root / lib / hypervisor / hv_lxc.py @ b8aa46ed
History | View | Annotate | Download (13.1 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2010 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 |
"""LXC hypervisor
|
23 |
|
24 |
"""
|
25 |
|
26 |
import os |
27 |
import os.path |
28 |
import time |
29 |
import logging |
30 |
|
31 |
from ganeti import constants |
32 |
from ganeti import errors # pylint: disable-msg=W0611 |
33 |
from ganeti import utils |
34 |
from ganeti.hypervisor import hv_base |
35 |
from ganeti.errors import HypervisorError |
36 |
|
37 |
|
38 |
class LXCHypervisor(hv_base.BaseHypervisor): |
39 |
"""LXC-based virtualization.
|
40 |
|
41 |
Since current (Spring 2010) distributions are not yet ready for
|
42 |
running under a container, the following changes must be done
|
43 |
manually:
|
44 |
- remove udev
|
45 |
- disable the kernel log component of sysklogd/rsyslog/etc.,
|
46 |
otherwise they will fail to read the log, and at least rsyslog
|
47 |
will fill the filesystem with error messages
|
48 |
|
49 |
TODO:
|
50 |
- move hardcoded parameters into hypervisor parameters, once we
|
51 |
have the container-parameter support
|
52 |
- implement memory limits, but only optionally, depending on host
|
53 |
kernel support
|
54 |
|
55 |
Problems/issues:
|
56 |
- LXC is very temperamental; in daemon mode, it succeeds or fails
|
57 |
in launching the instance silently, without any error
|
58 |
indication, and when failing it can leave network interfaces
|
59 |
around, and future successful startups will list the instance
|
60 |
twice
|
61 |
- shutdown sequence of containers leaves the init 'dead', and the
|
62 |
container effectively stopped, but LXC still believes the
|
63 |
container to be running; need to investigate using the
|
64 |
notify_on_release and release_agent feature of cgroups
|
65 |
|
66 |
"""
|
67 |
_ROOT_DIR = constants.RUN_GANETI_DIR + "/lxc"
|
68 |
_DEVS = [ |
69 |
"c 1:3", # /dev/null |
70 |
"c 1:5", # /dev/zero |
71 |
"c 1:7", # /dev/full |
72 |
"c 1:8", # /dev/random |
73 |
"c 1:9", # /dev/urandom |
74 |
"c 1:10", # /dev/aio |
75 |
"c 5:0", # /dev/tty |
76 |
"c 5:1", # /dev/console |
77 |
"c 5:2", # /dev/ptmx |
78 |
"c 136:*", # first block of Unix98 PTY slaves |
79 |
] |
80 |
_DENIED_CAPABILITIES = [ |
81 |
"mac_override", # Allow MAC configuration or state changes |
82 |
# TODO: remove sys_admin too, for safety
|
83 |
#"sys_admin", # Perform a range of system administration operations
|
84 |
"sys_boot", # Use reboot(2) and kexec_load(2) |
85 |
"sys_module", # Load and unload kernel modules |
86 |
"sys_time", # Set system clock, set real-time (hardware) clock |
87 |
] |
88 |
_DIR_MODE = 0755
|
89 |
|
90 |
PARAMETERS = { |
91 |
} |
92 |
|
93 |
def __init__(self): |
94 |
hv_base.BaseHypervisor.__init__(self)
|
95 |
utils.EnsureDirs([(self._ROOT_DIR, self._DIR_MODE)]) |
96 |
|
97 |
@staticmethod
|
98 |
def _GetMountSubdirs(path): |
99 |
"""Return the list of mountpoints under a given path.
|
100 |
|
101 |
"""
|
102 |
result = [] |
103 |
for _, mountpoint, _, _ in utils.GetMounts(): |
104 |
if (mountpoint.startswith(path) and |
105 |
mountpoint != path): |
106 |
result.append(mountpoint) |
107 |
|
108 |
result.sort(key=lambda x: x.count("/"), reverse=True) |
109 |
return result
|
110 |
|
111 |
@classmethod
|
112 |
def _InstanceDir(cls, instance_name): |
113 |
"""Return the root directory for an instance.
|
114 |
|
115 |
"""
|
116 |
return utils.PathJoin(cls._ROOT_DIR, instance_name)
|
117 |
|
118 |
@classmethod
|
119 |
def _InstanceConfFile(cls, instance_name): |
120 |
"""Return the configuration file for an instance.
|
121 |
|
122 |
"""
|
123 |
return utils.PathJoin(cls._ROOT_DIR, instance_name + ".conf") |
124 |
|
125 |
@classmethod
|
126 |
def _InstanceLogFile(cls, instance_name): |
127 |
"""Return the log file for an instance.
|
128 |
|
129 |
"""
|
130 |
return utils.PathJoin(cls._ROOT_DIR, instance_name + ".log") |
131 |
|
132 |
@classmethod
|
133 |
def _GetCgroupMountPoint(cls): |
134 |
for _, mountpoint, fstype, _ in utils.GetMounts(): |
135 |
if fstype == "cgroup": |
136 |
return mountpoint
|
137 |
raise errors.HypervisorError("The cgroup filesystem is not mounted") |
138 |
|
139 |
@classmethod
|
140 |
def _GetCgroupCpuList(cls, instance_name): |
141 |
"""Return the list of CPU ids for an instance.
|
142 |
|
143 |
"""
|
144 |
cgroup = cls._GetCgroupMountPoint() |
145 |
try:
|
146 |
cpus = utils.ReadFile(utils.PathJoin(cgroup, |
147 |
instance_name, |
148 |
"cpuset.cpus"))
|
149 |
except EnvironmentError, err: |
150 |
raise errors.HypervisorError("Getting CPU list for instance" |
151 |
" %s failed: %s" % (instance_name, err))
|
152 |
# cpuset.cpus format: comma-separated list of CPU ids
|
153 |
# or dash-separated id ranges
|
154 |
# Example: "0-1,3"
|
155 |
cpu_list = [] |
156 |
for range_def in cpus.split(","): |
157 |
boundaries = range_def.split("-")
|
158 |
n_elements = len(boundaries)
|
159 |
lower = int(boundaries[0]) |
160 |
higher = int(boundaries[n_elements - 1]) |
161 |
cpu_list.extend(range(lower, higher + 1)) |
162 |
return cpu_list
|
163 |
|
164 |
def ListInstances(self): |
165 |
"""Get the list of running instances.
|
166 |
|
167 |
"""
|
168 |
result = utils.RunCmd(["lxc-ls"])
|
169 |
if result.failed:
|
170 |
raise errors.HypervisorError("Running lxc-ls failed: %s" % result.output) |
171 |
return result.stdout.splitlines()
|
172 |
|
173 |
def GetInstanceInfo(self, instance_name): |
174 |
"""Get instance properties.
|
175 |
|
176 |
@type instance_name: string
|
177 |
@param instance_name: the instance name
|
178 |
|
179 |
@return: (name, id, memory, vcpus, stat, times)
|
180 |
|
181 |
"""
|
182 |
# TODO: read container info from the cgroup mountpoint
|
183 |
|
184 |
result = utils.RunCmd(["lxc-info", "-n", instance_name]) |
185 |
if result.failed:
|
186 |
raise errors.HypervisorError("Running lxc-info failed: %s" % |
187 |
result.output) |
188 |
# lxc-info output examples:
|
189 |
# 'ganeti-lxc-test1' is STOPPED
|
190 |
# 'ganeti-lxc-test1' is RUNNING
|
191 |
_, state = result.stdout.rsplit(None, 1) |
192 |
if state != "RUNNING": |
193 |
return None |
194 |
|
195 |
cpu_list = self._GetCgroupCpuList(instance_name)
|
196 |
return (instance_name, 0, 0, len(cpu_list), 0, 0) |
197 |
|
198 |
def GetAllInstancesInfo(self): |
199 |
"""Get properties of all instances.
|
200 |
|
201 |
@return: [(name, id, memory, vcpus, stat, times),...]
|
202 |
|
203 |
"""
|
204 |
data = [] |
205 |
for name in self.ListInstances(): |
206 |
data.append(self.GetInstanceInfo(name))
|
207 |
return data
|
208 |
|
209 |
def _CreateConfigFile(self, instance, root_dir): |
210 |
"""Create an lxc.conf file for an instance"""
|
211 |
out = [] |
212 |
# hostname
|
213 |
out.append("lxc.utsname = %s" % instance.name)
|
214 |
|
215 |
# separate pseudo-TTY instances
|
216 |
out.append("lxc.pts = 255")
|
217 |
# standard TTYs
|
218 |
out.append("lxc.tty = 6")
|
219 |
# console log file
|
220 |
console_log = utils.PathJoin(self._ROOT_DIR, instance.name + ".console") |
221 |
try:
|
222 |
utils.WriteFile(console_log, data="", mode=constants.SECURE_FILE_MODE)
|
223 |
except EnvironmentError, err: |
224 |
raise errors.HypervisorError("Creating console log file %s for" |
225 |
" instance %s failed: %s" %
|
226 |
(console_log, instance.name, err)) |
227 |
out.append("lxc.console = %s" % console_log)
|
228 |
|
229 |
# root FS
|
230 |
out.append("lxc.rootfs = %s" % root_dir)
|
231 |
|
232 |
# TODO: additional mounts, if we disable CAP_SYS_ADMIN
|
233 |
|
234 |
# Device control
|
235 |
# deny direct device access
|
236 |
out.append("lxc.cgroup.devices.deny = a")
|
237 |
for devinfo in self._DEVS: |
238 |
out.append("lxc.cgroup.devices.allow = %s rw" % devinfo)
|
239 |
|
240 |
# Networking
|
241 |
for idx, nic in enumerate(instance.nics): |
242 |
out.append("# NIC %d" % idx)
|
243 |
mode = nic.nicparams[constants.NIC_MODE] |
244 |
link = nic.nicparams[constants.NIC_LINK] |
245 |
if mode == constants.NIC_MODE_BRIDGED:
|
246 |
out.append("lxc.network.type = veth")
|
247 |
out.append("lxc.network.link = %s" % link)
|
248 |
else:
|
249 |
raise errors.HypervisorError("LXC hypervisor only supports" |
250 |
" bridged mode (NIC %d has mode %s)" %
|
251 |
(idx, mode)) |
252 |
out.append("lxc.network.hwaddr = %s" % nic.mac)
|
253 |
out.append("lxc.network.flags = up")
|
254 |
|
255 |
# Capabilities
|
256 |
for cap in self._DENIED_CAPABILITIES: |
257 |
out.append("lxc.cap.drop = %s" % cap)
|
258 |
|
259 |
return "\n".join(out) + "\n" |
260 |
|
261 |
def StartInstance(self, instance, block_devices): |
262 |
"""Start an instance.
|
263 |
|
264 |
For LCX, we try to mount the block device and execute 'lxc-start
|
265 |
start' (we use volatile containers).
|
266 |
|
267 |
"""
|
268 |
root_dir = self._InstanceDir(instance.name)
|
269 |
try:
|
270 |
utils.EnsureDirs([(root_dir, self._DIR_MODE)])
|
271 |
except errors.GenericError, err:
|
272 |
raise HypervisorError("Creating instance directory failed: %s", str(err)) |
273 |
|
274 |
conf_file = self._InstanceConfFile(instance.name)
|
275 |
utils.WriteFile(conf_file, data=self._CreateConfigFile(instance, root_dir))
|
276 |
|
277 |
log_file = self._InstanceLogFile(instance.name)
|
278 |
if not os.path.exists(log_file): |
279 |
try:
|
280 |
utils.WriteFile(log_file, data="", mode=constants.SECURE_FILE_MODE)
|
281 |
except EnvironmentError, err: |
282 |
raise errors.HypervisorError("Creating hypervisor log file %s for" |
283 |
" instance %s failed: %s" %
|
284 |
(log_file, instance.name, err)) |
285 |
|
286 |
if not os.path.ismount(root_dir): |
287 |
if not block_devices: |
288 |
raise HypervisorError("LXC needs at least one disk") |
289 |
|
290 |
sda_dev_path = block_devices[0][1] |
291 |
result = utils.RunCmd(["mount", sda_dev_path, root_dir])
|
292 |
if result.failed:
|
293 |
raise HypervisorError("Mounting the root dir of LXC instance %s" |
294 |
" failed: %s" % (instance.name, result.output))
|
295 |
result = utils.RunCmd(["lxc-start", "-n", instance.name, |
296 |
"-o", log_file,
|
297 |
"-l", "DEBUG", |
298 |
"-f", conf_file, "-d"]) |
299 |
if result.failed:
|
300 |
raise HypervisorError("Running the lxc-start script failed: %s" % |
301 |
result.output) |
302 |
|
303 |
def StopInstance(self, instance, force=False, retry=False, name=None): |
304 |
"""Stop an instance.
|
305 |
|
306 |
This method has complicated cleanup tests, as we must:
|
307 |
- try to kill all leftover processes
|
308 |
- try to unmount any additional sub-mountpoints
|
309 |
- finally unmount the instance dir
|
310 |
|
311 |
"""
|
312 |
if name is None: |
313 |
name = instance.name |
314 |
|
315 |
root_dir = self._InstanceDir(name)
|
316 |
if not os.path.exists(root_dir): |
317 |
return
|
318 |
|
319 |
if name in self.ListInstances(): |
320 |
# Signal init to shutdown; this is a hack
|
321 |
if not retry and not force: |
322 |
result = utils.RunCmd(["chroot", root_dir, "poweroff"]) |
323 |
if result.failed:
|
324 |
raise HypervisorError("Running 'poweroff' on the instance" |
325 |
" failed: %s" % result.output)
|
326 |
time.sleep(2)
|
327 |
result = utils.RunCmd(["lxc-stop", "-n", name]) |
328 |
if result.failed:
|
329 |
logging.warning("Error while doing lxc-stop for %s: %s", name,
|
330 |
result.output) |
331 |
|
332 |
for mpath in self._GetMountSubdirs(root_dir): |
333 |
result = utils.RunCmd(["umount", mpath])
|
334 |
if result.failed:
|
335 |
logging.warning("Error while umounting subpath %s for instance %s: %s",
|
336 |
mpath, name, result.output) |
337 |
|
338 |
result = utils.RunCmd(["umount", root_dir])
|
339 |
if result.failed and force: |
340 |
msg = ("Processes still alive in the chroot: %s" %
|
341 |
utils.RunCmd("fuser -vm %s" % root_dir).output)
|
342 |
logging.error(msg) |
343 |
raise HypervisorError("Unmounting the chroot dir failed: %s (%s)" % |
344 |
(result.output, msg)) |
345 |
|
346 |
def RebootInstance(self, instance): |
347 |
"""Reboot an instance.
|
348 |
|
349 |
This is not (yet) implemented (in Ganeti) for the LXC hypervisor.
|
350 |
|
351 |
"""
|
352 |
# TODO: implement reboot
|
353 |
raise HypervisorError("The LXC hypervisor doesn't implement the" |
354 |
" reboot functionality")
|
355 |
|
356 |
def GetNodeInfo(self): |
357 |
"""Return information about the node.
|
358 |
|
359 |
This is just a wrapper over the base GetLinuxNodeInfo method.
|
360 |
|
361 |
@return: a dict with the following keys (values in MiB):
|
362 |
- memory_total: the total memory size on the node
|
363 |
- memory_free: the available memory on the node for instances
|
364 |
- memory_dom0: the memory used by the node itself, if available
|
365 |
|
366 |
"""
|
367 |
return self.GetLinuxNodeInfo() |
368 |
|
369 |
@classmethod
|
370 |
def GetShellCommandForConsole(cls, instance, hvparams, beparams): |
371 |
"""Return a command for connecting to the console of an instance.
|
372 |
|
373 |
"""
|
374 |
return "lxc-console -n %s" % instance.name |
375 |
|
376 |
def Verify(self): |
377 |
"""Verify the hypervisor.
|
378 |
|
379 |
For the chroot manager, it just checks the existence of the base dir.
|
380 |
|
381 |
"""
|
382 |
if not os.path.exists(self._ROOT_DIR): |
383 |
return "The required directory '%s' does not exist." % self._ROOT_DIR |
384 |
|
385 |
@classmethod
|
386 |
def PowercycleNode(cls): |
387 |
"""LXC powercycle, just a wrapper over Linux powercycle.
|
388 |
|
389 |
"""
|
390 |
cls.LinuxPowercycle() |
391 |
|
392 |
def MigrateInstance(self, instance, target, live): |
393 |
"""Migrate an instance.
|
394 |
|
395 |
@type instance: L{objects.Instance}
|
396 |
@param instance: the instance to be migrated
|
397 |
@type target: string
|
398 |
@param target: hostname (usually ip) of the target node
|
399 |
@type live: boolean
|
400 |
@param live: whether to do a live or non-live migration
|
401 |
|
402 |
"""
|
403 |
raise HypervisorError("Migration is not supported by the LXC hypervisor") |