openstack創建實例的流程

workflow: 1. 用戶向nova-api發送請求
用戶發送請求到nova-api,這裡有兩種：
a.通過openstack api
從 server.py's controller.create():

self.helper.create_instance(req, body, self.compute_api.create)  

create_instance_helper.CreateInstanceHelper() 查表獲取基本信息
b.通過ec2 api
從cloud.py.run_instances()

統一調 computer.api.create() 將新的數據插回去

self._ask_scheduler_to_create_instance(context, base_options,  
                                       instance_type, zone_blob,  
                                       availability_zone, injected_files,  
                                       admin_password, image,  
                                       instance_id=instance_id,  
                                       requested_networks=requested_networks)  

2. API 將處理好的數據通過MQ 轉發給scheduler .(code from Computer.api)

rpc.cast(context,  
            FLAGS.scheduler_topic,  
            {"method": "run_instance",  
             "args": {"topic": FLAGS.compute_topic,  
                      "instance_id": instance_id,  
                      "request_spec": request_spec,  
                      "availability_zone": availability_zone,  
                      "admin_password": admin_password,  
                      "injected_files": injected_files,  
                      "requested_networks": requested_networks}}) 

3. Scheduler 獲取信息並作出決定 哪一個host 可以來run instance.

def __getattr__(self, key):  
    return functools.partial(self._schedule, key) 

def _schedule(self, method, context, topic, *args, **kwargs):  
    .......  
    rpc.cast(context,  
             db.queue_get_for(context, topic, host),  
             {"method": method,  
              "args": kwargs})  
    LOG.debug(_("Casted to %(topic)s %(host)s for %(method)s") % locals())

4. Computer 從池中獲取信息 並讓 Networker 去准備一個ip ,讓volume 准備卷, 然後初始化相應的信息,例如創建image,映射device,創建domain,
並將domain 放入running pool中 然後就進入等待直到instance的狀態變為running.
a. networker 分配ip

network_info = self.network_api.allocate_for_instance(context,  
                          instance, vpn=is_vpn,  
                          requested_networks=requested_networks)

def allocate_floating_ip(self, context):  
   return rpc.call(context,  
                   FLAGS.network_topic,  
                   {'method': 'allocate_floating_ip',  
                    'args': {'project_id': context.project_id}}) 

b 讓 volume 准備卷

bd_mapping = self._setup_block_device_mapping(context, instance_id)  
def create(self, context, size, snapshot_id, name, description,  
          volume_type=None, metadata=None, availability_zone=None):  
         rpc.cast(context,  
                      FLAGS.scheduler_topic,  
                              {"method": "create_volume",  
                                   "args": {"topic": FLAGS.volume_topic,  
                                   "volume_id": volume['id'],  
                                    "snapshot_id": snapshot_id}}) 

c call nova.virt.libvirt.firewall.IptablesFirewallDriver 建立網絡規則
這裡是重頭戲，單獨開個貼記錄下....
d call libvirt 創建domian 並launch

domain = self._create_new_domain(xml)  
def _create_new_domain(self, xml, persistent=True, launch_flags=0):  
    if persistent:  
               # To create a persistent domain, first define it, then launch it.  
               domain = self._conn.defineXML(xml)  
               domain.createWithFlags(launch_flags)  
   else:  
               # createXML call creates a transient domain  
               domain = self._conn.createXML(xml, launch_flags)  
   return domain

e call virt.libvirt.connetion.spwan 等待

def spawn(self, context, instance, network_info,  
      block_device_info=None):  
       ..........  
       def _wait_for_boot():  
             instance_name = instance['name']  
            try:  
                    state = self.get_info(instance_name)['state']  
            except exception.NotFound:  
                    msg = _("During reboot, %s disappeared.") % instance_name  
                     LOG.error(msg)  
             raise utils.LoopingCallDone  
  
            if state == power_state.RUNNING:  
                msg = _("Instance %s spawned successfully.") % instance_name  
                LOG.info(msg)  
                raise utils.LoopingCallDone  
  
timer = utils.LoopingCall(_wait_for_boot)  
return timer.start(interval=0.5, now=True) 

5 一旦instance的狀態改變至running,他就會去通過networker獲取網絡信息, 這裡有幾種方式,取決於你的networkManager