mgr/cephadm: block removing last instance of _admin label

Fixes: https://tracker.ceph.com/issues/54425
Signed-off-by: Adam King <adking@redhat.com>

diff --git a/src/pybind/mgr/cephadm/module.py b/src/pybind/mgr/cephadm/module.py
index ae428249e4601ae14f94d54fe18c018a7595d318..4757c2e182f143ae22c2d20ff0b03211dd322dc5 100644 (file)
--- a/src/pybind/mgr/cephadm/module.py
+++ b/src/pybind/mgr/cephadm/module.py
@@ -1520,7 +1520,19 @@ Then run the following:
         return 'Added label %s to host %s' % (label, host)
 
     @handle_orch_error
-    def remove_host_label(self, host: str, label: str) -> str:
+    def remove_host_label(self, host: str, label: str, force: bool = False) -> str:
+        # if we remove the _admin label from the only host that has it we could end up
+        # removing the only instance of the config and keyring and cause issues
+        if not force and label == '_admin':
+            p = PlacementSpec(label='_admin')
+            admin_hosts = p.filter_matching_hostspecs(self.inventory.all_specs())
+            if len(admin_hosts) == 1 and admin_hosts[0] == host:
+                raise OrchestratorValidationError(f"Host {host} is the last host with the '_admin'"
+                                                  " label.\nRemoving the _admin label from this host could cause the removal"
+                                                  " of the last cluster config/keyring managed by cephadm.\n"
+                                                  "It is recommended to add the _admin label to another host"
+                                                  " before completing this operation.\nIf you're certain this is"
+                                                  " what you want rerun this command with --force.")
         self.inventory.rm_label(host, label)
         self.log.info('Removed label %s to host %s' % (label, host))
         self._kick_serve_loop()

diff --git a/src/pybind/mgr/orchestrator/_interface.py b/src/pybind/mgr/orchestrator/_interface.py
index 344dea13854e8f08038baf4d8a4cff791648e46c..2ed881f6ab533ca70a76e75f9638f1ded85b5b3b 100644 (file)
--- a/src/pybind/mgr/orchestrator/_interface.py
+++ b/src/pybind/mgr/orchestrator/_interface.py
@@ -392,7 +392,7 @@ class Orchestrator(object):
         """
         raise NotImplementedError()
 
-    def remove_host_label(self, host: str, label: str) -> OrchResult[str]:
+    def remove_host_label(self, host: str, label: str, force: bool = False) -> OrchResult[str]:
         """
         Remove a host label
         """

diff --git a/src/pybind/mgr/orchestrator/module.py b/src/pybind/mgr/orchestrator/module.py
index cdf93f7a4bd4608175a6c4714a7de4c1369851a3..88f24365294b56bb62bc863840b994aa0483c3c5 100644 (file)
--- a/src/pybind/mgr/orchestrator/module.py
+++ b/src/pybind/mgr/orchestrator/module.py
@@ -420,9 +420,9 @@ class OrchestratorCli(OrchestratorClientMixin, MgrModule,
         return HandleCommandResult(stdout=completion.result_str())
 
     @_cli_write_command('orch host label rm')
-    def _host_label_rm(self, hostname: str, label: str) -> HandleCommandResult:
+    def _host_label_rm(self, hostname: str, label: str, force: bool = False) -> HandleCommandResult:
         """Remove a host label"""
-        completion = self.remove_host_label(hostname, label)
+        completion = self.remove_host_label(hostname, label, force)
         raise_if_exception(completion)
         return HandleCommandResult(stdout=completion.result_str())
 

diff --git a/src/pybind/mgr/rook/module.py b/src/pybind/mgr/rook/module.py
index 22433d44cca71d03960588e170cc09ffadc44076..4300e98d502b1b39d71cf01de6bfa12a9068afb1 100644 (file)
--- a/src/pybind/mgr/rook/module.py
+++ b/src/pybind/mgr/rook/module.py
@@ -633,7 +633,7 @@ class RookOrchestrator(MgrModule, orchestrator.Orchestrator):
     def add_host_label(self, host: str, label: str) -> OrchResult[str]:
         return self.rook_cluster.add_host_label(host, label)
     
-    def remove_host_label(self, host: str, label: str) -> OrchResult[str]:
+    def remove_host_label(self, host: str, label: str, force: bool = False) -> OrchResult[str]:
         return self.rook_cluster.remove_host_label(host, label)
     """
     @handle_orch_error