cephadm: fix exiting maintenance when systemd target doesn't exist

If the systemd target doesn't exist we need to just bypass enabling
it and return success or the host will just be stuck in maitnenance
mode.

Signed-off-by: Adam King <adking@redhat.com>
(cherry picked from commit 3a15f1dc81dfd5f963e1bf36316fa5b289a898b0)

diff --git a/src/cephadm/cephadm b/src/cephadm/cephadm
index fbf5a81d22ac1f14cb4a54920a0b3c4f2575accc..477719bbec679fde93ef01de31940ea778bd3110 100755 (executable)
--- a/src/cephadm/cephadm
+++ b/src/cephadm/cephadm
@@ -7685,11 +7685,11 @@ def command_exporter(ctx: CephadmContext) -> None:
 ##################################
 
 
-def systemd_target_state(target_name: str, subsystem: str = 'ceph') -> bool:
+def systemd_target_state(ctx: CephadmContext, target_name: str, subsystem: str = 'ceph') -> bool:
     # TODO: UNITTEST
     return os.path.exists(
         os.path.join(
-            UNIT_DIR,
+            ctx.unit_dir,
             f'{subsystem}.target.wants',
             target_name
         )
@@ -7699,13 +7699,13 @@ def systemd_target_state(target_name: str, subsystem: str = 'ceph') -> bool:
 @infer_fsid
 def command_maintenance(ctx: CephadmContext) -> str:
     if not ctx.fsid:
-        raise Error('must pass --fsid to specify cluster')
+        raise Error('failed - must pass --fsid to specify cluster')
 
     target = f'ceph-{ctx.fsid}.target'
 
     if ctx.maintenance_action.lower() == 'enter':
         logger.info('Requested to place host into maintenance')
-        if systemd_target_state(target):
+        if systemd_target_state(ctx, target):
             _out, _err, code = call(ctx,
                                     ['systemctl', 'disable', target],
                                     verbosity=CallVerbosity.DEBUG)
@@ -7728,8 +7728,14 @@ def command_maintenance(ctx: CephadmContext) -> str:
 
     else:
         logger.info('Requested to exit maintenance state')
+        # if we've never deployed a daemon on this host there will be no systemd
+        # target to disable so attempting a disable will fail. We still need to
+        # return success here or host will be permanently stuck in maintenance mode
+        # as no daemons can be deployed so no systemd target will ever exist to disable.
+        if not os.path.exists(ctx.unit_dir + '/ceph.target'):
+            return 'skipped - systemd target not present on this host. Host removed from maintenance mode.'
         # exit maintenance request
-        if not systemd_target_state(target):
+        if not systemd_target_state(ctx, target):
             _out, _err, code = call(ctx,
                                     ['systemctl', 'enable', target],
                                     verbosity=CallVerbosity.DEBUG)

diff --git a/src/cephadm/tests/test_cephadm.py b/src/cephadm/tests/test_cephadm.py
index 16e32833f636458330a53533d419387e29e43ba7..43a0e4a35ed4091e92fffd7de5788870806e73aa 100644 (file)
--- a/src/cephadm/tests/test_cephadm.py
+++ b/src/cephadm/tests/test_cephadm.py
@@ -901,14 +901,16 @@ class TestMaintenance:
         wants.mkdir()
         target = wants / TestMaintenance.systemd_target
         target.touch()
-        cd.UNIT_DIR = str(base)
+        ctx = cd.CephadmContext()
+        ctx.unit_dir = str(base)
 
-        assert cd.systemd_target_state(target.name)
+        assert cd.systemd_target_state(ctx, target.name)
 
     def test_systemd_target_NOTOK(self, tmp_path):
         base = tmp_path 
-        cd.UNIT_DIR = str(base)
-        assert not cd.systemd_target_state(TestMaintenance.systemd_target)
+        ctx = cd.CephadmContext()
+        ctx.unit_dir = str(base)
+        assert not cd.systemd_target_state(ctx, TestMaintenance.systemd_target)
 
     def test_parser_OK(self):
         args = cd._parse_args(['host-maintenance', 'enter'])