qa: misc cephfs test improvements

qa/suites/multimds/basic/q_check_counter/check_counter.yaml

@@ -0,0 +1,8 @@
+
+tasks:
+- check-counter:
+    counters:
+      mds:
+        - "mds.exported"
+        - "mds.imported"
+

qa/tasks/cephfs/test_exports.py

@@ -6,6 +6,8 @@
 log = logging.getLogger(__name__)
 
 class TestExports(CephFSTestCase):
+    MDSS_REQUIRED = 2
+
     def _wait_subtrees(self, status, rank, test):
         timeout = 30
         pause = 2
@@ -30,15 +32,15 @@ def test_export_pin(self):
         self._wait_subtrees(status, 0, [])
 
         # NOP
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "-1", "1"])
+        self.mount_a.setfattr("1", "ceph.dir.pin", "-1")
         self._wait_subtrees(status, 0, [])
 
         # NOP (rank < -1)
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "-2341", "1"])
+        self.mount_a.setfattr("1", "ceph.dir.pin", "-2341")
         self._wait_subtrees(status, 0, [])
 
         # pin /1 to rank 1
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "1", "1"])
+        self.mount_a.setfattr("1", "ceph.dir.pin", "1")
         self._wait_subtrees(status, 1, [('/1', 1)])
 
         # Check export_targets is set properly
@@ -48,34 +50,34 @@ def test_export_pin(self):
         self.assertTrue(sorted(r0['export_targets']) == [1])
 
         # redundant pin /1/2 to rank 1
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "1", "1/2"])
+        self.mount_a.setfattr("1/2", "ceph.dir.pin", "1")
         self._wait_subtrees(status, 1, [('/1', 1), ('/1/2', 1)])
 
         # change pin /1/2 to rank 0
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "0", "1/2"])
+        self.mount_a.setfattr("1/2", "ceph.dir.pin", "0")
         self._wait_subtrees(status, 1, [('/1', 1), ('/1/2', 0)])
         self._wait_subtrees(status, 0, [('/1', 1), ('/1/2', 0)])
 
         # change pin /1/2/3 to (presently) non-existent rank 2
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "2", "1/2/3"])
+        self.mount_a.setfattr("1/2/3", "ceph.dir.pin", "2")
         self._wait_subtrees(status, 0, [('/1', 1), ('/1/2', 0)])
         self._wait_subtrees(status, 1, [('/1', 1), ('/1/2', 0)])
 
         # change pin /1/2 back to rank 1
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "1", "1/2"])
+        self.mount_a.setfattr("1/2", "ceph.dir.pin", "1")
         self._wait_subtrees(status, 1, [('/1', 1), ('/1/2', 1)])
 
         # add another directory pinned to 1
         self.mount_a.run_shell(["mkdir", "-p", "1/4/5"])
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "1", "1/4/5"])
+        self.mount_a.setfattr("1/4/5", "ceph.dir.pin", "1")
         self._wait_subtrees(status, 1, [('/1', 1), ('/1/2', 1), ('/1/4/5', 1)])
 
         # change pin /1 to 0
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "0", "1"])
+        self.mount_a.setfattr("1", "ceph.dir.pin", "0")
         self._wait_subtrees(status, 0, [('/1', 0), ('/1/2', 1), ('/1/4/5', 1)])
 
         # change pin /1/2 to default (-1); does the subtree root properly respect it's parent pin?
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "-1", "1/2"])
+        self.mount_a.setfattr("1/2", "ceph.dir.pin", "-1")
         self._wait_subtrees(status, 0, [('/1', 0), ('/1/4/5', 1)])
 
         if len(list(status.get_standbys())):
@@ -95,8 +97,8 @@ def test_export_pin(self):
 
         # Test rename
         self.mount_a.run_shell(["mkdir", "-p", "a/b", "aa/bb"])
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "1", "a"])
-        self.mount_a.run_shell(["setfattr", "-n", "ceph.dir.pin", "-v", "0", "aa/bb"])
+        self.mount_a.setfattr("a", "ceph.dir.pin", "1")
+        self.mount_a.setfattr("aa/bb", "ceph.dir.pin", "0")
         self._wait_subtrees(status, 0, [('/1', 0), ('/1/4/5', 1), ('/1/2/3', 2), ('/a', 1), ('/aa/bb', 0)])
         self.mount_a.run_shell(["mv", "aa", "a/b/"])
         self._wait_subtrees(status, 0, [('/1', 0), ('/1/4/5', 1), ('/1/2/3', 2), ('/a', 1), ('/a/b/aa/bb', 0)])

qa/tasks/cephfs/test_journal_migration.py

@@ -9,19 +9,41 @@
 
 class TestJournalMigration(CephFSTestCase):
     CLIENTS_REQUIRED = 1
+    MDSS_REQUIRED = 2
 
     def test_journal_migration(self):
         old_journal_version = JOURNAL_FORMAT_LEGACY
         new_journal_version = JOURNAL_FORMAT_RESILIENT
 
-        self.fs.set_ceph_conf('mds', 'mds journal format', old_journal_version)
+        # Pick out two daemons to use
+        mds_a, mds_b = sorted(self.mds_cluster.mds_ids[0:2]) 
 
-        # Create a filesystem using the older journal format.
         self.mount_a.umount_wait()
         self.fs.mds_stop()
+
+        # Enable standby replay, to cover the bug case #8811 where
+        # a standby replay might mistakenly end up trying to rewrite
+        # the journal at the same time as an active daemon.
+        self.fs.set_ceph_conf('mds', 'mds standby replay', "true")
+        self.fs.set_ceph_conf('mds', 'mds standby for rank', "0")
+
+        # Create a filesystem using the older journal format.
+        self.fs.set_ceph_conf('mds', 'mds journal format', old_journal_version)
         self.fs.recreate()
-        self.fs.mds_restart()
+        self.fs.mds_restart(mds_id=mds_a)
         self.fs.wait_for_daemons()
+        self.assertEqual(self.fs.get_active_names(), [mds_a])
+
+        def replay_names():
+            return [s['name']
+                    for s in self.fs.status().get_replays(fscid = self.fs.id)]
+
+        # Start the standby and wait for it to come up
+        self.fs.mds_restart(mds_id=mds_b)
+        self.wait_until_equal(
+                replay_names,
+                [mds_b],
+                timeout = 30)
 
         # Do some client work so that the log is populated with something.
         with self.mount_a.mounted():
@@ -41,8 +63,8 @@ def test_journal_migration(self):
         self.fs.set_ceph_conf('mds', 'mds journal format', new_journal_version)
 
         # Restart the MDS.
-        self.fs.mds_fail_restart()
-        self.fs.wait_for_daemons()
+        self.fs.mds_fail_restart(mds_id=mds_a)
+        self.fs.mds_fail_restart(mds_id=mds_b)
 
         # This ensures that all daemons come up into a valid state
         self.fs.wait_for_daemons()
@@ -79,11 +101,18 @@ def test_journal_migration(self):
             # Approximate value of "lots", expected from having run fsstress
             raise RuntimeError("Unexpectedly few journal events: {0}".format(event_count))
 
-        # Do some client work so that the log is populated with something.
+        # Do some client work to check that writing the log is still working
         with self.mount_a.mounted():
             workunit(self.ctx, {
                 'clients': {
                     "client.{0}".format(self.mount_a.client_id): ["fs/misc/trivial_sync.sh"],
                 },
                 "timeout": "3h"
             })
+
+        # Check that both an active and a standby replay are still up
+        self.assertEqual(len(replay_names()), 1)
+        self.assertEqual(len(self.fs.get_active_names()), 1)
+        self.assertTrue(self.mds_cluster.mds_daemons[mds_a].running())
+        self.assertTrue(self.mds_cluster.mds_daemons[mds_b].running())
+