendor_dimension_super.vendor_name`. Note how this is the `_super` projection.
+# Then, our join looks for the match in `vendor_dimension_%.vendor_name`.
+# Note: This might not suit for all column scenarios, but currently we did not find a better way to join
+# v_catalog.comments with v_catalog.columns.
 VERTICA_GET_COLUMNS = textwrap.dedent(
     """
-        SELECT column_name, data_type, column_default, is_nullable, comment
-        FROM v_catalog.columns col left join v_catalog.comments com on col.table_id=com.object_id
-        and com.object_type='COLUMN' and col.column_name=com.child_object
+        SELECT
+          column_name,
+          data_type,
+          column_default,
+          is_nullable,
+          comment
+        FROM v_catalog.columns col
+        LEFT JOIN v_catalog.comments com
+          ON com.object_type = 'COLUMN'
+         AND com.object_name LIKE CONCAT(CONCAT(col.table_name, '_%.'), col.column_name)
         WHERE lower(table_name) = '{table}'
         AND {schema_condition}
         UNION ALL
-        SELECT column_name, data_type, '' as column_default, true as is_nullable, ''  as comment
+        SELECT
+          column_name,
+          data_type,
+          '' AS column_default,
+          true AS is_nullable,
+          ''  AS comment
         FROM v_catalog.view_columns
         WHERE lower(table_name) = '{table}'
         AND {schema_condition}
@@ -47,3 +69,16 @@ VERTICA_VIEW_DEFINITION = textwrap.dedent(
       AND {schema_condition}
     """
 )
+
+VERTICA_LIST_DATABASES = "SELECT database_name from v_catalog.databases"
+
+VERTICA_TABLE_COMMENTS = textwrap.dedent(
+    """
+    SELECT
+      object_schema as schema,
+      object_name as table_name,
+      comment as table_comment
+    FROM v_catalog.comments
+    WHERE object_type = 'TABLE';
+    """
+)

ingestion/src/metadata/orm_profiler/orm/functions/median.py

@@ -59,7 +59,7 @@ def _(elements, compiler, **kwargs):
 def _(elements, compiler, **kwargs):
     """Median computation for MSSQL"""
     col = elements.clauses.clauses[0].name
-    return "percentile_cont(0.5)  WITHIN GROUP (ORDER BY %s ASC) OVER()" % col
+    return "percentile_cont(0.5) WITHIN GROUP (ORDER BY %s ASC) OVER()" % col
 
 
 @compiles(MedianFn, Dialects.Hive)
@@ -70,7 +70,7 @@ def _(elements, compiler, **kwargs):
 
 
 @compiles(MedianFn, Dialects.MySQL)
-def _(elemenst, compiler, **kwargs):  # pylint: disable=unused-argument
+def _(elements, compiler, **kwargs):  # pylint: disable=unused-argument
     """Median computation for MySQL currently not supported
     Needs to be tackled in https://github.com/open-metadata/OpenMetadata/issues/6340
     """
@@ -93,3 +93,11 @@ def _(elements, compiler, **kwargs):  # pylint: disable=unused-argument
     """.format(
         col=col, table=table.value
     )
+
+
+@compiles(MedianFn, Dialects.Vertica)
+def _(elements, compiler, **kwargs):  # pylint: disable=unused-argument
+    col, table = list(elements.clauses)
+    return "(SELECT MEDIAN({col}) OVER() FROM {table} LIMIT 1)".format(
+        col=col, table=table.value
+    )

ingestion/src/metadata/orm_profiler/orm/functions/modulo.py

@@ -58,6 +58,7 @@ def _(element, compiler, **kw):
 @compiles(ModuloFn, Dialects.Trino)
 @compiles(ModuloFn, Dialects.IbmDbSa)
 @compiles(ModuloFn, Dialects.Db2)
+@compiles(ModuloFn, Dialects.Vertica)
 def _(element, compiler, **kw):
     """Modulo function for specific dialect"""
     value, base = validate_and_compile(element, compiler, **kw)

ingestion/src/metadata/orm_profiler/orm/functions/random_num.py

@@ -101,3 +101,12 @@ def _(*_, **__):
     from the already sampled results when executing row::MOD(0, 100) < profile_sample.
     """
     return "0"
+
+
+@compiles(RandomNumFn, Dialects.Vertica)
+def _(*_, **__):
+    """
+    Vertica RANDOM() returns a number 0 < n < 1 as a float.
+    We need to cast it to integer to perform the modulo
+    """
+    return "(RANDOM() * 100)::INTEGER"

openmetadata-docs/content/connectors/database/vertica/airflow.md

@@ -22,6 +22,17 @@ To deploy OpenMetadata, check the <a href="/deployment">Deployment</a> guides.
 To run the Ingestion via the UI you'll need to use the OpenMetadata Ingestion Container, which comes shipped with
 custom Airflow plugins to handle the workflow deployment.
 
+### Permissions
+
+To run the ingestion we need a user with `SELECT` grants on the schemas that you'd like to ingest, as well as to the
+`V_CATALOG` schema. You can grant those as follows for the schemas in your database:
+
+```sql
+CREATE USER openmetadata IDENTIFIED BY 'password';
+GRANT SELECT ON ALL TABLES IN SCHEMA PUBLIC TO openmetadata;
+GRANT SELECT ON ALL TABLES IN SCHEMA V_CATALOG TO openmetadata;
+```
+
 ### Python Requirements
 
 To run the Vertica ingestion, you will need to install:

openmetadata-docs/content/connectors/database/vertica/cli.md

@@ -22,6 +22,17 @@ To deploy OpenMetadata, check the <a href="/deployment">Deployment</a> guides.
 To run the Ingestion via the UI you'll need to use the OpenMetadata Ingestion Container, which comes shipped with
 custom Airflow plugins to handle the workflow deployment.
 
+### Permissions
+
+To run the ingestion we need a user with `SELECT` grants on the schemas that you'd like to ingest, as well as to the
+`V_CATALOG` schema. You can grant those as follows for the schemas in your database:
+
+```sql
+CREATE USER openmetadata IDENTIFIED BY 'password';
+GRANT SELECT ON ALL TABLES IN SCHEMA PUBLIC TO openmetadata;
+GRANT SELECT ON ALL TABLES IN SCHEMA V_CATALOG TO openmetadata;
+```
+
 ### Python Requirements
 
 To run the Vertica ingestion, you will need to install:

openmetadata-docs/content/connectors/database/vertica/index.md

@@ -43,6 +43,17 @@ To deploy OpenMetadata, check the <a href="/deployment">Deployment</a> guides.
 To run the Ingestion via the UI you'll need to use the OpenMetadata Ingestion Container, which comes shipped with
 custom Airflow plugins to handle the workflow deployment.
 
+### Permissions
+
+To run the ingestion we need a user with `SELECT` grants on the schemas that you'd like to ingest, as well as to the
+`V_CATALOG` schema. You can grant those as follows for the schemas in your database:
+
+```sql
+CREATE USER openmetadata IDENTIFIED BY 'password';
+GRANT SELECT ON ALL TABLES IN SCHEMA PUBLIC TO openmetadata;
+GRANT SELECT ON ALL TABLES IN SCHEMA V_CATALOG TO openmetadata;
+```
+
 ## Metadata Ingestion
 
 ### 1. Visit the Services Page

openmetadata-docs/content/connectors/database/vertica/troubleshooting.md

@@ -0,0 +1,16 @@
+---
+title: Vertica Connector Troubleshooting
+slug: /connectors/database/vertica/troubleshooting
+---
+
+# Troubleshooting
+
+Learn how to resolve the most common problems people encounter in the Vertica connector.
+
+## Profiler: New session rejected
+
+If you see the following error when computing the profiler `New session rejected due to limit, already XYZ sessions active`,
+it means that the number of threads configured in the profiler workflow is exceeding the connection limits of your
+Vertica instance.
+
+Note that by default the profiler runs with 5 threads. In case you see this error, you might need to reduce this number.

openmetadata-docs/content/menu.md

@@ -344,6 +344,8 @@ site_menu:
     url: /connectors/database/vertica/airflow
   - category: Connectors / Database / Vertica / CLI
     url: /connectors/database/vertica/cli
+  - category: Connectors / Database / Vertica / Troubleshooting
+    url: /connectors/database/vertica/troubleshooting
   - category: Connectors / Dashboard
     url: /connectors/dashboard
   - category: Connectors / Dashboard / Looker