novabootfromvolume在多主机zone下的坑-创新互联

测试环境：3个计算节点，分别属于3个zone

为丹徒等地区用户提供了全套网页设计制作服务，及丹徒网站建设行业解决方案。主营业务为成都做网站、网站制作、成都外贸网站建设、丹徒网站设计，以传统方式定制建设网站，并提供域名空间备案等一条龙服务，秉承以专业、用心的态度为用户提供真诚的服务。我们深信只要达到每一位用户的要求，就会得到认可，从而选择与我们长期合作。这样，我们也可以走得更远！

[root@controller2 ~(keystone_admin)]# nova availability-zone-list

+-----------------------+----------------------------------------+

| Name          | Status                 |

+-----------------------+----------------------------------------+

| internal        | available                |

| |- controller2     |                     |

| | |- nova-conductor  | enabled :-) 2016-08-20T14:57:07.000000 |

| | |- nova-scheduler  | enabled :-) 2016-08-20T14:57:06.000000 |

| | |- nova-consoleauth | enabled :-) 2016-08-20T14:57:08.000000 |

| | |- nova-cert     | enabled :-) 2016-08-20T14:57:07.000000 |

| nova          | available                |

| |- controller3     |                     |

| | |- nova-compute   | enabled :-) 2016-08-20T14:57:04.000000 |

| ag1          | available                |

| |- controller1     |                     |

| | |- nova-compute   | enabled :-) 2016-08-19T23:41:45.000000 |

| ag2          | available                |

| |- controller2     |                     |

| | |- nova-compute   | enabled :-) 2016-08-20T14:57:06.000000 |

+-----------------------+----------------------------------------+

测试方式： 启动虚拟机的时候选择Booting from p_w_picpath(creates a new volume)测试

产生的原因： cinder无法识别nova的多zone，cinder能获取的zone信息有：

                    1、cinder-volume所在的zone

                    2、cinder.conf配置文件中的两个参数storage_availability_zone = nova和default_availability_zone = nova

大致过下代码调用过程：(从nova那边调用cinder那部分开始)

1、nova/virt/block_device.py    ->    class DriverImageBlockDevice    def attach    vol = volume_api.create

2、nova/volume/cinder.py    ->  class API    def create    item = client.volumes.create

3、cinder/api/v2/volumes.py    ->    class VolumeController    def create    new_volume = self.volume_api.create

4、cinder/volume/api.py    ->    class API    def create    flow_engine = create_volume.get_flow

    在create函数中cinder获取能获得到的zone的信息：

        raw_zones = self.list_availability_zones(enable_cache=True)

    availability_zones = set([az['name'] for az in raw_zones])

    if CONF.storage_availability_zone:

      availability_zones.add(CONF.storage_availability_zone)

def list_availability_zones  services = objects.ServiceList.get_all_by_topic   ->

cinder/objects/service.py（def get_all_by_topic） db.service_get_all_by_topic  ->

cinder/db/api.py（def service_get_all_by_topic） IMPL.service_get_all_by_topic ->

cinder/db/sqlalchemy/api.py（def service_get_all_by_topic）

@require_admin_context
def service_get_all_by_topic(context, topic, disabled=None):
    query = model_query(
        context, models.Service, read_deleted="no").\ # models在这里cinder/db/sqlalchemy/models.py
        filter_by(topic=topic)                        # topic默认传过来的是cinder-volume
    if disabled is not None:
        query = query.filter_by(disabled=disabled)
    return query.all()

查询数据库找出cinder-volume所在的zone

5、cinder/volume/flows/api/create_volume.py  ->    def get_flow

# 这里是taskflow，只需关注add方法里面的东东
def get_flow(db_api, p_w_picpath_service_api, availability_zones, create_what,
             scheduler_rpcapi=None, volume_rpcapi=None):
    """Constructs and returns the api entrypoint flow.
    This flow will do the following:
    1. Inject keys & values for dependent tasks.
    2. Extracts and validates the input keys & values.
    3. Reserves the quota (reverts quota on any failures).
    4. Creates the database entry.
    5. Commits the quota.
    6. Casts to volume manager or scheduler for further processing.
    """
    flow_name = ACTION.replace(":", "_") + "_api"
    api_flow = linear_flow.Flow(flow_name)
    api_flow.add(ExtractVolumeRequestTask(
        p_w_picpath_service_api,
        availability_zones,
        rebind={'size': 'raw_size',
                'availability_zone': 'raw_availability_zone',
                'volume_type': 'raw_volume_type'}))
                
    api_flow.add(QuotaReserveTask(),
                 EntryCreateTask(db_api),
                 QuotaCommitTask())
                 
    if scheduler_rpcapi and volume_rpcapi:
        # This will cast it out to either the scheduler or volume manager via
        # the rpc apis provided.
        api_flow.add(VolumeCastTask(scheduler_rpcapi, volume_rpcapi, db_api))
    # Now load (but do not run) the flow using the provided initial data.
    
    return taskflow.engines.load(api_flow, store=create_what)

我们关心的在class ExtractVolumeRequestTask里

先看下入口execute方法，为什么看execute方法，你翻下taskflow怎么用的，就懂了

def execute(self, context, size, snapshot, p_w_picpath_id, source_volume,
                availability_zone, volume_type, metadata, key_manager,
                source_replica, consistencygroup, cgsnapshot):
        utils.check_exclusive_options(snapshot=snapshot,
                                      p_w_picpathRef=p_w_picpath_id,
                                      source_volume=source_volume)
        policy.enforce_action(context, ACTION)
        # TODO(harlowja): what guarantee is there that the snapshot or source
        # volume will remain available after we do this initial verification??
        snapshot_id = self._extract_snapshot(snapshot)
        source_volid = self._extract_source_volume(source_volume)
        source_replicaid = self._extract_source_replica(source_replica)
        size = self._extract_size(size, source_volume, snapshot)
        consistencygroup_id = self._extract_consistencygroup(consistencygroup)
        cgsnapshot_id = self._extract_cgsnapshot(cgsnapshot)
        self._check_p_w_picpath_metadata(context, p_w_picpath_id, size)
        availability_zone = self._extract_availability_zone(availability_zone,  # 关心的在这里
                                                            snapshot,
                                                            source_volume)

# _extract_availability_zone函数                                                          
def _extract_availability_zone(self, availability_zone, snapshot,
                                   source_volume):
        """Extracts and returns a validated availability zone.

        This function will extract the availability zone (if not provided) from
        the snapshot or source_volume and then performs a set of validation
        checks on the provided or extracted availability zone and then returns
        the validated availability zone.
        """

        # Try to extract the availability zone from the corresponding snapshot
        # or source volume if either is valid so that we can be in the same
        # availability zone as the source.
        if availability_zone is None:
            if snapshot:
                try:
                    availability_zone = snapshot['volume']['availability_zone']
                except (TypeError, KeyError):
                    pass
            if source_volume and availability_zone is None:
                try:
                    availability_zone = source_volume['availability_zone']
                except (TypeError, KeyError):
                    pass

        if availability_zone is None:
            if CONF.default_availability_zone:    # default_availability_zone判断
                availability_zone = CONF.default_availability_zone 
            else:
                # For backwards compatibility use the storage_availability_zone
                availability_zone = CONF.storage_availability_zone 

        if availability_zone not in self.availability_zones:     # self.availability_zones就是上述cinder-volume的zone外加cinder.conf中两个配置参数
            if CONF.allow_availability_zone_fallback:    # allow_availability_zone_fallback这个很关键
                original_az = availability_zone
                availability_zone = (
                    CONF.default_availability_zone or
                    CONF.storage_availability_zone)
                LOG.warning(_LW("Availability zone '%(s_az)s' "
                                "not found, falling back to "
                                "'%(s_fallback_az)s'."),
                            {'s_az': original_az,
                             's_fallback_az': availability_zone})
            else:
                msg = _("Availability zone '%(s_az)s' is invalid.")
                msg = msg % {'s_az': availability_zone}
                raise exception.InvalidInput(reason=msg)             #  没启动allow_availability_zone_fallback的话，就抛异常了

另外有需要云服务器可以了解下创新互联scvps.cn，海内外云服务器15元起步，三天无理由+7*72小时售后在线，公司持有idc许可证，提供“云服务器、裸金属服务器、高防服务器、香港服务器、美国服务器、虚拟主机、免备案服务器”等云主机租用服务以及企业上云的综合解决方案，具有“安全稳定、简单易用、服务可用性高、性价比高”等特点与优势，专为企业上云打造定制，能够满足用户丰富、多元化的应用场景需求。

文章标题：novabootfromvolume在多主机zone下的坑-创新互联
文章网址：http://wjwzjz.com/article/pjcdo.html