Merge PR ceph#19440 into master

* refs/pull/19440/head:
	mds: update mds option descriptions
	mds: obsolete MDSMap option configs
	mds: organize Filesystem class def

Reviewed-by: John Spray <[email protected]>

PendingReleaseNotes

@@ -20,6 +20,11 @@
       - mds remove_data_pool -> fs rm_data_pool
       - mds rm_data_pool -> fs rm_data_pool
 
+  * New CephFS file system attributes session_timeout and session_autoclose
+    are configurable via `ceph fs set`. The MDS config options
+    mds_session_timeout, mds_session_autoclose, and mds_max_file_size are now
+    obsolete.
+
 
 
 >= 12.2.2

doc/cephfs/dirfrags.rst

@@ -25,10 +25,9 @@ fragments may be *merged* to reduce the number of fragments in the directory.
 Splitting and merging
 =====================
 
-An MDS will only consider doing splits and merges if the ``mds_bal_frag``
-setting is true in the MDS's configuration file, and the allow_dirfrags
-setting is true in the filesystem map (set on the mons).  These settings
-are both true by default since the *Luminous* (12.2.x) release of Ceph.
+An MDS will only consider doing splits if the allow_dirfrags setting is true in
+the file system map (set on the mons).  This setting is true by default since
+the *Luminous* release (12.2.X).
 
 When an MDS identifies a directory fragment to be split, it does not
 do the split immediately.  Because splitting interrupts metadata IO,

doc/cephfs/eviction.rst

@@ -23,9 +23,9 @@ Automatic client eviction
 
 There are two situations in which a client may be evicted automatically:
 
-On an active MDS daemon, if a client has not communicated with the MDS for
-over ``mds_session_autoclose`` seconds (300 seconds by default), then it
-will be evicted automatically.
+On an active MDS daemon, if a client has not communicated with the MDS for over
+``session_autoclose`` (a file system variable) seconds (300 seconds by
+default), then it will be evicted automatically.
 
 During MDS startup (including on failover), the MDS passes through a
 state called ``reconnect``.  During this state, it waits for all the

doc/cephfs/health-messages.rst

@@ -69,7 +69,7 @@ are like locks.  Sometimes, for example when another client needs access,
 the MDS will request clients release their capabilities.  If the client
 is unresponsive or buggy, it might fail to do so promptly or fail to do
 so at all.  This message appears if a client has taken longer than
-``mds_session_timeout`` (default 60s) to comply.
+``session_timeout`` (default 60s) to comply.
 
 Message: "Client *name* failing to respond to cache pressure"
 Code: MDS_HEALTH_CLIENT_RECALL, MDS_HEALTH_CLIENT_RECALL_MANY

doc/cephfs/mds-config-ref.rst

@@ -10,15 +10,6 @@
 :Type: Boolean
 :Default: ``true`` 
 
-
-``mds max file size``
-
-:Description: The maximum allowed file size to set when creating a 
-              new file system.
-
-:Type:  64-bit Integer Unsigned
-:Default:  ``1ULL << 40``
-
 ``mds cache memory limit``
 
 :Description: The memory limit the MDS should enforce for its cache.
@@ -107,24 +98,6 @@
 :Default: ``24.0*60.0``
 
 
-``mds session timeout``
-
-:Description: The interval (in seconds) of client inactivity before Ceph 
-              times out capabilities and leases.
-
-:Type:  Float
-:Default: ``60``
-
-
-``mds session autoclose``
-
-:Description: The interval (in seconds) before Ceph closes 
-              a laggy client's session.
-
-:Type:  Float
-:Default: ``300``
-
-
 ``mds reconnect timeout``
 
 :Description: The interval (in seconds) to wait for clients to reconnect 
@@ -254,13 +227,6 @@
 :Default: ``0``
 
 
-``mds bal frag``
-
-:Description: Determines whether the MDS will fragment directories.
-:Type:  Boolean
-:Default:  ``false``
-
-
 ``mds bal split size``
 
 :Description: The maximum directory size before the MDS will split a directory 

qa/tasks/cephfs/filesystem.py

@@ -437,14 +437,18 @@ def deactivate(self, rank):
             raise RuntimeError("cannot deactivate rank 0")
         self.mon_manager.raw_cluster_cmd("mds", "deactivate", "%d:%d" % (self.id, rank))
 
+    def set_var(self, var, *args):
+        a = map(str, args)
+        self.mon_manager.raw_cluster_cmd("fs", "set", self.name, var, *a)
+
     def set_max_mds(self, max_mds):
-        self.mon_manager.raw_cluster_cmd("fs", "set", self.name, "max_mds", "%d" % max_mds)
+        self.set_var("max_mds", "%d" % max_mds)
 
     def set_allow_dirfrags(self, yes):
-        self.mon_manager.raw_cluster_cmd("fs", "set", self.name, "allow_dirfrags", str(yes).lower(), '--yes-i-really-mean-it')
+        self.set_var("allow_dirfrags", str(yes).lower(), '--yes-i-really-mean-it')
 
     def set_allow_new_snaps(self, yes):
-        self.mon_manager.raw_cluster_cmd("fs", "set", self.name, "allow_new_snaps", str(yes).lower(), '--yes-i-really-mean-it')
+        self.set_var("allow_new_snaps", str(yes).lower(), '--yes-i-really-mean-it')
 
     def get_pgs_per_fs_pool(self):
         """
@@ -559,6 +563,9 @@ def _df(self):
     def get_mds_map(self):
         return self.status().get_fsmap(self.id)['mdsmap']
 
+    def get_var(self, var):
+        return self.status().get_fsmap(self.id)['mdsmap'][var]
+
     def add_data_pool(self, name):
         self.mon_manager.raw_cluster_cmd('osd', 'pool', 'create', name, self.get_pgs_per_fs_pool().__str__())
         self.mon_manager.raw_cluster_cmd('fs', 'add_data_pool', self.name, name)

qa/tasks/cephfs/test_client_limits.py

@@ -134,10 +134,10 @@ def test_client_release_bug(self):
         # Client B tries to stat the file that client A created
         rproc = self.mount_b.write_background("file1")
 
-        # After mds_session_timeout, we should see a health warning (extra lag from
+        # After session_timeout, we should see a health warning (extra lag from
         # MDS beacon period)
-        mds_session_timeout = float(self.fs.get_config("mds_session_timeout"))
-        self.wait_for_health("MDS_CLIENT_LATE_RELEASE", mds_session_timeout + 10)
+        session_timeout = self.fs.get_var("session_timeout")
+        self.wait_for_health("MDS_CLIENT_LATE_RELEASE", session_timeout + 10)
 
         # Client B should still be stuck
         self.assertFalse(rproc.finished)

qa/tasks/cephfs/test_client_recovery.py

@@ -28,10 +28,9 @@ class TestClientNetworkRecovery(CephFSTestCase):
     REQUIRE_ONE_CLIENT_REMOTE = True
     CLIENTS_REQUIRED = 2
 
-    LOAD_SETTINGS = ["mds_session_timeout", "mds_reconnect_timeout", "ms_max_backoff"]
+    LOAD_SETTINGS = ["mds_reconnect_timeout", "ms_max_backoff"]
 
     # Environment references
-    mds_session_timeout = None
     mds_reconnect_timeout = None
     ms_max_backoff = None
 
@@ -43,6 +42,8 @@ def test_network_death(self):
         I/O after failure.
         """
 
+        session_timeout = self.fs.get_var("session_timeout")
+
         # We only need one client
         self.mount_b.umount_wait()
 
@@ -65,7 +66,7 @@ def test_network_death(self):
         # ...then it should block
         self.assertFalse(write_blocked.finished)
         self.assert_session_state(client_id, "open")
-        time.sleep(self.mds_session_timeout * 1.5)  # Long enough for MDS to consider session stale
+        time.sleep(session_timeout * 1.5)  # Long enough for MDS to consider session stale
         self.assertFalse(write_blocked.finished)
         self.assert_session_state(client_id, "stale")
 
@@ -85,10 +86,9 @@ class TestClientRecovery(CephFSTestCase):
     REQUIRE_KCLIENT_REMOTE = True
     CLIENTS_REQUIRED = 2
 
-    LOAD_SETTINGS = ["mds_session_timeout", "mds_reconnect_timeout", "ms_max_backoff"]
+    LOAD_SETTINGS = ["mds_reconnect_timeout", "ms_max_backoff"]
 
     # Environment references
-    mds_session_timeout = None
     mds_reconnect_timeout = None
     ms_max_backoff = None
 
@@ -212,6 +212,8 @@ def test_reconnect_eviction(self):
         self.mount_a.create_destroy()
 
     def test_stale_caps(self):
+        session_timeout = self.fs.get_var("session_timeout")
+
         # Capability release from stale session
         # =====================================
         cap_holder = self.mount_a.open_background()
@@ -224,7 +226,7 @@ def test_stale_caps(self):
         self.mount_a.kill()
 
         try:
-            # Now, after mds_session_timeout seconds, the waiter should
+            # Now, after session_timeout seconds, the waiter should
             # complete their operation when the MDS marks the holder's
             # session stale.
             cap_waiter = self.mount_b.write_background()
@@ -237,9 +239,9 @@ def test_stale_caps(self):
 
             cap_waited = b - a
             log.info("cap_waiter waited {0}s".format(cap_waited))
-            self.assertTrue(self.mds_session_timeout / 2.0 <= cap_waited <= self.mds_session_timeout * 2.0,
+            self.assertTrue(session_timeout / 2.0 <= cap_waited <= session_timeout * 2.0,
                             "Capability handover took {0}, expected approx {1}".format(
-                                cap_waited, self.mds_session_timeout
+                                cap_waited, session_timeout
                             ))
 
             cap_holder.stdin.close()
@@ -259,6 +261,8 @@ def test_evicted_caps(self):
         # Eviction while holding a capability
         # ===================================
 
+        session_timeout = self.fs.get_var("session_timeout")
+
         # Take out a write capability on a file on client A,
         # and then immediately kill it.
         cap_holder = self.mount_a.open_background()
@@ -288,9 +292,9 @@ def test_evicted_caps(self):
             log.info("cap_waiter waited {0}s".format(cap_waited))
             # This is the check that it happened 'now' rather than waiting
             # for the session timeout
-            self.assertLess(cap_waited, self.mds_session_timeout / 2.0,
+            self.assertLess(cap_waited, session_timeout / 2.0,
                             "Capability handover took {0}, expected less than {1}".format(
-                                cap_waited, self.mds_session_timeout / 2.0
+                                cap_waited, session_timeout / 2.0
                             ))
 
             cap_holder.stdin.close()

qa/tasks/cephfs/test_fragment.py

@@ -33,7 +33,6 @@ def _configure(self, **kwargs):
         Apply kwargs as MDS configuration settings, enable dirfrags
         and restart the MDSs.
         """
-        kwargs['mds_bal_frag'] = "true"
 
         for k, v in kwargs.items():
             self.ceph_cluster.set_ceph_conf("mds", k, v.__str__())

qa/tasks/cephfs/test_misc.py

@@ -11,9 +11,6 @@
 class TestMisc(CephFSTestCase):
     CLIENTS_REQUIRED = 2
 
-    LOAD_SETTINGS = ["mds_session_autoclose"]
-    mds_session_autoclose = None
-
     def test_getattr_caps(self):
         """
         Check if MDS recognizes the 'mask' parameter of open request.
@@ -104,14 +101,16 @@ def test_evict_client(self):
         only session
         """
 
+        session_autoclose = self.fs.get_var("session_autoclose")
+
         self.mount_b.umount_wait()
         ls_data = self.fs.mds_asok(['session', 'ls'])
         self.assert_session_count(1, ls_data)
 
         self.mount_a.kill()
         self.mount_a.kill_cleanup()
 
-        time.sleep(self.mds_session_autoclose * 1.5)
+        time.sleep(session_autoclose * 1.5)
         ls_data = self.fs.mds_asok(['session', 'ls'])
         self.assert_session_count(1, ls_data)
 
@@ -126,7 +125,7 @@ def test_evict_client(self):
         self.mount_a.kill()
         self.mount_a.kill_cleanup()
 
-        time.sleep(self.mds_session_autoclose * 1.5)
+        time.sleep(session_autoclose * 1.5)
         ls_data = self.fs.mds_asok(['session', 'ls'])
         self.assert_session_count(1, ls_data)
 

qa/tasks/cephfs/test_strays.py

@@ -962,7 +962,7 @@ def test_purge_queue_op_rate(self):
 
         max_purge_files = 2
 
-        self.set_conf('mds', 'mds_bal_frag', 'false')
+        self.fs.set_allow_dirfrags(True)
         self.set_conf('mds', 'mds_max_purge_files', "%d" % max_purge_files)
         self.fs.mds_fail_restart()
         self.fs.wait_for_daemons()

src/common/legacy_config_opts.h

@@ -398,7 +398,6 @@ OPTION(filer_max_purge_ops, OPT_U32)
 OPTION(filer_max_truncate_ops, OPT_U32)
 
 OPTION(mds_data, OPT_STR)
-OPTION(mds_max_file_size, OPT_U64) // Used when creating new CephFS. Change with 'ceph fs set <fs_name> max_file_size <size>' afterwards
 // max xattr kv pairs size for each dir/file
 OPTION(mds_max_xattr_pairs_size, OPT_U32)
 OPTION(mds_max_file_recover, OPT_U32)
@@ -409,17 +408,15 @@ OPTION(mds_beacon_interval, OPT_FLOAT)
 OPTION(mds_beacon_grace, OPT_FLOAT)
 OPTION(mds_enforce_unique_name, OPT_BOOL)
 
-OPTION(mds_session_timeout, OPT_FLOAT)    // cap bits and leases time out if client unresponsive or not returning its caps
 OPTION(mds_session_blacklist_on_timeout, OPT_BOOL)    // whether to blacklist clients whose sessions are dropped due to timeout
 OPTION(mds_session_blacklist_on_evict, OPT_BOOL)  // whether to blacklist clients whose sessions are dropped via admin commands
 
 OPTION(mds_sessionmap_keys_per_op, OPT_U32)    // how many sessions should I try to load/store in a single OMAP operation?
 OPTION(mds_recall_state_timeout, OPT_FLOAT)    // detect clients which aren't trimming caps
 OPTION(mds_freeze_tree_timeout, OPT_FLOAT)    // detecting freeze tree deadlock
-OPTION(mds_session_autoclose, OPT_FLOAT) // autoclose idle session
 OPTION(mds_health_summarize_threshold, OPT_INT) // collapse N-client health metrics to a single 'many'
 OPTION(mds_reconnect_timeout, OPT_FLOAT)  // seconds to wait for clients during mds restart
-	      //  make it (mds_session_timeout - mds_beacon_grace)
+	      //  make it (mdsmap.session_timeout - mds_beacon_grace)
 OPTION(mds_tick_interval, OPT_FLOAT)
 OPTION(mds_dirstat_min_interval, OPT_FLOAT)    // try to avoid propagating more often than this
 OPTION(mds_scatter_nudge_interval, OPT_FLOAT)  // how quickly dirstat changes propagate up the hierarchy
@@ -436,7 +433,6 @@ OPTION(mds_bal_export_pin, OPT_BOOL)  // allow clients to pin directory trees to
 OPTION(mds_bal_sample_interval, OPT_DOUBLE)  // every 3 seconds
 OPTION(mds_bal_replicate_threshold, OPT_FLOAT)
 OPTION(mds_bal_unreplicate_threshold, OPT_FLOAT)
-OPTION(mds_bal_frag, OPT_BOOL)
 OPTION(mds_bal_split_size, OPT_INT)
 OPTION(mds_bal_split_rd, OPT_FLOAT)
 OPTION(mds_bal_split_wr, OPT_FLOAT)