]>
Commit | Line | Data |
---|---|---|
d2e6a577 FG |
1 | from __future__ import print_function |
2 | import json | |
b32b8144 | 3 | import logging |
d2e6a577 FG |
4 | from textwrap import dedent |
5 | from ceph_volume.util import prepare as prepare_utils | |
b32b8144 | 6 | from ceph_volume.util import encryption as encryption_utils |
181888fb | 7 | from ceph_volume.util import system, disk |
3a9019d9 | 8 | from ceph_volume.util.arg_validators import exclude_group_options |
181888fb | 9 | from ceph_volume import conf, decorators, terminal |
3efd9988 | 10 | from ceph_volume.api import lvm as api |
b32b8144 | 11 | from .common import prepare_parser, rollback_osd |
d2e6a577 FG |
12 | |
13 | ||
b32b8144 FG |
14 | logger = logging.getLogger(__name__) |
15 | ||
16 | ||
17 | def prepare_dmcrypt(key, device, device_type, tags): | |
18 | """ | |
19 | Helper for devices that are encrypted. The operations needed for | |
20 | block, db, wal, or data/journal devices are all the same | |
21 | """ | |
22 | if not device: | |
23 | return '' | |
24 | tag_name = 'ceph.%s_uuid' % device_type | |
25 | uuid = tags[tag_name] | |
26 | # format data device | |
27 | encryption_utils.luks_format( | |
28 | key, | |
29 | device | |
30 | ) | |
31 | encryption_utils.luks_open( | |
32 | key, | |
33 | device, | |
34 | uuid | |
35 | ) | |
36 | ||
37 | return '/dev/mapper/%s' % uuid | |
38 | ||
39 | ||
40 | def prepare_filestore(device, journal, secrets, tags, osd_id, fsid): | |
d2e6a577 | 41 | """ |
3efd9988 | 42 | :param device: The name of the logical volume to work with |
d2e6a577 FG |
43 | :param journal: similar to device but can also be a regular/plain disk |
44 | :param secrets: A dict with the secrets needed to create the osd (e.g. cephx) | |
45 | :param id_: The OSD id | |
46 | :param fsid: The OSD fsid, also known as the OSD UUID | |
47 | """ | |
48 | cephx_secret = secrets.get('cephx_secret', prepare_utils.create_key()) | |
d2e6a577 | 49 | |
b32b8144 FG |
50 | # encryption-only operations |
51 | if secrets.get('dmcrypt_key'): | |
52 | # format and open ('decrypt' devices) and re-assign the device and journal | |
53 | # variables so that the rest of the process can use the mapper paths | |
54 | key = secrets['dmcrypt_key'] | |
55 | device = prepare_dmcrypt(key, device, 'data', tags) | |
56 | journal = prepare_dmcrypt(key, journal, 'journal', tags) | |
57 | ||
94b18763 FG |
58 | # vdo detection |
59 | is_vdo = api.is_vdo(device) | |
d2e6a577 | 60 | # create the directory |
3efd9988 | 61 | prepare_utils.create_osd_path(osd_id) |
d2e6a577 FG |
62 | # format the device |
63 | prepare_utils.format_device(device) | |
64 | # mount the data device | |
94b18763 | 65 | prepare_utils.mount_osd(device, osd_id, is_vdo=is_vdo) |
d2e6a577 FG |
66 | # symlink the journal |
67 | prepare_utils.link_journal(journal, osd_id) | |
68 | # get the latest monmap | |
69 | prepare_utils.get_monmap(osd_id) | |
70 | # prepare the osd filesystem | |
1adf2230 | 71 | prepare_utils.osd_mkfs_filestore(osd_id, fsid, cephx_secret) |
d2e6a577 FG |
72 | # write the OSD keyring if it doesn't exist already |
73 | prepare_utils.write_keyring(osd_id, cephx_secret) | |
b32b8144 FG |
74 | if secrets.get('dmcrypt_key'): |
75 | # if the device is going to get activated right away, this can be done | |
76 | # here, otherwise it will be recreated | |
77 | encryption_utils.write_lockbox_keyring( | |
78 | osd_id, | |
79 | fsid, | |
80 | tags['ceph.cephx_lockbox_secret'] | |
81 | ) | |
d2e6a577 FG |
82 | |
83 | ||
b32b8144 | 84 | def prepare_bluestore(block, wal, db, secrets, tags, osd_id, fsid): |
3efd9988 FG |
85 | """ |
86 | :param block: The name of the logical volume for the bluestore data | |
87 | :param wal: a regular/plain disk or logical volume, to be used for block.wal | |
88 | :param db: a regular/plain disk or logical volume, to be used for block.db | |
89 | :param secrets: A dict with the secrets needed to create the osd (e.g. cephx) | |
90 | :param id_: The OSD id | |
91 | :param fsid: The OSD fsid, also known as the OSD UUID | |
92 | """ | |
93 | cephx_secret = secrets.get('cephx_secret', prepare_utils.create_key()) | |
b32b8144 FG |
94 | # encryption-only operations |
95 | if secrets.get('dmcrypt_key'): | |
96 | # If encrypted, there is no need to create the lockbox keyring file because | |
97 | # bluestore re-creates the files and does not have support for other files | |
98 | # like the custom lockbox one. This will need to be done on activation. | |
99 | # format and open ('decrypt' devices) and re-assign the device and journal | |
100 | # variables so that the rest of the process can use the mapper paths | |
101 | key = secrets['dmcrypt_key'] | |
102 | block = prepare_dmcrypt(key, block, 'block', tags) | |
103 | wal = prepare_dmcrypt(key, wal, 'wal', tags) | |
104 | db = prepare_dmcrypt(key, db, 'db', tags) | |
3efd9988 | 105 | |
3efd9988 FG |
106 | # create the directory |
107 | prepare_utils.create_osd_path(osd_id, tmpfs=True) | |
108 | # symlink the block | |
109 | prepare_utils.link_block(block, osd_id) | |
110 | # get the latest monmap | |
111 | prepare_utils.get_monmap(osd_id) | |
112 | # write the OSD keyring if it doesn't exist already | |
113 | prepare_utils.write_keyring(osd_id, cephx_secret) | |
114 | # prepare the osd filesystem | |
115 | prepare_utils.osd_mkfs_bluestore( | |
116 | osd_id, fsid, | |
117 | keyring=cephx_secret, | |
118 | wal=wal, | |
119 | db=db | |
120 | ) | |
d2e6a577 FG |
121 | |
122 | ||
123 | class Prepare(object): | |
124 | ||
125 | help = 'Format an LVM device and associate it with an OSD' | |
126 | ||
127 | def __init__(self, argv): | |
128 | self.argv = argv | |
b32b8144 | 129 | self.osd_id = None |
d2e6a577 | 130 | |
3efd9988 | 131 | def get_ptuuid(self, argument): |
181888fb FG |
132 | uuid = disk.get_partuuid(argument) |
133 | if not uuid: | |
134 | terminal.error('blkid could not detect a PARTUUID for device: %s' % argument) | |
3efd9988 | 135 | raise RuntimeError('unable to use device') |
181888fb FG |
136 | return uuid |
137 | ||
92f5a8d4 | 138 | def setup_device(self, device_type, device_name, tags, size): |
3efd9988 FG |
139 | """ |
140 | Check if ``device`` is an lv, if so, set the tags, making sure to | |
141 | update the tags with the lv_uuid and lv_path which the incoming tags | |
142 | will not have. | |
143 | ||
144 | If the device is not a logical volume, then retrieve the partition UUID | |
145 | by querying ``blkid`` | |
146 | """ | |
147 | if device_name is None: | |
148 | return '', '', tags | |
149 | tags['ceph.type'] = device_type | |
94b18763 | 150 | tags['ceph.vdo'] = api.is_vdo(device_name) |
f6b5b4d7 TL |
151 | |
152 | try: | |
153 | vg_name, lv_name = device_name.split('/') | |
154 | lv = api.get_first_lv(filters={'lv_name': lv_name, | |
155 | 'vg_name': vg_name}) | |
156 | except ValueError: | |
157 | lv = None | |
158 | ||
3efd9988 FG |
159 | if lv: |
160 | uuid = lv.lv_uuid | |
161 | path = lv.lv_path | |
162 | tags['ceph.%s_uuid' % device_type] = uuid | |
163 | tags['ceph.%s_device' % device_type] = path | |
164 | lv.set_tags(tags) | |
92f5a8d4 TL |
165 | elif disk.is_device(device_name): |
166 | # We got a disk, create an lv | |
167 | lv_type = "osd-{}".format(device_type) | |
168 | uuid = system.generate_uuid() | |
169 | tags['ceph.{}_uuid'.format(device_type)] = uuid | |
170 | kwargs = { | |
171 | 'device': device_name, | |
172 | 'tags': tags, | |
173 | } | |
174 | if size != 0: | |
175 | kwargs['size'] = disk.Size.parse(size) | |
176 | lv = api.create_lv( | |
177 | lv_type, | |
178 | uuid, | |
179 | **kwargs) | |
180 | path = lv.lv_path | |
181 | tags['ceph.{}_device'.format(device_type)] = path | |
182 | lv.set_tags(tags) | |
3efd9988 FG |
183 | else: |
184 | # otherwise assume this is a regular disk partition | |
185 | uuid = self.get_ptuuid(device_name) | |
186 | path = device_name | |
187 | tags['ceph.%s_uuid' % device_type] = uuid | |
188 | tags['ceph.%s_device' % device_type] = path | |
189 | return path, uuid, tags | |
190 | ||
92f5a8d4 | 191 | def prepare_data_device(self, device_type, osd_uuid): |
3efd9988 FG |
192 | """ |
193 | Check if ``arg`` is a device or partition to create an LV out of it | |
194 | with a distinct volume group name, assigning LV tags on it and | |
195 | ultimately, returning the logical volume object. Failing to detect | |
196 | a device or partition will result in error. | |
197 | ||
198 | :param arg: The value of ``--data`` when parsing args | |
199 | :param device_type: Usually, either ``data`` or ``block`` (filestore vs. bluestore) | |
92f5a8d4 | 200 | :param osd_uuid: The OSD uuid |
3efd9988 | 201 | """ |
92f5a8d4 TL |
202 | device = self.args.data |
203 | if disk.is_partition(device) or disk.is_device(device): | |
3efd9988 | 204 | # we must create a vg, and then a single lv |
92f5a8d4 TL |
205 | lv_name_prefix = "osd-{}".format(device_type) |
206 | kwargs = {'device': device, | |
207 | 'tags': {'ceph.type': device_type}, | |
208 | } | |
209 | logger.debug('data device size: {}'.format(self.args.data_size)) | |
210 | if self.args.data_size != 0: | |
211 | kwargs['size'] = disk.Size.parse(self.args.data_size) | |
3efd9988 | 212 | return api.create_lv( |
92f5a8d4 TL |
213 | lv_name_prefix, |
214 | osd_uuid, | |
215 | **kwargs) | |
3efd9988 FG |
216 | else: |
217 | error = [ | |
92f5a8d4 | 218 | 'Cannot use device ({}).'.format(device), |
b32b8144 | 219 | 'A vg/lv path or an existing device is needed'] |
3efd9988 FG |
220 | raise RuntimeError(' '.join(error)) |
221 | ||
92f5a8d4 | 222 | raise RuntimeError('no data logical volume found with: {}'.format(device)) |
3efd9988 | 223 | |
91327a77 | 224 | def safe_prepare(self, args=None): |
b32b8144 FG |
225 | """ |
226 | An intermediate step between `main()` and `prepare()` so that we can | |
227 | capture the `self.osd_id` in case we need to rollback | |
91327a77 AA |
228 | |
229 | :param args: Injected args, usually from `lvm create` which compounds | |
230 | both `prepare` and `create` | |
b32b8144 | 231 | """ |
91327a77 AA |
232 | if args is not None: |
233 | self.args = args | |
f6b5b4d7 TL |
234 | |
235 | try: | |
236 | vgname, lvname = self.args.data.split('/') | |
237 | lv = api.get_first_lv(filters={'lv_name': lvname, | |
238 | 'vg_name': vgname}) | |
239 | except ValueError: | |
240 | lv = None | |
241 | ||
242 | if api.is_ceph_device(lv): | |
92f5a8d4 TL |
243 | logger.info("device {} is already used".format(self.args.data)) |
244 | raise RuntimeError("skipping {}, it is already prepared".format(self.args.data)) | |
b32b8144 | 245 | try: |
91327a77 | 246 | self.prepare() |
b32b8144 | 247 | except Exception: |
28e407b8 | 248 | logger.exception('lvm prepare was unable to complete') |
b32b8144 | 249 | logger.info('will rollback OSD ID creation') |
91327a77 | 250 | rollback_osd(self.args, self.osd_id) |
b32b8144 | 251 | raise |
91327a77 AA |
252 | terminal.success("ceph-volume lvm prepare successful for: %s" % self.args.data) |
253 | ||
254 | def get_cluster_fsid(self): | |
255 | """ | |
256 | Allows using --cluster-fsid as an argument, but can fallback to reading | |
257 | from ceph.conf if that is unset (the default behavior). | |
258 | """ | |
259 | if self.args.cluster_fsid: | |
260 | return self.args.cluster_fsid | |
261 | else: | |
262 | return conf.ceph.get('global', 'fsid') | |
b32b8144 | 263 | |
d2e6a577 | 264 | @decorators.needs_root |
91327a77 | 265 | def prepare(self): |
d2e6a577 FG |
266 | # FIXME we don't allow re-using a keyring, we always generate one for the |
267 | # OSD, this needs to be fixed. This could either be a file (!) or a string | |
268 | # (!!) or some flags that we would need to compound into a dict so that we | |
269 | # can convert to JSON (!!!) | |
270 | secrets = {'cephx_secret': prepare_utils.create_key()} | |
b32b8144 | 271 | cephx_lockbox_secret = '' |
91327a77 | 272 | encrypted = 1 if self.args.dmcrypt else 0 |
b32b8144 FG |
273 | cephx_lockbox_secret = '' if not encrypted else prepare_utils.create_key() |
274 | ||
275 | if encrypted: | |
276 | secrets['dmcrypt_key'] = encryption_utils.create_dmcrypt_key() | |
277 | secrets['cephx_lockbox_secret'] = cephx_lockbox_secret | |
d2e6a577 | 278 | |
91327a77 AA |
279 | cluster_fsid = self.get_cluster_fsid() |
280 | ||
281 | osd_fsid = self.args.osd_fsid or system.generate_uuid() | |
282 | crush_device_class = self.args.crush_device_class | |
b32b8144 FG |
283 | if crush_device_class: |
284 | secrets['crush_device_class'] = crush_device_class | |
285 | # reuse a given ID if it exists, otherwise create a new ID | |
91327a77 | 286 | self.osd_id = prepare_utils.create_id(osd_fsid, json.dumps(secrets), osd_id=self.args.osd_id) |
b32b8144 FG |
287 | tags = { |
288 | 'ceph.osd_fsid': osd_fsid, | |
289 | 'ceph.osd_id': self.osd_id, | |
290 | 'ceph.cluster_fsid': cluster_fsid, | |
291 | 'ceph.cluster_name': conf.cluster, | |
292 | 'ceph.crush_device_class': crush_device_class, | |
e306af50 | 293 | 'ceph.osdspec_affinity': prepare_utils.get_osdspec_affinity() |
b32b8144 | 294 | } |
91327a77 | 295 | if self.args.filestore: |
92f5a8d4 TL |
296 | #TODO: allow auto creation of journal on passed device, only works |
297 | # when physical device is passed, not LV | |
91327a77 | 298 | if not self.args.journal: |
b5b8bbf5 | 299 | raise RuntimeError('--journal is required when using --filestore') |
b5b8bbf5 | 300 | |
f6b5b4d7 TL |
301 | try: |
302 | vg_name, lv_name = self.args.data.split('/') | |
303 | data_lv = api.get_first_lv(filters={'lv_name': lv_name, | |
304 | 'vg_name': vg_name}) | |
305 | except ValueError: | |
306 | data_lv = None | |
307 | ||
3efd9988 | 308 | if not data_lv: |
92f5a8d4 | 309 | data_lv = self.prepare_data_device('data', osd_fsid) |
181888fb | 310 | |
b32b8144 FG |
311 | tags['ceph.data_device'] = data_lv.lv_path |
312 | tags['ceph.data_uuid'] = data_lv.lv_uuid | |
313 | tags['ceph.cephx_lockbox_secret'] = cephx_lockbox_secret | |
314 | tags['ceph.encrypted'] = encrypted | |
94b18763 | 315 | tags['ceph.vdo'] = api.is_vdo(data_lv.lv_path) |
3efd9988 | 316 | |
91327a77 | 317 | journal_device, journal_uuid, tags = self.setup_device( |
92f5a8d4 | 318 | 'journal', self.args.journal, tags, self.args.journal_size) |
3efd9988 FG |
319 | |
320 | tags['ceph.type'] = 'data' | |
321 | data_lv.set_tags(tags) | |
d2e6a577 FG |
322 | |
323 | prepare_filestore( | |
324 | data_lv.lv_path, | |
325 | journal_device, | |
326 | secrets, | |
b32b8144 FG |
327 | tags, |
328 | self.osd_id, | |
329 | osd_fsid, | |
d2e6a577 | 330 | ) |
91327a77 | 331 | elif self.args.bluestore: |
f6b5b4d7 TL |
332 | try: |
333 | vg_name, lv_name = self.args.data.split('/') | |
334 | block_lv = api.get_first_lv(filters={'lv_name': lv_name, | |
335 | 'vg_name': vg_name}) | |
336 | except ValueError: | |
337 | block_lv = None | |
338 | ||
3efd9988 | 339 | if not block_lv: |
92f5a8d4 | 340 | block_lv = self.prepare_data_device('block', osd_fsid) |
3efd9988 | 341 | |
b32b8144 FG |
342 | tags['ceph.block_device'] = block_lv.lv_path |
343 | tags['ceph.block_uuid'] = block_lv.lv_uuid | |
344 | tags['ceph.cephx_lockbox_secret'] = cephx_lockbox_secret | |
345 | tags['ceph.encrypted'] = encrypted | |
94b18763 | 346 | tags['ceph.vdo'] = api.is_vdo(block_lv.lv_path) |
3efd9988 | 347 | |
92f5a8d4 TL |
348 | wal_device, wal_uuid, tags = self.setup_device( |
349 | 'wal', self.args.block_wal, tags, self.args.block_wal_size) | |
350 | db_device, db_uuid, tags = self.setup_device( | |
351 | 'db', self.args.block_db, tags, self.args.block_db_size) | |
3efd9988 FG |
352 | |
353 | tags['ceph.type'] = 'block' | |
354 | block_lv.set_tags(tags) | |
355 | ||
356 | prepare_bluestore( | |
357 | block_lv.lv_path, | |
358 | wal_device, | |
359 | db_device, | |
360 | secrets, | |
b32b8144 FG |
361 | tags, |
362 | self.osd_id, | |
363 | osd_fsid, | |
3efd9988 | 364 | ) |
d2e6a577 FG |
365 | |
366 | def main(self): | |
367 | sub_command_help = dedent(""" | |
368 | Prepare an OSD by assigning an ID and FSID, registering them with the | |
369 | cluster with an ID and FSID, formatting and mounting the volume, and | |
370 | finally by adding all the metadata to the logical volumes using LVM | |
371 | tags, so that it can later be discovered. | |
372 | ||
373 | Once the OSD is ready, an ad-hoc systemd unit will be enabled so that | |
374 | it can later get activated and the OSD daemon can get started. | |
375 | ||
b32b8144 | 376 | Encryption is supported via dmcrypt and the --dmcrypt flag. |
d2e6a577 | 377 | |
91327a77 | 378 | Existing logical volume (lv): |
3efd9988 | 379 | |
91327a77 | 380 | ceph-volume lvm prepare --data {vg/lv} |
3efd9988 | 381 | |
92f5a8d4 | 382 | Existing block device (a logical volume will be created): |
d2e6a577 | 383 | |
91327a77 | 384 | ceph-volume lvm prepare --data /path/to/device |
d2e6a577 | 385 | |
92f5a8d4 TL |
386 | Optionally, can consume db and wal devices, partitions or logical |
387 | volumes. A device will get a logical volume, partitions and existing | |
388 | logical volumes will be used as is: | |
d2e6a577 | 389 | |
92f5a8d4 | 390 | ceph-volume lvm prepare --data {vg/lv} --block.wal {partition} --block.db {/path/to/device} |
d2e6a577 FG |
391 | """) |
392 | parser = prepare_parser( | |
393 | prog='ceph-volume lvm prepare', | |
394 | description=sub_command_help, | |
395 | ) | |
396 | if len(self.argv) == 0: | |
397 | print(sub_command_help) | |
398 | return | |
3a9019d9 | 399 | exclude_group_options(parser, argv=self.argv, groups=['filestore', 'bluestore']) |
91327a77 AA |
400 | self.args = parser.parse_args(self.argv) |
401 | # the unfortunate mix of one superset for both filestore and bluestore | |
402 | # makes this validation cumbersome | |
403 | if self.args.filestore: | |
404 | if not self.args.journal: | |
405 | raise SystemExit('--journal is required when using --filestore') | |
3efd9988 FG |
406 | # Default to bluestore here since defaulting it in add_argument may |
407 | # cause both to be True | |
91327a77 AA |
408 | if not self.args.bluestore and not self.args.filestore: |
409 | self.args.bluestore = True | |
92f5a8d4 | 410 | self.safe_prepare() |