摘要：標簽源碼閱讀在中虛擬機的創(chuàng)建無疑是非常重要的了解虛擬機創(chuàng)建流程并閱讀模塊關于創(chuàng)建虛擬機的源碼對開發(fā)有很很大幫助本篇文章將以版本為基礎講解創(chuàng)建虛擬機的源碼由于模塊代碼復雜而且閱讀源碼所需知識較多所以側(cè)重于流程邏輯源碼閱讀可能不夠詳盡指出模塊結(jié)

標簽： openstack nova 源碼閱讀

在openstack中,虛擬機的創(chuàng)建無疑是非常重要的,了解虛擬機創(chuàng)建流程并閱讀nova模塊關于創(chuàng)建虛擬機的源碼對opensatck開發(fā)有很很大幫助,本篇文章將以openstack queens版本為基礎.講解創(chuàng)建虛擬機的源碼.由于nova模塊代碼復雜,而且閱讀源碼所需知識較多,所以側(cè)重于流程邏輯,源碼閱讀可能不夠詳盡指出.

DB: 用于存儲nova服務相關數(shù)據(jù)的SQL數(shù)據(jù)庫，通常為MySQL

API：接收HTTP請求的組件,處理請求并通過oslo.messaging隊列或HTTP與其他組件通信

Scheduler: 決定虛擬機在哪個主機運行

Network: 管理IP轉(zhuǎn)發(fā),橋接和vlans

Compute: 管理虛擬機和虛擬機管理器直接的通信

Conductor: 處理需要協(xié)同的請求(創(chuàng)建/重建),作為數(shù)據(jù)庫代理,或者處理對象轉(zhuǎn)換

為了簡便,這里省略了認證過程,實際上一個請求發(fā)送過來,還要經(jīng)過認證和鑒權(quán)等過程,確保該用戶有權(quán)限創(chuàng)建虛擬機.
在openstack的wiki中給出創(chuàng)建一個虛擬的workflow,圖有點大,但對于每個組件的工作內(nèi)容寫的非常詳細

我們可以把創(chuàng)建流程分成幾個部分

用戶發(fā)送一個創(chuàng)建虛擬機的請求,Nova-Api接收到請求后,Nova-Api負責激活擴展插件,檢查虛擬機名稱,接收注入文件,提取新虛擬機的網(wǎng)絡設置,檢查配置和鏡像等工作.
然后Nova-Api將處理好的參數(shù)集以JSON文件通過HTTP請求發(fā)送給Nova的Compute－Api.然后向用戶發(fā)送帶有虛擬機預留ID的相應（這時的相應碼202,提示虛擬機創(chuàng)建成功,但實際上虛擬機還未真正創(chuàng)建成功)
Compute-Api在接收到請求后將會檢查創(chuàng)建政策,進一步檢查虛擬機,網(wǎng)絡,鏡像和配額,然后正式建立虛擬機的配置.
接著在數(shù)據(jù)庫中生成虛擬機的相應記錄,而后通過消息隊列發(fā)送請求讓scheduler選擇一個host來創(chuàng)建虛擬機.

scheduler接收到消息后根據(jù)消息中filters對所有host進行過濾,最后選出某個host,然后跟新數(shù)據(jù)庫,并通過消息隊列向被選定的host發(fā)送創(chuàng)建虛擬機消息
被選定的host接收到隊列的消息后在數(shù)據(jù)庫中更新虛擬機和任務的狀態(tài),然后通過nova network-api為虛擬機創(chuàng)建或獲取網(wǎng)絡.
接著通過nova volume-api為虛擬機創(chuàng)建新的卷,決定虛擬機的塊設備映射,并將卷掛載到虛擬機上
此時虛擬機的調(diào)度和資源準備都以完成.

被選定的host獲取鏡像,建立塊設備映射,最后生成libvirt.xml最后執(zhí)行hypervisor的spawn()方法,至此,虛擬機已經(jīng)在host上運行了.最后一步是跟新虛擬機和任務的狀態(tài).

下面將根據(jù)虛擬機創(chuàng)建時調(diào)用的模塊順序閱讀代碼

Nova-Api將處理不同類型的請求寫成了各種controler類,而處理創(chuàng)建虛擬機的類被稱為ServerControler

# nova/api/openstack/compute/servers.py
class ServersController(wsgi.Controller):
    
    # 為了簡明,省略了很多用于檢查的裝飾器
    @wsgi.response(202)
    def create(self, req, body):
         """Creates a new server for a given user."""
        context = req.environ["nova.context"]
        server_dict = body["server"]
        password = self._get_server_admin_password(server_dict)
        name = common.normalize_name(server_dict["name"])
        description = name
        # create方法干的是些從請求中提取并檢查參數(shù)的臟活
        # 省略...
        try:
            # 依然是構(gòu)建參數(shù)的代碼...
            
            # 調(diào)用compute_api創(chuàng)建虛擬機
            (instances, resv_id) = self.compute_api.create(context,
                            inst_type,
                            image_uuid,
                            display_name=name,
                            display_description=description,
                            availability_zone=availability_zone,
                            forced_host=host, forced_node=node,
                            metadata=server_dict.get("metadata", {}),
                            admin_password=password,
                            requested_networks=requested_networks,
                            check_server_group_quota=True,
                            supports_multiattach=supports_multiattach,
                            **create_kwargs)
        # 錯誤處理...

# nova/compute/api.py
class API(base.Base):
    def create(self, context, instance_type,
               image_href, kernel_id=None, ramdisk_id=None,
               min_count=None, max_count=None,
               display_name=None, display_description=None,
               key_name=None, key_data=None, security_groups=None,
               availability_zone=None, forced_host=None, forced_node=None,
               user_data=None, metadata=None, injected_files=None,
               admin_password=None, block_device_mapping=None,
               access_ip_v4=None, access_ip_v6=None, requested_networks=None,
               config_drive=None, auto_disk_config=None, scheduler_hints=None,
               legacy_bdm=True, shutdown_terminate=False,
               check_server_group_quota=False, tags=None,
               supports_multiattach=False):
        """準備實例創(chuàng)建工作，然后將實例信息發(fā)送至scheduler，
        由scheduler計算host上創(chuàng)建和在DB創(chuàng)建記錄。
        """
        # preparation
        
        # 為了簡介,所有參數(shù)簡略為args和kwargs
        self_create_instance(*args, **kwargs)
        
    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, filter_properties,
               reservation_id=None, legacy_bdm=True, shutdown_terminate=False,
               check_server_group_quota=False, tags=None,
               supports_multiattach=False):
        """核查所有參數(shù)"""
        # verifying
        pass
        
        # 獲取鏡像信息
        if image_href:
            # if image_href is provied, get image via glance api
            image_id, boot_meta = self._get_image(context, image_href)
        else:
            # if image_href is not proved, get image metadata from bdm
            image_id = None
            boot_meta = self._get_bdm_image_metadata(
                context, block_device_mapping, legacy_bdm)
        
        # 繼續(xù)檢查參數(shù)
        
        # 由于block device mapping有兩種版本,為了兼容,需要檢查并在必要時轉(zhuǎn)換
         block_device_mapping = self._check_and_transform_bdm(context,
            base_options, instance_type, boot_meta, min_count, max_count,
            block_device_mapping, legacy_bdm)
        
        # go on checking
        
        # 為了支持cell特性,參見cell wiki
        # https://docs.openstack.org/nova/ocata/cells.html
        if CONF.cells.enable:
            # 創(chuàng)建instance模型對象
            # 檢查quota
            
            # 調(diào)用rpc api將消息發(fā)送到隊列
            self.compute_task_api.build_instance(*args, **kwargs)
        
        else:
            compute_task_api.schedule_and_build_instances(*args, **kwargs)
        
        
        return instances, reservation_id

nova組件之間可以通過rpc api以消息隊列通信,而最后的真正執(zhí)行的任務的類都在manager.py文件中定義.這里我們方便理解省略調(diào)度代碼

# nova/conductor/api.py
class ComputeTaskAPI(object):
    def schedule_and_build_instance(self, *args, **kwargs):
        # very simple method
        # call rpc api only
        self.conductor_compute_rpc_api.schedule_and_build_instance(*args, **kwargs)

# nova/conductor/rpcapi.py
class ComputeTaskAPI(object):
    def schedule_and_build_instance(self, *args, **kwargs):
        # 構(gòu)建參數(shù)和api版本檢查
        # 最后將其發(fā)送到消息隊列
        cctxt.cast(context, "schedule_and_build_instance", **kwargs)

# nova/compute/manager.py
class ComputeManager(object):
    @wrap_exception()
    @reverts_task_state
    @wrap_instance_fault
    def build_and_run_instance(self, *args, **kwargs):
        # 給資源加鎖,避免競爭
        @utils.synchronized(instance.uuid)
        def _locked_do_build_and_run_instance(*args, **kwargs):
            with self._build_semaphore:
                try:
                    result = self._do_build_and_run_instance(*args, **kwargs)
                # handle exceptions
                pass
        # 由于創(chuàng)建虛擬機的工作可能會持續(xù)很長時間,為了避免進程阻塞
        # 將這個任務分發(fā)給某個worker
        utils.spawn_n(_locked_do_build_and_run_instance,
        context, instance, ...)

    def _do_build_and_run_instance(self, *args, **kwargs):
        # 更新虛擬機和任務狀態(tài)
        # 解碼注入文件
        try:
            with timeutils.StopWatch() as timer:
                self._build_and_run_instance(*args)
        # handle exceptions
    def _build_and_run_instance(self, *args, **kwargs):
        # 獲取 image ref
        try:
            scheduler_hints = self._get_scheduler_hints(filter_properties, request_spec)
            rt = self._get_resource_tracker()
            with rt.instance_claime(context, instance, node, limits):
                # 獲取群組策略和鏡像metadata

                # 通過調(diào)用_build_resources創(chuàng)建network和volume
                with self._build_resources(*args) as resources:
                    # handle vm and task state
                    # spawn instance on hypervisor
                    with timeuitls.StopWatch() as time:
                        # 通過driver創(chuàng)建xml,然后真正運行虛擬機
                        self.driver.spawn(*args, **kwargs)
        # handle exceptions