From: Shweta Bhosale <Shweta.Bhosale1@ibm.com>
Date: Thu, 23 Oct 2025 15:06:06 +0000 (+0530)
Subject: mgr/cephadm: Signal handling while daemon reconfig
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=8ef8a1e2f378f13afc695be74590d3ffc3454dcb;p=ceph.git

mgr/cephadm: Signal handling while daemon reconfig

Fixes: https://tracker.ceph.com/issues/73633
Signed-off-by: Shweta Bhosale <Shweta.Bhosale1@ibm.com>
---

diff --git a/src/cephadm/cephadm.py b/src/cephadm/cephadm.py
index 7cf65220c80..201b0319ba9 100755
--- a/src/cephadm/cephadm.py
+++ b/src/cephadm/cephadm.py
@@ -1202,10 +1202,15 @@ def deploy_daemon(
     # If this was a reconfig and the daemon is not a Ceph daemon, restart it
     # so it can pick up potential changes to its configuration files
     if deployment_type == DeploymentType.RECONFIG and daemon_type not in ceph_daemons():
-        # ceph daemons do not need a restart; others (presumably) do to pick
-        # up the new config
-        call_throws(ctx, ['systemctl', 'reset-failed', ident.unit_name])
-        call_throws(ctx, ['systemctl', 'restart', ident.unit_name])
+        if not ctx.skip_restart_for_reconfig:
+            # ceph daemons do not need a restart; others (presumably) do to pick
+            # up the new config
+            call_throws(ctx, ['systemctl', 'reset-failed', ident.unit_name])
+            call_throws(ctx, ['systemctl', 'restart', ident.unit_name])
+        elif ctx.send_signal_to_daemon:
+            ctx.signal_name = ctx.send_signal_to_daemon
+            ctx.signal_number = None
+            command_signal(ctx)
 
 
 def clean_cgroup(ctx: CephadmContext, fsid: str, unit_name: str) -> None:
@@ -4896,6 +4901,18 @@ def _add_deploy_parser_args(
         default=None,
         help='Time in seconds to wait for graceful service shutdown before forcefully killing it'
     )
+    parser_deploy.add_argument(
+        '--skip-restart-for-reconfig',
+        action='store_true',
+        default=False,
+        help='skip restart for non ceph daemons and perform default action'
+    )
+    parser_deploy.add_argument(
+        '--send-signal-to-daemon',
+        action='store_true',
+        default=False,
+        help='Send signal to daemon'
+    )
 
 
 def _name_opts(parser: argparse.ArgumentParser) -> None:
diff --git a/src/pybind/mgr/cephadm/module.py b/src/pybind/mgr/cephadm/module.py
index ef4699838d1..bad6568f500 100644
--- a/src/pybind/mgr/cephadm/module.py
+++ b/src/pybind/mgr/cephadm/module.py
@@ -2710,7 +2710,9 @@ Then run the following:
     def _daemon_action(self,
                        daemon_spec: CephadmDaemonDeploySpec,
                        action: str,
-                       image: Optional[str] = None) -> str:
+                       image: Optional[str] = None,
+                       skip_restart_for_reconfig: bool = False,
+                       send_signal_to_daemon: Optional[str] = None) -> str:
         self._daemon_action_set_image(action, image, daemon_spec.daemon_type,
                                       daemon_spec.daemon_id)
 
@@ -2735,7 +2737,9 @@ Then run the following:
                     daemon_spec)
             with self.async_timeout_handler(daemon_spec.host, f'cephadm deploy ({daemon_spec.daemon_type} daemon)'):
                 return self.wait_async(
-                    CephadmServe(self)._create_daemon(daemon_spec, reconfig=(action == 'reconfig')))
+                    CephadmServe(self)._create_daemon(daemon_spec, reconfig=(action == 'reconfig'),
+                                                      skip_restart_for_reconfig=skip_restart_for_reconfig,
+                                                      send_signal_to_daemon=send_signal_to_daemon))
 
         actions = {
             'start': ['reset-failed', 'start'],
diff --git a/src/pybind/mgr/cephadm/serve.py b/src/pybind/mgr/cephadm/serve.py
index 1f1b5c2fd6f..4f5d37250e2 100644
--- a/src/pybind/mgr/cephadm/serve.py
+++ b/src/pybind/mgr/cephadm/serve.py
@@ -1182,6 +1182,8 @@ class CephadmServe:
                     dd.hostname, dd.name()
                 )
             )
+            skip_restart_for_reconfig = False
+            send_signal_to_daemon = None
             if not last_config:
                 self.log.info('Reconfiguring %s (unknown last config time)...' % (
                     dd.name()))
@@ -1231,7 +1233,9 @@ class CephadmServe:
                     action = 'redeploy'
                 try:
                     daemon_spec = CephadmDaemonDeploySpec.from_daemon_description(dd)
-                    self.mgr._daemon_action(daemon_spec, action=action)
+                    self.mgr._daemon_action(daemon_spec, action=action,
+                                            skip_restart_for_reconfig=skip_restart_for_reconfig,
+                                            send_signal_to_daemon=send_signal_to_daemon)
                     if self.mgr.cache.rm_scheduled_daemon_action(dd.hostname, dd.name()):
                         self.mgr.cache.save_host(dd.hostname)
                 except OrchestratorError as e:
@@ -1411,6 +1415,8 @@ class CephadmServe:
                              daemon_spec: CephadmDaemonDeploySpec,
                              reconfig: bool = False,
                              osd_uuid_map: Optional[Dict[str, Any]] = None,
+                             skip_restart_for_reconfig: bool = False,
+                             send_signal_to_daemon: Optional[str] = None,
                              ) -> str:
 
         daemon_params: Dict[str, Any] = {}
@@ -1451,6 +1457,10 @@ class CephadmServe:
 
                 if reconfig:
                     daemon_params['reconfig'] = True
+                if skip_restart_for_reconfig:
+                    daemon_params['skip_restart_for_reconfig'] = True
+                if send_signal_to_daemon:
+                    daemon_params['send_signal_to_daemon'] = send_signal_to_daemon
                 if self.mgr.allow_ptrace:
                     daemon_params['allow_ptrace'] = True