Merge remote-tracking branch 'gh/next'

PendingReleaseNotes

@@ -4,3 +4,18 @@ v0.67
 * The output of 'ceph status --format=json' or 'ceph -s --format=json'
   has changed to return status information in a more structured and
   usable format.
+
+* The 'ceph pg dump_stuck [threshold]' command used to require a
+  --threshold or -t prefix to the threshold argument, but now does
+  not.
+
+* Many more ceph commands now output formatted information; select
+  with '--format=<format>', where <format> can be 'json', 'json-pretty',
+  'xml', or 'xml-pretty'.
+
+* ceph-rest-api, a wrapper around ceph_rest_api.py, can be used to start
+  up a test single-threaded HTTP server that provides access to cluster
+  information and administration in very similar ways to the ceph
+  commandline tool.  ceph_rest_api.py can be used as a WSGI application
+  for deployment in a more-capable web server.  See ceph-rest-api.8
+  for more.

qa/workunits/cephtool/test.sh

@@ -79,10 +79,8 @@ if ! grep "$mymsg" /tmp/$$; then
 fi
 kill $wpid
 
-ceph mds cluster_down --no-log-to-stderr 2>&1 | grep "marked mdsmap DOWN"
-expect_false ceph mds cluster_down
-ceph mds cluster_up --no-log-to-stderr 2>&1 | grep "unmarked mdsmap DOWN"
-expect_false ceph mds cluster_up
+ceph mds cluster_down
+ceph mds cluster_up
 
 ceph mds compat rm_incompat 4
 ceph mds compat rm_incompat 4
@@ -167,9 +165,20 @@ ceph osd getmaxosd | grep 'max_osd = 10'
 ceph osd setmaxosd $save
 ceph osd getmaxosd | grep "max_osd = $save"
 
+for id in `ceph osd ls` ; do
+	ceph tell osd.$id version
+done
+
 id=`ceph osd create`
 ceph osd lost $id --yes-i-really-mean-it
 ceph osd rm $id
+
+uuid=`uuidgen`
+id=`ceph osd create $uuid`
+id2=`ceph osd create $uuid`
+[ "$id" = "$id2" ]
+ceph osd rm $id
+
 ceph osd ls
 ceph osd lspools | grep data
 ceph osd map data foo | grep 'pool.*data.*object.*foo.*pg.*up.*acting'
@@ -191,11 +200,6 @@ ceph osd pool delete data3 data3 --yes-i-really-really-mean-it
 
 ceph osd stat | grep up,
 
-for id in `ceph osd ls` ; do
-	ceph tell osd.$id version
-done
-
-
 ceph pg debug unfound_objects_exist
 ceph pg debug degraded_pgs_exist
 ceph pg deep-scrub 0.0

src/auth/cephx/CephxKeyServer.h

@@ -268,6 +268,9 @@ class KeyServer : public KeyStore {
     map<EntityName, EntityAuth>::const_iterator b = data.secrets_begin();
     return (b != data.secrets_end());
   }
+  int get_num_secrets() {
+    return data.secrets.size();
+  }
 
   /*void add_rotating_secret(uint32_t service_id, ExpiringCryptoKey& key) {
     Mutex::Locker l(lock);

src/ceph-disk

@@ -198,69 +198,95 @@ def maybe_mkdir(*a, **kw):
             raise
 
 
+# a device "name" is something like
+#  sdb
+#  cciss!c0d1
+def get_dev_name(path):
+    """
+    get device name from path.  e.g., /dev/sda -> sdas, /dev/cciss/c0d1 -> cciss!c0d1
+    """
+    assert path.startswith('/dev/')
+    base = path[5:]
+    return base.replace('/', '!')
+
+# a device "path" is something like
+#  /dev/sdb
+#  /dev/cciss/c0d1
+def get_dev_path(name):
+    """
+    get a path (/dev/...) from a name (cciss!c0d1)
+    """
+    return '/dev/' + name.replace('!', '/')
+
+def get_dev_relpath(name):
+    """
+    get a relative path to /dev from a name (cciss!c0d1)
+    """
+    return name.replace('!', '/')
+
+
+def get_partition_dev(dev, pnum):
+    """
+    get the device name for a partition
+
+    assume that partitions are named like the base dev, with a number, and optionally
+    some intervening characters (like 'p').  e.g.,
+
+       sda 1 -> sda1
+       cciss/c0d1 1 -> cciss!c0d1p1
+    """
+    name = get_dev_name(os.path.realpath(dev))
+    partname = None
+    for f in os.listdir(os.path.join('/sys/block', name)):
+        if f.startswith(name) and f.endswith(str(pnum)):
+            # we want the shortest name that starts with the base name and ends with the partition number
+            if not partname or len(f) < len(partname):
+                partname = f
+    if partname:
+        return get_dev_path(partname)
+    else:
+        raise Error('partition %d for %s does not appear to exist' % (pnum, dev))
+
 def list_all_partitions():
     """
     Return a list of devices and partitions
     """
     dev_part_list = {}
-    for name in os.listdir('/dev/disk/by-path'):
-        target = os.readlink(os.path.join('/dev/disk/by-path', name))
-        dev = target.split('/')[-1]
-        #print "name %s target %s dev %s" % (name, target, dev)
-        (baser) = re.search('(.*)-part\d+$', name)
-        if baser is not None:
-            basename = baser.group(1)
-            #print 'basename %s' % basename
-            base = os.readlink(os.path.join('/dev/disk/by-path', basename)).split('/')[-1]
-            if base not in dev_part_list:
-                dev_part_list[base] = []
-            dev_part_list[base].append(dev)
-        else:
-            if dev not in dev_part_list:
-                dev_part_list[dev] = []
+    for name in os.listdir('/sys/block'):
+        if not os.path.exists(os.path.join('/sys/block', name, 'device')):
+            continue
+        dev_part_list[name] = list_partitions(name)
     return dev_part_list
 
-
-def list_partitions(disk):
+def list_partitions(basename):
     """
-    Return a list of partitions on the given device
+    Return a list of partitions on the given device name
     """
-    disk = os.path.realpath(disk)
-    assert not is_partition(disk)
-    assert disk.startswith('/dev/')
-    base = disk.split('/')[-1]
     partitions = []
-    for name in os.listdir(os.path.join('/sys/block', base)):
-        if name.startswith(base):
-            partitions.append('/dev/' + name)
+    for name in os.listdir(os.path.join('/sys/block', basename)):
+        if name.startswith(basename):
+            partitions.append(name)
     return partitions
 
 
 def is_partition(dev):
     """
-    Check whether a given device is a partition or a full disk.
+    Check whether a given device path is a partition or a full disk.
     """
     dev = os.path.realpath(dev)
     if not stat.S_ISBLK(os.lstat(dev).st_mode):
         raise Error('not a block device', dev)
 
-    # we can't tell just from the name of the device if it is a
-    # partition or not.  look in the by-path dir and see if the
-    # referring symlink ends in -partNNN.
-    name = dev.split('/')[-1]
-    for name in os.listdir('/dev/disk/by-path'):
-        target = os.readlink(os.path.join('/dev/disk/by-path', name))
-        cdev = target.split('/')[-1]
-        if '/dev/' + cdev != dev:
-            continue
-        (baser) = re.search('(.*)-part\d+$', name)
-        if baser is not None:
+    name = get_dev_name(dev)
+    if os.path.exists(os.path.join('/sys/block', name)):
+        return False
+
+    # make sure it is a partition of something else
+    for basename in os.listdir('/sys/block'):
+        if os.path.exists(os.path.join('/sys/block', basename, name)):
             return True
-        else:
-            return False
 
-    # hrm, don't know...
-    return False
+    raise Error('not a disk or partition', dev)
 
 
 def is_mounted(dev):
@@ -288,7 +314,7 @@ def is_held(dev):
     """
     assert os.path.exists(dev)
     dev = os.path.realpath(dev)
-    base = dev.split('/')[-1]
+    base = get_dev_name(dev)
 
     # full disk?
     directory = '/sys/block/{base}/holders'.format(base=base)
@@ -320,7 +346,9 @@ def verify_not_in_use(dev):
         if holders:
             raise Error('Device is in use by a device-mapper mapping (dm-crypt?)' % dev, ','.join(holders))
     else:
-        for partition in list_partitions(dev):
+        basename = get_dev_name(os.path.realpath(dev))
+        for partname in list_partitions(basename):
+            partition = get_dev_path(partname)
             if is_mounted(partition):
                 raise Error('Device is mounted', partition)
             holders = is_held(partition)
@@ -762,16 +790,6 @@ def zap(dev):
         raise Error(e)
 
 
-def get_udev_version():
-    version = _check_output(
-        args=[
-            'udevadm',
-            '--version',
-        ],
-    )   
-    return int(version)
-
-
 def prepare_journal_dev(
     data,
     journal,
@@ -840,30 +858,9 @@ def prepare_journal_dev(
                 ],
             )
 
-        if get_udev_version() >= 172:
-            journal_symlink = '/dev/disk/by-partuuid/{journal_uuid}'.format(
-                journal_uuid=journal_uuid,
-                )
-        else:
-            # udev prior to version 172 doesn't create by-partuuid directory
-            # use by-path symlink instead.  This is the third symlink returned
-            # by udevadm when queried.
-            LOG.debug('Using alternate persistant name for journal symlink')
-            symlinks = _check_output(
-                args=[
-                    'udevadm',
-                    'info',
-                    '--query=symlink',
-                    '--name={name}'.format(name=os.path.basename(journal)),
-                    ],
-                )   
-            journal_symlink = None
-            for udev_line in symlinks.split():
-                if 'by-path' in udev_line:
-                    journal_symlink = '/dev/{symlink}-part{num}'.format(symlink=str(udev_line), num=num)
-                    break
-            if not journal_symlink:
-               raise Error('Unable to get device by path from udev')
+        journal_symlink = '/dev/disk/by-partuuid/{journal_uuid}'.format(
+            journal_uuid=journal_uuid,
+            )
 
         journal_dmcrypt = None
         if journal_dm_keypath:
@@ -1042,7 +1039,7 @@ def prepare_dev(
         except subprocess.CalledProcessError as e:
             raise Error(e)
 
-        rawdev = '{data}1'.format(data=data)
+        rawdev = get_partition_dev(data, 1)
 
     dev = None
     if osd_dm_keypath:
@@ -2009,7 +2006,7 @@ def is_suppressed(path):
     try:
         if not disk.startswith('/dev/') or not stat.S_ISBLK(os.lstat(path).st_mode):
             return False
-        base = disk.split('/')[-1]
+        base = get_dev_name(disk)
         while len(base):
             if os.path.exists(SUPPRESS_PREFIX + base):
                 return True
@@ -2023,7 +2020,7 @@ def set_suppress(path):
         raise Error('does not exist', path)
     if not stat.S_ISBLK(os.lstat(path).st_mode):
         raise Error('not a block device', path)
-    base = disk.split('/')[-1]
+    base = get_dev_name(disk)
 
     with file(SUPPRESS_PREFIX + base, 'w') as f:
         pass
@@ -2036,7 +2033,7 @@ def unset_suppress(path):
     if not stat.S_ISBLK(os.lstat(path).st_mode):
         raise Error('not a block device', path)
     assert disk.startswith('/dev/')
-    base = disk.split('/')[-1]
+    base = get_dev_name(disk)
 
     fn = SUPPRESS_PREFIX + base
     if not os.path.exists(fn):

src/ceph.in

@@ -142,8 +142,6 @@ def parse_cmdargs(args=None, target=''):
 
     parser.add_argument('-f', '--format', choices=['json', 'json-pretty',
                         'xml', 'xml-pretty', 'plain'], dest='output_format')
-    # for pg dump_stuck
-    parser.add_argument('--threshold', type=int, help='number of seconds for a pg to be considered stuck for pg dump_stuck')
 
     # returns a Namespace with the parsed args, and a list of all extras
     parsed_args, extras = parser.parse_known_args(args)
@@ -683,8 +681,6 @@ def main():
             compat = True
             if parsed_args.output_format:
                 childargs.extend(['--format', parsed_args.output_format])
-            if parsed_args.threshold:
-                childargs.extend(['--threshold', parsed_args.threshold])
             ret, outbuf, outs = send_command(cluster_handle, target, childargs,
                                              inbuf)
         elif ret:
@@ -699,6 +695,15 @@ def main():
             ret, outbuf, outs = new_style_command(parsed_args, childargs, target,
                                                   sigdict, inbuf, verbose)
 
+            # debug tool: send any successful command *again* to
+            # verify that it is idempotent.
+            if not ret and 'CEPH_CLI_TEST_DUP_COMMAND' in os.environ:
+                ret, outbuf, outs = new_style_command(parsed_args, childargs, target,
+                                                      sigdict, inbuf, verbose)
+                if ret < 0:
+                    ret = -ret
+                    print >> sys.stderr, prefix + 'Second attempt of previously successful command failed with {0}: {1}'.format(errno.errorcode[ret], outs)
+
         if ret < 0:
             ret = -ret
             print >> sys.stderr, prefix + 'Error {0}: {1}'.format(errno.errorcode[ret], outs)

src/ceph_mon.cc

@@ -76,7 +76,8 @@ int obtain_monmap(MonitorDBStore &store, bufferlist &bl)
     }
   }
 
-  if (store.exists("mon_sync", "in_sync")) {
+  if (store.exists("mon_sync", "in_sync")
+      || store.exists("mon_sync", "force_sync")) {
     dout(10) << __func__ << " detected aborted sync" << dendl;
     if (store.exists("mon_sync", "latest_monmap")) {
       int err = store.get("mon_sync", "latest_monmap", bl);

src/client/Client.cc

@@ -1882,7 +1882,7 @@ void Client::handle_mds_map(MMDSMap* m)
     int newstate = mdsmap->get_state(p->first);
     if (!mdsmap->is_up(p->first) ||
 	mdsmap->get_inst(p->first) != p->second->inst) {
-      messenger->mark_down(p->second->inst.addr);
+      messenger->mark_down(p->second->con);
       if (mdsmap->is_up(p->first))
 	p->second->inst = mdsmap->get_inst(p->first);
     } else if (oldstate == newstate)