4 # Copyright (C) 2010 Google Inc.
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.
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.
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
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
38 class LXCHypervisor(hv_base.BaseHypervisor):
39 """LXC-based virtualization.
41 Since current (Spring 2010) distributions are not yet ready for
42 running under a container, the following changes must be done
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
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
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
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
67 _ROOT_DIR = constants.RUN_GANETI_DIR + "/lxc"
68 _LOG_FILE = constants.LOG_DIR + "hv_lxc.log"
73 "c 1:8", # /dev/random
74 "c 1:9", # /dev/urandom
77 "c 5:1", # /dev/console
79 "c 136:*", # first block of Unix98 PTY slaves
81 _DENIED_CAPABILITIES = [
82 "mac_override", # Allow MAC configuration or state changes
83 # TODO: remove sys_admin too, for safety
84 #"sys_admin", # Perform a range of system administration operations
85 "sys_boot", # Use reboot(2) and kexec_load(2)
86 "sys_module", # Load and unload kernel modules
87 "sys_time", # Set system clock, set real-time (hardware) clock
95 hv_base.BaseHypervisor.__init__(self)
96 utils.EnsureDirs([(self._ROOT_DIR, self._DIR_MODE)])
99 def _GetMountSubdirs(path):
100 """Return the list of mountpoints under a given path.
104 for _, mountpoint, _, _ in utils.GetMounts():
105 if (mountpoint.startswith(path) and
107 result.append(mountpoint)
109 result.sort(key=lambda x: x.count("/"), reverse=True)
113 def _InstanceDir(cls, instance_name):
114 """Return the root directory for an instance.
117 return utils.PathJoin(cls._ROOT_DIR, instance_name)
120 def _InstanceConfFile(cls, instance_name):
121 """Return the configuration file for an instance.
124 return utils.PathJoin(cls._ROOT_DIR, instance_name + ".conf")
127 def _GetCgroupMountPoint(cls):
128 for _, mountpoint, fstype, _ in utils.GetMounts():
129 if fstype == "cgroup":
131 raise errors.HypervisorError("The cgroup filesystem is not mounted")
134 def _GetCgroupCpuList(cls, instance_name):
135 """Return the list of CPU ids for an instance.
138 cgroup = cls._GetCgroupMountPoint()
140 cpus = utils.ReadFile(utils.PathJoin(cgroup,
143 except EnvironmentError, err:
144 raise errors.HypervisorError("Getting CPU list for instance"
145 " %s failed: %s" % (instance_name, err))
146 # cpuset.cpus format: comma-separated list of CPU ids
147 # or dash-separated id ranges
150 for range_def in cpus.split(","):
151 boundaries = range_def.split("-")
152 n_elements = len(boundaries)
153 lower = int(boundaries[0])
154 higher = int(boundaries[n_elements - 1])
155 cpu_list.extend(range(lower, higher + 1))
158 def ListInstances(self):
159 """Get the list of running instances.
162 result = utils.RunCmd(["lxc-ls"])
164 raise errors.HypervisorError("Can't run lxc-ls: %s" % result.output)
165 return result.stdout.splitlines()
167 def GetInstanceInfo(self, instance_name):
168 """Get instance properties.
170 @type instance_name: string
171 @param instance_name: the instance name
173 @return: (name, id, memory, vcpus, stat, times)
176 # TODO: read container info from the cgroup mountpoint
178 result = utils.RunCmd(["lxc-info", "-n", instance_name])
180 raise errors.HypervisorError("Can't run lxc-info: %s" % result.output)
181 # lxc-info output examples:
182 # 'ganeti-lxc-test1' is STOPPED
183 # 'ganeti-lxc-test1' is RUNNING
184 _, state = result.stdout.rsplit(None, 1)
186 cpu_list = self._GetCgroupCpuList(instance_name)
188 if state == "RUNNING":
189 return (instance_name, 0, 0, len(cpu_list), 0, 0)
192 def GetAllInstancesInfo(self):
193 """Get properties of all instances.
195 @return: [(name, id, memory, vcpus, stat, times),...]
199 for name in self.ListInstances():
200 data.append(self.GetInstanceInfo(name))
203 def _CreateConfigFile(self, instance, root_dir):
204 """Create an lxc.conf file for an instance"""
207 out.append("lxc.utsname = %s" % instance.name)
209 # separate pseudo-TTY instances
210 out.append("lxc.pts = 255")
212 out.append("lxc.tty = 6")
214 console_log = utils.PathJoin(self._ROOT_DIR, instance.name + ".console")
216 utils.WriteFile(console_log, data="", mode=constants.SECURE_FILE_MODE)
217 except EnvironmentError, err:
218 raise errors.HypervisorError("Creating console log file %s for"
219 " instance %s failed: %s" %
220 (console_log, instance.name, err))
221 out.append("lxc.console = %s" % console_log)
224 out.append("lxc.rootfs = %s" % root_dir)
226 # TODO: additional mounts, if we disable CAP_SYS_ADMIN
229 # deny direct device access
230 out.append("lxc.cgroup.devices.deny = a")
231 for devinfo in self._DEVS:
232 out.append("lxc.cgroup.devices.allow = %s rw" % devinfo)
235 for idx, nic in enumerate(instance.nics):
236 out.append("# NIC %d" % idx)
237 mode = nic.nicparams[constants.NIC_MODE]
238 link = nic.nicparams[constants.NIC_LINK]
239 if mode == constants.NIC_MODE_BRIDGED:
240 out.append("lxc.network.type = veth")
241 out.append("lxc.network.link = %s" % link)
243 raise errors.HypervisorError("LXC hypervisor only supports"
244 " bridged mode (NIC %d has mode %s)" %
246 out.append("lxc.network.hwaddr = %s" % nic.mac)
247 out.append("lxc.network.flags = up")
250 for cap in self._DENIED_CAPABILITIES:
251 out.append("lxc.cap.drop = %s" % cap)
253 return "\n".join(out) + "\n"
255 def StartInstance(self, instance, block_devices):
256 """Start an instance.
258 For LCX, we try to mount the block device and execute 'lxc-start
259 start' (we use volatile containers).
262 root_dir = self._InstanceDir(instance.name)
264 utils.EnsureDirs([(root_dir, self._DIR_MODE)])
265 except errors.GenericError, err:
266 raise HypervisorError("Cannot create instance directory: %s", str(err))
268 conf_file = self._InstanceConfFile(instance.name)
269 utils.WriteFile(conf_file, data=self._CreateConfigFile(instance, root_dir))
271 if not os.path.ismount(root_dir):
272 if not block_devices:
273 raise HypervisorError("LXC needs at least one disk")
275 sda_dev_path = block_devices[0][1]
276 result = utils.RunCmd(["mount", sda_dev_path, root_dir])
278 raise HypervisorError("Can't mount the chroot dir: %s" % result.output)
279 # TODO: replace the global log file with a per-instance log file
280 result = utils.RunCmd(["lxc-start", "-n", instance.name,
281 "-o", self._LOG_FILE, "-l", "DEBUG",
282 "-f", conf_file, "-d"])
284 raise HypervisorError("Running the lxc-start script failed: %s" %
287 def StopInstance(self, instance, force=False, retry=False, name=None):
290 This method has complicated cleanup tests, as we must:
291 - try to kill all leftover processes
292 - try to unmount any additional sub-mountpoints
293 - finally unmount the instance dir
299 root_dir = self._InstanceDir(name)
300 if not os.path.exists(root_dir):
303 if name in self.ListInstances():
304 # Signal init to shutdown; this is a hack
305 if not retry and not force:
306 result = utils.RunCmd(["chroot", root_dir, "poweroff"])
308 raise HypervisorError("Can't run 'poweroff' for the instance: %s" %
311 result = utils.RunCmd(["lxc-stop", "-n", name])
313 logging.warning("Error while doing lxc-stop for %s: %s", name,
316 for mpath in self._GetMountSubdirs(root_dir):
317 result = utils.RunCmd(["umount", mpath])
319 logging.warning("Error while umounting subpath %s for instance %s: %s",
320 mpath, name, result.output)
322 result = utils.RunCmd(["umount", root_dir])
323 if result.failed and force:
324 msg = ("Processes still alive in the chroot: %s" %
325 utils.RunCmd("fuser -vm %s" % root_dir).output)
327 raise HypervisorError("Can't umount the chroot dir: %s (%s)" %
328 (result.output, msg))
330 def RebootInstance(self, instance):
331 """Reboot an instance.
333 This is not (yet) implemented (in Ganeti) for the LXC hypervisor.
336 # TODO: implement reboot
337 raise HypervisorError("The LXC hypervisor doesn't implement the"
338 " reboot functionality")
340 def GetNodeInfo(self):
341 """Return information about the node.
343 This is just a wrapper over the base GetLinuxNodeInfo method.
345 @return: a dict with the following keys (values in MiB):
346 - memory_total: the total memory size on the node
347 - memory_free: the available memory on the node for instances
348 - memory_dom0: the memory used by the node itself, if available
351 return self.GetLinuxNodeInfo()
354 def GetShellCommandForConsole(cls, instance, hvparams, beparams):
355 """Return a command for connecting to the console of an instance.
358 return "lxc-console -n %s" % instance.name
361 """Verify the hypervisor.
363 For the chroot manager, it just checks the existence of the base dir.
366 if not os.path.exists(self._ROOT_DIR):
367 return "The required directory '%s' does not exist." % self._ROOT_DIR
370 def PowercycleNode(cls):
371 """LXC powercycle, just a wrapper over Linux powercycle.
374 cls.LinuxPowercycle()
376 def MigrateInstance(self, instance, target, live):
377 """Migrate an instance.
379 @type instance: L{objects.Instance}
380 @param instance: the instance to be migrated
382 @param target: hostname (usually ip) of the target node
384 @param live: whether to do a live or non-live migration
387 raise HypervisorError("Migration not supported by the LXC hypervisor")