OpenStack源码分析 Neutron源码分析（一）-----------Restful API篇

原文：https://blog.csdn.net/happyanger6/article/details/54586463

首先，先分析WSGI应用的实现。

由前面的文章http://blog.csdn.net/happyanger6/article/details/54518491可知，WSGI应用的构建过程主要就是通过paste库的loadapp加载，因此核心就是分析这个过程。我们从neutron-server的起始代码开始逐步分析。

neutron-server的入口是:

neutron/cmd/eventlet/server/__init__.py:main

def main():
    server.boot_server(_main_neutron_server)

boot_server在neutron/server/__init__.py中，它主要的功能就是解析命令行指定的配置文件，一般是"--config-file=/etc/neutron/neutron.conf"，然后就执行_main_neutron_server。

neutron/cmd/eventlet/server/__init__.py::_main_neutron_serve

def _main_neutron_server():
    if cfg.CONF.web_framework == 'legacy':
        wsgi_eventlet.eventlet_wsgi_server()
    else:
        wsgi_pecan.pecan_wsgi_server()

可以看到，接下来根据配置文件中配置的web框架方式，决定如何启动wsgi_server,传统的方式是通过eventlet，现在又新加入了pecan方式。默认情况下，还是使用的eventlet方式，因此接着分析eventlet_wsig_server。这并不响应我们分析WSGI应用的代码，因为这属于WSGI服务器的部分。

neutron/server/wsgi_eventlet.py:

def eventlet_wsgi_server():
    neutron_api = service.serve_wsgi(service.NeutronApiService)
    start_api_and_rpc_workers(neutron_api)

这里也能看到，核心功能一部分是WSGI，另一部分就是rpc部分。这里将Netron提供的API功能封装成了NeutronApiService类。我们来看看serve_wsgi:

neutron/service.py:

def serve_wsgi(cls):

    try:
        service = cls.create()
        service.start()
    except Exception:
        with excutils.save_and_reraise_exception():
            LOG.exception(_LE('Unrecoverable error: please check log '
                              'for details.'))

    return service

很明显，这是用NeutronApiService的类方法"create"来创建实例，然后"start"启动服务。接着分析下NeutronApiService的代码：

neutron/service.py:

class NeutronApiService(WsgiService):
    """Class for neutron-api service."""

    @classmethod
    def create(cls, app_name='neutron'):

        # Setup logging early, supplying both the CLI options and the
        # configuration mapping from the config file
        # We only update the conf dict for the verbose and debug
        # flags. Everything else must be set up in the conf file...
        # Log the options used when starting if we're in debug mode...

        config.setup_logging()
        service = cls(app_name)
        return service

可以看到NeutronApiService继承自"WsgiService"，表明其是一个WSGI服务。然后类方法"create"构造了其实例并返回。

class WsgiService(object):
    """Base class for WSGI based services.

    For each api you define, you must also define these flags:
    :<api>_listen: The address on which to listen
    :<api>_listen_port: The port on which to listen

    """

    def __init__(self, app_name):
        self.app_name = app_name
        self.wsgi_app = None

    def start(self):
        self.wsgi_app = _run_wsgi(self.app_name)

    def wait(self):
        self.wsgi_app.wait()

构造过程很简单，只是简单的记录app_name,这里是"neutron"，然后在start函数里真正加载WSGI APP,并运行服务，因此这才是我们分析的开始。

def _run_wsgi(app_name):
    app = config.load_paste_app(app_name)
    if not app:
        LOG.error(_LE('No known API applications configured.'))
        return
    return run_wsgi_app(app)

load_paste_app从函数名，也可以明白它的作用就是加载paste定义的WSGI应用。

neutron/commom/config.py:

def load_paste_app(app_name):
    """Builds and returns a WSGI app from a paste config file.

    :param app_name: Name of the application to load
    """
    loader = wsgi.Loader(cfg.CONF)
    app = loader.load_app(app_name)
    return app

wsgi.Loader是从neutron.conf中读取deploy配置文件的路径，然后根据指定的配置文件来加载app,默认是"/etc/neutron/api-paste.ini"。然后通过deploy.loadapp来加载app,这个deploy就是PasteDeploy。

oslo_service/wsgi.py:
def load_app(self, name):
    """Return the paste URLMap wrapped WSGI application.

    :param name: Name of the application to load.
    :returns: Paste URLMap object wrapping the requested application.
    :raises: PasteAppNotFound

    """
    try:
        LOG.debug("Loading app %(name)s from %(path)s",
                  {'name': name, 'path': self.config_path})
        return deploy.loadapp("config:%s" % self.config_path, name=name)
    except LookupError:
        LOG.exception(_LE("Couldn't lookup app: %s"), name)
        raise PasteAppNotFound(name=name, path=self.config_path)

分析到这里可知，后面app的加载过程就是PasteDeploy的加载过程，有了上篇http://blog.csdn.net/happyanger6/article/details/54518491文章中的基础，我们对着源码来理解:

先来看下配置文件"/etc/neutron/api-paste.ini":

[composite:neutron]
use = egg:Paste#urlmap
/: neutronversions
/v2.0: neutronapi_v2_0

[composite:neutronapi_v2_0]
use = call:neutron.auth:pipeline_factory
noauth = cors request_id catch_errors extensions neutronapiapp_v2_0
keystone = cors request_id catch_errors authtoken keystonecontext extensions neutronapiapp_v2_0

[filter:request_id]
paste.filter_factory = oslo_middleware:RequestId.factory

[filter:catch_errors]
paste.filter_factory = oslo_middleware:CatchErrors.factory

[filter:cors]
paste.filter_factory = oslo_middleware.cors:filter_factory
oslo_config_project = neutron

[filter:keystonecontext]
paste.filter_factory = neutron.auth:NeutronKeystoneContext.factory

[filter:authtoken]
paste.filter_factory = keystonemiddleware.auth_token:filter_factory

[filter:extensions]
paste.filter_factory = neutron.api.extensions:plugin_aware_extension_middleware_factory

[app:neutronversions]
paste.app_factory = neutron.api.versions:Versions.factory

[app:neutronapiapp_v2_0]
paste.app_factory = neutron.api.v2.router:APIRouter.factory

首先是一个组合类型的section,这个section表明用Paste.urlmap来构造应用,因此会将对"/"的访问交给另外一个app[app:nuetronversion],而将对"/v2.0"的访问交给另外一个组合[composite:neutronapi_v2_0]生成的app。

通过这2个就构造了所有的WSGI应用，其中对"/"的访问，而通过neutron.api,version:Versions.factory类方法来构造一个对象，然后将请求交于这个对象处理，

具体而言就是交于对象的__call__方法。我们来看下是如何构造的:

neutron/api/versinos.py:

class Versions(object):

    @classmethod
    def factory(cls, global_config, **local_config):
        return cls(app=None)

通过factory方法构造一个对象，这个对象就是一个WSGI应用。它就处理对"/"的方法，而根据WSGI规范，会调用这个对象的__call__方法:

@webob.dec.wsgify(RequestClass=wsgi.Request)
def __call__(self, req):
    """Respond to a request for all Neutron API versions."""
    version_objs = [
        {
            "id": "v2.0",
            "status": "CURRENT",
        },
    ]

    if req.path != '/':
        if self.app:
            return req.get_response(self.app)
        language = req.best_match_language()
        msg = _('Unknown API version specified')
        msg = oslo_i18n.translate(msg, language)
        return webob.exc.HTTPNotFound(explanation=msg)

..............

可以看到，通过@webob.dec.wsgify装饰器将__call__封装成符合WSGI规范的函数，这样"/"请求最终就是由"__call__"处理的。

这个"/"还比较简单，复杂的是对"/v2.0"的访问，这是大部分API的接口，我们看到这个组合段的app是用一个函数来构造的:

[composite:neutronapi_v2_0]
use = call:neutron.auth:pipeline_factory

use = call:...表示后面的是一个可调用对象，用它来构造最终的app.剩余的参数noauth,keystone等会作为参数传给pipeline_factory。

neutron/auth.py:

def pipeline_factory(loader, global_conf, **local_conf):
    """Create a paste pipeline based on the 'auth_strategy' config option."""
    pipeline = local_conf[cfg.CONF.auth_strategy]
    pipeline = pipeline.split()
    filters = [loader.get_filter(n) for n in pipeline[:-1]]
    app = loader.get_app(pipeline[-1])
    filters.reverse()
    for filter in filters:
        app = filter(app)
    return app

先从配置文件neutron.conf中读取auth策略,默认是"auth_strategy = keystone"，因此从api-paste.ini中取到的pipeline为"cors request_id catch_errors authtoken keystonecontext extensions neutronapiapp_v2_0"它们都定义在其它的"filter"或"app" section段中。

首先，从pipeline中获取最后一个app,即为"neutronapiapp_v2_0",从中加载app,然后依次用各个filter处理构造的app，并最终返回最后构造出的WSGI APP.

因此，我们按下面的顺序分析即可：

通过app_factory工厂方法来构造app，然后通过不同的filter_factory方法构造不同的filter对象，并将app依次通过filter对象处理。
[app:neutronapiapp_v2_0]
paste.app_factory = neutron.api.v2.router:APIRouter.factory

neutron/api/v2/router.py:

class APIRouter(base_wsgi.Router):

    @classmethod
    def factory(cls, global_config, **local_config):
        return cls(**local_config)

工厂方法构造了一个APIRouter对象作为app返回，因此分析其__init__方法:

def __init__(self, **local_config):
    mapper = routes_mapper.Mapper()
    plugin = manager.NeutronManager.get_plugin()
    ext_mgr = extensions.PluginAwareExtensionManager.get_instance()
    ext_mgr.extend_resources("2.0", attributes.RESOURCE_ATTRIBUTE_MAP)

    col_kwargs = dict(collection_actions=COLLECTION_ACTIONS,
                      member_actions=MEMBER_ACTIONS)
    def _map_resource(collection, resource, params, parent=None):
        allow_bulk = cfg.CONF.allow_bulk
        allow_pagination = cfg.CONF.allow_pagination
        allow_sorting = cfg.CONF.allow_sorting
        controller = base.create_resource(
            collection, resource, plugin, params, allow_bulk=allow_bulk,
            parent=parent, allow_pagination=allow_pagination,
            allow_sorting=allow_sorting)
        path_prefix = None
        if parent:
            path_prefix = "/%s/{%s_id}/%s" % (parent['collection_name'],
                                              parent['member_name'],
                                              collection)
        mapper_kwargs = dict(controller=controller,
                             requirements=REQUIREMENTS,
                             path_prefix=path_prefix,
                             **col_kwargs)
        return mapper.collection(collection, resource,
                                 **mapper_kwargs)

    mapper.connect('index', '/', controller=Index(RESOURCES))
    for resource in RESOURCES:
        _map_resource(RESOURCES[resource], resource,
                      attributes.RESOURCE_ATTRIBUTE_MAP.get(
                          RESOURCES[resource], dict()))
        resource_registry.register_resource_by_name(resource)

    for resource in SUB_RESOURCES:
        _map_resource(SUB_RESOURCES[resource]['collection_name'], resource,
                      attributes.RESOURCE_ATTRIBUTE_MAP.get(
                          SUB_RESOURCES[resource]['collection_name'],
                          dict()),
                      SUB_RESOURCES[resource]['parent'])

    # Certain policy checks require that the extensions are loaded
    # and the RESOURCE_ATTRIBUTE_MAP populated before they can be
    # properly initialized. This can only be claimed with certainty
    # once this point in the code has been reached. In the event
    # that the policies have been initialized before this point,
    # calling reset will cause the next policy check to
    # re-initialize with all of the required data in place.
    policy.reset()
    super(APIRouter, self).__init__(mapper)

这个属于核心API的构造，因此详细分析一下。

mapper = routes_mapper.Mapper()

首先，是声明一个routes.Mapper，这个上篇routes分析时讲过，用来构造URL和对应controller的映射，方便根据不同的URL路由给不同的controller处理。

plugin = manager.NeutronManager.get_plugin()

然后，先构造了一个NeutronManger的单例，这个对象构造的过程中会根据配置加载核心插件，一般就是"Ml2Plugin"，然后会加载以下几个默认的服务插件:

neutron/plugings/common/constants.py:

DEFAULT_SERVICE_PLUGINS = {
    'auto_allocate': 'auto-allocated-topology',
    'tag': 'tag',
    'timestamp_core': 'timestamp_core',
    'network_ip_availability': 'network-ip-availability'
}

然后是加载扩展插件：

extensions.PluginAwareExtensionManager.get_instance()

扩展插件的加载会从neutron/extensions目录下加载所有插件。

通过上面2步就加载完了核心插件，服务插件和扩展插件，然后就是构造不同URL的controller。

for resource in RESOURCES:
    _map_resource(RESOURCES[resource], resource,
                  attributes.RESOURCE_ATTRIBUTE_MAP.get(
                      RESOURCES[resource], dict()))

依次构造以下几个URL的controller."/networks","/subnets","/subnetpools","/ports"。

RESOURCES = {'network': 'networks',
             'subnet': 'subnets',
             'subnetpool': 'subnetpools',
             'port': 'ports'}
这个构造过程是通过_map_resource函数完成的，构造时会从配置文件中获取一些允许进行的操作，如"allow_bulk"
等。

在构造具体的mapper时，会传递以下参数:

col_kwargs = dict(collection_actions=COLLECTION_ACTIONS,
                  member_actions=MEMBER_ACTIONS)

COLLECTION_ACTIONS = ['index', 'create']
MEMBER_ACTIONS = ['show', 'update', 'delete']

这些就是可以对URL发起的操作类型，这些操作最终会根据访问的URL(/networks,ports)转换为create_network,update_port这些函数交给对应的controller处理。这些后面还会分析。

具体的controller是通过base.create_resource生成的，来看下代码:

neutron/api/v2/base.py:

class Controller(object):
    LIST = 'list'
    SHOW = 'show'
    CREATE = 'create'
    UPDATE = 'update'
    DELETE = 'delete'

..........

..........

def create_resource(collection, resource, plugin, params, allow_bulk=False,
                    member_actions=None, parent=None, allow_pagination=False,
                    allow_sorting=False):
    controller = Controller(plugin, collection, resource, params, allow_bulk,
                            member_actions=member_actions, parent=parent,
                            allow_pagination=allow_pagination,
                            allow_sorting=allow_sorting)
    return wsgi_resource.Resource(controller, FAULT_MAP)

可以看到，所有的Controller都是这个文件中定义的Controller类的实例对象，然后还会再将其调用wsgi_resouce.Resource.

neutron/api/v2/resouce.py:

def Resource(controller, faults=None, deserializers=None, serializers=None,
             action_status=None):
    """Represents an API entity resource and the associated serialization and
    deserialization logic
    """
    default_deserializers = {'application/json': wsgi.JSONDeserializer()}
    default_serializers = {'application/json': wsgi.JSONDictSerializer()}
    format_types = {'json': 'application/json'}
    action_status = action_status or dict(create=201, delete=204)

    default_deserializers.update(deserializers or {})
    default_serializers.update(serializers or {})

    deserializers = default_deserializers
    serializers = default_serializers
    faults = faults or {}

    @webob.dec.wsgify(RequestClass=Request)
    def resource(request):
        route_args = request.environ.get('wsgiorg.routing_args')
        if route_args:
            args = route_args[1].copy()
        else:
            args = {}

        # NOTE(jkoelker) by now the controller is already found, remove
        #                it from the args if it is in the matchdict
        args.pop('controller', None)
        fmt = args.pop('format', None)
        action = args.pop('action', None)
        content_type = format_types.get(fmt,
                                        request.best_match_content_type())
        language = request.best_match_language()
        deserializer = deserializers.get(content_type)
        serializer = serializers.get(content_type)

        try:
            if request.body:
                args['body'] = deserializer.deserialize(request.body)['body']

            method = getattr(controller, action)

            result = method(request=request, **args)
        except (exceptions.NeutronException,
                netaddr.AddrFormatError,
                oslo_policy.PolicyNotAuthorized) as e:
            for fault in faults:
                if isinstance(e, fault):
                    mapped_exc = faults[fault]
                    break
            else:
                mapped_exc = webob.exc.HTTPInternalServerError
            if 400 <= mapped_exc.code < 500:
                LOG.info(_LI('%(action)s failed (client error): %(exc)s'),
                         {'action': action, 'exc': e})
            else:
                LOG.exception(_LE('%s failed'), action)
            e = translate(e, language)
            body = serializer.serialize(
                {'NeutronError': get_exception_data(e)})
            kwargs = {'body': body, 'content_type': content_type}
            raise mapped_exc(**kwargs)
        except webob.exc.HTTPException as e:
            type_, value, tb = sys.exc_info()
            if hasattr(e, 'code') and 400 <= e.code < 500:
                LOG.info(_LI('%(action)s failed (client error): %(exc)s'),
                         {'action': action, 'exc': e})
            else:
                LOG.exception(_LE('%s failed'), action)
            translate(e, language)
            value.body = serializer.serialize(
                {'NeutronError': get_exception_data(e)})
            value.content_type = content_type
            six.reraise(type_, value, tb)
        except NotImplementedError as e:
            e = translate(e, language)
            # NOTE(armando-migliaccio): from a client standpoint
            # it makes sense to receive these errors, because
            # extensions may or may not be implemented by
            # the underlying plugin. So if something goes south,
            # because a plugin does not implement a feature,
            # returning 500 is definitely confusing.
            body = serializer.serialize(
                {'NotImplementedError': get_exception_data(e)})
            kwargs = {'body': body, 'content_type': content_type}
            raise webob.exc.HTTPNotImplemented(**kwargs)
        except Exception:
            # NOTE(jkoelker) Everything else is 500
            LOG.exception(_LE('%s failed'), action)
            # Do not expose details of 500 error to clients.
            msg = _('Request Failed: internal server error while '
                    'processing your request.')
            msg = translate(msg, language)
            body = serializer.serialize(
                {'NeutronError': get_exception_data(
                    webob.exc.HTTPInternalServerError(msg))})
            kwargs = {'body': body, 'content_type': content_type}
            raise webob.exc.HTTPInternalServerError(**kwargs)

        status = action_status.get(action, 200)
        body = serializer.serialize(result)
        # NOTE(jkoelker) Comply with RFC2616 section 9.7
        if status == 204:
            content_type = ''
            body = None

        return webob.Response(request=request, status=status,
                              content_type=content_type,
                              body=body)
    # NOTE(blogan): this is something that is needed for the transition to
    # pecan.  This will allow the pecan code to have a handle on the controller
    # for an extension so it can reuse the code instead of forcing every
    # extension to rewrite the code for use with pecan.
    setattr(resource, 'controller', controller)
    return resource

可以看到，所有的请求都会先交于resouce函数处理，进行反序列化和请求参数的获取，最终再交给controller处理。

action = args.pop('action', None)

method = getattr(controller, action)

result = method(request=request, **args)

这样对于"/networks","/subnets","/subnetpools","/ports"都会最终交于controller对应的action函数，以create_network为例:

def create(self, request, body=None, **kwargs):
    self._notifier.info(request.context,
                        self._resource + '.create.start',
                        body)
    return self._create(request, body, **kwargs)

@db_api.retry_db_errors
def _create(self, request, body, **kwargs):

       action = self._plugin_handlers[self.CREATE]
_create中会从selc._plugin_handlers里取对应操作映射的action,这个映射是在controller的构造函数里创建的:

self._plugin_handlers = {
    self.LIST: 'get%s_%s' % (parent_part, self._collection),
    self.SHOW: 'get%s_%s' % (parent_part, self._resource)
}
for action in [self.CREATE, self.UPDATE, self.DELETE]:
    self._plugin_handlers[action] = '%s%s_%s' % (action, parent_part,
                                                 self._resource)

self._resource为"network","port"这些RESOUCES，因此create对应的为create_network,create_port。
在_create中最终调用do_create:

obj_creator = getattr(self._plugin, action)
try:
    if emulated:
        return self._emulate_bulk_create(obj_creator, request,
                                         body, parent_id)
    else:
        if self._collection in body:
            # This is weird but fixing it requires changes to the
            # plugin interface
            kwargs.update({self._collection: body})
        else:
            kwargs.update({self._resource: body})
        return obj_creator(request.context, **kwargs)

可以看到会从self._plugin里获取对应的action,这个_plugin就是核心插件Ml2Plugin,因此所有的核心操作最终都
会交给Ml2Plugin的对应create_network,create_port等方法执行。这样就明白了所有核心资源的创建删除等
操作最终都会将给Ml2Plugin的对应方法处理。

那么Ml2Plugin插件的处理过程又是如何呢？我们先来看下其构造函数:

def __init__(self):
    # First load drivers, then initialize DB, then initialize drivers
    self.type_manager = managers.TypeManager()
    self.extension_manager = managers.ExtensionManager()
    self.mechanism_manager = managers.MechanismManager()
    super(Ml2Plugin, self).__init__()

可以看到它初始化了type_manager,mechanism_manager这2个管理器分别用来管理type和mechanism.其中不同的网络拓扑类型对应着Type Driver,而网络实现机制对应着Mechanism Driver。这两个管理器都是通过stevedor来管理的，这样就可以向查找标准库一样管理Type,Mechanism Driver了。

其中Type插件的加载会以'neutron.ml2.type_drivers'作为命名空间，Mechanism插件的加载会以'neutron.ml2.mechanism_drivers"作为命名空间。

这样实际上Ml2Plugin的不同操作会交给不同的type,mechanism插件处理，这样的架构十分灵活，比如:

def create_network(self, context, network):
    result, mech_context = self._create_network_db(context, network)
    try:
        self.mechanism_manager.create_network_postcommit(mech_context)

创建网络会交由mechanism_manager处理。

这样就是APIRouter构造出的app的全部内容了，对于核心URL会交由resource->Controller-->Ml2Plugin--->Type,Mechanism层层处理。也很方便我们根据需要自己实现不同的Type,Mechanism Driver.

然后就是将这个app交由不同的filter处理，我们继续看这些filter干了些什么。第一个filter是:

[filter:extensions]
paste.filter_factory = neutron.api.extensions:plugin_aware_extension_middleware_factory

neutron/api/extensions.py:

def plugin_aware_extension_middleware_factory(global_config, **local_config):
    """Paste factory."""
    def _factory(app):
        ext_mgr = PluginAwareExtensionManager.get_instance()
        return ExtensionMiddleware(app, ext_mgr=ext_mgr)
    return _factory

可以看到会用ExtensionMiddleware对象对app进行处理，这个处理和APIRouter的__init__函数处理类似，只不过这次是为扩展插件构造URL和Controller.这些扩展插件的Controller是ExtensionController。由于过程类似，就不再详细展开了，可以自行分析下。这样通过第一个filter就构造出了扩展插件的WSGI应用。

第二个filter:

[filter:authtoken]
paste.filter_factory = keystonemiddleware.auth_token:filter_factory

keystonemiddleware/auth_token/__init__.py:

def filter_factory(global_conf, **local_conf):
    """Returns a WSGI filter app for use with paste.deploy."""
    conf = global_conf.copy()
    conf.update(local_conf)

    def auth_filter(app):
        return AuthProtocol(app, conf)
    return auth_filter

可以看到对app封装了一个AuthProtocol对象。分析其代码不难看出其作用是对请求是否通过了认证进行检查，即是否携带合法token。这样后面的filter的作用也类似，就是对请求进行一些预处理，所有预处理都完成后再交由实际的Controller处理。

这样我们就分析完了整个WSGI应用的构造和处理过程，不难得出下面的处理流程：