Merge pull request ceph#61293 from adk3798/ceph-exporter-sock-dir

cephadm: fix handling of ceph-exporter sock-dir

Reviewed-by: Ernesto Puerta <[email protected]>

Author: adk3798

src/cephadm/cephadmlib/daemons/ceph.py

@@ -343,13 +343,9 @@ def get_daemon_args(self) -> List[str]:
             )
         return args
 
-    def validate(self) -> None:
-        if not os.path.isdir(self.sock_dir):
-            raise Error(
-                f'Desired sock dir for ceph-exporter is not directory: {self.sock_dir}'
-            )
-
     def container(self, ctx: CephadmContext) -> CephContainer:
+        uid, gid = self.uid_gid(ctx)
+        make_run_dir(ctx.fsid, uid, gid)
         ctr = daemon_to_container(ctx, self)
         return to_deployment_container(ctx, ctr)
 
@@ -366,6 +362,9 @@ def customize_container_mounts(
     ) -> None:
         cm = Ceph.get_ceph_mounts(ctx, self.identity)
         mounts.update(cm)
+        # we want to always mount /var/run/ceph/<fsid> to the sock
+        # dir within the container. See https://tracker.ceph.com/issues/69475
+        mounts.update({f'/var/run/ceph/{ctx.fsid}': self.sock_dir})
         if self.https_enabled:
             data_dir = self.identity.data_dir(ctx.data_dir)
             mounts.update({os.path.join(data_dir, 'etc/certs'): '/etc/certs'})
@@ -390,13 +389,6 @@ def customize_container_envs(
     def default_entrypoint(self) -> str:
         return self.entrypoint
 
-    def prepare_data_dir(self, data_dir: str, uid: int, gid: int) -> None:
-        if not os.path.exists(self.sock_dir):
-            os.mkdir(self.sock_dir)
-        # part of validation is for the sock dir, so we postpone
-        # it until now
-        self.validate()
-
     def create_daemon_dirs(self, data_dir: str, uid: int, gid: int) -> None:
         """Create files under the container data dir"""
         if not os.path.isdir(data_dir):

src/python-common/ceph/deployment/service_spec.py

@@ -2873,7 +2873,7 @@ def __init__(self,
             extra_entrypoint_args=extra_entrypoint_args)
 
         self.service_type = service_type
-        self.sock_dir = sock_dir
+        self.sock_dir = None
         self.addrs = addrs
         self.port = port
         self.prio_limit = prio_limit
@@ -2885,6 +2885,11 @@ def get_port_start(self) -> List[int]:
     def validate(self) -> None:
         super(CephExporterSpec, self).validate()
 
+        if self.sock_dir and self.sock_dir != '/var/run/ceph/':
+            raise SpecValidationError(
+                'sock_dir setting is deprecated and must be either unset or set to /var/run/ceph/'
+            )
+
         if not isinstance(self.prio_limit, int):
             raise SpecValidationError(
                     f'prio_limit must be an integer. Got {type(self.prio_limit)}')