mgr/cephadm: fix KeyError when host is removed during serve loop

Signed-off-by: bachmanity1 <bachmanity138@gmail.com>

diff --git a/src/pybind/mgr/cephadm/inventory.py b/src/pybind/mgr/cephadm/inventory.py
index aed6ba03efadb406ce9fc42a08864007bed7a1d2..3fda8c0e590701aa69c66870035731e47011ee80 100644 (file)
--- a/src/pybind/mgr/cephadm/inventory.py
+++ b/src/pybind/mgr/cephadm/inventory.py
@@ -1331,7 +1331,7 @@ class HostCache():
             if host in self.mgr.offline_hosts:
                 dd.status = orchestrator.DaemonDescriptionStatus.error
                 dd.status_desc = 'host is offline'
-            elif self.mgr.inventory._inventory[host].get("status", "").lower() == "maintenance":
+            elif self.mgr.inventory._inventory.get(host, {}).get("status", "").lower() == "maintenance":
                 # We do not refresh daemons on hosts in maintenance mode, so stored daemon statuses
                 # could be wrong. We must assume maintenance is working and daemons are stopped
                 dd.status = orchestrator.DaemonDescriptionStatus.stopped

diff --git a/src/pybind/mgr/cephadm/serve.py b/src/pybind/mgr/cephadm/serve.py
index 3f1c63bcda114821d6aecda836a2ebfbd1dcb017..22083ffbbacd2f12a6cd1216ca499bc9a74c3ce7 100644 (file)
--- a/src/pybind/mgr/cephadm/serve.py
+++ b/src/pybind/mgr/cephadm/serve.py
@@ -252,8 +252,11 @@ class CephadmServe:
         @forall_hosts
         def refresh(host: str) -> None:
 
-            # skip hosts that are in maintenance - they could be powered off
-            if self.mgr.inventory._inventory[host].get("status", "").lower() == "maintenance":
+            # skip hosts that were removed or are in maintenance - they could be powered off
+            host_info = self.mgr.inventory._inventory.get(host)
+            if host_info is None:
+                return
+            if host_info.get("status", "").lower() == "maintenance":
                 return
 
             if self.mgr.use_agent:
@@ -862,8 +865,13 @@ class CephadmServe:
 
         try:
             all_slots, slots_to_add, daemons_to_remove = ha.place()
-            daemons_to_remove = [d for d in daemons_to_remove if (d.hostname and self.mgr.inventory._inventory[d.hostname].get(
-                'status', '').lower() not in ['maintenance', 'offline'] and d.hostname not in self.mgr.offline_hosts)]
+            daemons_to_remove = [
+                d for d in daemons_to_remove if (
+                    d.hostname
+                    and d.hostname in self.mgr.inventory._inventory
+                    and self.mgr.inventory._inventory.get(d.hostname, {}).get(
+                        'status', '').lower() not in ['maintenance', 'offline']
+                    and d.hostname not in self.mgr.offline_hosts)]
             self.log.debug('Add %s, remove %s' % (slots_to_add, daemons_to_remove))
         except OrchestratorError as e:
             msg = f'Failed to apply {spec.service_name()} spec {spec}: {str(e)}'