root / lib / hypervisor / hv_base.py @ 874f6148
History | View | Annotate | Download (17.6 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 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 |
"""Base class for all hypervisors
|
23 |
|
24 |
The syntax for the _CHECK variables and the contents of the PARAMETERS
|
25 |
dict is the same, see the docstring for L{BaseHypervisor.PARAMETERS}.
|
26 |
|
27 |
@var _FILE_CHECK: stub for file checks, without the required flag
|
28 |
@var _DIR_CHECK: stub for directory checks, without the required flag
|
29 |
@var REQ_FILE_CHECK: mandatory file parameter
|
30 |
@var OPT_FILE_CHECK: optional file parameter
|
31 |
@var REQ_DIR_CHECK: mandatory directory parametr
|
32 |
@var OPT_DIR_CHECK: optional directory parameter
|
33 |
@var NO_CHECK: parameter without any checks at all
|
34 |
@var REQUIRED_CHECK: parameter required to exist (and non-false), but
|
35 |
without other checks; beware that this can't be used for boolean
|
36 |
parameters, where you should use NO_CHECK or a custom checker
|
37 |
|
38 |
"""
|
39 |
|
40 |
import os |
41 |
import re |
42 |
import logging |
43 |
|
44 |
|
45 |
from ganeti import errors |
46 |
from ganeti import utils |
47 |
from ganeti import constants |
48 |
|
49 |
|
50 |
def _IsCpuMaskWellFormed(cpu_mask): |
51 |
"""Verifies if the given single CPU mask is valid
|
52 |
|
53 |
The single CPU mask should be in the form "a,b,c,d", where each
|
54 |
letter is a positive number or range.
|
55 |
|
56 |
"""
|
57 |
try:
|
58 |
cpu_list = utils.ParseCpuMask(cpu_mask) |
59 |
except errors.ParseError, _:
|
60 |
return False |
61 |
return isinstance(cpu_list, list) and len(cpu_list) > 0 |
62 |
|
63 |
|
64 |
def _IsMultiCpuMaskWellFormed(cpu_mask): |
65 |
"""Verifies if the given multiple CPU mask is valid
|
66 |
|
67 |
A valid multiple CPU mask is in the form "a:b:c:d", where each
|
68 |
letter is a single CPU mask.
|
69 |
|
70 |
"""
|
71 |
try:
|
72 |
utils.ParseMultiCpuMask(cpu_mask) |
73 |
except errors.ParseError, _:
|
74 |
return False |
75 |
|
76 |
return True |
77 |
|
78 |
|
79 |
# Read the BaseHypervisor.PARAMETERS docstring for the syntax of the
|
80 |
# _CHECK values
|
81 |
|
82 |
# must be afile
|
83 |
_FILE_CHECK = (utils.IsNormAbsPath, "must be an absolute normalized path",
|
84 |
os.path.isfile, "not found or not a file")
|
85 |
|
86 |
# must be a directory
|
87 |
_DIR_CHECK = (utils.IsNormAbsPath, "must be an absolute normalized path",
|
88 |
os.path.isdir, "not found or not a directory")
|
89 |
|
90 |
# CPU mask must be well-formed
|
91 |
# TODO: implement node level check for the CPU mask
|
92 |
_CPU_MASK_CHECK = (_IsCpuMaskWellFormed, |
93 |
"CPU mask definition is not well-formed",
|
94 |
None, None) |
95 |
|
96 |
# Multiple CPU mask must be well-formed
|
97 |
_MULTI_CPU_MASK_CHECK = (_IsMultiCpuMaskWellFormed, |
98 |
"Multiple CPU mask definition is not well-formed",
|
99 |
None, None) |
100 |
|
101 |
# Check for validity of port number
|
102 |
_NET_PORT_CHECK = (lambda x: 0 < x < 65535, "invalid port number", |
103 |
None, None) |
104 |
|
105 |
# Check that an integer is non negative
|
106 |
_NONNEGATIVE_INT_CHECK = (lambda x: x >= 0, "cannot be negative", None, None) |
107 |
|
108 |
# nice wrappers for users
|
109 |
REQ_FILE_CHECK = (True, ) + _FILE_CHECK
|
110 |
OPT_FILE_CHECK = (False, ) + _FILE_CHECK
|
111 |
REQ_DIR_CHECK = (True, ) + _DIR_CHECK
|
112 |
OPT_DIR_CHECK = (False, ) + _DIR_CHECK
|
113 |
REQ_NET_PORT_CHECK = (True, ) + _NET_PORT_CHECK
|
114 |
OPT_NET_PORT_CHECK = (False, ) + _NET_PORT_CHECK
|
115 |
REQ_CPU_MASK_CHECK = (True, ) + _CPU_MASK_CHECK
|
116 |
OPT_CPU_MASK_CHECK = (False, ) + _CPU_MASK_CHECK
|
117 |
REQ_MULTI_CPU_MASK_CHECK = (True, ) + _MULTI_CPU_MASK_CHECK
|
118 |
OPT_MULTI_CPU_MASK_CHECK = (False, ) + _MULTI_CPU_MASK_CHECK
|
119 |
REQ_NONNEGATIVE_INT_CHECK = (True, ) + _NONNEGATIVE_INT_CHECK
|
120 |
OPT_NONNEGATIVE_INT_CHECK = (False, ) + _NONNEGATIVE_INT_CHECK
|
121 |
|
122 |
# no checks at all
|
123 |
NO_CHECK = (False, None, None, None, None) |
124 |
|
125 |
# required, but no other checks
|
126 |
REQUIRED_CHECK = (True, None, None, None, None) |
127 |
|
128 |
# migration type
|
129 |
MIGRATION_MODE_CHECK = (True, lambda x: x in constants.HT_MIGRATION_MODES, |
130 |
"invalid migration mode", None, None) |
131 |
|
132 |
|
133 |
def ParamInSet(required, my_set): |
134 |
"""Builds parameter checker for set membership.
|
135 |
|
136 |
@type required: boolean
|
137 |
@param required: whether this is a required parameter
|
138 |
@type my_set: tuple, list or set
|
139 |
@param my_set: allowed values set
|
140 |
|
141 |
"""
|
142 |
fn = lambda x: x in my_set |
143 |
err = ("The value must be one of: %s" % utils.CommaJoin(my_set))
|
144 |
return (required, fn, err, None, None) |
145 |
|
146 |
|
147 |
class BaseHypervisor(object): |
148 |
"""Abstract virtualisation technology interface
|
149 |
|
150 |
The goal is that all aspects of the virtualisation technology are
|
151 |
abstracted away from the rest of code.
|
152 |
|
153 |
@cvar PARAMETERS: a dict of parameter name: check type; the check type is
|
154 |
a five-tuple containing:
|
155 |
- the required flag (boolean)
|
156 |
- a function to check for syntax, that will be used in
|
157 |
L{CheckParameterSyntax}, in the master daemon process
|
158 |
- an error message for the above function
|
159 |
- a function to check for parameter validity on the remote node,
|
160 |
in the L{ValidateParameters} function
|
161 |
- an error message for the above function
|
162 |
@type CAN_MIGRATE: boolean
|
163 |
@cvar CAN_MIGRATE: whether this hypervisor can do migration (either
|
164 |
live or non-live)
|
165 |
|
166 |
"""
|
167 |
PARAMETERS = {} |
168 |
ANCILLARY_FILES = [] |
169 |
ANCILLARY_FILES_OPT = [] |
170 |
CAN_MIGRATE = False
|
171 |
|
172 |
def StartInstance(self, instance, block_devices, startup_paused): |
173 |
"""Start an instance."""
|
174 |
raise NotImplementedError |
175 |
|
176 |
def StopInstance(self, instance, force=False, retry=False, name=None, |
177 |
timeout=None):
|
178 |
"""Stop an instance
|
179 |
|
180 |
@type instance: L{objects.Instance}
|
181 |
@param instance: instance to stop
|
182 |
@type force: boolean
|
183 |
@param force: whether to do a "hard" stop (destroy)
|
184 |
@type retry: boolean
|
185 |
@param retry: whether this is just a retry call
|
186 |
@type name: string or None
|
187 |
@param name: if this parameter is passed, the the instance object
|
188 |
should not be used (will be passed as None), and the shutdown
|
189 |
must be done by name only
|
190 |
@type timeout: int or None
|
191 |
@param timeout: if the parameter is not None, a soft shutdown operation will
|
192 |
be killed after the specified number of seconds. A hard (forced)
|
193 |
shutdown cannot have a timeout
|
194 |
|
195 |
"""
|
196 |
raise NotImplementedError |
197 |
|
198 |
def CleanupInstance(self, instance_name): |
199 |
"""Cleanup after a stopped instance
|
200 |
|
201 |
This is an optional method, used by hypervisors that need to cleanup after
|
202 |
an instance has been stopped.
|
203 |
|
204 |
@type instance_name: string
|
205 |
@param instance_name: instance name to cleanup after
|
206 |
|
207 |
"""
|
208 |
pass
|
209 |
|
210 |
def RebootInstance(self, instance): |
211 |
"""Reboot an instance."""
|
212 |
raise NotImplementedError |
213 |
|
214 |
def ListInstances(self): |
215 |
"""Get the list of running instances."""
|
216 |
raise NotImplementedError |
217 |
|
218 |
def GetInstanceInfo(self, instance_name): |
219 |
"""Get instance properties.
|
220 |
|
221 |
@type instance_name: string
|
222 |
@param instance_name: the instance name
|
223 |
|
224 |
@return: tuple (name, id, memory, vcpus, state, times)
|
225 |
|
226 |
"""
|
227 |
raise NotImplementedError |
228 |
|
229 |
def GetAllInstancesInfo(self): |
230 |
"""Get properties of all instances.
|
231 |
|
232 |
@return: list of tuples (name, id, memory, vcpus, stat, times)
|
233 |
|
234 |
"""
|
235 |
raise NotImplementedError |
236 |
|
237 |
def GetNodeInfo(self): |
238 |
"""Return information about the node.
|
239 |
|
240 |
@return: a dict with the following keys (values in MiB):
|
241 |
- memory_total: the total memory size on the node
|
242 |
- memory_free: the available memory on the node for instances
|
243 |
- memory_dom0: the memory used by the node itself, if available
|
244 |
|
245 |
"""
|
246 |
raise NotImplementedError |
247 |
|
248 |
@classmethod
|
249 |
def GetInstanceConsole(cls, instance, hvparams, beparams): |
250 |
"""Return information for connecting to the console of an instance.
|
251 |
|
252 |
"""
|
253 |
raise NotImplementedError |
254 |
|
255 |
@classmethod
|
256 |
def GetAncillaryFiles(cls): |
257 |
"""Return a list of ancillary files to be copied to all nodes as ancillary
|
258 |
configuration files.
|
259 |
|
260 |
@rtype: (list of absolute paths, list of absolute paths)
|
261 |
@return: (all files, optional files)
|
262 |
|
263 |
"""
|
264 |
# By default we return a member variable, so that if an hypervisor has just
|
265 |
# a static list of files it doesn't have to override this function.
|
266 |
assert set(cls.ANCILLARY_FILES).issuperset(cls.ANCILLARY_FILES_OPT), \ |
267 |
"Optional ancillary files must be a subset of ancillary files"
|
268 |
|
269 |
return (cls.ANCILLARY_FILES, cls.ANCILLARY_FILES_OPT)
|
270 |
|
271 |
def Verify(self): |
272 |
"""Verify the hypervisor.
|
273 |
|
274 |
@return: Problem description if something is wrong, C{None} otherwise
|
275 |
|
276 |
"""
|
277 |
raise NotImplementedError |
278 |
|
279 |
def MigrationInfo(self, instance): # pylint: disable=R0201,W0613 |
280 |
"""Get instance information to perform a migration.
|
281 |
|
282 |
By default assume no information is needed.
|
283 |
|
284 |
@type instance: L{objects.Instance}
|
285 |
@param instance: instance to be migrated
|
286 |
@rtype: string/data (opaque)
|
287 |
@return: instance migration information - serialized form
|
288 |
|
289 |
"""
|
290 |
return "" |
291 |
|
292 |
def AcceptInstance(self, instance, info, target): |
293 |
"""Prepare to accept an instance.
|
294 |
|
295 |
By default assume no preparation is needed.
|
296 |
|
297 |
@type instance: L{objects.Instance}
|
298 |
@param instance: instance to be accepted
|
299 |
@type info: string/data (opaque)
|
300 |
@param info: migration information, from the source node
|
301 |
@type target: string
|
302 |
@param target: target host (usually ip), on this node
|
303 |
|
304 |
"""
|
305 |
pass
|
306 |
|
307 |
def BalloonInstanceMemory(self, instance, mem): |
308 |
"""Balloon an instance memory to a certain value.
|
309 |
|
310 |
@type instance: L{objects.Instance}
|
311 |
@param instance: instance to be accepted
|
312 |
@type mem: int
|
313 |
@param mem: actual memory size to use for instance runtime
|
314 |
|
315 |
"""
|
316 |
raise NotImplementedError |
317 |
|
318 |
def FinalizeMigrationDst(self, instance, info, success): |
319 |
"""Finalize the instance migration on the target node.
|
320 |
|
321 |
Should finalize or revert any preparation done to accept the instance.
|
322 |
Since by default we do no preparation, we also don't have anything to do
|
323 |
|
324 |
@type instance: L{objects.Instance}
|
325 |
@param instance: instance whose migration is being finalized
|
326 |
@type info: string/data (opaque)
|
327 |
@param info: migration information, from the source node
|
328 |
@type success: boolean
|
329 |
@param success: whether the migration was a success or a failure
|
330 |
|
331 |
"""
|
332 |
pass
|
333 |
|
334 |
def MigrateInstance(self, instance, target, live): |
335 |
"""Migrate an instance.
|
336 |
|
337 |
@type instance: L{objects.Instance}
|
338 |
@param instance: the instance to be migrated
|
339 |
@type target: string
|
340 |
@param target: hostname (usually ip) of the target node
|
341 |
@type live: boolean
|
342 |
@param live: whether to do a live or non-live migration
|
343 |
|
344 |
"""
|
345 |
raise NotImplementedError |
346 |
|
347 |
def FinalizeMigrationSource(self, instance, success, live): |
348 |
"""Finalize the instance migration on the source node.
|
349 |
|
350 |
@type instance: L{objects.Instance}
|
351 |
@param instance: the instance that was migrated
|
352 |
@type success: bool
|
353 |
@param success: whether the migration succeeded or not
|
354 |
@type live: bool
|
355 |
@param live: whether the user requested a live migration or not
|
356 |
|
357 |
"""
|
358 |
pass
|
359 |
|
360 |
def GetMigrationStatus(self, instance): |
361 |
"""Get the migration status
|
362 |
|
363 |
@type instance: L{objects.Instance}
|
364 |
@param instance: the instance that is being migrated
|
365 |
@rtype: L{objects.MigrationStatus}
|
366 |
@return: the status of the current migration (one of
|
367 |
L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
|
368 |
progress info that can be retrieved from the hypervisor
|
369 |
|
370 |
"""
|
371 |
raise NotImplementedError |
372 |
|
373 |
def _InstanceStartupMemory(self, instance): |
374 |
"""Get the correct startup memory for an instance
|
375 |
|
376 |
This function calculates how much memory an instance should be started
|
377 |
with, making sure it's a value between the minimum and the maximum memory,
|
378 |
but also trying to use no more than the current free memory on the node.
|
379 |
|
380 |
@type instance: L{objects.Instance}
|
381 |
@param instance: the instance that is being started
|
382 |
@rtype: integer
|
383 |
@return: memory the instance should be started with
|
384 |
|
385 |
"""
|
386 |
free_memory = self.GetNodeInfo()["memory_free"] |
387 |
max_start_mem = min(instance.beparams[constants.BE_MAXMEM], free_memory)
|
388 |
start_mem = max(instance.beparams[constants.BE_MINMEM], max_start_mem)
|
389 |
return start_mem
|
390 |
|
391 |
@classmethod
|
392 |
def CheckParameterSyntax(cls, hvparams): |
393 |
"""Check the given parameters for validity.
|
394 |
|
395 |
This should check the passed set of parameters for
|
396 |
validity. Classes should extend, not replace, this function.
|
397 |
|
398 |
@type hvparams: dict
|
399 |
@param hvparams: dictionary with parameter names/value
|
400 |
@raise errors.HypervisorError: when a parameter is not valid
|
401 |
|
402 |
"""
|
403 |
for key in hvparams: |
404 |
if key not in cls.PARAMETERS: |
405 |
raise errors.HypervisorError("Parameter '%s' is not supported" % key) |
406 |
|
407 |
# cheap tests that run on the master, should not access the world
|
408 |
for name, (required, check_fn, errstr, _, _) in cls.PARAMETERS.items(): |
409 |
if name not in hvparams: |
410 |
raise errors.HypervisorError("Parameter '%s' is missing" % name) |
411 |
value = hvparams[name] |
412 |
if not required and not value: |
413 |
continue
|
414 |
if not value: |
415 |
raise errors.HypervisorError("Parameter '%s' is required but" |
416 |
" is currently not defined" % (name, ))
|
417 |
if check_fn is not None and not check_fn(value): |
418 |
raise errors.HypervisorError("Parameter '%s' fails syntax" |
419 |
" check: %s (current value: '%s')" %
|
420 |
(name, errstr, value)) |
421 |
|
422 |
@classmethod
|
423 |
def ValidateParameters(cls, hvparams): |
424 |
"""Check the given parameters for validity.
|
425 |
|
426 |
This should check the passed set of parameters for
|
427 |
validity. Classes should extend, not replace, this function.
|
428 |
|
429 |
@type hvparams: dict
|
430 |
@param hvparams: dictionary with parameter names/value
|
431 |
@raise errors.HypervisorError: when a parameter is not valid
|
432 |
|
433 |
"""
|
434 |
for name, (required, _, _, check_fn, errstr) in cls.PARAMETERS.items(): |
435 |
value = hvparams[name] |
436 |
if not required and not value: |
437 |
continue
|
438 |
if check_fn is not None and not check_fn(value): |
439 |
raise errors.HypervisorError("Parameter '%s' fails" |
440 |
" validation: %s (current value: '%s')" %
|
441 |
(name, errstr, value)) |
442 |
|
443 |
@classmethod
|
444 |
def PowercycleNode(cls): |
445 |
"""Hard powercycle a node using hypervisor specific methods.
|
446 |
|
447 |
This method should hard powercycle the node, using whatever
|
448 |
methods the hypervisor provides. Note that this means that all
|
449 |
instances running on the node must be stopped too.
|
450 |
|
451 |
"""
|
452 |
raise NotImplementedError |
453 |
|
454 |
@staticmethod
|
455 |
def GetLinuxNodeInfo(meminfo="/proc/meminfo", cpuinfo="/proc/cpuinfo"): |
456 |
"""For linux systems, return actual OS information.
|
457 |
|
458 |
This is an abstraction for all non-hypervisor-based classes, where
|
459 |
the node actually sees all the memory and CPUs via the /proc
|
460 |
interface and standard commands. The other case if for example
|
461 |
xen, where you only see the hardware resources via xen-specific
|
462 |
tools.
|
463 |
|
464 |
@param meminfo: name of the file containing meminfo
|
465 |
@type meminfo: string
|
466 |
@param cpuinfo: name of the file containing cpuinfo
|
467 |
@type cpuinfo: string
|
468 |
@return: a dict with the following keys (values in MiB):
|
469 |
- memory_total: the total memory size on the node
|
470 |
- memory_free: the available memory on the node for instances
|
471 |
- memory_dom0: the memory used by the node itself, if available
|
472 |
|
473 |
"""
|
474 |
try:
|
475 |
data = utils.ReadFile(meminfo).splitlines() |
476 |
except EnvironmentError, err: |
477 |
raise errors.HypervisorError("Failed to list node info: %s" % (err,)) |
478 |
|
479 |
result = {} |
480 |
sum_free = 0
|
481 |
try:
|
482 |
for line in data: |
483 |
splitfields = line.split(":", 1) |
484 |
|
485 |
if len(splitfields) > 1: |
486 |
key = splitfields[0].strip()
|
487 |
val = splitfields[1].strip()
|
488 |
if key == "MemTotal": |
489 |
result["memory_total"] = int(val.split()[0]) / 1024 |
490 |
elif key in ("MemFree", "Buffers", "Cached"): |
491 |
sum_free += int(val.split()[0]) / 1024 |
492 |
elif key == "Active": |
493 |
result["memory_dom0"] = int(val.split()[0]) / 1024 |
494 |
except (ValueError, TypeError), err: |
495 |
raise errors.HypervisorError("Failed to compute memory usage: %s" % |
496 |
(err,)) |
497 |
result["memory_free"] = sum_free
|
498 |
|
499 |
cpu_total = 0
|
500 |
try:
|
501 |
fh = open(cpuinfo)
|
502 |
try:
|
503 |
cpu_total = len(re.findall("(?m)^processor\s*:\s*[0-9]+\s*$", |
504 |
fh.read())) |
505 |
finally:
|
506 |
fh.close() |
507 |
except EnvironmentError, err: |
508 |
raise errors.HypervisorError("Failed to list node info: %s" % (err,)) |
509 |
result["cpu_total"] = cpu_total
|
510 |
# FIXME: export correct data here
|
511 |
result["cpu_nodes"] = 1 |
512 |
result["cpu_sockets"] = 1 |
513 |
|
514 |
return result
|
515 |
|
516 |
@classmethod
|
517 |
def LinuxPowercycle(cls): |
518 |
"""Linux-specific powercycle method.
|
519 |
|
520 |
"""
|
521 |
try:
|
522 |
fd = os.open("/proc/sysrq-trigger", os.O_WRONLY)
|
523 |
try:
|
524 |
os.write(fd, "b")
|
525 |
finally:
|
526 |
fd.close() |
527 |
except OSError: |
528 |
logging.exception("Can't open the sysrq-trigger file")
|
529 |
result = utils.RunCmd(["reboot", "-n", "-f"]) |
530 |
if not result: |
531 |
logging.error("Can't run shutdown: %s", result.output)
|
532 |
|
533 |
@staticmethod
|
534 |
def _FormatVerifyResults(msgs): |
535 |
"""Formats the verification results, given a list of errors.
|
536 |
|
537 |
@param msgs: list of errors, possibly empty
|
538 |
@return: overall problem description if something is wrong,
|
539 |
C{None} otherwise
|
540 |
|
541 |
"""
|
542 |
if msgs:
|
543 |
return "; ".join(msgs) |
544 |
else:
|
545 |
return None |