mgr/cephadm: add `orch host ok-to-stop` command

$ ceph orch host ok-to-stop host1
It is presumed safe to stop host host1

Signed-off-by: Michael Fritch <mfritch@suse.com>

diff --git a/src/pybind/mgr/cephadm/module.py b/src/pybind/mgr/cephadm/module.py
index e025a2091866adde2aa56fcdd90bd505494af768..d6ade94fac4cb801a8c905d9aaa0249fc5be347d 100644 (file)
--- a/src/pybind/mgr/cephadm/module.py
+++ b/src/pybind/mgr/cephadm/module.py
@@ -1143,6 +1143,29 @@ you may want to run:
         self.log.info('Removed label %s to host %s' % (label, host))
         return 'Removed label %s from host %s' % (label, host)
 
+    @trivial_completion
+    def host_ok_to_stop(self, hostname: str):
+        if hostname not in self.cache.get_hosts():
+            raise OrchestratorError(f'Cannot find host "{hostname}"')
+
+        daemons = self.cache.get_daemons()
+        daemon_map = defaultdict(lambda: [])
+        for dd in daemons:
+            if dd.hostname == hostname:
+                daemon_map[dd.daemon_type].append(dd.daemon_id)
+
+        for daemon_type,daemon_ids in daemon_map.items():
+            r = self.cephadm_services[daemon_type].ok_to_stop(daemon_ids)
+            if r.retval:
+                self.log.error(f'It is NOT safe to stop host {hostname}')
+                raise orchestrator.OrchestratorError(
+                        r.stderr,
+                        errno=r.retval)
+
+        msg = f'It is presumed safe to stop host {hostname}'
+        self.log.info(msg)
+        return msg
+
     def update_osdspec_previews(self, search_host: str = ''):
         # Set global 'pending' flag for host
         self.cache.loading_osdspec_preview.add(search_host)

diff --git a/src/pybind/mgr/orchestrator/_interface.py b/src/pybind/mgr/orchestrator/_interface.py
index f40157c4acb1440416b7420ce3d510465491d6a3..2bbf89b02a09d56a8e8e3f79bf56a3727db54fc1 100644 (file)
--- a/src/pybind/mgr/orchestrator/_interface.py
+++ b/src/pybind/mgr/orchestrator/_interface.py
@@ -845,6 +845,14 @@ class Orchestrator(object):
         """
         raise NotImplementedError()
 
+    def host_ok_to_stop(self, hostname:str) -> Completion:
+        """
+        Check if the specified host can be safely stopped without reducing availability
+
+        :param host: hostname
+        """
+        raise NotImplementedError()
+
     def get_inventory(self, host_filter=None, refresh=False):
         # type: (Optional[InventoryFilter], bool) -> Completion
         """

diff --git a/src/pybind/mgr/orchestrator/module.py b/src/pybind/mgr/orchestrator/module.py
index d4abda8e5dd2f983f391614c432c5ce234968798..aae20f7e8f39bdd401ab8f00101c2721103aa8f1 100644 (file)
--- a/src/pybind/mgr/orchestrator/module.py
+++ b/src/pybind/mgr/orchestrator/module.py
@@ -348,6 +348,16 @@ class OrchestratorCli(OrchestratorClientMixin, MgrModule,
         raise_if_exception(completion)
         return HandleCommandResult(stdout=completion.result_str())
 
+    @_cli_write_command(
+        'orch host ok-to-stop',
+        'name=hostname,type=CephString',
+        desc='Check if the specified host can be safely stopped without reducing availability')
+    def _host_ok_to_stop(self, hostname: str):
+        completion = self.host_ok_to_stop(hostname)
+        self._orchestrator_wait([completion])
+        raise_if_exception(completion)
+        return HandleCommandResult(stdout=completion.result_str())
+
     @_cli_read_command(
         'orch device ls',
         "name=hostname,type=CephString,n=N,req=false "