Add support to ingest sys.configurations for SQL Server instances

postgres/datadog_checks/postgres/metadata.py

@@ -1,7 +1,6 @@
 # (C) Datadog, Inc. 2023-present
 # All rights reserved
 # Licensed under a 3-clause BSD style license (see LICENSE)
-import json
 import time
 from typing import Dict, Optional, Tuple  # noqa: F401
 
@@ -15,6 +14,7 @@
 
 from datadog_checks.base import is_affirmative
 from datadog_checks.base.utils.db.utils import DBMAsyncJob, default_json_event_encoding
+from datadog_checks.base.utils.serialization import json
 from datadog_checks.base.utils.tracking import tracked_method
 
 # default pg_settings collection interval in seconds

sqlserver/assets/configuration/spec.yaml

@@ -236,6 +236,22 @@ files:
         type: boolean
         example: false
         display_default: false
+    - name: collect_settings
+      description: Configure collection of sys.configurations. This is an alpha feature.
+      options:
+        - name: enabled
+          description: |
+            Enable collection of sys.configurations. Requires `dbm: true`.
+          value:
+            type: boolean
+            example: false
+        - name: collection_interval
+          description: |
+            Set the database settings collection interval (in seconds). Each collection involves a single query to
+            `sys.configurations`.
+          value:
+            type: number
+            example: 600
     - name: query_metrics
       description: Configure collection of query metrics
       options:

sqlserver/datadog_checks/sqlserver/config_models/instance.py

@@ -37,6 +37,15 @@ class Azure(BaseModel):
     fully_qualified_domain_name: Optional[str] = None
 
 
+class CollectSettings(BaseModel):
+    model_config = ConfigDict(
+        arbitrary_types_allowed=True,
+        frozen=True,
+    )
+    collection_interval: Optional[float] = None
+    enabled: Optional[bool] = None
+
+
 class CustomQuery(BaseModel):
     model_config = ConfigDict(
         arbitrary_types_allowed=True,
@@ -115,6 +124,7 @@ class InstanceConfig(BaseModel):
     availability_group: Optional[str] = None
     aws: Optional[Aws] = None
     azure: Optional[Azure] = None
+    collect_settings: Optional[CollectSettings] = None
     command_timeout: Optional[int] = None
     connection_string: Optional[str] = None
     connector: Optional[str] = None

sqlserver/datadog_checks/sqlserver/data/conf.yaml.example

@@ -222,6 +222,21 @@ instances:
     #
     # dbm: false
 
+    ## Configure collection of sys.configurations. This is an alpha feature.
+    #
+    # collect_settings:
+
+        ## @param enabled - boolean - optional - default: false
+        ## Enable collection of sys.configurations. Requires `dbm: true`.
+        #
+        # enabled: false
+
+        ## @param collection_interval - number - optional - default: 600
+        ## Set the database settings collection interval (in seconds). Each collection involves a single query to
+        ## `sys.configurations`.
+        #
+        # collection_interval: 600
+
     ## Configure collection of query metrics
     #
     # query_metrics:

sqlserver/datadog_checks/sqlserver/metadata.py

@@ -0,0 +1,151 @@
+# (C) Datadog, Inc. 2023-present
+# All rights reserved
+# Licensed under a 3-clause BSD style license (see LICENSE)
+import time
+
+from datadog_checks.base import is_affirmative
+from datadog_checks.base.utils.db.utils import (
+    DBMAsyncJob,
+    default_json_event_encoding,
+)
+from datadog_checks.base.utils.serialization import json
+from datadog_checks.base.utils.tracking import tracked_method
+
+try:
+    import datadog_agent
+except ImportError:
+    from ..stubs import datadog_agent
+
+from datadog_checks.sqlserver.const import STATIC_INFO_ENGINE_EDITION, STATIC_INFO_VERSION
+
+# default settings collection interval in seconds
+DEFAULT_SETTINGS_COLLECTION_INTERVAL = 600
+
+SETTINGS_QUERY = """\
+SELECT {columns} FROM sys.configurations
+"""
+
+SQL_SERVER_SETTINGS_COLUMNS = [
+    "name",
+    "value",
+    "minimum",
+    "maximum",
+    "value_in_use",
+    "is_dynamic",
+    "is_advanced",
+]
+
+# some columns use the sql_varient type, which isn't supported
+# by most pyodbc drivers, instead we can cast these values to VARCHAR
+SQL_COLS_CAST_TYPE = {
+    "minimum": "varchar(max)",
+    "maximum": "varchar(max)",
+    "value_in_use": "varchar(max)",
+    "value": "varchar(max)",
+}
+
+
+def agent_check_getter(self):
+    return self.check
+
+
+class SqlserverMetadata(DBMAsyncJob):
+    """
+    Collects database metadata. Supports:
+        1. collection of sqlserver instance settings
+    """
+
+    def __init__(self, check):
+        self.check = check
+        # do not emit any dd.internal metrics for DBM specific check code
+        self.tags = [t for t in self.check.tags if not t.startswith('dd.internal')]
+        self.log = check.log
+        self.collection_interval = check.settings_config.get(
+            'collection_interval', DEFAULT_SETTINGS_COLLECTION_INTERVAL
+        )
+
+        super(SqlserverMetadata, self).__init__(
+            check,
+            run_sync=is_affirmative(check.settings_config.get('run_sync', False)),
+            enabled=is_affirmative(check.settings_config.get('enabled', False)),
+            expected_db_exceptions=(),
+            min_collection_interval=check.min_collection_interval,
+            dbms="sqlserver",
+            rate_limit=1 / float(self.collection_interval),
+            job_name="database-metadata",
+            shutdown_callback=self._close_db_conn,
+        )
+        self.disable_secondary_tags = is_affirmative(
+            check.statement_metrics_config.get('disable_secondary_tags', False)
+        )
+        self._conn_key_prefix = "dbm-metadata-"
+        self._settings_query = None
+        self._time_since_last_settings_query = 0
+        self._max_query_metrics = check.statement_metrics_config.get("max_queries", 250)
+
+    def _close_db_conn(self):
+        pass
+
+    def run_job(self):
+        self.report_sqlserver_metadata()
+
+    def _get_available_settings_columns(self, cursor, all_expected_columns):
+        cursor.execute("select top 0 * from sys.configurations")
+        all_columns = {i[0] for i in cursor.description}
+        available_columns = [c for c in all_expected_columns if c in all_columns]
+        missing_columns = set(all_expected_columns) - set(available_columns)
+        if missing_columns:
+            self.log.debug(
+                "missing the following expected settings columns from sys.configurations: %s", missing_columns
+            )
+        self.log.debug("found available sys.configurations columns: %s", available_columns)
+        return available_columns
+
+    def _get_settings_query_cached(self, cursor):
+        if self._settings_query:
+            return self._settings_query
+        available_columns = self._get_available_settings_columns(cursor, SQL_SERVER_SETTINGS_COLUMNS)
+        formatted_columns = []
+        for column in available_columns:
+            if column in SQL_COLS_CAST_TYPE:
+                formatted_columns.append(f"CAST({column} AS {SQL_COLS_CAST_TYPE[column]}) AS {column}")
+            else:
+                formatted_columns.append(column)
+        self._settings_query = SETTINGS_QUERY.format(
+            columns=', '.join(formatted_columns),
+        )
+        return self._settings_query
+
+    @tracked_method(agent_check_getter=agent_check_getter, track_result_length=True)
+    def _load_settings_rows(self, cursor):
+        self.log.debug("collecting sql server instance settings")
+        query = self._get_settings_query_cached(cursor)
+        self.log.debug("Running query [%s] %s", query)
+        cursor.execute(query)
+        columns = [i[0] for i in cursor.description]
+        # construct row dicts manually as there's no DictCursor for pyodbc
+        rows = [dict(zip(columns, row)) for row in cursor.fetchall()]
+        self.log.debug("loaded sql server settings len(rows)=%s", len(rows))
+        return rows
+
+    @tracked_method(agent_check_getter=agent_check_getter)
+    def report_sqlserver_metadata(self):
+        with self.check.connection.open_managed_default_connection(key_prefix=self._conn_key_prefix):
+            with self.check.connection.get_managed_cursor(key_prefix=self._conn_key_prefix) as cursor:
+                settings_rows = self._load_settings_rows(cursor)
+                event = {
+                    "host": self.check.resolved_hostname,
+                    "agent_version": datadog_agent.get_version(),
+                    "dbms": "sqlserver",
+                    "kind": "sqlserver_configs",
+                    "collection_interval": self.collection_interval,
+                    'dbms_version': "{},{}".format(
+                        self.check.static_info_cache.get(STATIC_INFO_VERSION, ""),
+                        self.check.static_info_cache.get(STATIC_INFO_ENGINE_EDITION, ""),
+                    ),
+                    "tags": self.tags,
+                    "timestamp": time.time() * 1000,
+                    "cloud_metadata": self.check.cloud_metadata,
+                    "metadata": settings_rows,
+                }
+                self._check.database_monitoring_metadata(json.dumps(event, default=default_json_event_encoding))

sqlserver/datadog_checks/sqlserver/sqlserver.py

@@ -17,6 +17,7 @@
 from datadog_checks.base.utils.db.utils import resolve_db_host
 from datadog_checks.base.utils.serialization import json
 from datadog_checks.sqlserver.activity import SqlserverActivity
+from datadog_checks.sqlserver.metadata import SqlserverMetadata
 from datadog_checks.sqlserver.statements import SqlserverStatementMetrics
 from datadog_checks.sqlserver.utils import Database, parse_sqlserver_major_version
 
@@ -120,7 +121,9 @@ def __init__(self, name, init_config, instances):
         # DBM
         self.dbm_enabled = self.instance.get('dbm', False)
         self.statement_metrics_config = self.instance.get('query_metrics', {}) or {}
+        self.settings_config = self.instance.get('collect_settings', {}) or {}
         self.statement_metrics = SqlserverStatementMetrics(self)
+        self.sql_metadata = SqlserverMetadata(self)
         self.activity_config = self.instance.get('query_activity', {}) or {}
         self.activity = SqlserverActivity(self)
         self.cloud_metadata = {}
@@ -179,6 +182,7 @@ def __init__(self, name, init_config, instances):
     def cancel(self):
         self.statement_metrics.cancel()
         self.activity.cancel()
+        self.sql_metadata.cancel()
 
     def config_checks(self):
         if self.autodiscovery and self.instance.get('database'):
@@ -741,6 +745,7 @@ def check(self, _):
             if self.dbm_enabled:
                 self.statement_metrics.run_job_loop(self.tags)
                 self.activity.run_job_loop(self.tags)
+                self.sql_metadata.run_job_loop(self.tags)
         else:
             self.log.debug("Skipping check")
 

sqlserver/tests/test_high_cardinality.py

@@ -19,6 +19,7 @@ def dbm_instance(instance_docker):
     instance_docker['dbm'] = True
     instance_docker['query_metrics'] = {'enabled': True, 'run_sync': True}
     instance_docker['query_activity'] = {'enabled': True, 'run_sync': True}
+    instance_docker['collect_settings'] = {'enabled': False}
     return copy(instance_docker)
 
 

sqlserver/tests/test_metadata.py

@@ -0,0 +1,88 @@
+# (C) Datadog, Inc. 2023-present
+# All rights reserved
+# Licensed under a 3-clause BSD style license (see LICENSE)
+
+from __future__ import unicode_literals
+
+import logging
+from copy import copy
+
+import pytest
+
+from datadog_checks.sqlserver import SQLServer
+
+from .common import CHECK_NAME
+
+try:
+    import pyodbc
+except ImportError:
+    pyodbc = None
+
+
+@pytest.fixture
+def dbm_instance(instance_docker):
+    instance_docker['dbm'] = True
+    instance_docker['min_collection_interval'] = 1
+    # set a very small collection interval so the tests go fast
+    instance_docker['collect_settings'] = {
+        'enabled': True,
+        'run_sync': True,
+        'collection_interval': 0.1,
+    }
+    return copy(instance_docker)
+
+
+@pytest.mark.integration
+@pytest.mark.usefixtures('dd_environment')
+@pytest.mark.parametrize(
+    "expected_columns,available_columns",
+    [
+        [
+            ["name", "value"],
+            ["name", "value"],
+        ],
+        [
+            ["name", "value", "some_missing_column"],
+            ["name", "value"],
+        ],
+    ],
+)
+def test_get_available_settings_columns(dbm_instance, expected_columns, available_columns):
+    check = SQLServer(CHECK_NAME, {}, [dbm_instance])
+    check.initialize_connection()
+    _conn_key_prefix = "dbm-metadata-"
+    with check.connection.open_managed_default_connection(key_prefix=_conn_key_prefix):
+        with check.connection.get_managed_cursor(key_prefix=_conn_key_prefix) as cursor:
+            result_available_columns = check.sql_metadata._get_available_settings_columns(cursor, expected_columns)
+            assert result_available_columns == available_columns
+
+
+@pytest.mark.integration
+@pytest.mark.usefixtures('dd_environment')
+def test_get_settings_query_cached(dbm_instance, caplog):
+    caplog.set_level(logging.DEBUG)
+    check = SQLServer(CHECK_NAME, {}, [dbm_instance])
+    check.initialize_connection()
+    _conn_key_prefix = "dbm-metadata"
+    with check.connection.open_managed_default_connection(key_prefix=_conn_key_prefix):
+        with check.connection.get_managed_cursor(key_prefix=_conn_key_prefix) as cursor:
+            for _ in range(3):
+                query = check.sql_metadata._get_settings_query_cached(cursor)
+                assert query, "query should be non-empty"
+    times_columns_loaded = 0
+    for r in caplog.records:
+        if r.message.startswith("found available sys.configurations columns"):
+            times_columns_loaded += 1
+    assert times_columns_loaded == 1, "columns should have been loaded only once"
+
+
+def test_sqlserver_collect_settings(aggregator, dd_run_check, dbm_instance):
+    check = SQLServer(CHECK_NAME, {}, [dbm_instance])
+    # dd_run_check(check)
+    check.initialize_connection()
+    check.check(dbm_instance)
+    dbm_metadata = aggregator.get_event_platform_events("dbm-metadata")
+    event = dbm_metadata[0]
+    assert event['dbms'] == "sqlserver"
+    assert event['kind'] == "sqlserver_configs"
+    assert len(event["metadata"]) > 0