Trash RBD image on namespace deletion.

Fixes ceph#953

Signed-off-by: Gil Bregman <[email protected]>

.github/workflows/build-container.yml

@@ -139,7 +139,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        test: ["cli", "cli_change_lb", "cli_change_keys", "cli_change_ns_visibility", "state", "multi_gateway", "server", "grpc", "omap_lock", "log_files", "nsid", "psk", "dhchap", "subsys_grp_name_append"]
+        test: ["cli", "cli_change_lb", "cli_change_keys", "cli_change_ns_visibility", "cli_trash_rbd", "state", "multi_gateway", "server", "grpc", "omap_lock", "log_files", "nsid", "psk", "dhchap", "subsys_grp_name_append"]
     runs-on: ubuntu-latest
     env:
       HUGEPAGES: 512  # for multi gateway test, approx 256 per gateway instance

ceph-nvmeof.conf

@@ -85,6 +85,7 @@ timeout = 60.0
 
 # Example value: {"max_queue_depth" : 16, "max_io_size" : 4194304, "io_unit_size" : 1048576, "zcopy" : false}
 transport_tcp_options = {"in_capsule_data_size" : 8192, "max_io_qpairs_per_ctrlr" : 7}
+#iobuf_options = {"small_pool_count" : 8192, "large_pool_count" : 1024, "small_bufsize" : 8192, "large_bufsize" : 135168}
 
 [monitor]
 #timeout = 1.0

control/cephutils.py

@@ -82,8 +82,8 @@ def get_number_created_gateways(self, pool, group, caching=True):
                     self.rebalance_supported = True
                     self.rebalance_ana_group = data.get("rebalance_ana_group", None)
                     self.num_gws = data.get("num gws", None)
-                    self.logger.info(f"Rebalance ana_group: {self.rebalance_ana_group},\
-                                      num-gws: {self.num_gws} ")
+                    self.logger.info(f"Rebalance ana_group: {self.rebalance_ana_group}, "
+                                     f"num-gws: {self.num_gws}")
                 else:
                     self.rebalance_supported = False
                 pos = conv_str.find("[")
@@ -191,6 +191,28 @@ def create_image(self, pool_name, image_name, size) -> bool:
 
         return True
 
+    def delete_image(self, pool_name, image_name) -> bool:
+        if not pool_name and not image_name:
+            return True
+
+        if not self.pool_exists(pool_name):
+            self.logger.warning(f"Pool {pool_name} doesn't exist, can't delete RBD image")
+            return True
+
+        with rados.Rados(conffile=self.ceph_conf, rados_id=self.rados_id) as cluster:
+            with cluster.open_ioctx(pool_name) as ioctx:
+                rbd_inst = rbd.RBD()
+                try:
+                    rbd_inst.remove(ioctx, image_name)
+                except rbd.ImageNotFound:
+                    self.logger.warning(f"Image {pool_name}/{image_name} is not found")
+                    return True
+                except (rbd.ImageBusy, rbd.ImageHasSnapshots):
+                    self.logger.exception(f"Can't delete image {pool_name}/{image_name}")
+                    return False
+
+        return True
+
     def get_image_size(self, pool_name, image_name) -> int:
         image_size = 0
         if not self.pool_exists(pool_name):

control/cli.py

@@ -1726,6 +1726,10 @@ def ns_add(self, args):
                 self.cli.parser.error("--size argument is not allowed for add command when "
                                       "RBD image creation is disabled")
 
+        if args.rbd_trash_image_on_delete and not args.rbd_create_image:
+            self.cli.parser.error("Can't trash associated RBD image on delete if it wasn't "
+                                  "created automatically by the gateway")
+
         req = pb2.namespace_add_req(rbd_pool_name=args.rbd_pool,
                                     rbd_image_name=args.rbd_image,
                                     subsystem_nqn=args.subsystem,
@@ -1736,7 +1740,8 @@ def ns_add(self, args):
                                     create_image=args.rbd_create_image,
                                     size=img_size,
                                     force=args.force,
-                                    no_auto_visible=args.no_auto_visible)
+                                    no_auto_visible=args.no_auto_visible,
+                                    trash_image=args.rbd_trash_image_on_delete)
         try:
             ret = self.stub.namespace_add(req)
         except Exception as ex:
@@ -1776,7 +1781,7 @@ def ns_del(self, args):
 
         try:
             ret = self.stub.namespace_delete(pb2.namespace_delete_req(
-                subsystem_nqn=args.subsystem, nsid=args.nsid))
+                subsystem_nqn=args.subsystem, nsid=args.nsid, are_you_sure=args.are_you_sure))
         except Exception as ex:
             ret = pb2.req_status(status=errno.EINVAL,
                                  error_message=f"Failure deleting namespace:\n{ex}")
@@ -1925,15 +1930,15 @@ def ns_list(self, args):
                 namespaces_list = []
                 for ns in namespaces_info.namespaces:
                     if args.subsystem == GatewayUtils.ALL_SUBSYSTEMS:
-                        if not ns.subsystem_nqn:
+                        if not ns.ns_subsystem_nqn:
                             err_func(f"Got namespace with ID {ns.nsid} on an unknown subsystem")
                             subsys_nqn = "<n/a>"
                         else:
-                            subsys_nqn = ns.subsystem_nqn
+                            subsys_nqn = ns.ns_subsystem_nqn
                     else:
-                        if ns.subsystem_nqn and ns.subsystem_nqn != args.subsystem:
+                        if ns.ns_subsystem_nqn and ns.ns_subsystem_nqn != args.subsystem:
                             err_func(f"Got a namespace with ID {ns.nsid} in subsystem "
-                                     f"{ns.subsystem_nqn} which is different than the "
+                                     f"{ns.ns_subsystem_nqn} which is different than the "
                                      f"requested one {args.subsystem}")
                             return errno.ENODEV
                         subsys_nqn = namespaces_info.subsystem_nqn
@@ -1960,10 +1965,11 @@ def ns_list(self, args):
                         else:
                             visibility = "Restrictive"
 
+                    trash_msg = "\n(trash on deletion)" if ns.trash_image else ""
                     namespaces_list.append([subsys_nqn,
                                             ns.nsid,
                                             break_string(ns.bdev_name, "-", 2),
-                                            f"{ns.rbd_pool_name}/{ns.rbd_image_name}",
+                                            f"{ns.rbd_pool_name}/{ns.rbd_image_name}{trash_msg}",
                                             self.format_size(ns.rbd_image_size),
                                             self.format_size(ns.block_size),
                                             break_string(ns.uuid, "-", 3),
@@ -2454,12 +2460,20 @@ def ns_change_visibility(self, args):
                  help="Make the namespace visible only to specific hosts",
                  action='store_true',
                  required=False),
+        argument("--rbd-trash-image-on-delete",
+                 help="When deleting the namespace, trash associated RBD image. "
+                      "Only applies to images created automatically by the gateway",
+                 action='store_true',
+                 required=False),
     ]
     ns_del_args_list = ns_common_args + [
         argument("--nsid",
                  help="Namespace ID",
                  type=int,
                  required=True),
+        argument("--are-you-sure",
+                 help="If you choose to delete the associated RBD image, set this to \"yes\"",
+                 required=False),
     ]
     ns_resize_args_list = ns_common_args + [
         argument("--nsid",