root / lib / hypervisor.py @ 523687d7
History | View | Annotate | Download (14.7 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007 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 |
"""Module that abstracts the virtualisation interface
|
23 |
|
24 |
"""
|
25 |
|
26 |
import time |
27 |
import os |
28 |
from cStringIO import StringIO |
29 |
|
30 |
from ganeti import utils |
31 |
from ganeti import logger |
32 |
from ganeti import ssconf |
33 |
from ganeti import constants |
34 |
from ganeti.errors import HypervisorError |
35 |
|
36 |
_HT_XEN30 = "xen-3.0"
|
37 |
_HT_FAKE = "fake"
|
38 |
|
39 |
VALID_HTYPES = (_HT_XEN30, _HT_FAKE) |
40 |
|
41 |
def GetHypervisor(): |
42 |
"""Return a Hypervisor instance.
|
43 |
|
44 |
This function parses the cluster hypervisor configuration file and
|
45 |
instantiates a class based on the value of this file.
|
46 |
|
47 |
"""
|
48 |
ht_kind = ssconf.SimpleStore().GetHypervisorType() |
49 |
if ht_kind == _HT_XEN30:
|
50 |
cls = XenHypervisor |
51 |
elif ht_kind == _HT_FAKE:
|
52 |
cls = FakeHypervisor |
53 |
else:
|
54 |
raise HypervisorError("Unknown hypervisor type '%s'" % ht_kind) |
55 |
return cls()
|
56 |
|
57 |
|
58 |
class BaseHypervisor(object): |
59 |
"""Abstract virtualisation technology interface
|
60 |
|
61 |
The goal is that all aspects of the virtualisation technology must
|
62 |
be abstracted away from the rest of code.
|
63 |
|
64 |
"""
|
65 |
def __init__(self): |
66 |
pass
|
67 |
|
68 |
def StartInstance(self, instance, block_devices, extra_args): |
69 |
"""Start an instance."""
|
70 |
raise NotImplementedError |
71 |
|
72 |
def StopInstance(self, instance, force=False): |
73 |
"""Stop an instance."""
|
74 |
raise NotImplementedError |
75 |
|
76 |
def RebootInstance(self, instance): |
77 |
"""Reboot an instance."""
|
78 |
raise NotImplementedError |
79 |
|
80 |
def ListInstances(self): |
81 |
"""Get the list of running instances."""
|
82 |
raise NotImplementedError |
83 |
|
84 |
def GetInstanceInfo(self, instance_name): |
85 |
"""Get instance properties.
|
86 |
|
87 |
Args:
|
88 |
instance_name: the instance name
|
89 |
|
90 |
Returns:
|
91 |
(name, id, memory, vcpus, state, times)
|
92 |
|
93 |
"""
|
94 |
raise NotImplementedError |
95 |
|
96 |
def GetAllInstancesInfo(self): |
97 |
"""Get properties of all instances.
|
98 |
|
99 |
Returns:
|
100 |
[(name, id, memory, vcpus, stat, times),...]
|
101 |
"""
|
102 |
raise NotImplementedError |
103 |
|
104 |
def GetNodeInfo(self): |
105 |
"""Return information about the node.
|
106 |
|
107 |
The return value is a dict, which has to have the following items:
|
108 |
(all values in MiB)
|
109 |
- memory_total: the total memory size on the node
|
110 |
- memory_free: the available memory on the node for instances
|
111 |
- memory_dom0: the memory used by the node itself, if available
|
112 |
|
113 |
"""
|
114 |
raise NotImplementedError |
115 |
|
116 |
@staticmethod
|
117 |
def GetShellCommandForConsole(instance_name): |
118 |
"""Return a command for connecting to the console of an instance.
|
119 |
|
120 |
"""
|
121 |
raise NotImplementedError |
122 |
|
123 |
def Verify(self): |
124 |
"""Verify the hypervisor.
|
125 |
|
126 |
"""
|
127 |
raise NotImplementedError |
128 |
|
129 |
|
130 |
class XenHypervisor(BaseHypervisor): |
131 |
"""Xen hypervisor interface"""
|
132 |
|
133 |
@staticmethod
|
134 |
def _WriteConfigFile(instance, block_devices, extra_args): |
135 |
"""Create a Xen 3.0 config file.
|
136 |
|
137 |
"""
|
138 |
config = StringIO() |
139 |
config.write("# this is autogenerated by Ganeti, please do not edit\n#\n")
|
140 |
config.write("kernel = '%s'\n" % constants.XEN_KERNEL)
|
141 |
if os.path.exists(constants.XEN_INITRD):
|
142 |
config.write("ramdisk = '%s'\n" % constants.XEN_INITRD)
|
143 |
config.write("memory = %d\n" % instance.memory)
|
144 |
config.write("vcpus = %d\n" % instance.vcpus)
|
145 |
config.write("name = '%s'\n" % instance.name)
|
146 |
|
147 |
vif_data = [] |
148 |
for nic in instance.nics: |
149 |
nic_str = "mac=%s, bridge=%s" % (nic.mac, nic.bridge)
|
150 |
ip = getattr(nic, "ip", None) |
151 |
if ip is not None: |
152 |
nic_str += ", ip=%s" % ip
|
153 |
vif_data.append("'%s'" % nic_str)
|
154 |
|
155 |
config.write("vif = [%s]\n" % ",".join(vif_data)) |
156 |
|
157 |
disk_data = ["'phy:%s,%s,w'" % (rldev.dev_path, cfdev.iv_name)
|
158 |
for cfdev, rldev in block_devices] |
159 |
config.write("disk = [%s]\n" % ",".join(disk_data)) |
160 |
|
161 |
config.write("root = '/dev/sda ro'\n")
|
162 |
config.write("on_poweroff = 'destroy'\n")
|
163 |
config.write("on_reboot = 'restart'\n")
|
164 |
config.write("on_crash = 'restart'\n")
|
165 |
if extra_args:
|
166 |
config.write("extra = '%s'\n" % extra_args)
|
167 |
# just in case it exists
|
168 |
utils.RemoveFile("/etc/xen/auto/%s" % instance.name)
|
169 |
f = open("/etc/xen/%s" % instance.name, "w") |
170 |
f.write(config.getvalue()) |
171 |
f.close() |
172 |
return True |
173 |
|
174 |
@staticmethod
|
175 |
def _RemoveConfigFile(instance): |
176 |
"""Remove the xen configuration file.
|
177 |
|
178 |
"""
|
179 |
utils.RemoveFile("/etc/xen/%s" % instance.name)
|
180 |
|
181 |
@staticmethod
|
182 |
def _GetXMList(include_node): |
183 |
"""Return the list of running instances.
|
184 |
|
185 |
If the `include_node` argument is True, then we return information
|
186 |
for dom0 also, otherwise we filter that from the return value.
|
187 |
|
188 |
The return value is a list of (name, id, memory, vcpus, state, time spent)
|
189 |
|
190 |
"""
|
191 |
for dummy in range(5): |
192 |
result = utils.RunCmd(["xm", "list"]) |
193 |
if not result.failed: |
194 |
break
|
195 |
logger.Error("xm list failed (%s): %s" % (result.fail_reason,
|
196 |
result.output)) |
197 |
time.sleep(1)
|
198 |
|
199 |
if result.failed:
|
200 |
raise HypervisorError("xm list failed, retries exceeded (%s): %s" % |
201 |
(result.fail_reason, result.stderr)) |
202 |
|
203 |
# skip over the heading and the domain 0 line (optional)
|
204 |
if include_node:
|
205 |
to_skip = 1
|
206 |
else:
|
207 |
to_skip = 2
|
208 |
lines = result.stdout.splitlines()[to_skip:] |
209 |
result = [] |
210 |
for line in lines: |
211 |
# The format of lines is:
|
212 |
# Name ID Mem(MiB) VCPUs State Time(s)
|
213 |
# Domain-0 0 3418 4 r----- 266.2
|
214 |
data = line.split() |
215 |
if len(data) != 6: |
216 |
raise HypervisorError("Can't parse output of xm list, line: %s" % line) |
217 |
try:
|
218 |
data[1] = int(data[1]) |
219 |
data[2] = int(data[2]) |
220 |
data[3] = int(data[3]) |
221 |
data[5] = float(data[5]) |
222 |
except ValueError, err: |
223 |
raise HypervisorError("Can't parse output of xm list," |
224 |
" line: %s, error: %s" % (line, err))
|
225 |
result.append(data) |
226 |
return result
|
227 |
|
228 |
def ListInstances(self): |
229 |
"""Get the list of running instances.
|
230 |
|
231 |
"""
|
232 |
xm_list = self._GetXMList(False) |
233 |
names = [info[0] for info in xm_list] |
234 |
return names
|
235 |
|
236 |
def GetInstanceInfo(self, instance_name): |
237 |
"""Get instance properties.
|
238 |
|
239 |
Args:
|
240 |
instance_name: the instance name
|
241 |
|
242 |
Returns:
|
243 |
(name, id, memory, vcpus, stat, times)
|
244 |
"""
|
245 |
xm_list = self._GetXMList(instance_name=="Domain-0") |
246 |
result = None
|
247 |
for data in xm_list: |
248 |
if data[0] == instance_name: |
249 |
result = data |
250 |
break
|
251 |
return result
|
252 |
|
253 |
def GetAllInstancesInfo(self): |
254 |
"""Get properties of all instances.
|
255 |
|
256 |
Returns:
|
257 |
[(name, id, memory, vcpus, stat, times),...]
|
258 |
"""
|
259 |
xm_list = self._GetXMList(False) |
260 |
return xm_list
|
261 |
|
262 |
def StartInstance(self, instance, block_devices, extra_args): |
263 |
"""Start an instance."""
|
264 |
self._WriteConfigFile(instance, block_devices, extra_args)
|
265 |
result = utils.RunCmd(["xm", "create", instance.name]) |
266 |
|
267 |
if result.failed:
|
268 |
raise HypervisorError("Failed to start instance %s: %s (%s)" % |
269 |
(instance.name, result.fail_reason, result.output)) |
270 |
|
271 |
def StopInstance(self, instance, force=False): |
272 |
"""Stop an instance."""
|
273 |
self._RemoveConfigFile(instance)
|
274 |
if force:
|
275 |
command = ["xm", "destroy", instance.name] |
276 |
else:
|
277 |
command = ["xm", "shutdown", instance.name] |
278 |
result = utils.RunCmd(command) |
279 |
|
280 |
if result.failed:
|
281 |
raise HypervisorError("Failed to stop instance %s: %s" % |
282 |
(instance.name, result.fail_reason)) |
283 |
|
284 |
def RebootInstance(self, instance): |
285 |
"""Reboot an instance."""
|
286 |
result = utils.RunCmd(["xm", "reboot", instance.name]) |
287 |
|
288 |
if result.failed:
|
289 |
raise HypervisorError("Failed to reboot instance %s: %s" % |
290 |
(instance.name, result.fail_reason)) |
291 |
|
292 |
def GetNodeInfo(self): |
293 |
"""Return information about the node.
|
294 |
|
295 |
The return value is a dict, which has to have the following items:
|
296 |
(all values in MiB)
|
297 |
- memory_total: the total memory size on the node
|
298 |
- memory_free: the available memory on the node for instances
|
299 |
- memory_dom0: the memory used by the node itself, if available
|
300 |
|
301 |
"""
|
302 |
# note: in xen 3, memory has changed to total_memory
|
303 |
result = utils.RunCmd(["xm", "info"]) |
304 |
if result.failed:
|
305 |
logger.Error("Can't run 'xm info': %s" % result.fail_reason)
|
306 |
return None |
307 |
|
308 |
xmoutput = result.stdout.splitlines() |
309 |
result = {} |
310 |
for line in xmoutput: |
311 |
splitfields = line.split(":", 1) |
312 |
|
313 |
if len(splitfields) > 1: |
314 |
key = splitfields[0].strip()
|
315 |
val = splitfields[1].strip()
|
316 |
if key == 'memory' or key == 'total_memory': |
317 |
result['memory_total'] = int(val) |
318 |
elif key == 'free_memory': |
319 |
result['memory_free'] = int(val) |
320 |
dom0_info = self.GetInstanceInfo("Domain-0") |
321 |
if dom0_info is not None: |
322 |
result['memory_dom0'] = dom0_info[2] |
323 |
|
324 |
return result
|
325 |
|
326 |
@staticmethod
|
327 |
def GetShellCommandForConsole(instance_name): |
328 |
"""Return a command for connecting to the console of an instance.
|
329 |
|
330 |
"""
|
331 |
return "xm console %s" % instance_name |
332 |
|
333 |
|
334 |
def Verify(self): |
335 |
"""Verify the hypervisor.
|
336 |
|
337 |
For Xen, this verifies that the xend process is running.
|
338 |
|
339 |
"""
|
340 |
if not utils.CheckDaemonAlive('/var/run/xend.pid', 'xend'): |
341 |
return "xend daemon is not running" |
342 |
|
343 |
|
344 |
class FakeHypervisor(BaseHypervisor): |
345 |
"""Fake hypervisor interface.
|
346 |
|
347 |
This can be used for testing the ganeti code without having to have
|
348 |
a real virtualisation software installed.
|
349 |
|
350 |
"""
|
351 |
_ROOT_DIR = "/var/run/ganeti-fake-hypervisor"
|
352 |
|
353 |
def __init__(self): |
354 |
BaseHypervisor.__init__(self)
|
355 |
if not os.path.exists(self._ROOT_DIR): |
356 |
os.mkdir(self._ROOT_DIR)
|
357 |
|
358 |
def ListInstances(self): |
359 |
"""Get the list of running instances.
|
360 |
|
361 |
"""
|
362 |
return os.listdir(self._ROOT_DIR) |
363 |
|
364 |
def GetInstanceInfo(self, instance_name): |
365 |
"""Get instance properties.
|
366 |
|
367 |
Args:
|
368 |
instance_name: the instance name
|
369 |
|
370 |
Returns:
|
371 |
(name, id, memory, vcpus, stat, times)
|
372 |
"""
|
373 |
file_name = "%s/%s" % (self._ROOT_DIR, instance_name) |
374 |
if not os.path.exists(file_name): |
375 |
return None |
376 |
try:
|
377 |
fh = file(file_name, "r") |
378 |
try:
|
379 |
inst_id = fh.readline().strip() |
380 |
memory = fh.readline().strip() |
381 |
vcpus = fh.readline().strip() |
382 |
stat = "---b-"
|
383 |
times = "0"
|
384 |
return (instance_name, inst_id, memory, vcpus, stat, times)
|
385 |
finally:
|
386 |
fh.close() |
387 |
except IOError, err: |
388 |
raise HypervisorError("Failed to list instance %s: %s" % |
389 |
(instance_name, err)) |
390 |
|
391 |
def GetAllInstancesInfo(self): |
392 |
"""Get properties of all instances.
|
393 |
|
394 |
Returns:
|
395 |
[(name, id, memory, vcpus, stat, times),...]
|
396 |
"""
|
397 |
data = [] |
398 |
for file_name in os.listdir(self._ROOT_DIR): |
399 |
try:
|
400 |
fh = file(self._ROOT_DIR+"/"+file_name, "r") |
401 |
inst_id = "-1"
|
402 |
memory = "0"
|
403 |
stat = "-----"
|
404 |
times = "-1"
|
405 |
try:
|
406 |
inst_id = fh.readline().strip() |
407 |
memory = fh.readline().strip() |
408 |
vcpus = fh.readline().strip() |
409 |
stat = "---b-"
|
410 |
times = "0"
|
411 |
finally:
|
412 |
fh.close() |
413 |
data.append((file_name, inst_id, memory, vcpus, stat, times)) |
414 |
except IOError, err: |
415 |
raise HypervisorError("Failed to list instances: %s" % err) |
416 |
return data
|
417 |
|
418 |
def StartInstance(self, instance, force, extra_args): |
419 |
"""Start an instance.
|
420 |
|
421 |
For the fake hypervisor, it just creates a file in the base dir,
|
422 |
creating an exception if it already exists. We don't actually
|
423 |
handle race conditions properly, since these are *FAKE* instances.
|
424 |
|
425 |
"""
|
426 |
file_name = self._ROOT_DIR + "/%s" % instance.name |
427 |
if os.path.exists(file_name):
|
428 |
raise HypervisorError("Failed to start instance %s: %s" % |
429 |
(instance.name, "already running"))
|
430 |
try:
|
431 |
fh = file(file_name, "w") |
432 |
try:
|
433 |
fh.write("0\n%d\n%d\n" % (instance.memory, instance.vcpus))
|
434 |
finally:
|
435 |
fh.close() |
436 |
except IOError, err: |
437 |
raise HypervisorError("Failed to start instance %s: %s" % |
438 |
(instance.name, err)) |
439 |
|
440 |
def StopInstance(self, instance, force=False): |
441 |
"""Stop an instance.
|
442 |
|
443 |
For the fake hypervisor, this just removes the file in the base
|
444 |
dir, if it exist, otherwise we raise an exception.
|
445 |
|
446 |
"""
|
447 |
file_name = self._ROOT_DIR + "/%s" % instance.name |
448 |
if not os.path.exists(file_name): |
449 |
raise HypervisorError("Failed to stop instance %s: %s" % |
450 |
(instance.name, "not running"))
|
451 |
utils.RemoveFile(file_name) |
452 |
|
453 |
def RebootInstance(self, instance): |
454 |
"""Reboot an instance.
|
455 |
|
456 |
For the fake hypervisor, this does nothing.
|
457 |
|
458 |
"""
|
459 |
return
|
460 |
|
461 |
def GetNodeInfo(self): |
462 |
"""Return information about the node.
|
463 |
|
464 |
The return value is a dict, which has to have the following items:
|
465 |
(all values in MiB)
|
466 |
- memory_total: the total memory size on the node
|
467 |
- memory_free: the available memory on the node for instances
|
468 |
- memory_dom0: the memory used by the node itself, if available
|
469 |
|
470 |
"""
|
471 |
# global ram usage from the xm info command
|
472 |
# memory : 3583
|
473 |
# free_memory : 747
|
474 |
# note: in xen 3, memory has changed to total_memory
|
475 |
try:
|
476 |
fh = file("/proc/meminfo") |
477 |
try:
|
478 |
data = fh.readlines() |
479 |
finally:
|
480 |
fh.close() |
481 |
except IOError, err: |
482 |
raise HypervisorError("Failed to list node info: %s" % err) |
483 |
|
484 |
result = {} |
485 |
sum_free = 0
|
486 |
for line in data: |
487 |
splitfields = line.split(":", 1) |
488 |
|
489 |
if len(splitfields) > 1: |
490 |
key = splitfields[0].strip()
|
491 |
val = splitfields[1].strip()
|
492 |
if key == 'MemTotal': |
493 |
result['memory_total'] = int(val.split()[0])/1024 |
494 |
elif key in ('MemFree', 'Buffers', 'Cached'): |
495 |
sum_free += int(val.split()[0])/1024 |
496 |
elif key == 'Active': |
497 |
result['memory_dom0'] = int(val.split()[0])/1024 |
498 |
|
499 |
result['memory_free'] = sum_free
|
500 |
return result
|
501 |
|
502 |
@staticmethod
|
503 |
def GetShellCommandForConsole(instance_name): |
504 |
"""Return a command for connecting to the console of an instance.
|
505 |
|
506 |
"""
|
507 |
return "echo Console not available for fake hypervisor" |
508 |
|
509 |
def Verify(self): |
510 |
"""Verify the hypervisor.
|
511 |
|
512 |
For the fake hypervisor, it just checks the existence of the base
|
513 |
dir.
|
514 |
|
515 |
"""
|
516 |
if not os.path.exists(self._ROOT_DIR): |
517 |
return "The required directory '%s' does not exist." % self._ROOT_DIR |