1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
|
import errno
import logging
import os
import subprocess
import tempfile
from typing import Dict, Tuple, Any, List, cast, Optional
from mgr_module import HandleCommandResult
from mgr_module import NFS_POOL_NAME as POOL_NAME
from ceph.deployment.service_spec import ServiceSpec, NFSServiceSpec
from orchestrator import DaemonDescription
from cephadm.services.cephadmservice import AuthEntity, CephadmDaemonDeploySpec, CephService
logger = logging.getLogger(__name__)
class NFSService(CephService):
TYPE = 'nfs'
def ranked(self) -> bool:
return True
def fence(self, daemon_id: str) -> None:
logger.info(f'Fencing old nfs.{daemon_id}')
ret, out, err = self.mgr.mon_command({
'prefix': 'auth rm',
'entity': f'client.nfs.{daemon_id}',
})
# TODO: block/fence this entity (in case it is still running somewhere)
def fence_old_ranks(self,
spec: ServiceSpec,
rank_map: Dict[int, Dict[int, Optional[str]]],
num_ranks: int) -> None:
for rank, m in list(rank_map.items()):
if rank >= num_ranks:
for daemon_id in m.values():
if daemon_id is not None:
self.fence(daemon_id)
del rank_map[rank]
nodeid = f'{spec.service_name()}.{rank}'
self.mgr.log.info(f'Removing {nodeid} from the ganesha grace table')
self.run_grace_tool(cast(NFSServiceSpec, spec), 'remove', nodeid)
self.mgr.spec_store.save_rank_map(spec.service_name(), rank_map)
else:
max_gen = max(m.keys())
for gen, daemon_id in list(m.items()):
if gen < max_gen:
if daemon_id is not None:
self.fence(daemon_id)
del rank_map[rank][gen]
self.mgr.spec_store.save_rank_map(spec.service_name(), rank_map)
def config(self, spec: NFSServiceSpec) -> None: # type: ignore
from nfs.cluster import create_ganesha_pool
assert self.TYPE == spec.service_type
create_ganesha_pool(self.mgr)
def prepare_create(self, daemon_spec: CephadmDaemonDeploySpec) -> CephadmDaemonDeploySpec:
assert self.TYPE == daemon_spec.daemon_type
daemon_spec.final_config, daemon_spec.deps = self.generate_config(daemon_spec)
return daemon_spec
def generate_config(self, daemon_spec: CephadmDaemonDeploySpec) -> Tuple[Dict[str, Any], List[str]]:
assert self.TYPE == daemon_spec.daemon_type
daemon_type = daemon_spec.daemon_type
daemon_id = daemon_spec.daemon_id
host = daemon_spec.host
spec = cast(NFSServiceSpec, self.mgr.spec_store[daemon_spec.service_name].spec)
deps: List[str] = []
nodeid = f'{daemon_spec.service_name}.{daemon_spec.rank}'
# create the RADOS recovery pool keyring
rados_user = f'{daemon_type}.{daemon_id}'
rados_keyring = self.create_keyring(daemon_spec)
# ensure rank is known to ganesha
self.mgr.log.info(f'Ensuring {nodeid} is in the ganesha grace table')
self.run_grace_tool(spec, 'add', nodeid)
# create the rados config object
self.create_rados_config_obj(spec)
# create the RGW keyring
rgw_user = f'{rados_user}-rgw'
rgw_keyring = self.create_rgw_keyring(daemon_spec)
# generate the ganesha config
def get_ganesha_conf() -> str:
context = {
"user": rados_user,
"nodeid": nodeid,
"pool": POOL_NAME,
"namespace": spec.service_id,
"rgw_user": rgw_user,
"url": f'rados://{POOL_NAME}/{spec.service_id}/{spec.rados_config_name()}',
# fall back to default NFS port if not present in daemon_spec
"port": daemon_spec.ports[0] if daemon_spec.ports else 2049,
"bind_addr": daemon_spec.ip if daemon_spec.ip else '',
}
return self.mgr.template.render('services/nfs/ganesha.conf.j2', context)
# generate the cephadm config json
def get_cephadm_config() -> Dict[str, Any]:
config: Dict[str, Any] = {}
config['pool'] = POOL_NAME
config['namespace'] = spec.service_id
config['userid'] = rados_user
config['extra_args'] = ['-N', 'NIV_EVENT']
config['files'] = {
'ganesha.conf': get_ganesha_conf(),
}
config.update(
self.get_config_and_keyring(
daemon_type, daemon_id,
keyring=rados_keyring,
host=host
)
)
config['rgw'] = {
'cluster': 'ceph',
'user': rgw_user,
'keyring': rgw_keyring,
}
logger.debug('Generated cephadm config-json: %s' % config)
return config
return get_cephadm_config(), deps
def create_rados_config_obj(self,
spec: NFSServiceSpec,
clobber: bool = False) -> None:
objname = spec.rados_config_name()
cmd = [
'rados',
'-n', f"mgr.{self.mgr.get_mgr_id()}",
'-k', str(self.mgr.get_ceph_option('keyring')),
'-p', POOL_NAME,
'--namespace', cast(str, spec.service_id),
]
result = subprocess.run(
cmd + ['get', objname, '-'],
stdout=subprocess.PIPE, stderr=subprocess.PIPE,
timeout=10)
if not result.returncode and not clobber:
logger.info('Rados config object exists: %s' % objname)
else:
logger.info('Creating rados config object: %s' % objname)
result = subprocess.run(
cmd + ['put', objname, '-'],
stdout=subprocess.PIPE, stderr=subprocess.PIPE,
timeout=10)
if result.returncode:
self.mgr.log.warning(
f'Unable to create rados config object {objname}: {result.stderr.decode("utf-8")}'
)
raise RuntimeError(result.stderr.decode("utf-8"))
def create_keyring(self, daemon_spec: CephadmDaemonDeploySpec) -> str:
daemon_id = daemon_spec.daemon_id
spec = cast(NFSServiceSpec, self.mgr.spec_store[daemon_spec.service_name].spec)
entity: AuthEntity = self.get_auth_entity(daemon_id)
osd_caps = 'allow rw pool=%s namespace=%s' % (POOL_NAME, spec.service_id)
logger.info('Creating key for %s' % entity)
keyring = self.get_keyring_with_caps(entity,
['mon', 'allow r',
'osd', osd_caps])
return keyring
def create_rgw_keyring(self, daemon_spec: CephadmDaemonDeploySpec) -> str:
daemon_id = daemon_spec.daemon_id
entity: AuthEntity = self.get_auth_entity(f'{daemon_id}-rgw')
logger.info('Creating key for %s' % entity)
keyring = self.get_keyring_with_caps(entity,
['mon', 'allow r',
'osd', 'allow rwx tag rgw *=*'])
return keyring
def run_grace_tool(self,
spec: NFSServiceSpec,
action: str,
nodeid: str) -> None:
# write a temp keyring and referencing config file. this is a kludge
# because the ganesha-grace-tool can only authenticate as a client (and
# not a mgr). Also, it doesn't allow you to pass a keyring location via
# the command line, nor does it parse the CEPH_ARGS env var.
tmp_id = f'mgr.nfs.grace.{spec.service_name()}'
entity = AuthEntity(f'client.{tmp_id}')
keyring = self.get_keyring_with_caps(
entity,
['mon', 'allow r', 'osd', f'allow rwx pool {POOL_NAME}']
)
tmp_keyring = tempfile.NamedTemporaryFile(mode='w', prefix='mgr-grace-keyring')
os.fchmod(tmp_keyring.fileno(), 0o600)
tmp_keyring.write(keyring)
tmp_keyring.flush()
tmp_conf = tempfile.NamedTemporaryFile(mode='w', prefix='mgr-grace-conf')
tmp_conf.write(self.mgr.get_minimal_ceph_conf())
tmp_conf.write(f'\tkeyring = {tmp_keyring.name}\n')
tmp_conf.flush()
try:
cmd: List[str] = [
'ganesha-rados-grace',
'--cephconf', tmp_conf.name,
'--userid', tmp_id,
'--pool', POOL_NAME,
'--ns', cast(str, spec.service_id),
action, nodeid,
]
self.mgr.log.debug(cmd)
result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE,
timeout=10)
if result.returncode:
self.mgr.log.warning(
f'ganesha-rados-grace tool failed: {result.stderr.decode("utf-8")}'
)
raise RuntimeError(f'grace tool failed: {result.stderr.decode("utf-8")}')
finally:
self.mgr.check_mon_command({
'prefix': 'auth rm',
'entity': entity,
})
def remove_rgw_keyring(self, daemon: DaemonDescription) -> None:
assert daemon.daemon_id is not None
daemon_id: str = daemon.daemon_id
entity: AuthEntity = self.get_auth_entity(f'{daemon_id}-rgw')
logger.info(f'Removing key for {entity}')
self.mgr.check_mon_command({
'prefix': 'auth rm',
'entity': entity,
})
def post_remove(self, daemon: DaemonDescription, is_failed_deploy: bool) -> None:
super().post_remove(daemon, is_failed_deploy=is_failed_deploy)
self.remove_rgw_keyring(daemon)
def ok_to_stop(self,
daemon_ids: List[str],
force: bool = False,
known: Optional[List[str]] = None) -> HandleCommandResult:
# if only 1 nfs, alert user (this is not passable with --force)
warn, warn_message = self._enough_daemons_to_stop(self.TYPE, daemon_ids, 'NFS', 1, True)
if warn:
return HandleCommandResult(-errno.EBUSY, '', warn_message)
# if reached here, there is > 1 nfs daemon.
if force:
return HandleCommandResult(0, warn_message, '')
# if reached here, > 1 nfs daemon and no force flag.
# Provide warning
warn_message = "WARNING: Removing NFS daemons can cause clients to lose connectivity. "
return HandleCommandResult(-errno.EBUSY, '', warn_message)
def purge(self, service_name: str) -> None:
if service_name not in self.mgr.spec_store:
return
spec = cast(NFSServiceSpec, self.mgr.spec_store[service_name].spec)
logger.info(f'Removing grace file for {service_name}')
cmd = [
'rados',
'-n', f"mgr.{self.mgr.get_mgr_id()}",
'-k', str(self.mgr.get_ceph_option('keyring')),
'-p', POOL_NAME,
'--namespace', cast(str, spec.service_id),
'rm', 'grace',
]
subprocess.run(
cmd,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
timeout=10
)
|
