qa/cephfs: support CephFS recovery pools

Add support for testing recovery of CephFS metadata into an alternate
RADOS pool, useful as a disaster recovery mechanism that avoids
modifying the metadata in-place.

Signed-off-by: Douglas Fuller <dfuller@redhat.com>
(cherry picked from commit 8f9a25202093339afb5308051d354d3ae79c6b2d)

diff --git a/qa/tasks/cephfs/cephfs_test_case.py b/qa/tasks/cephfs/cephfs_test_case.py
index 7f5c6e9d834cdbab431f9a7905d1213d0b6a02f7..801d0d3114c4adb6971cee81cabce98d0921eeda 100644 (file)
--- a/qa/tasks/cephfs/cephfs_test_case.py
+++ b/qa/tasks/cephfs/cephfs_test_case.py
@@ -43,6 +43,7 @@ class CephFSTestCase(CephTestCase):
     # FIXME weird explicit naming
     mount_a = None
     mount_b = None
+    recovery_mount = None
 
     # Declarative test requirements: subclasses should override these to indicate
     # their special needs.  If not met, tests will be skipped.
@@ -55,6 +56,9 @@ class CephFSTestCase(CephTestCase):
     # Whether to create the default filesystem during setUp
     REQUIRE_FILESYSTEM = True
 
+    # requires REQUIRE_FILESYSTEM = True
+    REQUIRE_RECOVERY_FILESYSTEM = False
+
     LOAD_SETTINGS = []
 
     def setUp(self):
@@ -158,6 +162,20 @@ class CephFSTestCase(CephTestCase):
                 self.mounts[i].mount()
                 self.mounts[i].wait_until_mounted()
 
+        if self.REQUIRE_RECOVERY_FILESYSTEM:
+            if not self.REQUIRE_FILESYSTEM:
+                raise case.SkipTest("Recovery filesystem requires a primary filesystem as well")
+            self.fs.mon_manager.raw_cluster_cmd('fs', 'flag', 'set',
+                                                'enable_multiple', 'true',
+                                                '--yes-i-really-mean-it')
+            self.recovery_fs = self.mds_cluster.newfs(name="recovery_fs", create=False)
+            self.recovery_fs.set_metadata_overlay(True)
+            self.recovery_fs.set_data_pool_name(self.fs.get_data_pool_name())
+            self.recovery_fs.create()
+            self.recovery_fs.getinfo(refresh=True)
+            self.recovery_fs.mds_restart()
+            self.recovery_fs.wait_for_daemons()
+
         # Load an config settings of interest
         for setting in self.LOAD_SETTINGS:
             setattr(self, setting, float(self.fs.mds_asok(

diff --git a/qa/tasks/cephfs/filesystem.py b/qa/tasks/cephfs/filesystem.py
index f96b6c0afffd0bad69e2dc58a50a63cf776b4afc..ad15e909c45041cf19a940f518b49f0cf5960760 100644 (file)
--- a/qa/tasks/cephfs/filesystem.py
+++ b/qa/tasks/cephfs/filesystem.py
@@ -259,6 +259,9 @@ class MDSCluster(CephCluster):
     def newfs(self, name='cephfs', create=True):
         return Filesystem(self._ctx, name=name, create=create)
 
+    def newfs_raw(self, name):
+        return Filesystem(self._ctx, create=None, name=name)
+
     def status(self):
         return FSStatus(self.mon_manager)
 
@@ -367,8 +370,9 @@ class Filesystem(MDSCluster):
 
         self.name = name
         self.id = None
-        self.name = None
         self.metadata_pool_name = None
+        self.metadata_overlay = False
+        self.data_pool_name = None
         self.data_pools = None
 
         client_list = list(misc.all_roles_of_type(self._ctx.cluster, 'client'))
@@ -411,7 +415,7 @@ class Filesystem(MDSCluster):
         return status
 
     def set_metadata_overlay(self, overlay):
-        if fscid is not None:
+        if self.id is not None:
             raise RuntimeError("cannot specify fscid when configuring overlay")
         self.metadata_overlay = overlay
 
@@ -444,7 +448,10 @@ class Filesystem(MDSCluster):
             self.name = "cephfs"
         if self.metadata_pool_name is None:
             self.metadata_pool_name = "{0}_metadata".format(self.name)
-        data_pool_name = "{0}_data".format(self.name)
+        if self.data_pool_name is None:
+            data_pool_name = "{0}_data".format(self.name)
+        else:
+            data_pool_name = self.data_pool_name
 
         log.info("Creating filesystem '{0}'".format(self.name))
 
@@ -452,10 +459,15 @@ class Filesystem(MDSCluster):
 
         self.mon_manager.raw_cluster_cmd('osd', 'pool', 'create',
                                          self.metadata_pool_name, pgs_per_fs_pool.__str__())
-        self.mon_manager.raw_cluster_cmd('osd', 'pool', 'create',
-                                         data_pool_name, pgs_per_fs_pool.__str__())
-        self.mon_manager.raw_cluster_cmd('fs', 'new',
-                                         self.name, self.metadata_pool_name, data_pool_name)
+        if self.metadata_overlay:
+            self.mon_manager.raw_cluster_cmd('fs', 'new',
+                                             self.name, self.metadata_pool_name, data_pool_name,
+                                             '--allow-dangerous-metadata-overlay')
+        else:
+            self.mon_manager.raw_cluster_cmd('osd', 'pool', 'create',
+                                             data_pool_name, pgs_per_fs_pool.__str__())
+            self.mon_manager.raw_cluster_cmd('fs', 'new',
+                                             self.name, self.metadata_pool_name, data_pool_name)
         self.check_pool_application(self.metadata_pool_name)
         self.check_pool_application(data_pool_name)
         # Turn off spurious standby count warnings from modifying max_mds in tests.
@@ -568,6 +580,11 @@ class Filesystem(MDSCluster):
     def get_metadata_pool_name(self):
         return self.metadata_pool_name
 
+    def set_data_pool_name(self, name):
+        if self.id is not None:
+            raise RuntimeError("can't set filesystem name if its fscid is set")
+        self.data_pool_name = name
+
     def get_namespace_id(self):
         return self.id