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"
72 "c 1:8", # /dev/random
73 "c 1:9", # /dev/urandom
76 "c 5:1", # /dev/console
78 "c 136:*", # first block of Unix98 PTY slaves
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
94 hv_base.BaseHypervisor.__init__(self)
95 utils.EnsureDirs([(self._ROOT_DIR, self._DIR_MODE)])
98 def _GetMountSubdirs(path):
99 """Return the list of mountpoints under a given path.
103 for _, mountpoint, _, _ in utils.GetMounts():
104 if (mountpoint.startswith(path) and
106 result.append(mountpoint)
108 result.sort(key=lambda x: x.count("/"), reverse=True)
112 def _InstanceDir(cls, instance_name):
113 """Return the root directory for an instance.
116 return utils.PathJoin(cls._ROOT_DIR, instance_name)
119 def _InstanceConfFile(cls, instance_name):
120 """Return the configuration file for an instance.
123 return utils.PathJoin(cls._ROOT_DIR, instance_name + ".conf")
126 def _InstanceLogFile(cls, instance_name):
127 """Return the log file for an instance.
130 return utils.PathJoin(cls._ROOT_DIR, instance_name + ".log")
133 def _GetCgroupMountPoint(cls):
134 for _, mountpoint, fstype, _ in utils.GetMounts():
135 if fstype == "cgroup":
137 raise errors.HypervisorError("The cgroup filesystem is not mounted")
140 def _GetCgroupCpuList(cls, instance_name):
141 """Return the list of CPU ids for an instance.
144 cgroup = cls._GetCgroupMountPoint()
146 cpus = utils.ReadFile(utils.PathJoin(cgroup,
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
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))
164 def ListInstances(self):
165 """Get the list of running instances.
168 result = utils.RunCmd(["lxc-ls"])
170 raise errors.HypervisorError("Can't run lxc-ls: %s" % result.output)
171 return result.stdout.splitlines()
173 def GetInstanceInfo(self, instance_name):
174 """Get instance properties.
176 @type instance_name: string
177 @param instance_name: the instance name
179 @return: (name, id, memory, vcpus, stat, times)
182 # TODO: read container info from the cgroup mountpoint
184 result = utils.RunCmd(["lxc-info", "-n", instance_name])
186 raise errors.HypervisorError("Can't run lxc-info: %s" % result.output)
187 # lxc-info output examples:
188 # 'ganeti-lxc-test1' is STOPPED
189 # 'ganeti-lxc-test1' is RUNNING
190 _, state = result.stdout.rsplit(None, 1)
192 cpu_list = self._GetCgroupCpuList(instance_name)
194 if state == "RUNNING":
195 return (instance_name, 0, 0, len(cpu_list), 0, 0)
198 def GetAllInstancesInfo(self):
199 """Get properties of all instances.
201 @return: [(name, id, memory, vcpus, stat, times),...]
205 for name in self.ListInstances():
206 data.append(self.GetInstanceInfo(name))
209 def _CreateConfigFile(self, instance, root_dir):
210 """Create an lxc.conf file for an instance"""
213 out.append("lxc.utsname = %s" % instance.name)
215 # separate pseudo-TTY instances
216 out.append("lxc.pts = 255")
218 out.append("lxc.tty = 6")
220 console_log = utils.PathJoin(self._ROOT_DIR, instance.name + ".console")
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)
230 out.append("lxc.rootfs = %s" % root_dir)
232 # TODO: additional mounts, if we disable CAP_SYS_ADMIN
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)
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)
249 raise errors.HypervisorError("LXC hypervisor only supports"
250 " bridged mode (NIC %d has mode %s)" %
252 out.append("lxc.network.hwaddr = %s" % nic.mac)
253 out.append("lxc.network.flags = up")
256 for cap in self._DENIED_CAPABILITIES:
257 out.append("lxc.cap.drop = %s" % cap)
259 return "\n".join(out) + "\n"
261 def StartInstance(self, instance, block_devices):
262 """Start an instance.
264 For LCX, we try to mount the block device and execute 'lxc-start
265 start' (we use volatile containers).
268 root_dir = self._InstanceDir(instance.name)
270 utils.EnsureDirs([(root_dir, self._DIR_MODE)])
271 except errors.GenericError, err:
272 raise HypervisorError("Cannot create instance directory: %s", str(err))
274 conf_file = self._InstanceConfFile(instance.name)
275 utils.WriteFile(conf_file, data=self._CreateConfigFile(instance, root_dir))
277 log_file = self._InstanceLogFile(instance.name)
278 if not os.path.exists(log_file):
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))
286 if not os.path.ismount(root_dir):
287 if not block_devices:
288 raise HypervisorError("LXC needs at least one disk")
290 sda_dev_path = block_devices[0][1]
291 result = utils.RunCmd(["mount", sda_dev_path, root_dir])
293 raise HypervisorError("Can't mount the chroot dir: %s" % result.output)
294 result = utils.RunCmd(["lxc-start", "-n", instance.name,
297 "-f", conf_file, "-d"])
299 raise HypervisorError("Running the lxc-start script failed: %s" %
302 def StopInstance(self, instance, force=False, retry=False, name=None):
305 This method has complicated cleanup tests, as we must:
306 - try to kill all leftover processes
307 - try to unmount any additional sub-mountpoints
308 - finally unmount the instance dir
314 root_dir = self._InstanceDir(name)
315 if not os.path.exists(root_dir):
318 if name in self.ListInstances():
319 # Signal init to shutdown; this is a hack
320 if not retry and not force:
321 result = utils.RunCmd(["chroot", root_dir, "poweroff"])
323 raise HypervisorError("Can't run 'poweroff' for the instance: %s" %
326 result = utils.RunCmd(["lxc-stop", "-n", name])
328 logging.warning("Error while doing lxc-stop for %s: %s", name,
331 for mpath in self._GetMountSubdirs(root_dir):
332 result = utils.RunCmd(["umount", mpath])
334 logging.warning("Error while umounting subpath %s for instance %s: %s",
335 mpath, name, result.output)
337 result = utils.RunCmd(["umount", root_dir])
338 if result.failed and force:
339 msg = ("Processes still alive in the chroot: %s" %
340 utils.RunCmd("fuser -vm %s" % root_dir).output)
342 raise HypervisorError("Can't umount the chroot dir: %s (%s)" %
343 (result.output, msg))
345 def RebootInstance(self, instance):
346 """Reboot an instance.
348 This is not (yet) implemented (in Ganeti) for the LXC hypervisor.
351 # TODO: implement reboot
352 raise HypervisorError("The LXC hypervisor doesn't implement the"
353 " reboot functionality")
355 def GetNodeInfo(self):
356 """Return information about the node.
358 This is just a wrapper over the base GetLinuxNodeInfo method.
360 @return: a dict with the following keys (values in MiB):
361 - memory_total: the total memory size on the node
362 - memory_free: the available memory on the node for instances
363 - memory_dom0: the memory used by the node itself, if available
366 return self.GetLinuxNodeInfo()
369 def GetShellCommandForConsole(cls, instance, hvparams, beparams):
370 """Return a command for connecting to the console of an instance.
373 return "lxc-console -n %s" % instance.name
376 """Verify the hypervisor.
378 For the chroot manager, it just checks the existence of the base dir.
381 if not os.path.exists(self._ROOT_DIR):
382 return "The required directory '%s' does not exist." % self._ROOT_DIR
385 def PowercycleNode(cls):
386 """LXC powercycle, just a wrapper over Linux powercycle.
389 cls.LinuxPowercycle()
391 def MigrateInstance(self, instance, target, live):
392 """Migrate an instance.
394 @type instance: L{objects.Instance}
395 @param instance: the instance to be migrated
397 @param target: hostname (usually ip) of the target node
399 @param live: whether to do a live or non-live migration
402 raise HypervisorError("Migration not supported by the LXC hypervisor")