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>

diff --git a/src/cephadm/cephadm b/src/cephadm/cephadm
index 585c0f3c7802cdef84a31f97a53af6626e66b41d..3735d096880243d6f0c6b9340b58fa52dc687698 100755 (executable)
--- a/src/cephadm/cephadm
+++ b/src/cephadm/cephadm
@@ -7711,11 +7711,11 @@ 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
         )
@@ -7725,13 +7725,13 @@ 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)
@@ -7754,8 +7754,14 @@ 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 03d872881d376e5f69317539d5a7fb4369164c28..4dedeeefbeea999e4343c3c2b78455e2d6d4dc63 100644 (file)
--- a/src/cephadm/tests/test_cephadm.py
+++ b/src/cephadm/tests/test_cephadm.py
@@ -921,14 +921,16 @@ 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'])