openstack ice resize 详解(三)

感谢朋友支持本博客，欢迎共同探讨交流，由于能力和时间有限，错误之处在所难免，欢迎指正！

如有转载，请保留源作者博客信息。

Better Me的博客：blog.csdn.net/tantexian

如需交流，欢迎大家博客留言。

由于篇幅较长：上接之前博文：openstack ice resize 详解(二)

三、resize完毕后的确认confirm_resize代码详解分析

1、/nova/api/openstack/compute/servers.py

@wsgi.response(202)

    @wsgi.serializers(xml=FullServerTemplate)

    @wsgi.deserializers(xml=ActionDeserializer)

    @wsgi.action('confirmResize')

    def _action_confirm_resize(self, req, id, body):

        context = req.environ['nova.context']

        instance = self._get_server(context, req, id) #从数据库获取实例信息

        try:

            self.compute_api.confirm_resize(context, instance)#跟进到2

        except exception.MigrationNotFound:

            msg = _("Instance has not been resized.")

            raise exc.HTTPBadRequest(explanation=msg)

        except exception.InstanceIsLocked as e:

            raise exc.HTTPConflict(explanation=e.format_message())

        except exception.InstanceInvalidState as state_error:

            common.raise_http_conflict_for_instance_invalid_state(state_error,

                    'confirmResize')

        return exc.HTTPNoContent()

2、/nova/compute/api.py

    @wrap_check_policy

    @check_instance_lock

    @check_instance_cell

    @check_instance_state(vm_state=[vm_states.RESIZED])

    def confirm_resize(self, context, instance, migration=None):

        """Confirms a migration/resize and deletes the 'old' instance."""

        elevated = context.elevated()

        if migration is None:

            migration = migration_obj.Migration.get_by_instance_and_status(

                elevated, instance.uuid, 'finished')

        # reserve quota only for any decrease in resource usage

        deltas = self._downsize_quota_delta(context, instance)

        quotas = self._reserve_quota_delta(context, deltas)

        migration.status = 'confirming' #设置迁移状态为正在confirming

        migration.save() #保存

        # With cells, the best we can do right now is commit the reservations

        # immediately...

        if CONF.cells.enable:

            quotas.commit(context)

        self._record_action_start(context, instance, #通知实例确认resize

                                  instance_actions.CONFIRM_RESIZE)

        self.compute_rpcapi.confirm_resize(context, #跟进代码至3

                                           instance,

                                           migration,

                                           migration.source_compute,

                                           quotas.reservations or [])

3、/nova/compute/rpcapi.py

    def confirm_resize(self, ctxt, instance, migration, host,

            reservations=None, cast=True):

        # NOTE(russellb) Havana compat

        version = self._get_compat_version('3.0', '2.39')

        cctxt = self.client.prepare(server=_compute_host(host, instance),

                version=version)

        rpc_method = cctxt.cast if cast else cctxt.call

        return rpc_method(ctxt, 'confirm_resize', #rpc调用confirm_resize,跟进到4

                          instance=instance, migration=migration,

                          reservations=reservations)

4、/nova/compute/manage.py

    @wrap_exception()

    @wrap_instance_event

    @wrap_instance_fault

    def confirm_resize(self, context, instance, reservations, migration):

        @utils.synchronized(instance['uuid'])

        def do_confirm_resize(context, instance, migration_id):

            # NOTE(wangpan): Get the migration status from db, if it has been

            #                confirmed, we do nothing and return here

            LOG.debug(_("Going to confirm migration %s") % migration_id,

                        context=context, instance=instance)

            try:

                # TODO(russellb) Why are we sending the migration object just

                # to turn around and look it up from the db again?

                #从migrate表中获取该instance resize记录

                migration = migration_obj.Migration.get_by_id(

                                    context.elevated(), migration_id)

            except exception.MigrationNotFound:

                LOG.error(_("Migration %s is not found during confirmation") %

                            migration_id, context=context, instance=instance)

                return

            if migration.status == 'confirmed': #如果已经被确认，则不能再次确认

                LOG.info(_("Migration %s is already confirmed") %

                            migration_id, context=context, instance=instance)

                return

            #如果迁移状态不是完成或者正在迁移，则数据有异常，直接return退出

            elif migration.status not in ('finished', 'confirming'):

                LOG.warn(_("Unexpected confirmation status '%(status)s' of "

                           "migration %(id)s, exit confirmation process") %

                           {"status": migration.status, "id": migration_id},

                           context=context, instance=instance)

                return

            # NOTE(wangpan): Get the instance from db, if it has been

            #                deleted, we do nothing and return here

            expected_attrs = ['metadata', 'system_metadata']

            try: #根据uuid获取instance信息

                instance = instance_obj.Instance.get_by_uuid(context,

                                instance.uuid, expected_attrs=expected_attrs)

            except exception.InstanceNotFound:

                LOG.info(_("Instance is not found during confirmation"),

                            context=context, instance=instance)

                return

           #跟进到5

            self._confirm_resize(context, instance, reservations=reservations,

                                 migration=migration)

        do_confirm_resize(context, instance, migration.id)#调用自身函数

5、/nova/compute/manage.py

    def _confirm_resize(self, context, instance, reservations=None,

                        migration=None):

        """Destroys the source instance."""

        #通知实例确认resize开始

        self._notify_about_instance_usage(context, instance,

                                          "resize.confirm.start")

        with self._error_out_instance_on_exception(context, instance['uuid'],

                                                   reservations):

            #删除之前保存的迁移信息，并更新instance为新的flavor数据

            # NOTE(danms): delete stashed migration information

            sys_meta, instance_type = self._cleanup_stored_instance_types(

                migration, instance)

            sys_meta.pop('old_vm_state', None)

            instance.system_metadata = sys_meta

            instance.save()

            # NOTE(tr3buchet): tear down networks on source host

            #清理源主机的网络信息

            self.network_api.setup_networks_on_host(context, instance,

                               migration.source_compute, teardown=True)

           #获取当前实例的网络信息

            network_info = self._get_instance_nw_info(context, instance)

            self.driver.confirm_migration(migration, instance, #跟进到代码6

                                          network_info)

            migration.status = 'confirmed' #设置迁移状态为确认完成

            migration.save(context.elevated())

            rt = self._get_resource_tracker(migration.source_node)

            #清理resize过程中claim的migrate信息

            rt.drop_resize_claim(instance, prefix='old_')

            # NOTE(mriedem): The old_vm_state could be STOPPED but the user

            # might have manually powered up the instance to confirm the

            # resize/migrate, so we need to check the current power state

            # on the instance and set the vm_state appropriately. We default

            # to ACTIVE because if the power state is not SHUTDOWN, we

            # assume _sync_instance_power_state will clean it up.

            p_state = instance.power_state #根据迁移前虚拟机状态设置vm_state 

            vm_state = None

            if p_state == power_state.SHUTDOWN:

                vm_state = vm_states.STOPPED

                LOG.debug(_("Resized/migrated instance is powered off. "

                          "Setting vm_state to '%s'."), vm_state,

                          instance=instance)

            else:

                vm_state = vm_states.ACTIVE

            instance.vm_state = vm_state

            instance.task_state = None

            instance.save(expected_task_state=[None, task_states.DELETING])

            self._notify_about_instance_usage( #通知实例resize确认完成

                context, instance, "resize.confirm.end",

                network_info=network_info)

            self._quota_commit(context, reservations)#更新配额信息

6、/nova/virt/libvirt/driver.py

    def confirm_migration(self, migration, instance, network_info):

        """Confirms a resize, destroying the source VM."""

        self._cleanup_resize(instance, network_info)#代码跟进到7

7、/nova/virt/libvirt/driver.py

    def _cleanup_resize(self, instance, network_info):

        target = libvirt_utils.get_instance_path(instance) + "_resize"

        if os.path.exists(target):

            # Deletion can fail over NFS, so retry the deletion as required.

            # Set maximum attempt as 5, most test can remove the directory

            # for the second time.

            #删除_resize文件夹，为了避免失败，重试5次

            utils.execute('rm', '-rf', target, delay_on_retry=True,

                          attempts=5)

        if instance['host'] != CONF.host:

            #调用libvirt接口，将该虚拟机undefine

            self._undefine_domain(instance)

            #网络及防火墙信息、配置清理

            self.unplug_vifs(instance, network_info)

            self.firewall_driver.unfilter_instance(instance, network_info)

确认confirm_resize完成。

四、resize完毕后的确认revert_resize代码详解分析

1、/nova/api/openstack/compute/servers.py

    @wsgi.response(202)

    @wsgi.serializers(xml=FullServerTemplate)

    @wsgi.deserializers(xml=ActionDeserializer)

    @wsgi.action('revertResize')

    def _action_revert_resize(self, req, id, body):

        context = req.environ['nova.context']

        instance = self._get_server(context, req, id) #从数据库获取instance信息

        try:

            self.compute_api.revert_resize(context, instance) #跟进到2

        except exception.MigrationNotFound:

            msg = _("Instance has not been resized.")

            raise exc.HTTPBadRequest(explanation=msg)

        except exception.FlavorNotFound:

            msg = _("Flavor used by the instance could not be found.")

            raise exc.HTTPBadRequest(explanation=msg)

        except exception.InstanceIsLocked as e:

            raise exc.HTTPConflict(explanation=e.format_message())

        except exception.InstanceInvalidState as state_error:

            common.raise_http_conflict_for_instance_invalid_state(state_error,

                    'revertResize')

        return webob.Response(status_int=202)

2、/nova/compute/api.py

    @wrap_check_policy

    @check_instance_lock

    @check_instance_cell

    @check_instance_state(vm_state=[vm_states.RESIZED])

    def revert_resize(self, context, instance):

        """Reverts a resize, deleting the 'new' instance in the process."""

        #回滚resize，并将new实例删除

        elevated = context.elevated()

        #从数据库migration 表中，获取迁移信息

        migration = migration_obj.Migration.get_by_instance_and_status(

            elevated, instance.uuid, 'finished') #

        # reverse quota reservation for increased resource usage

        #回滚配额信息

        deltas = self._reverse_upsize_quota_delta(context, migration)

        quotas = self._reserve_quota_delta(context, deltas)

        #实例状态设置为回滚

        instance.task_state = task_states.RESIZE_REVERTING

        try:

            instance.save(expected_task_state=[None])

        except Exception:

            with excutils.save_and_reraise_exception():

                quotas.rollback(context)

        migration.status = 'reverting' #迁移信息设置为回滚

        migration.save()

      &.bsp; # With cells, the best we can do right now is commit the reservations

        # immediate,y...

      &n"sp; if CONF.cells.enable:

            quotas.commit(context)

        self._record_action_start(context, instance, #通知实例回滚resize

                                  instance_actions.REVERT_RESIZE)

        self.compute_rpcapi.revert_resize(context, instance, #跟进到3

                                          migration,

                                          migration.dest_compute,

                                          quotas.reservations or [])

3、/nova/compute/rpcapi.py

    def revert_resize(self, ctxt, instance, migration, host,

                      reservations=None):

        # NOTE(russellb) Havana compat

        version = self._get_compat_version('3.0', '2.39')

        cctxt = self.client.prepare(server=_compute_host(host, instance),

                version=version)

        cctxt.cast(ctxt, 'revert_resize', #rpc调用，跟进到4

                   instance=instance, migration=migration,

                   reservations=reservations)

4、/nova/compute/manage.py

    @wrap_exception()

    @reverts_task_state

    @wrap_instance_event

    @wrap_instance_fault

    def revert_resize(self, context, instance, migration, reservations):

        """Destroys the new instance on the destination machine.

        Reverts the model changes, and powers on the old instance on the

        source machine.

        """

        #删除目的机器的new实例，回滚resize改变、将源主机的实例断电

        # NOTE(comstud): A revert_resize is essentially a resize back to

        # the old size, so we need to send a usage event here.

        self.conductor_api.notify_usage_exists(

                context, instance, current_period=True)

        with self._error_out_instance_on_exception(context, instance['uuid'],

                                                   reservations):

            # NOTE(tr3buchet): tear down networks on destination host

            #目标主机的网络信息清除

            self.network_api.setup_networks_on_host(context, instance,

                                                    teardown=True)

           #获取原始的实例，迁移信息

            instance_p = obj_base.obj_to_primitive(instance)

            migration_p = obj_base.obj_to_primitive(migration)

            #网络迁移开始

            self.conductor_api.network_migrate_instance_start(context,

                                                              instance_p,

                                                              migration_p)

           #获取实例网络及磁盘设备信息

            network_info = self._get_instance_nw_info(context, instance)

            bdms = (block_device_obj.BlockDeviceMappingList.

                    get_by_instance_uuid(context, instance.uuid))

            block_device_info = self._get_instance_volume_block_device_info(

                                context, instance, bdms=bdms)

           #销毁实例

            self.driver.destroy(context, instance, network_info,

                                block_device_info)

           #断开与卷连接

            self._terminate_volume_connections(context, instance, bdms)

           #设置迁移状态

            migration.status = 'reverted'

            migration.save(context.elevated())

            rt = self._get_resource_tracker(instance.node)

            rt.drop_resize_claim(instance)

        %r6nbsp;  #跟进到5

            self.compute_rpcapi.finish_revert_resize(context, instance,

                    migration, migration.source_compute,

                    reservations=reservations)

5、/nova/compute/rpcapi.py

    def finish_revert_resize(self, ctxt, instance, migration, host,

                             reservations=None):

        # NOTE(russellb) Havana compat

        version = self._get_compat_version('3.0', '2.47')

        cctxt = self.client.prepare(server=host, version=version)

        cctxt.cast(ctxt, 'finish_revert_resize', #rpc调用，跟进到6

                   instance=instance, migration=migration,

                   reservations=reservations)

6、/nova/compute/manage.py

    @wrap_exception()

    @reverts_task_state

    @wrap_instance_event

    @wrap_instance_fault

    def finish_revert_resize(self, context, instance, reservations, migration):

        """Finishes the second half of reverting a resize.

        Bring the original source instance state back (active/shutoff) and

        revert the resized attributes in the database.

        """

        with self._error_out_instance_on_exception(context, instance.uuid,

                              %2vnbsp;                    reservations):

            network_info = self._get_instance_nw_info(context, instance)

            self._notify_about_instance_usage( #通知resize回滚开始

                    context, instance, "resize.revert.start")

            sys_meta, instance_type = self._cleanup_stored_instance_types(

                migration, instance, True)

            # NOTE(mriedem): delete stashed old_vm_state information; we

            # default to ACTIVE for backwards compatibility if old_vm_state

            # is not set

            old_vm_state = sys_meta.pop('old_vm_state', vm_states.ACTIVE)

           #回滚数据库实例信息

            instance.system_metadata = sys_meta

            instance.memory_mb = instance_type['memory_mb']

            instance.vcpus = instance_type['vcpus']

            instance.root_gb = instance_type['root_gb']

            instance.ephemeral_gb = instance_type['ephemeral_gb']

            instance.instance_type_id = instance_type['id']

            instance.host = migration['source_compute']

            instance.node = migration[gsource_node']

        &n"sp;   instance.save()

            #源主机上创建网络

            self.network_api.setup_networks_on_host(context, instance,

                                            migration['source_compute'])

            #获取磁盘信息

            block_device_info = self._get_instance_volume_block_device_info(

                    context, instance, refresh_conn_info=True)

            power_on = old_vm_state != vm_states.STOPPED

            #完成回滚迁移，跟进到7

            self.driver.finish_revert_migration(context, instance,

                                       network_info,

                                       block_device_info, power_on)

            #设置虚拟机状态信息

            instance.launched_at = timeutils.utcnow()

            instance.save(expected_task_state=task_states.RESIZE_REVERTING)

            instance_p = obj_base.obj_to_primitive(instance)

            migration_p = obj_base.obj_to_primitive(migration)

            #迁移虚拟机网络

            self.conductor_api.network_migrate_instance_finish(context,

                                                               instance_p,

                                                               migration_p)

            # if the original vm state was STOPPED, set it back to STOPPED

            LOG.info(_("Updating instance to original state: '%s'") %

                     old_vm_state)

            #设置虚拟机相应运行状态

            if power_on:

                instance.vm_state = vm_states.ACTIVE

                instance.task_state = None

                instance.save()

            else:

                instance.task_state = task_states.POWERING_OFF

                instance.save()

                self.stop_instance(context, instance=instance)

            self._notify_about_instance_usage(#通知回滚完成

                    context, instance, "resize.revert.end")

            self._quota_commit(context, reservations)#配额更新

7、/nova/compute/virt/libvirt/driver.py

    def finish_revert_migration(self, context, instance, network_info,

                                block_device_info=None, power_on=True):

        LOG.debug(_("Starting finish_revert_migration"),

                   instance=instance)

       #获取实例路径及resize路径

        inst_base = libvirt_utils.get_instance_path(instance)

        inst_base_resize = inst_base + "_resize"

        # NOTE(danms): if we're recovering from a failed migration,

        # make sure we don't have a left-over same-host base directory

        # that would conflict. Also, don't fail on the rename if the

        # failure happened early.

        if os.path.exists(inst_base_resize): #inst_base_resize 路径存在，清理

            self._cleanup_failed_migration(inst_base)

            utils.execute('mv', inst_base_resize, inst_base)

       #获取磁盘信息

        disk_info = blockinfo.get_disk_info(CONF.libvirt.virt_type,

                                            instance,

                                            block_device_info)

        #根据上述信息及配置，生成xml文件，创建虚拟机

        xml = self.to_xml(context, instance, network_info, disk_info,

                          block_device_info=block_device_info)

        self._create_domain_and_network(context, xml, instance, network_info,

                                        block_device_info, power_on)

        if power_on:

            timer = loopingcall.FixedIntervalLoopingCall(

                                                    self._wait_for_running,

                                                    instance)

            timer.start(interval=0.5).wait()

至此回滚resize分析结束。