OpenStack之虚机热迁移代码解析
话说虚机迁移分为冷迁移以及热迁移，所谓热迁移用度娘的话说即是：热迁移（Live Migration，又叫动态迁移、实时迁移），即虚机保存/恢复(Save/Restore)：将整个虚拟机的运行状态完整保存下来，同时可以快速的恢复到原有硬件平台甚至是不同硬件平台上。恢复以后，虚机仍旧平滑运行，用户不会察觉到任何差异。OpenStack的虚机迁移是基于Libvirt实现的，下面来看看Openstack虚机热迁移的具体代码实现。
首先，由API入口进入到nova/api/openstack/compute/contrib/admin_actions.py
1 @wsgi.action(‘os-migrateLive‘)
2 def _migrate_live(self, req, id, body):
3 """Permit admins to (live) migrate a server to a new host."""
4 context = req.environ["nova.context"]
5 authorize(context, ‘migrateLive‘)
6
7 try:
8 block_migration = body["os-migrateLive"]["block_migration"]
9 disk_over_commit = body["os-migrateLive"]["disk_over_commit"]
10 host = body["os-migrateLive"]["host"]
11 except (TypeError, KeyError):
12 msg = _("host, block_migration and disk_over_commit must "
13 "be specified for live migration.")
14 raise exc.HTTPBadRequest(explanation=msg)
15
16 try:
17 block_migration = strutils.bool_from_string(block_migration,
18 strict=True)
19 disk_over_commit = strutils.bool_from_string(disk_over_commit,
20 strict=True)
21 except ValueError as err:
22 raise exc.HTTPBadRequest(explanation=str(err))
23
24 try:
25 instance = self.compute_api.get(context, id, want_objects=True)
26 self.compute_api.live_migrate(context, instance, block_migration,
27 disk_over_commit, host)
28 except (exception.ComputeServiceUnavailable,
29 exception.InvalidHypervisorType,
30 exception.UnableToMigrateToSelf,
31 exception.DestinationHypervisorTooOld,
32 exception.NoValidHost,
33 exception.InvalidLocalStorage,
34 exception.InvalidSharedStorage,
35 exception.MigrationPreCheckError) as ex:
36 raise exc.HTTPBadRequest(explanation=ex.format_message())
37 except exception.InstanceNotFound as e:
38 raise exc.HTTPNotFound(explanation=e.format_message())
39 except exception.InstanceInvalidState as state_error:
40 common.raise_http_conflict_for_instance_invalid_state(state_error,
41 ‘os-migrateLive‘)
42 except Exception:
43 if host is None:
44 msg = _("Live migration of instance %s to another host "
45 "failed") % id
46 else:
47 msg = _("Live migration of instance %(id)s to host %(host)s "
48 "failed") % {‘id‘: id, ‘host‘: host}
49 LOG.exception(msg)
50 # Return messages from scheduler
51 raise exc.HTTPBadRequest(explanation=msg)
52
53 return webob.Response(status_int=202)
这里第一行可以看到是与API文档的第二行照应的：
1 {
2 "os-migrateLive": {
3 "host": "0443e9a1254044d8b99f35eace132080",
4 "block_migration": false,
5 "disk_over_commit": false
6 }
7 }
好了，源码中其实执行迁移工作的就是第26、27行的一条语句：
1 self.compute_api.live_migrate(context, instance, block_migration, 2 disk_over_commit, host)
由这句进入到nova/compute/api.py中，源码如下：
1 @check_instance_cell
2 @check_instance_state(vm_state=[vm_states.ACTIVE])
3 def live_migrate(self, context, instance, block_migration,
4 disk_over_commit, host_name):
5 """Migrate a server lively to a new host."""
6 LOG.debug(_("Going to try to live migrate instance to %s"),
7 host_name or "another host", instance=instance)
8
9 instance.task_state = task_states.MIGRATING
10 instance.save(expected_task_state=[None])
11
12 self.compute_task_api.live_migrate_instance(context, instance,
13 host_name, block_migration=block_migration,
14 disk_over_commit=disk_over_commit)
第2行是一个装饰器，用于在进入API方法之前，检测虚拟机和/或任务的状态， 如果实例处于错误的状态，将会引发异常；接下来实时迁移虚机到新的主机，并将虚机状态置于“migrating”,然后由12行进入nova/conductor/api.py
1 def live_migrate_instance(self, context, instance, host_name,
2 block_migration, disk_over_commit):
3 scheduler_hint = {‘host‘: host_name}
4 self._manager.migrate_server(
5 context, instance, scheduler_hint, True, False, None,
6 block_migration, disk_over_commit, None)
将主机名存入字典scheduler_hint中，然后调用nova/conductor/manager.py方法migrate_server,
1 def migrate_server(self, context, instance, scheduler_hint, live, rebuild, 2 flavor, block_migration, disk_over_commit, reservations=None): 3 if instance and not isinstance(instance, instance_obj.Instance): 4 # NOTE(danms): Until v2 of the RPC API, we need to tolerate 5 # old-world instance objects here 6 attrs = [‘metadata‘, ‘system_metadata‘, ‘info_cache‘, 7 ‘security_groups‘] 8 instance = instance_obj.Instance._from_db_object( 9 context, instance_obj.Instance(), instance, 10 expected_attrs=attrs) 11 if live and not rebuild and not flavor: 12 self._live_migrate(context, instance, scheduler_hint, 13 block_migration, disk_over_commit) 14 elif not live and not rebuild and flavor: 15 instance_uuid = instance[‘uuid‘] 16 with compute_utils.EventReporter(context, self.db, 17 ‘cold_migrate‘, instance_uuid): 18 self._cold_migrate(context, instance, flavor, 19 scheduler_hint[‘filter_properties‘], 20 reservations) 21 else: 22 raise NotImplementedError()
由于在nova/conductor/api.py中传过来的参数是
1 self._manager.migrate_server( 2 context, instance, scheduler_hint, True, False, None, 3 block_migration, disk_over_commit, None)
因此live是True，rebuild是Flase,flavor是None，执行第12、13行代码：
1 if live and not rebuild and not flavor: 2 self._live_migrate(context, instance, scheduler_hint, 3 block_migration, disk_over_commit)
_live_migrate代码如下：
1 def _live_migrate(self, context, instance, scheduler_hint,
2 block_migration, disk_over_commit):
3 destination = scheduler_hint.get("host")
4 try:
5 live_migrate.execute(context, instance, destination,
6 block_migration, disk_over_commit)
7 except (exception.NoValidHost,
8 exception.ComputeServiceUnavailable,
9 exception.InvalidHypervisorType,
10 exception.InvalidCPUInfo,
11 exception.UnableToMigrateToSelf,
12 exception.DestinationHypervisorTooOld,
13 exception.InvalidLocalStorage,
14 exception.InvalidSharedStorage,
15 exception.HypervisorUnavailable,
16 exception.MigrationPreCheckError) as ex:
17 with excutils.save_and_reraise_exception():
18 #TODO(johngarbutt) - eventually need instance actions here
19 request_spec = {‘instance_properties‘: {
20 ‘uuid‘: instance[‘uuid‘], },
21 }
22 scheduler_utils.set_vm_state_and_notify(context,
23 ‘compute_task‘, ‘migrate_server‘,
24 dict(vm_state=instance[‘vm_state‘],
25 task_state=None,
26 expected_task_state=task_states.MIGRATING,),
27 ex, request_spec, self.db)
28 except Exception as ex:
29 LOG.error(_(‘Migration of instance %(instance_id)s to host‘
30 ‘ %(dest)s unexpectedly failed.‘),
31 {‘instance_id‘: instance[‘uuid‘], ‘dest‘: destination},
32 exc_info=True)
33 raise exception.MigrationError(reason=ex)
首先，第三行中将主机名赋给destination,然后执行迁移，后面的都是异常的捕捉，执行迁移的代码分为两部分，先看第一部分，在nova/conductor/tasks/live_migrate.py的184行左右：
1 def execute(context, instance, destination, 2 block_migration, disk_over_commit): 3 task = LiveMigrationTask(context, instance, 4 destination, 5 block_migration, 6 disk_over_commit) 7 #TODO(johngarbutt) create a superclass that contains a safe_execute call 8 return task.execute()
先创建包含安全执行回调的超类，然后返回如下函数也即执行迁移的第二部分代码，在54行左右：
1 def execute(self): 2 self._check_instance_is_running() 3 self._check_host_is_up(self.source) 4 5 if not self.destination: 6 self.destination = self._find_destination() 7 else: 8 self._check_requested_destination() 9 10 #TODO(johngarbutt) need to move complexity out of compute manager 11 return self.compute_rpcapi.live_migration(self.context, 12 host=self.source, 13 instance=self.instance, 14 dest=self.destination, 15 block_migration=self.block_migration, 16 migrate_data=self.migrate_data) 17 #TODO(johngarbutt) disk_over_commit?
这里有三部分内容：
- 如果目前主机不存在，则由调度算法选取一个目标主机，并且进行相关的检测，确保能够进行实时迁移操作；
- 如果目标主机存在，则直接进行相关的检测操作，确保能够进行实时迁移操作；
- 执行迁移操作。
前两部分不再赘述，直接看第三部分代码，在nova/compute/rpcapi.py中：
1 def live_migration(self, ctxt, instance, dest, block_migration, host, 2 migrate_data=None): 3 # NOTE(russellb) Havana compat 4 version = self._get_compat_version(‘3.0‘, ‘2.0‘) 5 instance_p = jsonutils.to_primitive(instance) 6 cctxt = self.client.prepare(server=host, version=version) 7 cctxt.cast(ctxt, ‘live_migration‘, instance=instance_p, 8 dest=dest, block_migration=block_migration, 9 migrate_data=migrate_data)
热迁移开始执行：
1 def live_migration(self, context, instance, dest, 2 post_method, recover_method, block_migration=False, 3 migrate_data=None): 4 """Spawning live_migration operation for distributing high-load. 5 6 :param context: security context 7 :param instance: 8 nova.db.sqlalchemy.models.Instance object 9 instance object that is migrated. 10 :param dest: destination host 11 :param post_method: 12 post operation method. 13 expected nova.compute.manager.post_live_migration. 14 :param recover_method: 15 recovery method when any exception occurs. 16 expected nova.compute.manager.recover_live_migration. 17 :param block_migration: if true, do block migration. 18 :param migrate_data: implementation specific params 19 20 """ 21 22 greenthread.spawn(self._live_migration, context, instance, dest, 23 post_method, recover_method, block_migration, 24 migrate_data)
这个方法中建立一个绿色线程来运行方法_live_migration，来执行实时迁移； 主要是调用libvirt python接口方法virDomainMigrateToURI，来实现从当前主机迁移domain对象到给定的目标主机；
spawn：建立一个绿色线程来运行方法“func(*args, **kwargs)”，这里就是来运行方法_live_migration；
_live_migration：执行实时迁移； 主要是调用libvirt python接口方法virDomainMigrateToURI，来实现从当前主机迁移domain对象到给定的目标主机；
接着在绿色线程中调用_live_migration方法：
1 def _live_migration(self, context, instance, dest, post_method,
2 recover_method, block_migration=False,
3 migrate_data=None):
4 """Do live migration.
5
6 :param context: security context
7 :param instance:
8 nova.db.sqlalchemy.models.Instance object
9 instance object that is migrated.
10 :param dest: destination host
11 :param post_method:
12 post operation method.
13 expected nova.compute.manager.post_live_migration.
14 :param recover_method:
15 recovery method when any exception occurs.
16 expected nova.compute.manager.recover_live_migration.
17 :param block_migration: if true, do block migration.
18 :param migrate_data: implementation specific params
19 """
20
21 # Do live migration.
22 try:
23 if block_migration:
24 flaglist = CONF.libvirt.block_migration_flag.split(‘,‘)
25 else:
26 flaglist = CONF.libvirt.live_migration_flag.split(‘,‘)
27 flagvals = [getattr(libvirt, x.strip()) for x in flaglist]
28 logical_sum = reduce(lambda x, y: x | y, flagvals)
29
30 dom = self._lookup_by_name(instance["name"])
31 dom.migrateToURI(CONF.libvirt.live_migration_uri % dest,
32 logical_sum,
33 None,
34 CONF.libvirt.live_migration_bandwidth)
35
36 except Exception as e:
37 with excutils.save_and_reraise_exception():
38 LOG.error(_("Live Migration failure: %s"), e,
39 instance=instance)
40 recover_method(context, instance, dest, block_migration)
41
42 # Waiting for completion of live_migration.
43 timer = loopingcall.FixedIntervalLoopingCall(f=None)
1 if block_migration: 2 flaglist = CONF.libvirt.block_migration_flag.split(‘,‘)
这个获取块迁移标志列表，block_migration_flag：这个参数定义了为块迁移设置迁移标志。
1 else: 2 flaglist = CONF.libvirt.live_migration_flag.split(‘,‘) 3 flagvals = [getattr(libvirt, x.strip()) for x in flaglist] 4 logical_sum = reduce(lambda x, y: x | y, flagvals)
这部分获取实时迁移标志列表，live_migration_flag这个参数定义了实时迁移的迁移标志。
1 dom = self._lookup_by_name(instance["name"])
根据给定的实例名称检索libvirt域对象。
1 timer = loopingcall.FixedIntervalLoopingCall(f=None)
获取等待完成实时迁移的时间。
热迁移代码部分至此结束。
时间： 2024-08-04 09:45:48