mgr/cephadm: add ability to zap OSDs' devices while draining host

Currently, when cephadm drains a host, it will remove all OSDs on
the host, but provides no option to zap the OSD's devices afterwards.
Given users are draining the host likely to remove it from the cluster,
it makes sense some users would want to clean up the devices on the
host that were being used for OSDs. Cephadm already supports zapping
devices outside of host draining, so it makes shouldn't take much to
add that functionality to the host drain as well.

Fixes: https://tracker.ceph.com/issues/61593
Signed-off-by: Adam King <adking@redhat.com>

diff --git a/doc/cephadm/host-management.rst b/doc/cephadm/host-management.rst
index 781e6378e15bc88a46274aa974d8c0b400b3f148..4b964c5f455a7f57a8366e2b4f57a62c184db301 100644 (file)
--- a/doc/cephadm/host-management.rst
+++ b/doc/cephadm/host-management.rst
@@ -123,6 +123,14 @@ All OSDs on the host will be scheduled to be removed. You can check the progress
 
 See :ref:`cephadm-osd-removal` for more details about OSD removal.
 
+The ``orch host drain`` command also supports a ``--zap-osd-devices``
+flag. Setting this flag while draining a host will cause cephadm to zap
+the devices of the OSDs it is removing as part of the drain process
+
+.. prompt:: bash #
+
+   ceph orch host drain *<host>* --zap-osd-devices
+
 Use the following command to determine whether any daemons are still on the
 host:
 

diff --git a/src/pybind/mgr/cephadm/module.py b/src/pybind/mgr/cephadm/module.py
index ae8801e29109c76432efe8f5abc05a3bdb9604a8..912f070d3d5e01fe194d5ec79c78a083aab7b58b 100644 (file)
--- a/src/pybind/mgr/cephadm/module.py
+++ b/src/pybind/mgr/cephadm/module.py
@@ -3355,7 +3355,7 @@ Then run the following:
         return self.to_remove_osds.all_osds()
 
     @handle_orch_error
-    def drain_host(self, hostname: str, force: bool = False, keep_conf_keyring: bool = False) -> str:
+    def drain_host(self, hostname: str, force: bool = False, keep_conf_keyring: bool = False, zap_osd_devices: bool = False) -> str:
         """
         Drain all daemons from a host.
         :param host: host name
@@ -3381,7 +3381,7 @@ Then run the following:
         daemons: List[orchestrator.DaemonDescription] = self.cache.get_daemons_by_host(hostname)
 
         osds_to_remove = [d.daemon_id for d in daemons if d.daemon_type == 'osd']
-        self.remove_osds(osds_to_remove)
+        self.remove_osds(osds_to_remove, zap=zap_osd_devices)
 
         daemons_table = ""
         daemons_table += "{:<20} {:<15}\n".format("type", "id")

diff --git a/src/pybind/mgr/cephadm/tests/test_cephadm.py b/src/pybind/mgr/cephadm/tests/test_cephadm.py
index 6cc06cedcf47ae354ededa46085c8efc9adf2e0a..e40de481ff8704daca8949c5ee6cc8f675934a86 100644 (file)
--- a/src/pybind/mgr/cephadm/tests/test_cephadm.py
+++ b/src/pybind/mgr/cephadm/tests/test_cephadm.py
@@ -2539,3 +2539,14 @@ Traceback (most recent call last):
         with pytest.raises(OrchestratorError, match=r'Command "very slow" timed out on host hostC \(non-default 999 second timeout\)'):
             with cephadm_module.async_timeout_handler('hostC', 'very slow', 999):
                 cephadm_module.wait_async(_timeout())
+
+    @mock.patch("cephadm.CephadmOrchestrator.remove_osds")
+    @mock.patch("cephadm.CephadmOrchestrator.add_host_label", lambda *a, **kw: None)
+    @mock.patch("cephadm.inventory.HostCache.get_daemons_by_host", lambda *a, **kw: [])
+    def test_host_drain_zap(self, _rm_osds, cephadm_module):
+        # pass force=true in these tests to bypass _admin label check
+        cephadm_module.drain_host('host1', force=True, zap_osd_devices=False)
+        assert _rm_osds.called_with([], zap=False)
+
+        cephadm_module.drain_host('host1', force=True, zap_osd_devices=True)
+        assert _rm_osds.called_with([], zap=True)

diff --git a/src/pybind/mgr/orchestrator/_interface.py b/src/pybind/mgr/orchestrator/_interface.py
index 3ce81255432dab2c9ff3fc25253c65cdac524179..2c7776280028c15506f0a891f1ebafbd75b6bf41 100644 (file)
--- a/src/pybind/mgr/orchestrator/_interface.py
+++ b/src/pybind/mgr/orchestrator/_interface.py
@@ -367,7 +367,7 @@ class Orchestrator(object):
         """
         raise NotImplementedError()
 
-    def drain_host(self, hostname: str, force: bool = False, keep_conf_keyring: bool = False) -> OrchResult[str]:
+    def drain_host(self, hostname: str, force: bool = False, keep_conf_keyring: bool = False, zap_osd_devices: bool = False) -> OrchResult[str]:
         """
         drain all daemons from a host
 

diff --git a/src/pybind/mgr/orchestrator/module.py b/src/pybind/mgr/orchestrator/module.py
index a4ca7704f041a9a8055ca2cc4887b0f3b43a9fc5..24e7d8d4eab182c465c9a79fe51d90a84de1927d 100644 (file)
--- a/src/pybind/mgr/orchestrator/module.py
+++ b/src/pybind/mgr/orchestrator/module.py
@@ -478,9 +478,9 @@ class OrchestratorCli(OrchestratorClientMixin, MgrModule,
         return HandleCommandResult(stdout=completion.result_str())
 
     @_cli_write_command('orch host drain')
-    def _drain_host(self, hostname: str, force: bool = False, keep_conf_keyring: bool = False) -> HandleCommandResult:
+    def _drain_host(self, hostname: str, force: bool = False, keep_conf_keyring: bool = False, zap_osd_devices: bool = False) -> HandleCommandResult:
         """drain all daemons from a host"""
-        completion = self.drain_host(hostname, force, keep_conf_keyring)
+        completion = self.drain_host(hostname, force, keep_conf_keyring, zap_osd_devices)
         raise_if_exception(completion)
         return HandleCommandResult(stdout=completion.result_str())