Add uptime metric to windows services (DataDog#21778)

* Add uptime metric to windows services

* Update unit tests

* Add docs

* Update manifest.json

* Fix metadata.csv

Author: briantu

windows_service/README.md

@@ -95,7 +95,7 @@ Beginning with Agent version 7.73, the check automatically adds a `windows_servi
 
 ### Metrics
 
-The Windows Service check does not include any metrics.
+See [metadata.csv][19] for a list of metrics provided by this integration.
 
 ### Events
 
@@ -141,3 +141,4 @@ If the service is present in the output, permissions are the issue. To give the
 [16]: https://learn.microsoft.com/en-US/troubleshoot/windows-server/group-policy/configure-group-policies-set-security
 [17]: https://learn.microsoft.com/en-us/windows/win32/services/service-trigger-events
 [18]: /integrations/windows-service?search=windows%20service
+[19]: https://github.com/DataDog/integrations-core/blob/master/windows_service/metadata.csv

windows_service/changelog.d/21778.added

@@ -0,0 +1 @@
+Add uptime metric to windows services

windows_service/datadog_checks/windows_service/windows_service.py

@@ -3,7 +3,9 @@
 # Licensed under a 3-clause BSD style license (see LICENSE)
 import ctypes
 import re
+import time
 
+import psutil
 import pywintypes
 import win32service
 import winerror
@@ -142,6 +144,8 @@ def __str__(self):
     @property
     def hSvc(self):
         if self._hSvc is None:
+            # Handle will automatically be closed by pywin32
+            # https://mhammond.github.io/pywin32/PySC_HANDLE.html
             self._hSvc = win32service.OpenService(self.scm_handle, self.name, win32service.SERVICE_QUERY_CONFIG)
         return self._hSvc
 
@@ -204,6 +208,21 @@ def startup_type_string(self):
         return startup_type_string
 
 
+def _build_process_cache() -> dict[int, "psutil.Process"]:
+    process_cache_dict = {}
+    for proc in psutil.process_iter(attrs=['pid', 'create_time']):
+        process_cache_dict[proc.pid] = proc
+    return process_cache_dict
+
+
+def _get_process_uptime_from_cache(pid: int, process_cache: dict[int, "psutil.Process"]) -> int:
+    process = process_cache.get(pid)
+    if process is None:
+        return 0
+
+    return int(time.time() - process.create_time())
+
+
 class WindowsService(AgentCheck):
     SERVICE_CHECK_NAME = 'windows_service.state'
     # https://docs.microsoft.com/en-us/windows/win32/api/winsvc/ns-winsvc-service_status_process
@@ -225,6 +244,7 @@ class WindowsService(AgentCheck):
         win32service.SERVICE_PAUSE_PENDING: "pause_pending",
         win32service.SERVICE_PAUSED: "paused",
     }
+    UNKNOWN_LITERAL = "unknown"
 
     def check(self, instance):
         services = instance.get('services', [])
@@ -243,14 +263,21 @@ def check(self, instance):
         services_unseen = {f.name for f in service_filters if f.name is not None}
 
         try:
+            # Handle will automatically be closed by pywin32
+            # https://mhammond.github.io/pywin32/PySC_HANDLE.html
             scm_handle = win32service.OpenSCManager(None, None, win32service.SC_MANAGER_ENUMERATE_SERVICE)
         except Exception as e:  # no cov
             raise Exception('Unable to open SCManager: {}'.format(e))
 
-        type_filter = win32service.SERVICE_WIN32
-        state_filter = win32service.SERVICE_STATE_ALL
+        service_status_process_enums = win32service.EnumServicesStatusEx(
+            scm_handle,
+            win32service.SERVICE_WIN32,
+            win32service.SERVICE_STATE_ALL,
+            None,
+            win32service.SC_ENUM_PROCESS_INFO,
+        )
 
-        service_statuses = win32service.EnumServicesStatus(scm_handle, type_filter, state_filter)
+        process_cache = _build_process_cache()
 
         # Sort service filters in reverse order on the regex pattern so more specific (longer)
         # regex patterns are tested first. This is to handle cases when a pattern is a prefix of
@@ -261,8 +288,13 @@ def check(self, instance):
         # See test_name_regex_order()
         service_filters = sorted(service_filters, reverse=True, key=lambda x: len(x.name or ""))
 
-        for short_name, display_name, service_status in service_statuses:
-            service_view = ServiceView(scm_handle, short_name)
+        for service_status_process_enum in service_status_process_enums:
+            service_name = service_status_process_enum["ServiceName"]
+            display_name = service_status_process_enum["DisplayName"]
+            state = service_status_process_enum["CurrentState"]
+            service_pid = service_status_process_enum["ProcessId"]
+
+            service_view = ServiceView(scm_handle, service_name)
 
             if 'ALL' not in services:
                 for service_filter in service_filters:
@@ -276,16 +308,20 @@ def check(self, instance):
                     except (pywintypes.error, OSError) as e:
                         self.log.exception("Exception at service match for %s", service_filter)
                         self.warning(
-                            "Failed to query %s service config for filter %s: %s", short_name, service_filter, str(e)
+                            "Failed to query %s service config for filter %s: %s", service_name, service_filter, str(e)
                         )
                 else:
                     continue
 
-            state = service_status[1]
+            service_uptime = 0
+            # If service_pid is 0, the service is not running
+            if service_pid != 0:
+                service_uptime = _get_process_uptime_from_cache(service_pid, process_cache)
+
             status = self.STATE_TO_STATUS.get(state, self.UNKNOWN)
-            state_string = self.STATE_TO_STRING.get(state, "unknown")
+            state_string = self.STATE_TO_STRING.get(state, self.UNKNOWN_LITERAL)
 
-            tags = ['windows_service:{}'.format(short_name), 'windows_service_state:{}'.format(state_string)]
+            tags = ['windows_service:{}'.format(service_name), 'windows_service_state:{}'.format(state_string)]
             tags.extend(custom_tags)
 
             if instance.get('collect_display_name_as_tag', False):
@@ -297,22 +333,23 @@ def check(self, instance):
                 except pywintypes.error as e:
                     self.log.exception("Exception at windows_service_startup_type tag for %s", service_filter)
                     self.warning(
-                        "Failed to query %s service config for filter %s: %s", short_name, service_filter, str(e)
+                        "Failed to query %s service config for filter %s: %s", service_name, service_filter, str(e)
                     )
 
             if not instance.get('disable_legacy_service_tag', False):
                 self._log_deprecation('service_tag', 'windows_service')
-                tags.append('service:{}'.format(short_name))
+                tags.append('service:{}'.format(service_name))
 
             self.service_check(self.SERVICE_CHECK_NAME, status, tags=tags)
-            self.log.debug('service state for %s %s', short_name, status)
+            self.log.debug('service state for %s %s', service_name, status)
+            self.gauge('windows_service.uptime', service_uptime, tags=tags)
 
         if 'ALL' not in services:
             for service in services_unseen:
                 # if a name doesn't match anything (wrong name or no permission to access the service), report UNKNOWN
                 status = self.UNKNOWN
 
-                tags = ['windows_service:{}'.format(service), 'windows_service_state:{}'.format("unknown")]
+                tags = ['windows_service:{}'.format(service), 'windows_service_state:{}'.format(self.UNKNOWN_LITERAL)]
 
                 tags.extend(custom_tags)
 

windows_service/manifest.json

@@ -48,6 +48,11 @@
       "events": {
         "creates_events": false
       },
+      "metrics": {
+        "prefix": "windows_service.",
+        "check": "windows_service.uptime",
+        "metadata_path": "metadata.csv"
+      },
       "service_checks": {
         "metadata_path": "assets/service_checks.json"
       },

windows_service/metadata.csv

@@ -0,0 +1,2 @@
+metric_name,metric_type,interval,unit_name,per_unit_name,description,orientation,integration,short_name,curated_metric,sample_tags
+windows_service.uptime,gauge,,second,,The uptime (in seconds) of the host process of the Windows service,0,windows_service,uptime,,

windows_service/pyproject.toml

@@ -37,6 +37,7 @@ dynamic = [
 [project.optional-dependencies]
 deps = [
     "pywin32==311; sys_platform == 'win32'",
+    "psutil==6.0.0",
 ]
 
 [project.urls]

windows_service/tests/test_bench.py

@@ -1,6 +1,7 @@
 # (C) Datadog, Inc. 2018-present
 # All rights reserved
 # Licensed under a 3-clause BSD style license (see LICENSE)
+
 from datadog_checks.windows_service import WindowsService