服务器之家:专注于服务器技术及软件下载分享
分类导航

云服务器|WEB服务器|FTP服务器|邮件服务器|虚拟主机|服务器安全|DNS服务器|服务器知识|Nginx|IIS|Tomcat|

服务器之家 - 服务器技术 - 服务器知识 - OpenStack之虚机热迁移的代码详细解析

OpenStack之虚机热迁移的代码详细解析

2020-12-21 16:49尛鱼 服务器知识

本篇文章主要介绍了OpenStack之虚机热迁移,有需要的朋友可以了解下。

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
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
@wsgi.action('os-migrateLive')
  def _migrate_live(self, req, id, body):
    """Permit admins to (live) migrate a server to a new host."""
    context = req.environ["nova.context"]
    authorize(context, 'migrateLive')
 
    try:
      block_migration = body["os-migrateLive"]["block_migration"]
      disk_over_commit = body["os-migrateLive"]["disk_over_commit"]
      host = body["os-migrateLive"]["host"]
    except (TypeError, KeyError):
      msg = _("host, block_migration and disk_over_commit must "
          "be specified for live migration.")
      raise exc.HTTPBadRequest(explanation=msg)
 
    try:
      block_migration = strutils.bool_from_string(block_migration,
                            strict=True)
      disk_over_commit = strutils.bool_from_string(disk_over_commit,
                             strict=True)
    except ValueError as err:
      raise exc.HTTPBadRequest(explanation=str(err))
 
    try:
      instance = self.compute_api.get(context, id, want_objects=True)
      self.compute_api.live_migrate(context, instance, block_migration,
                     disk_over_commit, host)
    except (exception.ComputeServiceUnavailable,
        exception.InvalidHypervisorType,
        exception.UnableToMigrateToSelf,
        exception.DestinationHypervisorTooOld,
        exception.NoValidHost,
        exception.InvalidLocalStorage,
        exception.InvalidSharedStorage,
        exception.MigrationPreCheckError) as ex:
      raise exc.HTTPBadRequest(explanation=ex.format_message())
    except exception.InstanceNotFound as e:
      raise exc.HTTPNotFound(explanation=e.format_message())
    except exception.InstanceInvalidState as state_error:
      common.raise_http_conflict_for_instance_invalid_state(state_error,
          'os-migrateLive')
    except Exception:
      if host is None:
        msg = _("Live migration of instance %s to another host "
            "failed") % id
      else:
        msg = _("Live migration of instance %(id)s to host %(host)s "
            "failed") % {'id': id, 'host': host}
      LOG.exception(msg)
      # Return messages from scheduler
      raise exc.HTTPBadRequest(explanation=msg)
 
    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
3
4
5
6
7
{
  "os-migrateLive": {
    "host": "0443e9a1254044d8b99f35eace132080",
    "block_migration": false,
    "disk_over_commit": false
  }
}

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
2
self.compute_api.live_migrate(context, instance, block_migration,
                 disk_over_commit, host)

由这句进入到nova/compute/api.py中,源码如下:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
@check_instance_cell
  @check_instance_state(vm_state=[vm_states.ACTIVE])
  def live_migrate(self, context, instance, block_migration,
           disk_over_commit, host_name):
    """Migrate a server lively to a new host."""
    LOG.debug(_("Going to try to live migrate instance to %s"),
         host_name or "another host", instance=instance)
 
    instance.task_state = task_states.MIGRATING
    instance.save(expected_task_state=[None])
 
    self.compute_task_api.live_migrate_instance(context, instance,
        host_name, block_migration=block_migration,
        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
2
3
4
5
6
def live_migrate_instance(self, context, instance, host_name,
                block_migration, disk_over_commit):
     scheduler_hint = {'host': host_name}
     self._manager.migrate_server(
       context, instance, scheduler_hint, True, False, None,
       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
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
def migrate_server(self, context, instance, scheduler_hint, live, rebuild,
      flavor, block_migration, disk_over_commit, reservations=None):
    if instance and not isinstance(instance, instance_obj.Instance):
      # NOTE(danms): Until v2 of the RPC API, we need to tolerate
      # old-world instance objects here
      attrs = ['metadata', 'system_metadata', 'info_cache',
           'security_groups']
      instance = instance_obj.Instance._from_db_object(
        context, instance_obj.Instance(), instance,
        expected_attrs=attrs)
    if live and not rebuild and not flavor:
      self._live_migrate(context, instance, scheduler_hint,
                block_migration, disk_over_commit)
    elif not live and not rebuild and flavor:
      instance_uuid = instance['uuid']
      with compute_utils.EventReporter(context, self.db,
                     'cold_migrate', instance_uuid):
        self._cold_migrate(context, instance, flavor,
                  scheduler_hint['filter_properties'],
                  reservations)
    else:
      raise NotImplementedError()

由于在nova/conductor/api.py中传过来的参数是

?
1
2
3
self._manager.migrate_server(
       context, instance, scheduler_hint, True, False, None,
       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
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
if live and not rebuild and not flavor:
       self._live_migrate(context, instance, scheduler_hint,
                block_migration, disk_over_commit)
 _live_migrate代码如下:
def _live_migrate(self, context, instance, scheduler_hint,
           block_migration, disk_over_commit):
    destination = scheduler_hint.get("host")
    try:
      live_migrate.execute(context, instance, destination,
               block_migration, disk_over_commit)
    except (exception.NoValidHost,
        exception.ComputeServiceUnavailable,
        exception.InvalidHypervisorType,
        exception.InvalidCPUInfo,
        exception.UnableToMigrateToSelf,
        exception.DestinationHypervisorTooOld,
        exception.InvalidLocalStorage,
        exception.InvalidSharedStorage,
        exception.HypervisorUnavailable,
        exception.MigrationPreCheckError) as ex:
      with excutils.save_and_reraise_exception():
        #TODO(johngarbutt) - eventually need instance actions here
        request_spec = {'instance_properties': {
          'uuid': instance['uuid'], },
        }
        scheduler_utils.set_vm_state_and_notify(context,
            'compute_task', 'migrate_server',
            dict(vm_state=instance['vm_state'],
               task_state=None,
               expected_task_state=task_states.MIGRATING,),
            ex, request_spec, self.db)
    except Exception as ex:
      LOG.error(_('Migration of instance %(instance_id)s to host'
            ' %(dest)s unexpectedly failed.'),
            {'instance_id': instance['uuid'], 'dest': destination},
            exc_info=True)
      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
2
3
4
5
6
7
8
def execute(context, instance, destination,
      block_migration, disk_over_commit):
  task = LiveMigrationTask(context, instance,
               destination,
               block_migration,
               disk_over_commit)
  #TODO(johngarbutt) create a superclass that contains a safe_execute call
  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
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
def execute(self):
    self._check_instance_is_running()
    self._check_host_is_up(self.source)
 
    if not self.destination:
      self.destination = self._find_destination()
    else:
      self._check_requested_destination()
 
    #TODO(johngarbutt) need to move complexity out of compute manager
    return self.compute_rpcapi.live_migration(self.context,
        host=self.source,
        instance=self.instance,
        dest=self.destination,
        block_migration=self.block_migration,
        migrate_data=self.migrate_data)
        #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
2
3
4
5
6
7
8
9
def live_migration(self, ctxt, instance, dest, block_migration, host,
            migrate_data=None):
    # NOTE(russellb) Havana compat
    version = self._get_compat_version('3.0', '2.0')
    instance_p = jsonutils.to_primitive(instance)
    cctxt = self.client.prepare(server=host, version=version)
    cctxt.cast(ctxt, 'live_migration', instance=instance_p,
          dest=dest, block_migration=block_migration,
          migrate_data=migrate_data)

热迁移开始执行:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
def live_migration(self, context, instance, dest,
            post_method, recover_method, block_migration=False,
            migrate_data=None):
    """Spawning live_migration operation for distributing high-load.
 
    :param context: security context
    :param instance:
      nova.db.sqlalchemy.models.Instance object
      instance object that is migrated.
    :param dest: destination host
    :param post_method:
      post operation method.
      expected nova.compute.manager.post_live_migration.
    :param recover_method:
      recovery method when any exception occurs.
      expected nova.compute.manager.recover_live_migration.
    :param block_migration: if true, do block migration.
    :param migrate_data: implementation specific params
 
    """
 
    greenthread.spawn(self._live_migration, context, instance, dest,
             post_method, recover_method, block_migration,
             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
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
def _live_migration(self, context, instance, dest, post_method,
            recover_method, block_migration=False,
            migrate_data=None):
    """Do live migration.
 
    :param context: security context
    :param instance:
      nova.db.sqlalchemy.models.Instance object
      instance object that is migrated.
    :param dest: destination host
    :param post_method:
      post operation method.
      expected nova.compute.manager.post_live_migration.
    :param recover_method:
      recovery method when any exception occurs.
      expected nova.compute.manager.recover_live_migration.
    :param block_migration: if true, do block migration.
    :param migrate_data: implementation specific params
    """
 
    # Do live migration.
    try:
      if block_migration:
        flaglist = CONF.libvirt.block_migration_flag.split(',')
      else:
        flaglist = CONF.libvirt.live_migration_flag.split(',')
      flagvals = [getattr(libvirt, x.strip()) for x in flaglist]
      logical_sum = reduce(lambda x, y: x | y, flagvals)
 
      dom = self._lookup_by_name(instance["name"])
      dom.migrateToURI(CONF.libvirt.live_migration_uri % dest,
               logical_sum,
               None,
               CONF.libvirt.live_migration_bandwidth)
 
    except Exception as e:
      with excutils.save_and_reraise_exception():
        LOG.error(_("Live Migration failure: %s"), e,
             instance=instance)
        recover_method(context, instance, dest, block_migration)
 
    # Waiting for completion of live_migration.
    timer = loopingcall.FixedIntervalLoopingCall(f=None)
?
1
2
if block_migration:
         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
2
3
4
else:
         flaglist = CONF.libvirt.live_migration_flag.split(',')
       flagvals = [getattr(libvirt, x.strip()) for x in flaglist]
      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。

以上就是本文的全部内容,希望对大家的学习有所帮助,也希望大家多多支持服务器之家。 

延伸 · 阅读

精彩推荐