import shlex
import pwd
import grp
+import syslog
"""
Prepare:
- if encrypted, map the dmcrypt volume
- mount the volume in a temp location
- allocate an osd id (if needed)
- - if deactived, change the gpt partition info correctly
+ - if deactived, no-op (to activate with --reactivate flag)
- remount in the correct location /var/lib/ceph/osd/$cluster-$id
- - remove the deactive flag
+ - remove the deactive flag (with --reactivate flag)
- start ceph-osd
- triggered by udev when it sees the OSD gpt partition type
Deactivate:
- stop ceph-osd service if needed (make osd out with option --mark-out)
- remove 'ready', 'active', and INIT-specific files
- - remove gpt partition type and change partition name (prevent triggered by udev)
- create deactive flag
- umount device and remove mount point
CEPH_OSD_ONDISK_MAGIC = 'ceph osd volume v026'
-LINUX_RESERVED_TYPE = '8da63339-0007-60c0-c436-083ac8230908'
JOURNAL_UUID = '45b0969e-9b03-4f30-b4c6-b4b80ceff106'
MPATH_JOURNAL_UUID = '45b0969e-8ae0-4982-bf9d-5a8d867af560'
DMCRYPT_JOURNAL_UUID = '45b0969e-9b03-4f30-b4c6-5ec00ceff106'
init,
dmcrypt,
dmcrypt_key_dir,
+ reactivate=False,
):
if dmcrypt:
# check if the disk is deactive, change the journal owner, group
# mode for correct user and group.
if os.path.exists(os.path.join(path, 'deactive')):
+ # logging to syslog will help us easy to know udev triggered failure
+ if not reactivate:
+ unmount(path)
+ syslog.syslog(syslog.LOG_ERR, 'OSD deactivated! reactivate with: --reactivate')
+ raise Error('OSD deactivated! reactivate with: --reactivate')
# flag to activate a deactive osd.
deactive = True
journal_dev = os.path.realpath(os.path.join(path,'journal'))
(osd_id, cluster) = activate(path, activate_key_template, init)
# Now active successfully
- # change the gpt partition type for bootup (meet the udev rules)
- if deactive:
- # Change OSD gpt partition type
- if is_mpath(dev):
- type_code = MPATH_OSD_UUID
- else:
- type_code = OSD_UUID
- _change_gpt_partition_info(dev, type_code)
-
- # Change Journal gpt partition type
- if is_mpath(journal_dev):
- type_code = MPATH_JOURNAL_UUID
- else:
- type_code = JOURNAL_UUID
- _change_gpt_partition_info(journal_dev, type_code)
-
- # Remove the deactive flag
+ # If we got reactivate and deactive, remove the deactive file
+ if deactive and reactivate:
try:
os.remove(os.path.join(path, 'deactive'))
LOG.info('Remove `deactive` file.')
init=args.mark_init,
dmcrypt=args.dmcrypt,
dmcrypt_key_dir=args.dmcrypt_key_dir,
+ reactivate=args.reactivate,
)
osd_data = get_mount_point(cluster, osd_id)
11(3) : means OSD IN AND UP
"""
LOG.info("Checking osd id: %s ..." % osd_id)
+ found = False
status_code = 0
try:
out, ret = command(
except subprocess.CalledProcessError as e:
raise Error(e)
out_json = json.loads(out)
- if out_json[u'osds'][int(osd_id)][u'in'] is 1:
- status_code += 2
- if out_json[u'osds'][int(osd_id)][u'up'] is 1:
- status_code += 1
+ for item in out_json[u'osds']:
+ if item.get(u'osd') is int(osd_id):
+ found = True
+ if item.get(u'in') is 1:
+ status_code += 2
+ if item.get(u'up') is 1:
+ status_code += 1
+ if not found:
+ raise Error('Could not osd.%s in osd tree!' % osd_id)
return status_code
pass
-def _change_gpt_partition_info(device_part, type_code=LINUX_RESERVED_TYPE):
- """
- Due to udev rule 95-ceph-osd.rules, we need to remove the
- gpt partition type to prevent trigger ceph-disk-activate.
-
- Also change partition name for zap in destroy stage
- """
-
- (device, part_num) = split_dev_base_partnum(device_part)
-
- part_name = get_partition_name(device_part)
-
- if type_code is LINUX_RESERVED_TYPE:
- part_name = part_name + ' (deactive)'
-
- if type_code is MPATH_JOURNAL_UUID or type_code is JOURNAL_UUID or \
- type_code is MPATH_OSD_UUID or type_code is OSD_UUID:
- part_name = part_name.replace(" (deactive)", "")
-
- try:
- command_check_call(
- [
- 'sgdisk',
- '--change-name=%s:%s' % (part_num, part_name),
- '--typecode=%s:%s' % (part_num, type_code),
- '--',
- device,
- ],
- )
- except subprocess.CalledProcessError as e:
- raise Error(e)
-
-
def main_deactivate(args):
+ mount_info = []
if args.cluster is None:
args.cluster = 'ceph'
- if args.osd_id is None:
- raise Error("osd id can not be zero. Try to use --osd-id <OSDID>.")
+
+ if args.deactivate_by_id:
+ osd_id = args.deactivate_by_id
+ else:
+ if not os.path.exists(args.path):
+ raise Error('%s does not exist' % args.path)
+ else:
+ mounted_path = is_mounted(args.path)
+ if mounted_path is None:
+ raise Error('%s is not mounted' % args.path)
+ osd_id = get_oneliner(mounted_path, 'whoami')
+ mount_info.append(args.path)
+ mount_info.append(mounted_path)
+
# Do not do anything if osd is already down.
- status_code = _check_osd_status(args.cluster, args.osd_id)
+ status_code = _check_osd_status(args.cluster, osd_id)
if status_code == OSD_STATUS_IN_UP:
if args.mark_out is True:
- _mark_osd_out(args.cluster, args.osd_id)
- stop_daemon(args.cluster, args.osd_id)
+ _mark_osd_out(args.cluster, osd_id)
+ stop_daemon(args.cluster, osd_id)
elif status_code == OSD_STATUS_IN_DOWN:
if args.mark_out is True:
- _mark_osd_out(args.cluster, args.osd_id)
+ _mark_osd_out(args.cluster, osd_id)
+ LOG.info("OSD already out/down. Do not do anything now.")
+ return
elif status_code == OSD_STATUS_OUT_UP:
- stop_daemon(args.cluster, args.osd_id)
+ stop_daemon(args.cluster, osd_id)
elif status_code == OSD_STATUS_OUT_DOWN:
LOG.info("OSD already out/down. Do not do anything now.")
return
# GET the mounted device and mount point.
- mount_info = convert_osd_id(args.cluster, args.osd_id)
+ # If we already get mount_info (with specific parameter), pass this stage
+ if not mount_info:
+ mount_info = convert_osd_id(args.cluster, osd_id)
# remove 'ready', 'active', and INIT-specific files.
_remove_osd_directory_files(mount_info[1], args.cluster)
- # Remove filesystem gpt partition type
- _change_gpt_partition_info(mount_info[0], LINUX_RESERVED_TYPE)
-
- # Check journal
- # if journal is exist, remove the gpt partition type
- journal_path = os.path.join(mount_info[1], 'journal')
- if os.path.exists(journal_path) and os.path.islink(journal_path):
- _change_gpt_partition_info(os.path.realpath(journal_path), \
- LINUX_RESERVED_TYPE)
- else:
- LOG.info('Journal is not exist on osd.%s (or not symlink).', \
- args.osd_id)
-
# Write deactivate to osd directory!
with file(os.path.join(mount_info[1], 'deactive'), 'w'):
path_set_context(os.path.join(mount_info[1], 'deactive'))
raise Error(e)
def main_destroy(args):
+ mount_info = []
if args.cluster is None:
args.cluster = 'ceph'
- if args.osd_id is None:
- raise Error("osd id can not be zero. Try to use --osd-id <OSDID>.")
+
+ if args.destroy_by_id:
+ osd_id = args.destroy_by_id
+ else:
+ if not os.path.exists(args.path):
+ raise Error('%s does not exist' % args.path)
+ else:
+ # mount point is removed, try to mount to tmp.folder
+ try:
+ fs_type = get_dev_fs(args.path)
+ if fs_type != None:
+ tpath = mount(dev=args.path, fstype=fs_type, options='')
+ if tpath:
+ try:
+ osd_id = get_oneliner(tpath, 'whoami')
+ finally:
+ unmount(tpath)
+ except MountError:
+ pass
+ mount_info.append(args.path)
+
# Before osd deactivate, we cannot destroy it
- status_code = _check_osd_status(args.cluster, args.osd_id)
- if status_code != OSD_STATUS_OUT_DOWN:
+ status_code = _check_osd_status(args.cluster, osd_id)
+ if status_code != OSD_STATUS_OUT_DOWN and \
+ status_code != OSD_STATUS_IN_DOWN:
raise Error("Could not destroy the active osd. (osd-id: %s)" % \
- args.osd_id)
-
- # GET the mounted device and mount point.
- mount_info = convert_osd_id(args.cluster, args.osd_id)
+ osd_id)
# Remove OSD from crush map
- _remove_from_crush_map(args.cluster, args.osd_id)
+ _remove_from_crush_map(args.cluster, osd_id)
# Remove OSD cephx key
- _delete_osd_auth_key(args.cluster, args.osd_id)
+ _delete_osd_auth_key(args.cluster, osd_id)
# Deallocate OSD ID
- _deallocate_osd_id(args.cluster, args.osd_id)
+ _deallocate_osd_id(args.cluster, osd_id)
# Check zap flag. If we found zap flag, we need to find device for
# destroy this osd data.
if args.zap is True:
- # try to find osd data device.
- partmap = list_all_partitions(None)
- # list all partition which have the partition name with
- # deactive flag
- devtocheck = []
- for base, parts in sorted(partmap.iteritems()):
- if not parts:
- continue
- for p in parts:
- (dev, p_num) = split_dev_base_partnum(os.path.join("/dev", p))
- part_name = get_partition_name(os.path.join("/dev", p))
- LOG.debug("device: %s, p_num: %s" % (dev, p_num))
- LOG.debug("part_name: %s" % part_name)
- if part_name == "ceph data (deactive)" or \
- part_name == "ceph journal (deactive)":
+ # easy to do when we get device
+ if mount_info:
+ base_dev = get_partition_base(mount_info[0])
+ else:
+ # try to find osd data device.
+ partmap = list_all_partitions(None)
+ # list all partition which have the partition name with
+ # deactive flag
+ devtocheck = []
+ found = False
+ for base, parts in sorted(partmap.iteritems()):
+ if not parts:
+ continue
+ for p in parts:
+ (dev, p_num) = split_dev_base_partnum(os.path.join("/dev", p))
+ LOG.debug("device: %s, p_num: %s" % (dev, p_num))
devtocheck.append(os.path.join("/dev", p))
- # check all above device's osd_id
- # if the osd_id is correct, zap it.
- for item in devtocheck:
- try:
- fs_type = get_dev_fs(item)
- if fs_type != None:
- tpath = mount(dev=item, fstype=fs_type, options='')
- if tpath:
- try:
- whoami = get_oneliner(tpath, 'whoami')
- finally:
- unmount(tpath)
- if whoami is args.osd_id:
- (dev, part_num) = split_dev_base_partnum(item)
- except MountError:
- pass
+ # check all above device's osd_id
+ # if the osd_id is correct, zap it.
+ for item in devtocheck:
+ try:
+ whoami = -1
+ fs_type = get_dev_fs(item)
+ if fs_type != None:
+ tpath = mount(dev=item, fstype=fs_type, options='')
+ if tpath:
+ try:
+ whoami = get_oneliner(tpath, 'whoami')
+ finally:
+ unmount(tpath)
+ if whoami is osd_id:
+ found = True
+ (base_dev, part_num) = split_dev_base_partnum(item)
+ except MountError:
+ pass
+ if not found:
+ raise Error('Could not find the partition of osd.%s!' % osd_id)
# earse the osd data
- LOG.info("Prepare to zap the device %s" % dev)
- zap(dev)
+ LOG.info("Prepare to zap the device %s" % base_dev)
+ zap(base_dev)
return
def get_partition_uuid(part):
return get_sgdisk_partition_info(part, 'Partition unique GUID: (\S+)')
-def get_partition_name(part):
- regexp = "Partition name: \'*([A-Za-z ]+[ ()A-Za-z]*)\'*"
- return get_sgdisk_partition_info(part, regexp)
-
def get_sgdisk_partition_info(dev, regexp):
(base, partnum) = split_dev_base_partnum(dev)
out, _ = command(['sgdisk', '-i', partnum, base])
default='/etc/ceph/dmcrypt-keys',
help='directory where dm-crypt keys are stored',
)
+ activate_parser.add_argument(
+ '--reactivate',
+ action='store_true', default=False,
+ help='activate the deactived OSD',
+ )
activate_parser.set_defaults(
activate_key_template='{statedir}/bootstrap-osd/{cluster}.keyring',
func=main_activate,
help='cluster name to assign this disk to',
)
deactivate_parser.add_argument(
- '--osd-id',
- metavar='OSDID',
- help='ID of OSD to deactivate'
+ 'path',
+ metavar='PATH',
+ nargs='?',
+ help='path to block device or directory',
+ )
+ deactivate_parser.add_argument(
+ '--deactivate-by-id',
+ metavar='<id>',
+ help='ID of OSD to deactive'
)
deactivate_parser.add_argument(
'--mark-out',
action='store_true', default=False,
- help='option to mark this osd out',
+ help='option to mark the osd out',
)
deactivate_parser.set_defaults(
func=main_deactivate,
help='cluster name to assign this disk to',
)
destroy_parser.add_argument(
- '--osd-id',
- metavar='OSDID',
+ 'path',
+ metavar='PATH',
+ nargs='?',
+ help='path to block device or directory',
+ )
+ destroy_parser.add_argument(
+ '--destroy-by-id',
+ metavar='<id>',
help='ID of OSD to destroy'
)
destroy_parser.add_argument(
projects / ceph.git / commitdiff