Revision 814386b7
b/Makefile.am | ||
---|---|---|
321 | 321 |
lib/cmdlib/operating_system.py \ |
322 | 322 |
lib/cmdlib/tags.py \ |
323 | 323 |
lib/cmdlib/network.py \ |
324 |
lib/cmdlib/misc.py \ |
|
324 | 325 |
lib/cmdlib/test.py |
325 | 326 |
|
326 | 327 |
hypervisor_PYTHON = \ |
b/lib/cmdlib/__init__.py | ||
---|---|---|
28 | 28 |
|
29 | 29 |
# C0302: since we have waaaay too many lines in this module |
30 | 30 |
|
31 |
import time |
|
32 |
import logging |
|
33 |
|
|
34 |
from ganeti import utils |
|
35 |
from ganeti import errors |
|
36 |
from ganeti import locking |
|
37 |
from ganeti import constants |
|
38 |
from ganeti import compat |
|
39 |
from ganeti import query |
|
40 |
from ganeti import qlang |
|
41 |
|
|
42 | 31 |
from ganeti.cmdlib.base import ResultWithJobs, LogicalUnit, NoHooksLU, \ |
43 | 32 |
Tasklet, _QueryBase |
44 | 33 |
from ganeti.cmdlib.common import INSTANCE_DOWN, INSTANCE_ONLINE, \ |
... | ... | |
86 | 75 |
from ganeti.cmdlib.tags import LUTagsGet, LUTagsSearch, LUTagsSet, LUTagsDel |
87 | 76 |
from ganeti.cmdlib.network import LUNetworkAdd, LUNetworkRemove, \ |
88 | 77 |
LUNetworkSetParams, LUNetworkQuery, LUNetworkConnect, LUNetworkDisconnect |
78 |
from ganeti.cmdlib.misc import LUOobCommand, LUExtStorageDiagnose, \ |
|
79 |
LURestrictedCommand |
|
89 | 80 |
from ganeti.cmdlib.test import LUTestDelay, LUTestJqueue, LUTestAllocator |
90 |
|
|
91 |
|
|
92 |
class LUOobCommand(NoHooksLU): |
|
93 |
"""Logical unit for OOB handling. |
|
94 |
|
|
95 |
""" |
|
96 |
REQ_BGL = False |
|
97 |
_SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE) |
|
98 |
|
|
99 |
def ExpandNames(self): |
|
100 |
"""Gather locks we need. |
|
101 |
|
|
102 |
""" |
|
103 |
if self.op.node_names: |
|
104 |
self.op.node_names = _GetWantedNodes(self, self.op.node_names) |
|
105 |
lock_names = self.op.node_names |
|
106 |
else: |
|
107 |
lock_names = locking.ALL_SET |
|
108 |
|
|
109 |
self.needed_locks = { |
|
110 |
locking.LEVEL_NODE: lock_names, |
|
111 |
} |
|
112 |
|
|
113 |
self.share_locks[locking.LEVEL_NODE_ALLOC] = 1 |
|
114 |
|
|
115 |
if not self.op.node_names: |
|
116 |
# Acquire node allocation lock only if all nodes are affected |
|
117 |
self.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET |
|
118 |
|
|
119 |
def CheckPrereq(self): |
|
120 |
"""Check prerequisites. |
|
121 |
|
|
122 |
This checks: |
|
123 |
- the node exists in the configuration |
|
124 |
- OOB is supported |
|
125 |
|
|
126 |
Any errors are signaled by raising errors.OpPrereqError. |
|
127 |
|
|
128 |
""" |
|
129 |
self.nodes = [] |
|
130 |
self.master_node = self.cfg.GetMasterNode() |
|
131 |
|
|
132 |
assert self.op.power_delay >= 0.0 |
|
133 |
|
|
134 |
if self.op.node_names: |
|
135 |
if (self.op.command in self._SKIP_MASTER and |
|
136 |
self.master_node in self.op.node_names): |
|
137 |
master_node_obj = self.cfg.GetNodeInfo(self.master_node) |
|
138 |
master_oob_handler = _SupportsOob(self.cfg, master_node_obj) |
|
139 |
|
|
140 |
if master_oob_handler: |
|
141 |
additional_text = ("run '%s %s %s' if you want to operate on the" |
|
142 |
" master regardless") % (master_oob_handler, |
|
143 |
self.op.command, |
|
144 |
self.master_node) |
|
145 |
else: |
|
146 |
additional_text = "it does not support out-of-band operations" |
|
147 |
|
|
148 |
raise errors.OpPrereqError(("Operating on the master node %s is not" |
|
149 |
" allowed for %s; %s") % |
|
150 |
(self.master_node, self.op.command, |
|
151 |
additional_text), errors.ECODE_INVAL) |
|
152 |
else: |
|
153 |
self.op.node_names = self.cfg.GetNodeList() |
|
154 |
if self.op.command in self._SKIP_MASTER: |
|
155 |
self.op.node_names.remove(self.master_node) |
|
156 |
|
|
157 |
if self.op.command in self._SKIP_MASTER: |
|
158 |
assert self.master_node not in self.op.node_names |
|
159 |
|
|
160 |
for (node_name, node) in self.cfg.GetMultiNodeInfo(self.op.node_names): |
|
161 |
if node is None: |
|
162 |
raise errors.OpPrereqError("Node %s not found" % node_name, |
|
163 |
errors.ECODE_NOENT) |
|
164 |
else: |
|
165 |
self.nodes.append(node) |
|
166 |
|
|
167 |
if (not self.op.ignore_status and |
|
168 |
(self.op.command == constants.OOB_POWER_OFF and not node.offline)): |
|
169 |
raise errors.OpPrereqError(("Cannot power off node %s because it is" |
|
170 |
" not marked offline") % node_name, |
|
171 |
errors.ECODE_STATE) |
|
172 |
|
|
173 |
def Exec(self, feedback_fn): |
|
174 |
"""Execute OOB and return result if we expect any. |
|
175 |
|
|
176 |
""" |
|
177 |
master_node = self.master_node |
|
178 |
ret = [] |
|
179 |
|
|
180 |
for idx, node in enumerate(utils.NiceSort(self.nodes, |
|
181 |
key=lambda node: node.name)): |
|
182 |
node_entry = [(constants.RS_NORMAL, node.name)] |
|
183 |
ret.append(node_entry) |
|
184 |
|
|
185 |
oob_program = _SupportsOob(self.cfg, node) |
|
186 |
|
|
187 |
if not oob_program: |
|
188 |
node_entry.append((constants.RS_UNAVAIL, None)) |
|
189 |
continue |
|
190 |
|
|
191 |
logging.info("Executing out-of-band command '%s' using '%s' on %s", |
|
192 |
self.op.command, oob_program, node.name) |
|
193 |
result = self.rpc.call_run_oob(master_node, oob_program, |
|
194 |
self.op.command, node.name, |
|
195 |
self.op.timeout) |
|
196 |
|
|
197 |
if result.fail_msg: |
|
198 |
self.LogWarning("Out-of-band RPC failed on node '%s': %s", |
|
199 |
node.name, result.fail_msg) |
|
200 |
node_entry.append((constants.RS_NODATA, None)) |
|
201 |
else: |
|
202 |
try: |
|
203 |
self._CheckPayload(result) |
|
204 |
except errors.OpExecError, err: |
|
205 |
self.LogWarning("Payload returned by node '%s' is not valid: %s", |
|
206 |
node.name, err) |
|
207 |
node_entry.append((constants.RS_NODATA, None)) |
|
208 |
else: |
|
209 |
if self.op.command == constants.OOB_HEALTH: |
|
210 |
# For health we should log important events |
|
211 |
for item, status in result.payload: |
|
212 |
if status in [constants.OOB_STATUS_WARNING, |
|
213 |
constants.OOB_STATUS_CRITICAL]: |
|
214 |
self.LogWarning("Item '%s' on node '%s' has status '%s'", |
|
215 |
item, node.name, status) |
|
216 |
|
|
217 |
if self.op.command == constants.OOB_POWER_ON: |
|
218 |
node.powered = True |
|
219 |
elif self.op.command == constants.OOB_POWER_OFF: |
|
220 |
node.powered = False |
|
221 |
elif self.op.command == constants.OOB_POWER_STATUS: |
|
222 |
powered = result.payload[constants.OOB_POWER_STATUS_POWERED] |
|
223 |
if powered != node.powered: |
|
224 |
logging.warning(("Recorded power state (%s) of node '%s' does not" |
|
225 |
" match actual power state (%s)"), node.powered, |
|
226 |
node.name, powered) |
|
227 |
|
|
228 |
# For configuration changing commands we should update the node |
|
229 |
if self.op.command in (constants.OOB_POWER_ON, |
|
230 |
constants.OOB_POWER_OFF): |
|
231 |
self.cfg.Update(node, feedback_fn) |
|
232 |
|
|
233 |
node_entry.append((constants.RS_NORMAL, result.payload)) |
|
234 |
|
|
235 |
if (self.op.command == constants.OOB_POWER_ON and |
|
236 |
idx < len(self.nodes) - 1): |
|
237 |
time.sleep(self.op.power_delay) |
|
238 |
|
|
239 |
return ret |
|
240 |
|
|
241 |
def _CheckPayload(self, result): |
|
242 |
"""Checks if the payload is valid. |
|
243 |
|
|
244 |
@param result: RPC result |
|
245 |
@raises errors.OpExecError: If payload is not valid |
|
246 |
|
|
247 |
""" |
|
248 |
errs = [] |
|
249 |
if self.op.command == constants.OOB_HEALTH: |
|
250 |
if not isinstance(result.payload, list): |
|
251 |
errs.append("command 'health' is expected to return a list but got %s" % |
|
252 |
type(result.payload)) |
|
253 |
else: |
|
254 |
for item, status in result.payload: |
|
255 |
if status not in constants.OOB_STATUSES: |
|
256 |
errs.append("health item '%s' has invalid status '%s'" % |
|
257 |
(item, status)) |
|
258 |
|
|
259 |
if self.op.command == constants.OOB_POWER_STATUS: |
|
260 |
if not isinstance(result.payload, dict): |
|
261 |
errs.append("power-status is expected to return a dict but got %s" % |
|
262 |
type(result.payload)) |
|
263 |
|
|
264 |
if self.op.command in [ |
|
265 |
constants.OOB_POWER_ON, |
|
266 |
constants.OOB_POWER_OFF, |
|
267 |
constants.OOB_POWER_CYCLE, |
|
268 |
]: |
|
269 |
if result.payload is not None: |
|
270 |
errs.append("%s is expected to not return payload but got '%s'" % |
|
271 |
(self.op.command, result.payload)) |
|
272 |
|
|
273 |
if errs: |
|
274 |
raise errors.OpExecError("Check of out-of-band payload failed due to %s" % |
|
275 |
utils.CommaJoin(errs)) |
|
276 |
|
|
277 |
|
|
278 |
class _ExtStorageQuery(_QueryBase): |
|
279 |
FIELDS = query.EXTSTORAGE_FIELDS |
|
280 |
|
|
281 |
def ExpandNames(self, lu): |
|
282 |
# Lock all nodes in shared mode |
|
283 |
# Temporary removal of locks, should be reverted later |
|
284 |
# TODO: reintroduce locks when they are lighter-weight |
|
285 |
lu.needed_locks = {} |
|
286 |
#self.share_locks[locking.LEVEL_NODE] = 1 |
|
287 |
#self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET |
|
288 |
|
|
289 |
# The following variables interact with _QueryBase._GetNames |
|
290 |
if self.names: |
|
291 |
self.wanted = self.names |
|
292 |
else: |
|
293 |
self.wanted = locking.ALL_SET |
|
294 |
|
|
295 |
self.do_locking = self.use_locking |
|
296 |
|
|
297 |
def DeclareLocks(self, lu, level): |
|
298 |
pass |
|
299 |
|
|
300 |
@staticmethod |
|
301 |
def _DiagnoseByProvider(rlist): |
|
302 |
"""Remaps a per-node return list into an a per-provider per-node dictionary |
|
303 |
|
|
304 |
@param rlist: a map with node names as keys and ExtStorage objects as values |
|
305 |
|
|
306 |
@rtype: dict |
|
307 |
@return: a dictionary with extstorage providers as keys and as |
|
308 |
value another map, with nodes as keys and tuples of |
|
309 |
(path, status, diagnose, parameters) as values, eg:: |
|
310 |
|
|
311 |
{"provider1": {"node1": [(/usr/lib/..., True, "", [])] |
|
312 |
"node2": [(/srv/..., False, "missing file")] |
|
313 |
"node3": [(/srv/..., True, "", [])] |
|
314 |
} |
|
315 |
|
|
316 |
""" |
|
317 |
all_es = {} |
|
318 |
# we build here the list of nodes that didn't fail the RPC (at RPC |
|
319 |
# level), so that nodes with a non-responding node daemon don't |
|
320 |
# make all OSes invalid |
|
321 |
good_nodes = [node_name for node_name in rlist |
|
322 |
if not rlist[node_name].fail_msg] |
|
323 |
for node_name, nr in rlist.items(): |
|
324 |
if nr.fail_msg or not nr.payload: |
|
325 |
continue |
|
326 |
for (name, path, status, diagnose, params) in nr.payload: |
|
327 |
if name not in all_es: |
|
328 |
# build a list of nodes for this os containing empty lists |
|
329 |
# for each node in node_list |
|
330 |
all_es[name] = {} |
|
331 |
for nname in good_nodes: |
|
332 |
all_es[name][nname] = [] |
|
333 |
# convert params from [name, help] to (name, help) |
|
334 |
params = [tuple(v) for v in params] |
|
335 |
all_es[name][node_name].append((path, status, diagnose, params)) |
|
336 |
return all_es |
|
337 |
|
|
338 |
def _GetQueryData(self, lu): |
|
339 |
"""Computes the list of nodes and their attributes. |
|
340 |
|
|
341 |
""" |
|
342 |
# Locking is not used |
|
343 |
assert not (compat.any(lu.glm.is_owned(level) |
|
344 |
for level in locking.LEVELS |
|
345 |
if level != locking.LEVEL_CLUSTER) or |
|
346 |
self.do_locking or self.use_locking) |
|
347 |
|
|
348 |
valid_nodes = [node.name |
|
349 |
for node in lu.cfg.GetAllNodesInfo().values() |
|
350 |
if not node.offline and node.vm_capable] |
|
351 |
pol = self._DiagnoseByProvider(lu.rpc.call_extstorage_diagnose(valid_nodes)) |
|
352 |
|
|
353 |
data = {} |
|
354 |
|
|
355 |
nodegroup_list = lu.cfg.GetNodeGroupList() |
|
356 |
|
|
357 |
for (es_name, es_data) in pol.items(): |
|
358 |
# For every provider compute the nodegroup validity. |
|
359 |
# To do this we need to check the validity of each node in es_data |
|
360 |
# and then construct the corresponding nodegroup dict: |
|
361 |
# { nodegroup1: status |
|
362 |
# nodegroup2: status |
|
363 |
# } |
|
364 |
ndgrp_data = {} |
|
365 |
for nodegroup in nodegroup_list: |
|
366 |
ndgrp = lu.cfg.GetNodeGroup(nodegroup) |
|
367 |
|
|
368 |
nodegroup_nodes = ndgrp.members |
|
369 |
nodegroup_name = ndgrp.name |
|
370 |
node_statuses = [] |
|
371 |
|
|
372 |
for node in nodegroup_nodes: |
|
373 |
if node in valid_nodes: |
|
374 |
if es_data[node] != []: |
|
375 |
node_status = es_data[node][0][1] |
|
376 |
node_statuses.append(node_status) |
|
377 |
else: |
|
378 |
node_statuses.append(False) |
|
379 |
|
|
380 |
if False in node_statuses: |
|
381 |
ndgrp_data[nodegroup_name] = False |
|
382 |
else: |
|
383 |
ndgrp_data[nodegroup_name] = True |
|
384 |
|
|
385 |
# Compute the provider's parameters |
|
386 |
parameters = set() |
|
387 |
for idx, esl in enumerate(es_data.values()): |
|
388 |
valid = bool(esl and esl[0][1]) |
|
389 |
if not valid: |
|
390 |
break |
|
391 |
|
|
392 |
node_params = esl[0][3] |
|
393 |
if idx == 0: |
|
394 |
# First entry |
|
395 |
parameters.update(node_params) |
|
396 |
else: |
|
397 |
# Filter out inconsistent values |
|
398 |
parameters.intersection_update(node_params) |
|
399 |
|
|
400 |
params = list(parameters) |
|
401 |
|
|
402 |
# Now fill all the info for this provider |
|
403 |
info = query.ExtStorageInfo(name=es_name, node_status=es_data, |
|
404 |
nodegroup_status=ndgrp_data, |
|
405 |
parameters=params) |
|
406 |
|
|
407 |
data[es_name] = info |
|
408 |
|
|
409 |
# Prepare data in requested order |
|
410 |
return [data[name] for name in self._GetNames(lu, pol.keys(), None) |
|
411 |
if name in data] |
|
412 |
|
|
413 |
|
|
414 |
class LUExtStorageDiagnose(NoHooksLU): |
|
415 |
"""Logical unit for ExtStorage diagnose/query. |
|
416 |
|
|
417 |
""" |
|
418 |
REQ_BGL = False |
|
419 |
|
|
420 |
def CheckArguments(self): |
|
421 |
self.eq = _ExtStorageQuery(qlang.MakeSimpleFilter("name", self.op.names), |
|
422 |
self.op.output_fields, False) |
|
423 |
|
|
424 |
def ExpandNames(self): |
|
425 |
self.eq.ExpandNames(self) |
|
426 |
|
|
427 |
def Exec(self, feedback_fn): |
|
428 |
return self.eq.OldStyleQuery(self) |
|
429 |
|
|
430 |
|
|
431 |
class LURestrictedCommand(NoHooksLU): |
|
432 |
"""Logical unit for executing restricted commands. |
|
433 |
|
|
434 |
""" |
|
435 |
REQ_BGL = False |
|
436 |
|
|
437 |
def ExpandNames(self): |
|
438 |
if self.op.nodes: |
|
439 |
self.op.nodes = _GetWantedNodes(self, self.op.nodes) |
|
440 |
|
|
441 |
self.needed_locks = { |
|
442 |
locking.LEVEL_NODE: self.op.nodes, |
|
443 |
} |
|
444 |
self.share_locks = { |
|
445 |
locking.LEVEL_NODE: not self.op.use_locking, |
|
446 |
} |
|
447 |
|
|
448 |
def CheckPrereq(self): |
|
449 |
"""Check prerequisites. |
|
450 |
|
|
451 |
""" |
|
452 |
|
|
453 |
def Exec(self, feedback_fn): |
|
454 |
"""Execute restricted command and return output. |
|
455 |
|
|
456 |
""" |
|
457 |
owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE)) |
|
458 |
|
|
459 |
# Check if correct locks are held |
|
460 |
assert set(self.op.nodes).issubset(owned_nodes) |
|
461 |
|
|
462 |
rpcres = self.rpc.call_restricted_command(self.op.nodes, self.op.command) |
|
463 |
|
|
464 |
result = [] |
|
465 |
|
|
466 |
for node_name in self.op.nodes: |
|
467 |
nres = rpcres[node_name] |
|
468 |
if nres.fail_msg: |
|
469 |
msg = ("Command '%s' on node '%s' failed: %s" % |
|
470 |
(self.op.command, node_name, nres.fail_msg)) |
|
471 |
result.append((False, msg)) |
|
472 |
else: |
|
473 |
result.append((True, nres.payload)) |
|
474 |
|
|
475 |
return result |
b/lib/cmdlib/misc.py | ||
---|---|---|
1 |
# |
|
2 |
# |
|
3 |
|
|
4 |
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 Google Inc. |
|
5 |
# |
|
6 |
# This program is free software; you can redistribute it and/or modify |
|
7 |
# it under the terms of the GNU General Public License as published by |
|
8 |
# the Free Software Foundation; either version 2 of the License, or |
|
9 |
# (at your option) any later version. |
|
10 |
# |
|
11 |
# This program is distributed in the hope that it will be useful, but |
|
12 |
# WITHOUT ANY WARRANTY; without even the implied warranty of |
|
13 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
14 |
# General Public License for more details. |
|
15 |
# |
|
16 |
# You should have received a copy of the GNU General Public License |
|
17 |
# along with this program; if not, write to the Free Software |
|
18 |
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA |
|
19 |
# 02110-1301, USA. |
|
20 |
|
|
21 |
|
|
22 |
"""Miscellaneous logical units that don't fit into any category.""" |
|
23 |
|
|
24 |
import logging |
|
25 |
import time |
|
26 |
|
|
27 |
from ganeti import compat |
|
28 |
from ganeti import constants |
|
29 |
from ganeti import errors |
|
30 |
from ganeti import locking |
|
31 |
from ganeti import qlang |
|
32 |
from ganeti import query |
|
33 |
from ganeti import utils |
|
34 |
from ganeti.cmdlib.base import NoHooksLU, _QueryBase |
|
35 |
from ganeti.cmdlib.common import _GetWantedNodes, _SupportsOob |
|
36 |
|
|
37 |
|
|
38 |
class LUOobCommand(NoHooksLU): |
|
39 |
"""Logical unit for OOB handling. |
|
40 |
|
|
41 |
""" |
|
42 |
REQ_BGL = False |
|
43 |
_SKIP_MASTER = (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE) |
|
44 |
|
|
45 |
def ExpandNames(self): |
|
46 |
"""Gather locks we need. |
|
47 |
|
|
48 |
""" |
|
49 |
if self.op.node_names: |
|
50 |
self.op.node_names = _GetWantedNodes(self, self.op.node_names) |
|
51 |
lock_names = self.op.node_names |
|
52 |
else: |
|
53 |
lock_names = locking.ALL_SET |
|
54 |
|
|
55 |
self.needed_locks = { |
|
56 |
locking.LEVEL_NODE: lock_names, |
|
57 |
} |
|
58 |
|
|
59 |
self.share_locks[locking.LEVEL_NODE_ALLOC] = 1 |
|
60 |
|
|
61 |
if not self.op.node_names: |
|
62 |
# Acquire node allocation lock only if all nodes are affected |
|
63 |
self.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET |
|
64 |
|
|
65 |
def CheckPrereq(self): |
|
66 |
"""Check prerequisites. |
|
67 |
|
|
68 |
This checks: |
|
69 |
- the node exists in the configuration |
|
70 |
- OOB is supported |
|
71 |
|
|
72 |
Any errors are signaled by raising errors.OpPrereqError. |
|
73 |
|
|
74 |
""" |
|
75 |
self.nodes = [] |
|
76 |
self.master_node = self.cfg.GetMasterNode() |
|
77 |
|
|
78 |
assert self.op.power_delay >= 0.0 |
|
79 |
|
|
80 |
if self.op.node_names: |
|
81 |
if (self.op.command in self._SKIP_MASTER and |
|
82 |
self.master_node in self.op.node_names): |
|
83 |
master_node_obj = self.cfg.GetNodeInfo(self.master_node) |
|
84 |
master_oob_handler = _SupportsOob(self.cfg, master_node_obj) |
|
85 |
|
|
86 |
if master_oob_handler: |
|
87 |
additional_text = ("run '%s %s %s' if you want to operate on the" |
|
88 |
" master regardless") % (master_oob_handler, |
|
89 |
self.op.command, |
|
90 |
self.master_node) |
|
91 |
else: |
|
92 |
additional_text = "it does not support out-of-band operations" |
|
93 |
|
|
94 |
raise errors.OpPrereqError(("Operating on the master node %s is not" |
|
95 |
" allowed for %s; %s") % |
|
96 |
(self.master_node, self.op.command, |
|
97 |
additional_text), errors.ECODE_INVAL) |
|
98 |
else: |
|
99 |
self.op.node_names = self.cfg.GetNodeList() |
|
100 |
if self.op.command in self._SKIP_MASTER: |
|
101 |
self.op.node_names.remove(self.master_node) |
|
102 |
|
|
103 |
if self.op.command in self._SKIP_MASTER: |
|
104 |
assert self.master_node not in self.op.node_names |
|
105 |
|
|
106 |
for (node_name, node) in self.cfg.GetMultiNodeInfo(self.op.node_names): |
|
107 |
if node is None: |
|
108 |
raise errors.OpPrereqError("Node %s not found" % node_name, |
|
109 |
errors.ECODE_NOENT) |
|
110 |
else: |
|
111 |
self.nodes.append(node) |
|
112 |
|
|
113 |
if (not self.op.ignore_status and |
|
114 |
(self.op.command == constants.OOB_POWER_OFF and not node.offline)): |
|
115 |
raise errors.OpPrereqError(("Cannot power off node %s because it is" |
|
116 |
" not marked offline") % node_name, |
|
117 |
errors.ECODE_STATE) |
|
118 |
|
|
119 |
def Exec(self, feedback_fn): |
|
120 |
"""Execute OOB and return result if we expect any. |
|
121 |
|
|
122 |
""" |
|
123 |
master_node = self.master_node |
|
124 |
ret = [] |
|
125 |
|
|
126 |
for idx, node in enumerate(utils.NiceSort(self.nodes, |
|
127 |
key=lambda node: node.name)): |
|
128 |
node_entry = [(constants.RS_NORMAL, node.name)] |
|
129 |
ret.append(node_entry) |
|
130 |
|
|
131 |
oob_program = _SupportsOob(self.cfg, node) |
|
132 |
|
|
133 |
if not oob_program: |
|
134 |
node_entry.append((constants.RS_UNAVAIL, None)) |
|
135 |
continue |
|
136 |
|
|
137 |
logging.info("Executing out-of-band command '%s' using '%s' on %s", |
|
138 |
self.op.command, oob_program, node.name) |
|
139 |
result = self.rpc.call_run_oob(master_node, oob_program, |
|
140 |
self.op.command, node.name, |
|
141 |
self.op.timeout) |
|
142 |
|
|
143 |
if result.fail_msg: |
|
144 |
self.LogWarning("Out-of-band RPC failed on node '%s': %s", |
|
145 |
node.name, result.fail_msg) |
|
146 |
node_entry.append((constants.RS_NODATA, None)) |
|
147 |
else: |
|
148 |
try: |
|
149 |
self._CheckPayload(result) |
|
150 |
except errors.OpExecError, err: |
|
151 |
self.LogWarning("Payload returned by node '%s' is not valid: %s", |
|
152 |
node.name, err) |
|
153 |
node_entry.append((constants.RS_NODATA, None)) |
|
154 |
else: |
|
155 |
if self.op.command == constants.OOB_HEALTH: |
|
156 |
# For health we should log important events |
|
157 |
for item, status in result.payload: |
|
158 |
if status in [constants.OOB_STATUS_WARNING, |
|
159 |
constants.OOB_STATUS_CRITICAL]: |
|
160 |
self.LogWarning("Item '%s' on node '%s' has status '%s'", |
|
161 |
item, node.name, status) |
|
162 |
|
|
163 |
if self.op.command == constants.OOB_POWER_ON: |
|
164 |
node.powered = True |
|
165 |
elif self.op.command == constants.OOB_POWER_OFF: |
|
166 |
node.powered = False |
|
167 |
elif self.op.command == constants.OOB_POWER_STATUS: |
|
168 |
powered = result.payload[constants.OOB_POWER_STATUS_POWERED] |
|
169 |
if powered != node.powered: |
|
170 |
logging.warning(("Recorded power state (%s) of node '%s' does not" |
|
171 |
" match actual power state (%s)"), node.powered, |
|
172 |
node.name, powered) |
|
173 |
|
|
174 |
# For configuration changing commands we should update the node |
|
175 |
if self.op.command in (constants.OOB_POWER_ON, |
|
176 |
constants.OOB_POWER_OFF): |
|
177 |
self.cfg.Update(node, feedback_fn) |
|
178 |
|
|
179 |
node_entry.append((constants.RS_NORMAL, result.payload)) |
|
180 |
|
|
181 |
if (self.op.command == constants.OOB_POWER_ON and |
|
182 |
idx < len(self.nodes) - 1): |
|
183 |
time.sleep(self.op.power_delay) |
|
184 |
|
|
185 |
return ret |
|
186 |
|
|
187 |
def _CheckPayload(self, result): |
|
188 |
"""Checks if the payload is valid. |
|
189 |
|
|
190 |
@param result: RPC result |
|
191 |
@raises errors.OpExecError: If payload is not valid |
|
192 |
|
|
193 |
""" |
|
194 |
errs = [] |
|
195 |
if self.op.command == constants.OOB_HEALTH: |
|
196 |
if not isinstance(result.payload, list): |
|
197 |
errs.append("command 'health' is expected to return a list but got %s" % |
|
198 |
type(result.payload)) |
|
199 |
else: |
|
200 |
for item, status in result.payload: |
|
201 |
if status not in constants.OOB_STATUSES: |
|
202 |
errs.append("health item '%s' has invalid status '%s'" % |
|
203 |
(item, status)) |
|
204 |
|
|
205 |
if self.op.command == constants.OOB_POWER_STATUS: |
|
206 |
if not isinstance(result.payload, dict): |
|
207 |
errs.append("power-status is expected to return a dict but got %s" % |
|
208 |
type(result.payload)) |
|
209 |
|
|
210 |
if self.op.command in [ |
|
211 |
constants.OOB_POWER_ON, |
|
212 |
constants.OOB_POWER_OFF, |
|
213 |
constants.OOB_POWER_CYCLE, |
|
214 |
]: |
|
215 |
if result.payload is not None: |
|
216 |
errs.append("%s is expected to not return payload but got '%s'" % |
|
217 |
(self.op.command, result.payload)) |
|
218 |
|
|
219 |
if errs: |
|
220 |
raise errors.OpExecError("Check of out-of-band payload failed due to %s" % |
|
221 |
utils.CommaJoin(errs)) |
|
222 |
|
|
223 |
|
|
224 |
class _ExtStorageQuery(_QueryBase): |
|
225 |
FIELDS = query.EXTSTORAGE_FIELDS |
|
226 |
|
|
227 |
def ExpandNames(self, lu): |
|
228 |
# Lock all nodes in shared mode |
|
229 |
# Temporary removal of locks, should be reverted later |
|
230 |
# TODO: reintroduce locks when they are lighter-weight |
|
231 |
lu.needed_locks = {} |
|
232 |
#self.share_locks[locking.LEVEL_NODE] = 1 |
|
233 |
#self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET |
|
234 |
|
|
235 |
# The following variables interact with _QueryBase._GetNames |
|
236 |
if self.names: |
|
237 |
self.wanted = self.names |
|
238 |
else: |
|
239 |
self.wanted = locking.ALL_SET |
|
240 |
|
|
241 |
self.do_locking = self.use_locking |
|
242 |
|
|
243 |
def DeclareLocks(self, lu, level): |
|
244 |
pass |
|
245 |
|
|
246 |
@staticmethod |
|
247 |
def _DiagnoseByProvider(rlist): |
|
248 |
"""Remaps a per-node return list into an a per-provider per-node dictionary |
|
249 |
|
|
250 |
@param rlist: a map with node names as keys and ExtStorage objects as values |
|
251 |
|
|
252 |
@rtype: dict |
|
253 |
@return: a dictionary with extstorage providers as keys and as |
|
254 |
value another map, with nodes as keys and tuples of |
|
255 |
(path, status, diagnose, parameters) as values, eg:: |
|
256 |
|
|
257 |
{"provider1": {"node1": [(/usr/lib/..., True, "", [])] |
|
258 |
"node2": [(/srv/..., False, "missing file")] |
|
259 |
"node3": [(/srv/..., True, "", [])] |
|
260 |
} |
|
261 |
|
|
262 |
""" |
|
263 |
all_es = {} |
|
264 |
# we build here the list of nodes that didn't fail the RPC (at RPC |
|
265 |
# level), so that nodes with a non-responding node daemon don't |
|
266 |
# make all OSes invalid |
|
267 |
good_nodes = [node_name for node_name in rlist |
|
268 |
if not rlist[node_name].fail_msg] |
|
269 |
for node_name, nr in rlist.items(): |
|
270 |
if nr.fail_msg or not nr.payload: |
|
271 |
continue |
|
272 |
for (name, path, status, diagnose, params) in nr.payload: |
|
273 |
if name not in all_es: |
|
274 |
# build a list of nodes for this os containing empty lists |
|
275 |
# for each node in node_list |
|
276 |
all_es[name] = {} |
|
277 |
for nname in good_nodes: |
|
278 |
all_es[name][nname] = [] |
|
279 |
# convert params from [name, help] to (name, help) |
|
280 |
params = [tuple(v) for v in params] |
|
281 |
all_es[name][node_name].append((path, status, diagnose, params)) |
|
282 |
return all_es |
|
283 |
|
|
284 |
def _GetQueryData(self, lu): |
|
285 |
"""Computes the list of nodes and their attributes. |
|
286 |
|
|
287 |
""" |
|
288 |
# Locking is not used |
|
289 |
assert not (compat.any(lu.glm.is_owned(level) |
|
290 |
for level in locking.LEVELS |
|
291 |
if level != locking.LEVEL_CLUSTER) or |
|
292 |
self.do_locking or self.use_locking) |
|
293 |
|
|
294 |
valid_nodes = [node.name |
|
295 |
for node in lu.cfg.GetAllNodesInfo().values() |
|
296 |
if not node.offline and node.vm_capable] |
|
297 |
pol = self._DiagnoseByProvider(lu.rpc.call_extstorage_diagnose(valid_nodes)) |
|
298 |
|
|
299 |
data = {} |
|
300 |
|
|
301 |
nodegroup_list = lu.cfg.GetNodeGroupList() |
|
302 |
|
|
303 |
for (es_name, es_data) in pol.items(): |
|
304 |
# For every provider compute the nodegroup validity. |
|
305 |
# To do this we need to check the validity of each node in es_data |
|
306 |
# and then construct the corresponding nodegroup dict: |
|
307 |
# { nodegroup1: status |
|
308 |
# nodegroup2: status |
|
309 |
# } |
|
310 |
ndgrp_data = {} |
|
311 |
for nodegroup in nodegroup_list: |
|
312 |
ndgrp = lu.cfg.GetNodeGroup(nodegroup) |
|
313 |
|
|
314 |
nodegroup_nodes = ndgrp.members |
|
315 |
nodegroup_name = ndgrp.name |
|
316 |
node_statuses = [] |
|
317 |
|
|
318 |
for node in nodegroup_nodes: |
|
319 |
if node in valid_nodes: |
|
320 |
if es_data[node] != []: |
|
321 |
node_status = es_data[node][0][1] |
|
322 |
node_statuses.append(node_status) |
|
323 |
else: |
|
324 |
node_statuses.append(False) |
|
325 |
|
|
326 |
if False in node_statuses: |
|
327 |
ndgrp_data[nodegroup_name] = False |
|
328 |
else: |
|
329 |
ndgrp_data[nodegroup_name] = True |
|
330 |
|
|
331 |
# Compute the provider's parameters |
|
332 |
parameters = set() |
|
333 |
for idx, esl in enumerate(es_data.values()): |
|
334 |
valid = bool(esl and esl[0][1]) |
|
335 |
if not valid: |
|
336 |
break |
|
337 |
|
|
338 |
node_params = esl[0][3] |
|
339 |
if idx == 0: |
|
340 |
# First entry |
|
341 |
parameters.update(node_params) |
|
342 |
else: |
|
343 |
# Filter out inconsistent values |
|
344 |
parameters.intersection_update(node_params) |
|
345 |
|
|
346 |
params = list(parameters) |
|
347 |
|
|
348 |
# Now fill all the info for this provider |
|
349 |
info = query.ExtStorageInfo(name=es_name, node_status=es_data, |
|
350 |
nodegroup_status=ndgrp_data, |
|
351 |
parameters=params) |
|
352 |
|
|
353 |
data[es_name] = info |
|
354 |
|
|
355 |
# Prepare data in requested order |
|
356 |
return [data[name] for name in self._GetNames(lu, pol.keys(), None) |
|
357 |
if name in data] |
|
358 |
|
|
359 |
|
|
360 |
class LUExtStorageDiagnose(NoHooksLU): |
|
361 |
"""Logical unit for ExtStorage diagnose/query. |
|
362 |
|
|
363 |
""" |
|
364 |
REQ_BGL = False |
|
365 |
|
|
366 |
def CheckArguments(self): |
|
367 |
self.eq = _ExtStorageQuery(qlang.MakeSimpleFilter("name", self.op.names), |
|
368 |
self.op.output_fields, False) |
|
369 |
|
|
370 |
def ExpandNames(self): |
|
371 |
self.eq.ExpandNames(self) |
|
372 |
|
|
373 |
def Exec(self, feedback_fn): |
|
374 |
return self.eq.OldStyleQuery(self) |
|
375 |
|
|
376 |
|
|
377 |
class LURestrictedCommand(NoHooksLU): |
|
378 |
"""Logical unit for executing restricted commands. |
|
379 |
|
|
380 |
""" |
|
381 |
REQ_BGL = False |
|
382 |
|
|
383 |
def ExpandNames(self): |
|
384 |
if self.op.nodes: |
|
385 |
self.op.nodes = _GetWantedNodes(self, self.op.nodes) |
|
386 |
|
|
387 |
self.needed_locks = { |
|
388 |
locking.LEVEL_NODE: self.op.nodes, |
|
389 |
} |
|
390 |
self.share_locks = { |
|
391 |
locking.LEVEL_NODE: not self.op.use_locking, |
|
392 |
} |
|
393 |
|
|
394 |
def CheckPrereq(self): |
|
395 |
"""Check prerequisites. |
|
396 |
|
|
397 |
""" |
|
398 |
|
|
399 |
def Exec(self, feedback_fn): |
|
400 |
"""Execute restricted command and return output. |
|
401 |
|
|
402 |
""" |
|
403 |
owned_nodes = frozenset(self.owned_locks(locking.LEVEL_NODE)) |
|
404 |
|
|
405 |
# Check if correct locks are held |
|
406 |
assert set(self.op.nodes).issubset(owned_nodes) |
|
407 |
|
|
408 |
rpcres = self.rpc.call_restricted_command(self.op.nodes, self.op.command) |
|
409 |
|
|
410 |
result = [] |
|
411 |
|
|
412 |
for node_name in self.op.nodes: |
|
413 |
nres = rpcres[node_name] |
|
414 |
if nres.fail_msg: |
|
415 |
msg = ("Command '%s' on node '%s' failed: %s" % |
|
416 |
(self.op.command, node_name, nres.fail_msg)) |
|
417 |
result.append((False, msg)) |
|
418 |
else: |
|
419 |
result.append((True, nres.payload)) |
|
420 |
|
|
421 |
return result |
b/lib/cmdlib/query.py | ||
---|---|---|
24 | 24 |
from ganeti import constants |
25 | 25 |
from ganeti import errors |
26 | 26 |
from ganeti import query |
27 |
from ganeti.cmdlib import _ExtStorageQuery |
|
28 | 27 |
from ganeti.cmdlib.backup import _ExportQuery |
29 | 28 |
from ganeti.cmdlib.base import NoHooksLU |
30 | 29 |
from ganeti.cmdlib.cluster import _ClusterQuery |
31 | 30 |
from ganeti.cmdlib.group import _GroupQuery |
32 | 31 |
from ganeti.cmdlib.instance import _InstanceQuery |
32 |
from ganeti.cmdlib.misc import _ExtStorageQuery |
|
33 | 33 |
from ganeti.cmdlib.network import _NetworkQuery |
34 | 34 |
from ganeti.cmdlib.node import _NodeQuery |
35 | 35 |
from ganeti.cmdlib.operating_system import _OsQuery |
Also available in: Unified diff