openstack虚拟机启动过程源码分析

源码版本：H版

　　以nova-api为起点开始分析！ 

一、在nova-api进程中进行处理

　　根据对nova api的分析，当请求发过来的时候，由相应的Controller进行处理，此处如下：

nova/api/openstack/compute/servers.py

Controller类：
@wsgi.response(202)
@wsgi.serializers(xml=FullServerTemplate)
@wsgi.deserializers(xml=CreateDeserializer)
def create(self, req, body):
    try:
        _get_inst_type = flavors.get_flavor_by_flavor_id
        inst_type = _get_inst_type(flavor_id, ctxt=context,
                                   read_deleted="no")

        (instances, resv_id) = self.compute_api.create(context,
                        ...
                        legacy_bdm=legacy_bdm)
        ...

　　这里需要发现self.compute_api究竟为哪个类的对象。从Controller类的__init__函数中发现：

"""compute指nova.compute"""
self.compute_api = compute.API()

nova/compute/__init__.py

def API(*args, **kwargs):
    importutils = nova.openstack.common.importutils
    """此处class_name为nova.compute.api.API"""
    class_name = _get_compute_api_class_name()
    """创建nova.compute.api.API类对象并返回"""
    return importutils.import_object(class_name, *args, **kwargs)

　　所以self.compute_api即为nova.compute.api.API类的对象。 

1、接着看self.compute_api的create函数调用

nova/compute/api.py

API类:
@hooks.add_hook("create_instance")
def create(self, context, instance_type,
            ...):
    ...
    return self._create_instance(
                           ...)

def _create_instance(self, context, instance_type,
               image_href, kernel_id, ramdisk_id,
               min_count, max_count,
               display_name, display_description,
               key_name, key_data, security_groups,
               availability_zone, user_data, metadata,
               injected_files, admin_password,
               access_ip_v4, access_ip_v6,
               requested_networks, config_drive,
               block_device_mapping, auto_disk_config,
               reservation_id=None, scheduler_hints=None,
               legacy_bdm=True):
    ...

    self.compute_task_api.build_instances(context,
            instances=instances, image=boot_meta,
            filter_properties=filter_properties,
            admin_password=admin_password,
            injected_files=injected_files,
            requested_networks=requested_networks,
            security_groups=security_groups,
            block_device_mapping=block_device_mapping,
            legacy_bdm=False)

    return (instances, reservation_id)

　　这里需要分析self.compute_task_api为哪个类的对象，代码分析如下：

nova/compute/api.py

API类：
@property
def compute_task_api(self):
    if self._compute_task_api is None:
        from nova import conductor
        self._compute_task_api = conductor.ComputeTaskAPI()
    return self._compute_task_api

nova/conductor/__init__.py

def ComputeTaskAPI(*args, **kwargs):
    use_local = kwargs.pop('use_local', False)
　　if oslo.config.cfg.CONF.conductor.use_local or use_local:
　　　　api = conductor_api.LocalComputeTaskAPI
　　else:
　　　　"""conductor_api为nova.conductor.api"""
　　　　api = conductor_api.ComputeTaskAPI
　　return api(*args, **kwargs)

　　所以self.compute_task_api为nova.conductor.api.ComputeTaskAPI类的对象。 

2、接着看self.compute_task_api的build_instances函数

nova/conductor/api.py

ComputeTaskAPI类：
def build_instances(self, context, instances, image, filter_properties,
            admin_password, injected_files, requested_networks,
            security_groups, block_device_mapping, legacy_bdm=True):
    self.conductor_compute_rpcapi.build_instances(context,
            instances=instances, image=image,
            filter_properties=filter_properties,
            admin_password=admin_password, injected_files=injected_files,
            requested_networks=requested_networks,
            security_groups=security_groups,
            block_device_mapping=block_device_mapping,
            legacy_bdm=legacy_bdm)

　　其中，self.conductor_compute_rpcapi= rpcapi.ComputeTaskAPI() ! 

3、接着看self.conductor_compute_rpcapi的build_instances函数

nova/conductor/rpcapi.py

ComputeTaskAPI类：
def build_instances(self, context, instances, image, filter_properties,
            admin_password, injected_files, requested_networks,
            security_groups, block_device_mapping, legacy_bdm=True):
    instances_p = [jsonutils.to_primitive(inst) for inst in instances]
    image_p = jsonutils.to_primitive(image)
    cctxt = self.client.prepare(version='1.5')
    cctxt.cast(context, 'build_instances',
               instances=instances_p, image=image_p,
               filter_properties=filter_properties,
               admin_password=admin_password,
               injected_files=injected_files,
               requested_networks=requested_networks,
               security_groups=security_groups,
               block_device_mapping=block_device_mapping,
               legacy_bdm=legacy_bdm)

　　其中self.client = self.get_client(namespace=self.RPC_API_NAMESPACE)，即 cctxt = self.client.prepare(version='1.5')调用后cctxt为nova.rpcclient.RPCClient，根据http://www.cnblogs.com/littlebugfish/p/4058007.html中简述的消息队列调用原理，这里将转到nova-conductor进程中进行处理。

 

二、在nova-conductor进程中进行处理

nova/conductor/manager.py

ComputeTaskManager类：
def build_instances(self, context, instances, image, filter_properties,
            admin_password, injected_files, requested_networks,
            security_groups, block_device_mapping, legacy_bdm=True):
    ...
    self.scheduler_rpcapi.run_instance(context, request_spec=request_spec,
            admin_password=admin_password, injected_files=injected_files,
            requested_networks=requested_networks, is_first_time=True,
            filter_properties=filter_properties,
            legacy_bdm_in_spec=legacy_bdm)

　　其中，self.scheduler_rpcapi = scheduler_rpcapi.SchedulerAPI(),而scheduler_rpcapi为nova.scheduler.rpcapi,则self.scheduler_rpcapi为nova.scheduler.rpcapi.SchedulerAPI类的对象，接着看代码如下：

nova/scheduler/rpcapi.py

SchedulerAPI类:
def run_instance(self, ctxt, request_spec, admin_password,
            injected_files, requested_networks, is_first_time,
            filter_properties, legacy_bdm_in_spec=True):
    version = '2.0'
    msg_kwargs = {'request_spec': request_spec,
                  'admin_password': admin_password,
                  'injected_files': injected_files,
                  'requested_networks': requested_networks,
                  'is_first_time': is_first_time,
                  'filter_properties': filter_properties}
    if self.client.can_send_version('2.9'):
        version = '2.9'
        msg_kwargs['legacy_bdm_in_spec'] = legacy_bdm_in_spec
    cctxt = self.client.prepare(version=version)
    return cctxt.cast(ctxt, 'run_instance', **msg_kwargs)

　　根据http://www.cnblogs.com/littlebugfish/p/4058007.html中简述的消息队列调用原理，这里将转到nova-scheduler进程中进行处理。

 

三、在nova-scheduler进程中进行处理

nova/scheduler/manager.py

SchedulerManager类：
def run_instance(self, context, request_spec, admin_password,
            injected_files, requested_networks, is_first_time,
            filter_properties, legacy_bdm_in_spec=True):
        
    instance_uuids = request_spec['instance_uuids']
    with compute_utils.EventReporter(context, conductor_api.LocalAPI(),
                                     'schedule', *instance_uuids):
        try:
            return self.driver.schedule_run_instance(context,
                    request_spec, admin_password, injected_files,
                    requested_networks, is_first_time, filter_properties,
                    legacy_bdm_in_spec)

        ...

　　这里需要看self.driver究竟为哪个类的对象。在SchedulerManager类的__init__函数中，有如下代码：

if not scheduler_driver:
　　scheduler_driver = CONF.scheduler_driver
self.driver = importutils.import_object(scheduler_driver)

　　根据配置，CONF.scheduler_driver为nova.scheduler.filter_scheduler.FilterScheduler，所以self.driver为nova.scheduler.filter_scheduler.FilterScheduler类的对象，所以接下来看：

nova.scheduler.filter_scheduler.py

FilterScheduler类：
def schedule_run_instance(self, context, request_spec,
                              admin_password, injected_files,
                              requested_networks, is_first_time,
                              filter_properties, legacy_bdm_in_spec):
    ...
    """选用来创建虚拟机的主机"""
    weighed_hosts = self._schedule(context, request_spec,
                                   filter_properties, instance_uuids)
    ...

    for num, instance_uuid in enumerate(instance_uuids):
        request_spec['instance_properties']['launch_index'] = num

        try:
            try:
                weighed_host = weighed_hosts.pop(0)
                LOG.info(_("Choosing host %(weighed_host)s "
                            "for instance %(instance_uuid)s"),
                          {'weighed_host': weighed_host,
                           'instance_uuid': instance_uuid})
            except IndexError:
                raise exception.NoValidHost(reason="")
            """创建虚拟机"""
            self._provision_resource(context, weighed_host,
                                     request_spec,
                                     filter_properties,
                                     requested_networks,
                                     injected_files, admin_password,
                                     is_first_time,
                                     instance_uuid=instance_uuid,
                                     legacy_bdm_in_spec=legacy_bdm_in_spec)
        except Exception as ex:
            driver.handle_schedule_error(context, ex, instance_uuid,
                                         request_spec)
        retry = filter_properties.get('retry', {})
        retry['hosts'] = []
    self.notifier.info(context, 'scheduler.run_instance.end', payload) 

1、 选择用来创建虚拟机的主机，代码如下：

weighed_hosts = self._schedule(context, request_spec,
                               filter_properties, instance_uuids)

　　接着看_schedule函数，如下：

def _schedule(self, context, request_spec, filter_properties,
                  instance_uuids=None):
    ...
    hosts = self.host_manager.get_all_host_states(elevated)

    selected_hosts = []
    if instance_uuids:
        num_instances = len(instance_uuids)
    else:
        num_instances = request_spec.get('num_instances', 1)
    for num in xrange(num_instances):
        """根据规则过滤掉特定的主机。例如可以设定忽略某些主机，这样就不会选择这些主机来创建虚拟机"""
        hosts = self.host_manager.get_filtered_hosts(hosts,
                filter_properties, index=num)
        if not hosts:
            break

        LOG.debug(_("Filtered %(hosts)s"), {'hosts': hosts})
        """根据规则为候选主机分配权重以便选取最佳的主机创建虚拟机，此处暂时不详述权重分配的过程"""
        weighed_hosts = self.host_manager.get_weighed_hosts(hosts,
                filter_properties)

        LOG.debug(_("Weighed %(hosts)s"), {'hosts': weighed_hosts})

        scheduler_host_subset_size = CONF.scheduler_host_subset_size
        if scheduler_host_subset_size > len(weighed_hosts):
            scheduler_host_subset_size = len(weighed_hosts)
        if scheduler_host_subset_size < 1:
            scheduler_host_subset_size = 1

        chosen_host = random.choice(
            weighed_hosts[0:scheduler_host_subset_size])
        selected_hosts.append(chosen_host)
        chosen_host.obj.consume_from_instance(instance_properties)
        if update_group_hosts is True:
            filter_properties['group_hosts'].append(chosen_host.obj.host)
    return selected_hosts 

2、开始创建虚拟机，代码如下：

self._provision_resource(context, weighed_host,
                                 request_spec,
                                 filter_properties,
                                 requested_networks,
                                 injected_files, admin_password,
                                 is_first_time,
                                 instance_uuid=instance_uuid,
                                 legacy_bdm_in_spec=legacy_bdm_in_spec)

　　接着看_provision_resource函数，如下：

def _provision_resource(self, context, weighed_host, request_spec,
            filter_properties, requested_networks, injected_files,
            admin_password, is_first_time, instance_uuid=None,
            legacy_bdm_in_spec=True):
    ...
    try:
        """更新数据库"""
        updated_instance = driver.instance_update_db(context,
                instance_uuid, extra_values=values)

    except exception.InstanceNotFound:
        LOG.warning(_("Instance disappeared during scheduling"),
                    context=context, instance_uuid=instance_uuid)

    else:
        scheduler_utils.populate_filter_properties(filter_properties,
                weighed_host.obj)

        """这里self.compute_rpcapi为 nova.compute.rpcapi.ComputeAPI对象"""
        self.compute_rpcapi.run_instance(context,
                instance=updated_instance,
                host=weighed_host.obj.host,
                request_spec=request_spec,
                filter_properties=filter_properties,
                requested_networks=requested_networks,
                injected_files=injected_files,
                admin_password=admin_password, is_first_time=is_first_time,
                node=weighed_host.obj.nodename,
                legacy_bdm_in_spec=legacy_bdm_in_spec)

nova/compute/rpcapi.py

ComputeAPI类：
def run_instance(self, ctxt, instance, host, request_spec,
                     filter_properties, requested_networks,
                     injected_files, admin_password,
                     is_first_time, node=None, legacy_bdm_in_spec=True):
    instance_p = jsonutils.to_primitive(instance)
    msg_kwargs = {'instance': instance_p, 'request_spec': request_spec,
                  'filter_properties': filter_properties,
                  'requested_networks': requested_networks,
                  'injected_files': injected_files,
                  'admin_password': admin_password,
                  'is_first_time': is_first_time, 'node': node}

    if _icehouse_compat() or self.client.can_send_version('2.37'):
        version = _get_version('2.37')
        msg_kwargs['legacy_bdm_in_spec'] = legacy_bdm_in_spec
    else:
        version = '2.19'
    cctxt = self.client.prepare(server=host,
            version=version)
    cctxt.cast(ctxt, 'run_instance', **msg_kwargs)

　　根据http://www.cnblogs.com/littlebugfish/p/4058007.html中简述的消息队列调用原理，这里将转到nova-compute进程中进行处理。

 

四、在nova-compute进程中进行处理

nova/compute/manager.py

ComputeManager类：
@wrap_exception()
@reverts_task_state
@wrap_instance_event
@wrap_instance_fault
def run_instance(self, context, instance, request_spec=None,
                 filter_properties=None, requested_networks=None,
                 injected_files=None, admin_password=None,
                 is_first_time=False, node=None, legacy_bdm_in_spec=True):

    if filter_properties is None:
        filter_properties = {}

    @utils.synchronized(instance['uuid'])
    def do_run_instance():
        self._run_instance(context, request_spec,
                filter_properties, requested_networks, injected_files,
                admin_password, is_first_time, node, instance,
                legacy_bdm_in_spec)
    do_run_instance()

def _run_instance(self, context, request_spec,
                      filter_properties, requested_networks, injected_files,
                      admin_password, is_first_time, node, instance,
                      legacy_bdm_in_spec):
    ...

        instance, network_info = self._build_instance(context,
                request_spec, filter_properties, requested_networks,
                injected_files, admin_password, is_first_time, node,
                instance, image_meta, legacy_bdm_in_spec)
        notify("end", msg=_("Success"), network_info=network_info)
    ...

def _build_instance(self, context, request_spec, filter_properties,
            requested_networks, injected_files, admin_password, is_first_time,
            node, instance, image_meta, legacy_bdm_in_spec):
    context = context.elevated()

    ...
    try:
        limits = filter_properties.get('limits', {})
        with rt.instance_claim(context, instance, limits):
            ...

            instance = self._spawn(context, instance, image_meta,
                                   network_info, block_device_info,
                                   injected_files, admin_password,
                                   set_access_ip=set_access_ip)
    ...

    # spawn success
    return instance, network_info

def _spawn(self, context, instance, image_meta, network_info,
               block_device_info, injected_files, admin_password,
               set_access_ip=False):
    """Spawn an instance with error logging and update its power state."""
    instance = self._instance_update(context, instance['uuid'],
            vm_state=vm_states.BUILDING,
            task_state=task_states.SPAWNING,
            expected_task_state=task_states.BLOCK_DEVICE_MAPPING)
    try:
        self.driver.spawn(context, instance, image_meta,
                          injected_files, admin_password,
                          network_info,
                          block_device_info)
    ...
    return self._instance_update(context, instance['uuid'],
                                 **update_data)

　　这里将调用相应的Driver对象进行虚拟机的创建，此处以LibvirtDriver类对象为例，后续处理过程可以查看：http://www.cnblogs.com/littlebugfish/p/4058115.html中的分析！