mgr/cephadm: set health check warning in `_apply_service_config` for invalid config...

Fixes: https://tracker.ceph.com/issues/44414
Signed-off-by: Melissa Li <li.melissa.kun@gmail.com>

diff --git a/src/pybind/mgr/cephadm/serve.py b/src/pybind/mgr/cephadm/serve.py
index 282bcaf2c7f1206538f533ce53acaab6b8fe7cca..8b65b96588bbc7eba7535f664291921cdb3dde0b 100644 (file)
--- a/src/pybind/mgr/cephadm/serve.py
+++ b/src/pybind/mgr/cephadm/serve.py
@@ -583,16 +583,20 @@ class CephadmServe:
     def _apply_service_config(self, spec: ServiceSpec) -> None:
         if spec.config:
             section = utils.name_to_config_section(spec.service_name())
+            for name in ['CEPHADM_INVALID_CONFIG_OPTION', 'CEPHADM_FAILED_SET_OPTION']:
+                self.mgr.remove_health_warning(name)
+            invalid_config_options = []
+            options_failed_to_set = []
             for k, v in spec.config.items():
                 try:
                     current = self.mgr.get_foreign_ceph_option(section, k)
                 except KeyError:
-                    self.log.warning(
-                        f'Ignoring invalid {spec.service_name()} config option {k}'
-                    )
+                    msg = f'Ignoring invalid {spec.service_name()} config option {k}'
+                    self.log.warning(msg)
                     self.mgr.events.for_service(
                         spec, OrchestratorEvent.ERROR, f'Invalid config option {k}'
                     )
+                    invalid_config_options.append(msg)
                     continue
                 if current != v:
                     self.log.debug(f'setting [{section}] {k} = {v}')
@@ -604,9 +608,14 @@ class CephadmServe:
                             'who': section,
                         })
                     except MonCommandFailed as e:
-                        self.log.warning(
-                            f'Failed to set {spec.service_name()} option {k}: {e}'
-                        )
+                        msg = f'Failed to set {spec.service_name()} option {k}: {e}'
+                        self.log.warning(msg)
+                        options_failed_to_set.append(msg)
+
+            if invalid_config_options:
+                self.mgr.set_health_warning('CEPHADM_INVALID_CONFIG_OPTION', f'Ignoring {len(invalid_config_options)} invalid config option(s)', len(invalid_config_options), invalid_config_options)
+            if options_failed_to_set:
+                self.mgr.set_health_warning('CEPHADM_FAILED_SET_OPTION', f'Failed to set {len(options_failed_to_set)} option(s)', len(options_failed_to_set), options_failed_to_set)
 
     def _apply_service(self, spec: ServiceSpec) -> bool:
         """

diff --git a/src/pybind/mgr/cephadm/tests/test_cephadm.py b/src/pybind/mgr/cephadm/tests/test_cephadm.py
index 32c98598e49b3e57309562236d48b714d37ddeb6..1cbf009230e211bb3b1a027d32b651ab78f0332d 100644 (file)
--- a/src/pybind/mgr/cephadm/tests/test_cephadm.py
+++ b/src/pybind/mgr/cephadm/tests/test_cephadm.py
@@ -888,6 +888,19 @@ spec:
                             'entity': entity,
                         })
 
+    @mock.patch("cephadm.module.CephadmOrchestrator.get_foreign_ceph_option")
+    @mock.patch("cephadm.serve.CephadmServe._run_cephadm")
+    def test_invalid_config_option_health_warning(self, _run_cephadm, get_foreign_ceph_option, cephadm_module: CephadmOrchestrator):
+        _run_cephadm.return_value = ('{}', '', 0)
+        with with_host(cephadm_module, 'test'):
+            ps = PlacementSpec(hosts=['test:0.0.0.0=a'], count=1)
+            get_foreign_ceph_option.side_effect = KeyError
+            CephadmServe(cephadm_module)._apply_service_config(ServiceSpec('mgr', placement=ps, config={'test': 'foo'}))
+            assert cephadm_module.health_checks.get('CEPHADM_INVALID_CONFIG_OPTION') is not None
+            assert cephadm_module.health_checks['CEPHADM_INVALID_CONFIG_OPTION']['count'] == 1
+            assert 'Ignoring 1 invalid config option(s)' in cephadm_module.health_checks['CEPHADM_INVALID_CONFIG_OPTION']['summary']
+            assert 'Ignoring invalid mgr config option test' in cephadm_module.health_checks['CEPHADM_INVALID_CONFIG_OPTION']['detail']
+
     @mock.patch("cephadm.serve.CephadmServe._run_cephadm", _run_cephadm('{}'))
     @mock.patch("cephadm.services.nfs.NFSService.run_grace_tool", mock.MagicMock())
     @mock.patch("cephadm.services.nfs.NFSService.purge", mock.MagicMock())