root / lib / cmdlib / misc.py @ 5eacbcae
History | View | Annotate | Download (13.8 kB)
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
|