keystone在httpd的入口執行文件為/usr/bin/keystone-wsgi-public
查看文件/usr/bin/keystone-wsgi-public，根據代碼，看到主要是這幾行代碼對keystone的服務進行了初始化，keystone服務通過wsgiref.simple_server與httpd進行交互，初始化函數為initialize_public_application
1 from keystone.server.wsgi import initialize_public_application 2 3 if __name__ == "__main__": 4 ...... 5 import wsgiref.simple_server as wss 6 7 ...... 8 9 server = wss.make_server(args.host, args.port, initialize_public_application()) 10 11 ....... 12 server.serve_forever()
查看initialize_public_application
1 from keystone.server.flask import core as flask_core 2 def initialize_public_application(): 3 return flask_core.initialize_application(name='public', config_files=flask_core._get_config_files()) 4 initialize_admin_application = initialize_public_application
從這里可以看到在Rocky版本的keystone中，admin和public已經合並在了一起，不再區分admin和public，flask_core.initialize_application接受三個參數，原型如下
1 def initialize_application(name, post_log_configured_function=lambda: None, 2 config_files=None): 3 possible_topdir = os.path.normpath(os.path.join( 4 os.path.abspath(__file__), 5 os.pardir, 6 os.pardir, 7 os.pardir, 8 os.pardir)) 9 10 dev_conf = os.path.join(possible_topdir, 11 'etc', 12 'keystone.conf') 13 if not config_files: 14 config_files = None 15 if os.path.exists(dev_conf): 16 config_files = [dev_conf] 17 18 keystone.server.configure(config_files=config_files) 19 20 if CONF.debug: 21 CONF.log_opt_values(log.getLogger(CONF.prog), log.DEBUG) 22 23 post_log_configured_function() 24 25 def loadapp(): 26 app = application.application_factory(name) 27 return app 28 29 _unused, app = keystone.server.setup_backends( 30 startup_application_fn=loadapp) 31 32 profiler.setup(name) 33 34 return setup_app_middleware(app)
name 這個不需要講，但是name必須是public或者admin，在后面的初始化過程中有限制
post_log_configured_function 接收一個函數，會在initialize_application這個函數中調用，但是具體是什么作用，還沒有領悟出來，但是這個函數應該是可以做一些類似於初始化等等作用的
config_files 在開發模式下可以自定義keystone.conf，如果自己定義了keyston.conf這個文件，那么就會完全的替代/etc/keystone/keystone.conf中定義的一些配置參數，關於配置參數，后續會介紹
接着往下看這個函數
keystone.server.configure這個函數對keystone進行了一些配置
1 def configure(version=None, config_files=None, 2 pre_setup_logging_fn=lambda: None): 3 keystone.conf.configure() 4 sql.initialize() 5 keystone.conf.set_config_defaults() 6 7 CONF(project='keystone', version=version, 8 default_config_files=config_files) 9 10 pre_setup_logging_fn() 11 keystone.conf.setup_logging() 12 13 if CONF.insecure_debug: 14 LOG.warning( 15 'insecure_debug is enabled so responses may include sensitive ' 16 'information.')
keystone.conf.configure()這個函數添加了幾個參數，
standard-threads 可以自定義多少個線程
pydev-debug-host以及pydev-debug-port是對遠程debug的機器和port進行了定義，並在conf中注冊了它們三個參數
1 def configure(conf=None): 2 if conf is None: 3 conf = CONF 4 5 conf.register_cli_opt( 6 cfg.BoolOpt('standard-threads', default=False, 7 help='Do not monkey-patch threading system modules.')) 8 conf.register_cli_opt( 9 cfg.StrOpt('pydev-debug-host', 10 help='Host to connect to for remote debugger.')) 11 conf.register_cli_opt( 12 cfg.PortOpt('pydev-debug-port', 13 help='Port to connect to for remote debugger.')) 14 15 for module in conf_modules: 16 module.register_opts(conf) 17 18 # register any non-default auth methods here (used by extensions, etc) 19 auth.setup_authentication() 20 21 # add oslo.cache related config options 22 cache.configure(conf)
auth.setup_authentication()，配置了幾種認證方式，password,token等等，跟上述參數一樣，先定義，后注冊到conf中
cache.configure(conf)，對cache進行了定義和注冊，主要涉及到各種cache的定義等等
回到 keystone.server.configure
sql.initialize() 這個函數主要是對數據庫進行了初始化，導入配置文件中的參數
keystone.conf.set_config_defaults() 采用默認值對keystone的設置進行初始化
CONF(project='keystone', version=version, default_config_files=config_files) 這塊比較難理解，先說第三個參數，第三個參數采用指定的config_files對默認參數進行復寫
CONF的原型是 keystone.conf.CONF， 它采用了從文件中直接導入的方式，構造了一個單例的configure配置類 CONF = ConfigOpts()
CONF()調用了ConfigOpts.__call__()這個方法
pre_setup_logging_fn()這個和前文的post_log_configured_function是相似的，都是完成一些自定義的動作
keystone.conf.setup_logging()這個參數是對log進行配置
回到initialize_application
keystone.server.setup_backend(startup_application_fn=loadapp)
1 def setup_backends(load_extra_backends_fn=lambda: {}, 2 startup_application_fn=lambda: None): 3 drivers = backends.load_backends() 4 drivers.update(load_extra_backends_fn()) 5 res = startup_application_fn() 6 return drivers, res
1 def loadapp(): 2 app = application.application_factory(name) 3 return app
這兩段代碼中，setup_backends首先導入一些driver，導入方法主要是用stevedore的DriverManager類對一些方法進行導入，並將這些driver注冊到provider_api中，供后續調用
load_app采用了flask.application的工廠模式方法構造
1 def application_factory(name='public'): 2 if name not in ('admin', 'public'): 3 raise RuntimeError('Application name (for base_url lookup) must be ' 4 'either `admin` or `public`.') 5 6 app = flask.Flask(name) 7 app.after_request(_add_vary_x_auth_token_header) 8 9 app.config.update(PROPAGATE_EXCEPTIONS=True) 10 11 # TODO(morgan): Convert Subsystems over to Flask-Native, for now, we simply 12 dispatch_map = collections.OrderedDict() 13 14 hc_app = healthcheck.Healthcheck.app_factory( 15 {}, oslo_config_project='keystone') 16 dispatch_map['/healthcheck'] = hc_app 17 _routers = [] 18 sub_routers = [] 19 mapper = routes.Mapper() 20 for api_routers in ALL_API_ROUTERS: 21 moved_found = [pfx for 22 pfx in getattr(api_routers, '_path_prefixes', []) 23 if pfx in _MOVED_API_PREFIXES] 24 if moved_found: 25 raise RuntimeError('An API Router is trying to register path ' 26 'prefix(s) `%(pfx)s` that is handled by the ' 27 'native Flask app. Keystone cannot ' 28 'start.' % 29 {'pfx': ', '.join([p for p in moved_found])}) 30 31 routers_instance = api_routers.Routers() 32 _routers.append(routers_instance) 33 routers_instance.append_v3_routers(mapper, sub_routers) 34 35 keystone.api.discovery.register_version('v3') 36 for api in keystone.api.__apis__: 37 for api_bp in api.APIs: 38 api_bp.instantiate_and_register_to_app(app) 39 sub_routers.append(_ComposibleRouterStub(_routers)) 40 legacy_dispatcher = keystone_wsgi.ComposingRouter(mapper, sub_routers) 41 42 for pfx in itertools.chain(*[rtr.Routers._path_prefixes for 43 rtr in ALL_API_ROUTERS]): 44 dispatch_map['/v3/%s' % pfx] = legacy_dispatcher 45 46 app.wsgi_app = KeystoneDispatcherMiddleware( 47 app.wsgi_app, 48 dispatch_map) 49 return app
這塊限定了name只能是public or admin，這個函數主要是為keystone的請求添加路由的mapper以及中間件的調用，先由這個函數對請求的API的路徑進行解析， 最后KeystoneDispatcherMiddleware完成對路徑API的訪問，同樣是調用了__call__()方法
profiler.setup(name)是對信息的采集，對性能統計有作用
setup_app_middleware()同樣是通過stevedore.DriverManager進行中間件的導入
到此基本上整個keystone便啟動了~