¡@

Home 

OpenStack Study: messaging.py

OpenStack Index

**** CubicPower OpenStack Study ****

# Copyright (c) 2012 Rackspace Hosting

# All Rights Reserved.

# Copyright 2010 United States Government as represented by the

# Administrator of the National Aeronautics and Space Administration.

# All Rights Reserved.

# Copyright 2013 Red Hat, Inc.

#

# Licensed under the Apache License, Version 2.0 (the "License"); you may

# not use this file except in compliance with the License. You may obtain

# a copy of the License at

#

# http://www.apache.org/licenses/LICENSE-2.0

#

# Unless required by applicable law or agreed to in writing, software

# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT

# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the

# License for the specific language governing permissions and limitations

# under the License.

"""

Cell messaging module.

This module defines the different message types that are passed between

cells and the methods that they can call when the target cell has been

reached.

The interface into this module is the MessageRunner class.

"""

import sys

import traceback

from eventlet import queue

from oslo.config import cfg

from oslo import messaging

import six

from nova.cells import state as cells_state

from nova.cells import utils as cells_utils

from nova import compute

from nova.compute import rpcapi as compute_rpcapi

from nova.compute import task_states

from nova.compute import vm_states

from nova.consoleauth import rpcapi as consoleauth_rpcapi

from nova import context

from nova.db import base

from nova import exception

from nova.network import model as network_model

from nova.objects import base as objects_base

from nova.objects import instance as instance_obj

from nova.openstack.common import excutils

from nova.openstack.common.gettextutils import _

from nova.openstack.common import importutils

from nova.openstack.common import jsonutils

from nova.openstack.common import log as logging

from nova.openstack.common import timeutils

from nova.openstack.common import uuidutils

from nova import rpc

from nova import utils

cell_messaging_opts = [

cfg.IntOpt('max_hop_count',

default=10,

help='Maximum number of hops for cells routing.'),

cfg.StrOpt('scheduler',

default='nova.cells.scheduler.CellsScheduler',

help='Cells scheduler to use')]

CONF = cfg.CONF

CONF.import_opt('name', 'nova.cells.opts', group='cells')

CONF.import_opt('call_timeout', 'nova.cells.opts', group='cells')

CONF.register_opts(cell_messaging_opts, group='cells')

LOG = logging.getLogger(__name__)

# Separator used between cell names for the 'full cell name' and routing

# path.

_PATH_CELL_SEP = cells_utils.PATH_CELL_SEP

**** CubicPower OpenStack Study ****

def _reverse_path(path):

    """Reverse a path.  Used for sending responses upstream."""

    path_parts = path.split(_PATH_CELL_SEP)

    path_parts.reverse()

    return _PATH_CELL_SEP.join(path_parts)

**** CubicPower OpenStack Study ****

def _response_cell_name_from_path(routing_path, neighbor_only=False):

    """Reverse the routing_path.  If we only want to send to our parent,

    set neighbor_only to True.

    """

    path = _reverse_path(routing_path)

    if not neighbor_only or len(path) == 1:

        return path

    return _PATH_CELL_SEP.join(path.split(_PATH_CELL_SEP)[:2])

#

# Message classes.

#

**** CubicPower OpenStack Study ****

class _BaseMessage(object):

"""Base message class. It

**** CubicPower OpenStack Study ****

    def __init__(self, msg_runner, ctxt, method_name, method_kwargs,

            direction, need_response=False, fanout=False, uuid=None,

            routing_path=None, hop_count=0, max_hop_count=None,

            **kwargs):

        self.ctxt = ctxt

        self.resp_queue = None

        self.msg_runner = msg_runner

        self.state_manager = msg_runner.state_manager

        # Copy these.

        self.base_attrs_to_json = self.base_attrs_to_json[:]

        # Normally this would just be CONF.cells.name, but going through

        # the msg_runner allows us to stub it more easily.

        self.our_path_part = self.msg_runner.our_name

        self.uuid = uuid

        if self.uuid is None:

            self.uuid = uuidutils.generate_uuid()

        self.method_name = method_name

        self.method_kwargs = method_kwargs

        self.direction = direction

        self.need_response = need_response

        self.fanout = fanout

        self.routing_path = routing_path

        self.hop_count = hop_count

        if max_hop_count is None:

            max_hop_count = CONF.cells.max_hop_count

        self.max_hop_count = max_hop_count

        self.is_broadcast = False

        self._append_hop()

        # Each sub-class should set this when the message is inited

        self.next_hops = []

        self.resp_queue = None

        self.serializer = objects_base.NovaObjectSerializer()

**** CubicPower OpenStack Study ****

    def __repr__(self):

        _dict = self._to_dict()

        _dict.pop('method_kwargs')

        return "<%s: %s>" % (self.__class__.__name__, _dict)

**** CubicPower OpenStack Study ****

    def _append_hop(self):

        """Add our hop to the routing_path."""

        routing_path = (self.routing_path and

                self.routing_path + _PATH_CELL_SEP or '')

        self.routing_path = routing_path + self.our_path_part

        self.hop_count += 1

**** CubicPower OpenStack Study ****

    def _process_locally(self):

        """Its been determined that we should process this message in this

        cell.  Go through the MessageRunner to call the appropriate

        method for this message.  Catch the response and/or exception and

        encode it within a Response instance.  Return it so the caller

        can potentially return it to another cell... or return it to

        a caller waiting in this cell.

        """

        try:

            resp_value = self.msg_runner._process_message_locally(self)

            failure = False

        except Exception as exc:

            resp_value = sys.exc_info()

            failure = True

            LOG.exception(_("Error processing message locally: %(exc)s"),

                          {'exc': exc})

        return Response(self.routing_path, resp_value, failure)

**** CubicPower OpenStack Study ****

    def _setup_response_queue(self):

        """Shortcut to creating a response queue in the MessageRunner."""

        self.resp_queue = self.msg_runner._setup_response_queue(self)

**** CubicPower OpenStack Study ****

    def _cleanup_response_queue(self):

        """Shortcut to deleting a response queue in the MessageRunner."""

        if self.resp_queue:

            self.msg_runner._cleanup_response_queue(self)

            self.resp_queue = None

**** CubicPower OpenStack Study ****

    def _wait_for_json_responses(self, num_responses=1):

        """Wait for response(s) to be put into the eventlet queue.  Since

        each queue entry actually contains a list of JSON-ified responses,

        combine them all into a single list to return.

        Destroy the eventlet queue when done.

        """

        if not self.resp_queue:

            # Source is not actually expecting a response

            return

        responses = []

        wait_time = CONF.cells.call_timeout

        try:

            for x in xrange(num_responses):

                json_responses = self.resp_queue.get(timeout=wait_time)

                responses.extend(json_responses)

        except queue.Empty:

            raise exception.CellTimeout()

        finally:

            self._cleanup_response_queue()

        return responses

**** CubicPower OpenStack Study ****

    def _send_json_responses(self, json_responses, neighbor_only=False,

            fanout=False):

        """Send list of responses to this message.  Responses passed here

        are JSON-ified.  Targeted messages have a single response while

        Broadcast messages may have multiple responses.

        If this cell was the source of the message, these responses will

        be returned from self.process().

        Otherwise, we will route the response to the source of the

        request.  If 'neighbor_only' is True, the response will be sent

        to the neighbor cell, not the original requester.  Broadcast

        messages get aggregated at each hop, so neighbor_only will be

        True for those messages.

        """

        if not self.need_response:

            return

        if self.source_is_us():

            responses = []

            for json_response in json_responses:

                responses.append(Response.from_json(json_response))

            return responses

        direction = self.direction == 'up' and 'down' or 'up'

        response_kwargs = {'orig_message': self.to_json(),

                           'responses': json_responses}

        target_cell = _response_cell_name_from_path(self.routing_path,

                neighbor_only=neighbor_only)

        response = self.msg_runner._create_response_message(self.ctxt,

                direction, target_cell, self.uuid, response_kwargs,

                fanout=fanout)

        response.process()

**** CubicPower OpenStack Study ****

    def _send_response(self, response, neighbor_only=False):

        """Send a response to this message.  If the source of the

        request was ourselves, just return the response.  It'll be

        passed back to the caller of self.process().  See DocString for

        _send_json_responses() as it handles most of the real work for

        this method.

        'response' is an instance of Response class.

        """

        if not self.need_response:

            return

        if self.source_is_us():

            return response

        self._send_json_responses([response.to_json()],

                                  neighbor_only=neighbor_only)

**** CubicPower OpenStack Study ****

    def _send_response_from_exception(self, exc_info):

        """Take an exception as returned from sys.exc_info(), encode

        it in a Response, and send it.

        """

        response = Response(self.routing_path, exc_info, True)

        return self._send_response(response)

**** CubicPower OpenStack Study ****

    def _to_dict(self):

        """Convert a message to a dictionary.  Only used internally."""

        _dict = {}

        for key in self.base_attrs_to_json:

            _dict[key] = getattr(self, key)

        return _dict

**** CubicPower OpenStack Study ****

    def to_json(self):

        """Convert a message into JSON for sending to a sibling cell."""

        _dict = self._to_dict()

        # Convert context to dict.

        _dict['ctxt'] = _dict['ctxt'].to_dict()

        # NOTE(comstud): 'method_kwargs' needs special serialization

        # because it may contain objects.

        method_kwargs = _dict['method_kwargs']

        for k, v in method_kwargs.items():

            method_kwargs[k] = self.serializer.serialize_entity(self.ctxt, v)

        return jsonutils.dumps(_dict)

**** CubicPower OpenStack Study ****

    def source_is_us(self):

        """Did this cell create this message?"""

        return self.routing_path == self.our_path_part

**** CubicPower OpenStack Study ****

    def process(self):

        """Process a message.  Deal with it locally and/or forward it to a

        sibling cell.

        Override in a subclass.

        """

        raise NotImplementedError()

**** CubicPower OpenStack Study ****

class _TargetedMessage(_BaseMessage):

"""A targeted message is a message that is destined for a specific

single cell.

'target_cell' can be a full cell name like 'api!child-cell' or it can

be an instance of the CellState class if the target is a neighbor cell.

"""

message_type = 'targeted'

**** CubicPower OpenStack Study ****

    def __init__(self, msg_runner, ctxt, method_name, method_kwargs,

            direction, target_cell, **kwargs):

        super(_TargetedMessage, self).__init__(msg_runner, ctxt,

                method_name, method_kwargs, direction, **kwargs)

        if isinstance(target_cell, cells_state.CellState):

            # Neighbor cell or ourselves.  Convert it to a 'full path'.

            if target_cell.is_me:

                target_cell = self.our_path_part

            else:

                target_cell = '%s%s%s' % (self.our_path_part,

                                          _PATH_CELL_SEP,

                                          target_cell.name)

        self.target_cell = target_cell

        self.base_attrs_to_json.append('target_cell')

**** CubicPower OpenStack Study ****

    def _get_next_hop(self):

        """Return the cell name for the next hop.  If the next hop is

        the current cell, return None.

        """

        if self.target_cell == self.routing_path:

            return self.state_manager.my_cell_state

        target_cell = self.target_cell

        routing_path = self.routing_path

        current_hops = routing_path.count(_PATH_CELL_SEP)

        next_hop_num = current_hops + 1

        dest_hops = target_cell.count(_PATH_CELL_SEP)

        if dest_hops < current_hops:

            reason_args = {'target_cell': target_cell,

                           'routing_path': routing_path}

            reason = _("destination is %(target_cell)s but routing_path "

                       "is %(routing_path)s") % reason_args

            raise exception.CellRoutingInconsistency(reason=reason)

        dest_name_parts = target_cell.split(_PATH_CELL_SEP)

        if (_PATH_CELL_SEP.join(dest_name_parts[:next_hop_num]) !=

                routing_path):

            reason_args = {'target_cell': target_cell,

                           'routing_path': routing_path}

            reason = _("destination is %(target_cell)s but routing_path "

                       "is %(routing_path)s") % reason_args

            raise exception.CellRoutingInconsistency(reason=reason)

        next_hop_name = dest_name_parts[next_hop_num]

        if self.direction == 'up':

            next_hop = self.state_manager.get_parent_cell(next_hop_name)

        else:

            next_hop = self.state_manager.get_child_cell(next_hop_name)

        if not next_hop:

            cell_type = 'parent' if self.direction == 'up' else 'child'

            reason_args = {'cell_type': cell_type,

                           'target_cell': target_cell}

            reason = _("Unknown %(cell_type)s when routing to "

                       "%(target_cell)s") % reason_args

            raise exception.CellRoutingInconsistency(reason=reason)

        return next_hop

**** CubicPower OpenStack Study ****

    def process(self):

        """Process a targeted message.  This is called for all cells

        that touch this message.  If the local cell is the one that

        created this message, we reply directly with a Response instance.

        If the local cell is not the target, an eventlet queue is created

        and we wait for the response to show up via another thread

        receiving the Response back.

        Responses to targeted messages are routed directly back to the

        source.  No eventlet queues are created in intermediate hops.

        All exceptions for processing the message across the whole

        routing path are caught and encoded within the Response and

        returned to the caller.

        """

        try:

            next_hop = self._get_next_hop()

        except Exception as exc:

            exc_info = sys.exc_info()

            LOG.exception(_("Error locating next hop for message: %(exc)s"),

                          {'exc': exc})

            return self._send_response_from_exception(exc_info)

        if next_hop.is_me:

            # Final destination.

            response = self._process_locally()

            return self._send_response(response)

        # Need to forward via neighbor cell.

        if self.need_response and self.source_is_us():

            # A response is needed and the source of the message is

            # this cell.  Create the eventlet queue.

            self._setup_response_queue()

            wait_for_response = True

        else:

            wait_for_response = False

        try:

            # This is inside the try block, so we can encode the

            # exception and return it to the caller.

            if self.hop_count >= self.max_hop_count:

                raise exception.CellMaxHopCountReached(

                        hop_count=self.hop_count)

            next_hop.send_message(self)

        except Exception as exc:

            exc_info = sys.exc_info()

            err_str = _("Failed to send message to cell: %(next_hop)s: "

                        "%(exc)s")

            LOG.exception(err_str, {'exc': exc, 'next_hop': next_hop})

            self._cleanup_response_queue()

            return self._send_response_from_exception(exc_info)

        if wait_for_response:

            # Targeted messages only have 1 response.

            remote_response = self._wait_for_json_responses()[0]

            return Response.from_json(remote_response)

**** CubicPower OpenStack Study ****

class _BroadcastMessage(_BaseMessage):

"""A broadcast message. This means to call a method in every single

cell going in a certain direction.

"""

message_type = 'broadcast'

**** CubicPower OpenStack Study ****

    def __init__(self, msg_runner, ctxt, method_name, method_kwargs,

            direction, run_locally=True, **kwargs):

        super(_BroadcastMessage, self).__init__(msg_runner, ctxt,

                method_name, method_kwargs, direction, **kwargs)

        # The local cell creating this message has the option

        # to be able to process the message locally or not.

        self.run_locally = run_locally

        self.is_broadcast = True

**** CubicPower OpenStack Study ****

    def _get_next_hops(self):

        """Set the next hops and return the number of hops.  The next

        hops may include ourself.

        """

        if self.hop_count >= self.max_hop_count:

            return []

        if self.direction == 'down':

            return self.state_manager.get_child_cells()

        else:

            return self.state_manager.get_parent_cells()

**** CubicPower OpenStack Study ****

    def _send_to_cells(self, target_cells):

        """Send a message to multiple cells."""

        for cell in target_cells:

            cell.send_message(self)

**** CubicPower OpenStack Study ****

    def _send_json_responses(self, json_responses):

        """Responses to broadcast messages always need to go to the

        neighbor cell from which we received this message.  That

        cell aggregates the responses and makes sure to forward them

        to the correct source.

        """

        return super(_BroadcastMessage, self)._send_json_responses(

                json_responses, neighbor_only=True, fanout=True)

**** CubicPower OpenStack Study ****

    def process(self):

        """Process a broadcast message.  This is called for all cells

        that touch this message.

        The message is sent to all cells in the certain direction and

        the creator of this message has the option of whether or not

        to process it locally as well.

        If responses from all cells are required, each hop creates an

        eventlet queue and waits for responses from its immediate

        neighbor cells.  All responses are then aggregated into a

        single list and are returned to the neighbor cell until the

        source is reached.

        When the source is reached, a list of Response instances are

        returned to the caller.

        All exceptions for processing the message across the whole

        routing path are caught and encoded within the Response and

        returned to the caller.  It is possible to get a mix of

        successful responses and failure responses.  The caller is

        responsible for dealing with this.

        """

        try:

            next_hops = self._get_next_hops()

        except Exception as exc:

            exc_info = sys.exc_info()

            LOG.exception(_("Error locating next hops for message: %(exc)s"),

                          {'exc': exc})

            return self._send_response_from_exception(exc_info)

        # Short circuit if we don't need to respond

        if not self.need_response:

            if self.run_locally:

                self._process_locally()

            self._send_to_cells(next_hops)

            return

        # We'll need to aggregate all of the responses (from ourself

        # and our sibling cells) into 1 response

        try:

            self._setup_response_queue()

            self._send_to_cells(next_hops)

        except Exception as exc:

            # Error just trying to send to cells.  Send a single response

            # with the failure.

            exc_info = sys.exc_info()

            LOG.exception(_("Error sending message to next hops: %(exc)s"),

                          {'exc': exc})

            self._cleanup_response_queue()

            return self._send_response_from_exception(exc_info)

        if self.run_locally:

            # Run locally and store the Response.

            local_response = self._process_locally()

        else:

            local_response = None

        try:

            remote_responses = self._wait_for_json_responses(

                    num_responses=len(next_hops))

        except Exception as exc:

            # Error waiting for responses, most likely a timeout.

            # Send a single response back with the failure.

            exc_info = sys.exc_info()

            err_str = _("Error waiting for responses from neighbor cells: "

                        "%(exc)s")

            LOG.exception(err_str, {'exc': exc})

            return self._send_response_from_exception(exc_info)

        if local_response:

            remote_responses.append(local_response.to_json())

        return self._send_json_responses(remote_responses)

**** CubicPower OpenStack Study ****

class _ResponseMessage(_TargetedMessage):

"""A response message is really just a special targeted message,

saying to call 'parse_responses' when we reach the source of a 'call'.

The 'fanout' attribute on this message may be true if we're responding

to a broadcast or if we're about to respond to the source of an

original target message. Because multiple nova-cells services may

be running within a cell, we need to make sure the response gets

back to the correct one, so we have to fanout.

"""

message_type = 'response'

**** CubicPower OpenStack Study ****

    def __init__(self, msg_runner, ctxt, method_name, method_kwargs,

            direction, target_cell, response_uuid, **kwargs):

        super(_ResponseMessage, self).__init__(msg_runner, ctxt,

                method_name, method_kwargs, direction, target_cell, **kwargs)

        self.response_uuid = response_uuid

        self.base_attrs_to_json.append('response_uuid')

**** CubicPower OpenStack Study ****

    def process(self):

        """Process a response.  If the target is the local cell, process

        the response here.  Otherwise, forward it to where it needs to

        go.

        """

        next_hop = self._get_next_hop()

        if next_hop.is_me:

            self._process_locally()

            return

        if self.fanout is False:

            # Really there's 1 more hop on each of these below, but

            # it doesn't matter for this logic.

            target_hops = self.target_cell.count(_PATH_CELL_SEP)

            current_hops = self.routing_path.count(_PATH_CELL_SEP)

            if current_hops + 1 == target_hops:

                # Next hop is the target.. so we must fanout.  See

                # DocString above.

                self.fanout = True

        next_hop.send_message(self)

#

# Methods that may be called when processing messages after reaching

# a target cell.

#

**** CubicPower OpenStack Study ****

class _BaseMessageMethods(base.Base):

"""Base class for

**** CubicPower OpenStack Study ****

    def __init__(self, msg_runner):

        super(_BaseMessageMethods, self).__init__()

        self.msg_runner = msg_runner

        self.state_manager = msg_runner.state_manager

        self.compute_api = compute.API()

        self.compute_rpcapi = compute_rpcapi.ComputeAPI()

        self.consoleauth_rpcapi = consoleauth_rpcapi.ConsoleAuthAPI()

        self.host_api = compute.HostAPI()

**** CubicPower OpenStack Study ****

    def task_log_get_all(self, message, task_name, period_beginning,

                         period_ending, host, state):

        """Get task logs from the DB.  The message could have

        directly targeted this cell, or it could have been a broadcast

        message.

        If 'host' is not None, filter by host.

        If 'state' is not None, filter by state.

        """

        task_logs = self.db.task_log_get_all(message.ctxt, task_name,

                                             period_beginning,

                                             period_ending,

                                             host=host,

                                             state=state)

        return jsonutils.to_primitive(task_logs)

**** CubicPower OpenStack Study ****

class _ResponseMessageMethods(_BaseMessageMethods):

"""Methods that are called from a ResponseMessage. There's only

1 method (parse_responses) and it is called when the message reaches

the source of a 'call'. All we do is stuff the response into the

eventlet queue to signal the caller that's waiting.

"""

**** CubicPower OpenStack Study ****

    def parse_responses(self, message, orig_message, responses):

        self.msg_runner._put_response(message.response_uuid,

                responses)

**** CubicPower OpenStack Study ****

class _TargetedMessageMethods(_BaseMessageMethods):

"""These are the methods that can be called when routing a message

to a specific cell.

"""

**** CubicPower OpenStack Study ****

    def __init__(self, *args, **kwargs):

        super(_TargetedMessageMethods, self).__init__(*args, **kwargs)

**** CubicPower OpenStack Study ****

    def schedule_run_instance(self, message, host_sched_kwargs):

        """Parent cell told us to schedule new instance creation."""

        self.msg_runner.scheduler.run_instance(message, host_sched_kwargs)

**** CubicPower OpenStack Study ****

    def build_instances(self, message, build_inst_kwargs):

        """Parent cell told us to schedule new instance creation."""

        self.msg_runner.scheduler.build_instances(message, build_inst_kwargs)

**** CubicPower OpenStack Study ****

    def run_compute_api_method(self, message, method_info):

        """Run a method in the compute api class."""

        method = method_info['method']

        fn = getattr(self.compute_api, method, None)

        if not fn:

            detail = _("Unknown method '%(method)s' in compute API")

            raise exception.CellServiceAPIMethodNotFound(

                    detail=detail % {'method': method})

        args = list(method_info['method_args'])

        # 1st arg is instance_uuid that we need to turn into the

        # instance object.

        instance_uuid = args[0]

        try:

            instance = self.db.instance_get_by_uuid(message.ctxt,

                                                    instance_uuid)

        except exception.InstanceNotFound:

            with excutils.save_and_reraise_exception():

                # Must be a race condition.  Let's try to resolve it by

                # telling the top level cells that this instance doesn't

                # exist.

                instance = {'uuid': instance_uuid}

                self.msg_runner.instance_destroy_at_top(message.ctxt,

                                                        instance)

        # FIXME(comstud): This is temporary/transitional until I can

        # work out a better way to pass full objects down.

        EXPECTS_OBJECTS = ['start', 'stop', 'delete_instance_metadata',

                           'update_instance_metadata']

        if method in EXPECTS_OBJECTS:

            inst_obj = instance_obj.Instance()

            inst_obj._from_db_object(message.ctxt, inst_obj, instance)

            instance = inst_obj

        args[0] = instance

        return fn(message.ctxt, *args, **method_info['method_kwargs'])

**** CubicPower OpenStack Study ****

    def update_capabilities(self, message, cell_name, capabilities):

        """A child cell told us about their capabilities."""

        LOG.debug(_("Received capabilities from child cell "

                    "%(cell_name)s: %(capabilities)s"),

                  {'cell_name': cell_name, 'capabilities': capabilities})

        self.state_manager.update_cell_capabilities(cell_name,

                capabilities)

        # Go ahead and update our parents now that a child updated us

        self.msg_runner.tell_parents_our_capabilities(message.ctxt)

**** CubicPower OpenStack Study ****

    def update_capacities(self, message, cell_name, capacities):

        """A child cell told us about their capacity."""

        LOG.debug(_("Received capacities from child cell "

                    "%(cell_name)s: %(capacities)s"),

                  {'cell_name': cell_name, 'capacities': capacities})

        self.state_manager.update_cell_capacities(cell_name,

                capacities)

        # Go ahead and update our parents now that a child updated us

        self.msg_runner.tell_parents_our_capacities(message.ctxt)

**** CubicPower OpenStack Study ****

    def announce_capabilities(self, message):

        """A parent cell has told us to send our capabilities, so let's

        do so.

        """

        self.msg_runner.tell_parents_our_capabilities(message.ctxt)

**** CubicPower OpenStack Study ****

    def announce_capacities(self, message):

        """A parent cell has told us to send our capacity, so let's

        do so.

        """

        self.msg_runner.tell_parents_our_capacities(message.ctxt)

**** CubicPower OpenStack Study ****

    def service_get_by_compute_host(self, message, host_name):

        """Return the service entry for a compute host."""

        service = self.db.service_get_by_compute_host(message.ctxt,

                                                      host_name)

        return jsonutils.to_primitive(service)

**** CubicPower OpenStack Study ****

    def service_update(self, message, host_name, binary, params_to_update):

        """Used to enable/disable a service. For compute services, setting to

        disabled stops new builds arriving on that host.

        :param host_name: the name of the host machine that the service is

                          running

        :param binary: The name of the executable that the service runs as

        :param params_to_update: eg. {'disabled': True}

        """

        return jsonutils.to_primitive(

            self.host_api.service_update(message.ctxt, host_name, binary,

                                         params_to_update))

**** CubicPower OpenStack Study ****

    def service_delete(self, message, service_id):

        """Deletes the specified service."""

        self.host_api.service_delete(message.ctxt, service_id)

**** CubicPower OpenStack Study ****

    def proxy_rpc_to_manager(self, message, host_name, rpc_message,

                             topic, timeout):

        """Proxy RPC to the given compute topic."""

        # Check that the host exists.

        self.db.service_get_by_compute_host(message.ctxt, host_name)

        topic, _sep, server = topic.partition('.')

        cctxt = rpc.get_client(messaging.Target(topic=topic,

                                                server=server or None))

        method = rpc_message['method']

        kwargs = rpc_message['args']

        if message.need_response:

            cctxt = cctxt.prepare(timeout=timeout)

            return cctxt.call(message.ctxt, method, **kwargs)

        else:

            cctxt.cast(message.ctxt, method, **kwargs)

**** CubicPower OpenStack Study ****

    def compute_node_get(self, message, compute_id):

        """Get compute node by ID."""

        compute_node = self.db.compute_node_get(message.ctxt,

                                                compute_id)

        return jsonutils.to_primitive(compute_node)

**** CubicPower OpenStack Study ****

    def actions_get(self, message, instance_uuid):

        actions = self.db.actions_get(message.ctxt, instance_uuid)

        return jsonutils.to_primitive(actions)

**** CubicPower OpenStack Study ****

    def action_get_by_request_id(self, message, instance_uuid, request_id):

        action = self.db.action_get_by_request_id(message.ctxt, instance_uuid,

                                                  request_id)

        return jsonutils.to_primitive(action)

**** CubicPower OpenStack Study ****

    def action_events_get(self, message, action_id):

        action_events = self.db.action_events_get(message.ctxt, action_id)

        return jsonutils.to_primitive(action_events)

**** CubicPower OpenStack Study ****

    def validate_console_port(self, message, instance_uuid, console_port,

                              console_type):

        """Validate console port with child cell compute node."""

        # 1st arg is instance_uuid that we need to turn into the

        # instance object.

        try:

            instance = self.db.instance_get_by_uuid(message.ctxt,

                                                    instance_uuid)

        except exception.InstanceNotFound:

            with excutils.save_and_reraise_exception():

                # Must be a race condition.  Let's try to resolve it by

                # telling the top level cells that this instance doesn't

                # exist.

                instance = {'uuid': instance_uuid}

                self.msg_runner.instance_destroy_at_top(message.ctxt,

                                                        instance)

        return self.compute_rpcapi.validate_console_port(message.ctxt,

                instance, console_port, console_type)

**** CubicPower OpenStack Study ****

    def get_migrations(self, message, filters):

        return self.compute_api.get_migrations(message.ctxt, filters)

**** CubicPower OpenStack Study ****

    def instance_update_from_api(self, message, instance,

                                 expected_vm_state,

                                 expected_task_state,

                                 admin_state_reset):

        """Update an instance in this cell."""

        if not admin_state_reset:

            # NOTE(comstud): We don't want to nuke this cell's view

            # of vm_state and task_state unless it's a forced reset

            # via admin API.

            instance.obj_reset_changes(['vm_state', 'task_state'])

        # NOTE(alaski): A cell should be authoritative for its system_metadata

        # and metadata so we don't want to sync it down from the api.

        instance.obj_reset_changes(['metadata', 'system_metadata'])

        instance.save(message.ctxt, expected_vm_state=expected_vm_state,

                      expected_task_state=expected_task_state)

**** CubicPower OpenStack Study ****

    def _call_compute_api_with_obj(self, ctxt, instance, method, *args,

                                   **kwargs):

        try:

            # NOTE(comstud): We need to refresh the instance from this

            # cell's view in the DB.

            instance.refresh(ctxt)

        except exception.InstanceNotFound:

            with excutils.save_and_reraise_exception():

                # Must be a race condition.  Let's try to resolve it by

                # telling the top level cells that this instance doesn't

                # exist.

                instance = {'uuid': instance.uuid}

                self.msg_runner.instance_destroy_at_top(ctxt,

                                                        instance)

        fn = getattr(self.compute_api, method, None)

        return fn(ctxt, instance, *args, **kwargs)

**** CubicPower OpenStack Study ****

    def start_instance(self, message, instance):

        """Start an instance via compute_api.start()."""

        self._call_compute_api_with_obj(message.ctxt, instance, 'start')

**** CubicPower OpenStack Study ****

    def stop_instance(self, message, instance):

        """Stop an instance via compute_api.stop()."""

        do_cast = not message.need_response

        return self._call_compute_api_with_obj(message.ctxt, instance,

                                               'stop', do_cast=do_cast)

**** CubicPower OpenStack Study ****

    def reboot_instance(self, message, instance, reboot_type):

        """Reboot an instance via compute_api.reboot()."""

        self._call_compute_api_with_obj(message.ctxt, instance, 'reboot',

                                        reboot_type=reboot_type)

**** CubicPower OpenStack Study ****

    def suspend_instance(self, message, instance):

        """Suspend an instance via compute_api.suspend()."""

        self._call_compute_api_with_obj(message.ctxt, instance, 'suspend')

**** CubicPower OpenStack Study ****

    def resume_instance(self, message, instance):

        """Resume an instance via compute_api.suspend()."""

        self._call_compute_api_with_obj(message.ctxt, instance, 'resume')

**** CubicPower OpenStack Study ****

    def get_host_uptime(self, message, host_name):

        return self.host_api.get_host_uptime(message.ctxt, host_name)

**** CubicPower OpenStack Study ****

    def terminate_instance(self, message, instance):

        self._call_compute_api_with_obj(message.ctxt, instance, 'delete')

**** CubicPower OpenStack Study ****

    def soft_delete_instance(self, message, instance):

        self._call_compute_api_with_obj(message.ctxt, instance, 'soft_delete')

**** CubicPower OpenStack Study ****

    def pause_instance(self, message, instance):

        """Pause an instance via compute_api.pause()."""

        self._call_compute_api_with_obj(message.ctxt, instance, 'pause')

**** CubicPower OpenStack Study ****

    def unpause_instance(self, message, instance):

        """Unpause an instance via compute_api.pause()."""

        self._call_compute_api_with_obj(message.ctxt, instance, 'unpause')

**** CubicPower OpenStack Study ****

    def resize_instance(self, message, instance, flavor,

                        extra_instance_updates):

        """Resize an instance via compute_api.resize()."""

        self._call_compute_api_with_obj(message.ctxt, instance, 'resize',

                                        flavor_id=flavor['flavorid'],

                                        **extra_instance_updates)

**** CubicPower OpenStack Study ****

    def live_migrate_instance(self, message, instance, block_migration,

                              disk_over_commit, host_name):

        """Live migrate an instance via compute_api.live_migrate()."""

        self._call_compute_api_with_obj(message.ctxt, instance,

                                        'live_migrate', block_migration,

                                        disk_over_commit, host_name)

**** CubicPower OpenStack Study ****

    def revert_resize(self, message, instance):

        """Revert a resize for an instance in its cell."""

        self._call_compute_api_with_obj(message.ctxt, instance,

                                        'revert_resize')

**** CubicPower OpenStack Study ****

    def confirm_resize(self, message, instance):

        """Confirm a resize for an instance in its cell."""

        self._call_compute_api_with_obj(message.ctxt, instance,

                                        'confirm_resize')

**** CubicPower OpenStack Study ****

    def reset_network(self, message, instance):

        """Reset networking for an instance in its cell."""

        self._call_compute_api_with_obj(message.ctxt, instance,

                                        'reset_network')

**** CubicPower OpenStack Study ****

    def inject_network_info(self, message, instance):

        """Inject networking for an instance in its cell."""

        self._call_compute_api_with_obj(message.ctxt, instance,

                                        'inject_network_info')

**** CubicPower OpenStack Study ****

    def snapshot_instance(self, message, instance, image_id):

        """Snapshot an instance in its cell."""

        instance.refresh()

        instance.task_state = task_states.IMAGE_SNAPSHOT_PENDING

        instance.save(expected_task_state=[None])

        self.compute_rpcapi.snapshot_instance(message.ctxt,

                                              instance,

                                              image_id)

**** CubicPower OpenStack Study ****

    def backup_instance(self, message, instance, image_id,

                        backup_type, rotation):

        """Backup an instance in its cell."""

        instance.refresh()

        instance.task_state = task_states.IMAGE_BACKUP

        instance.save(expected_task_state=[None])

        self.compute_rpcapi.backup_instance(message.ctxt,

                                            instance,

                                            image_id,

                                            backup_type,

                                            rotation)

**** CubicPower OpenStack Study ****

    def rebuild_instance(self, message, instance, image_href, admin_password,

                         files_to_inject, preserve_ephemeral, kwargs):

        kwargs['preserve_ephemeral'] = preserve_ephemeral

        self._call_compute_api_with_obj(message.ctxt, instance, 'rebuild',

                                        image_href, admin_password,

                                        files_to_inject, **kwargs)

**** CubicPower OpenStack Study ****

class _BroadcastMessageMethods(_BaseMessageMethods):

"""These are the methods that can be called as a part of a broadcast

message.

"""

**** CubicPower OpenStack Study ****

    def _at_the_top(self):

        """Are we the API level?"""

        return not self.state_manager.get_parent_cells()

**** CubicPower OpenStack Study ****

    def _apply_expected_states(self, instance_info):

        """To attempt to address out-of-order messages, do some sanity

        checking on the VM and task states.  Add some requirements for

        vm_state and task_state to the instance_update() DB call if

        necessary.

        """

        expected_vm_state_map = {

                # For updates containing 'vm_state' of 'building',

                # only allow them to occur if the DB already says

                # 'building' or if the vm_state is None.  None

                # really shouldn't be possible as instances always

                # start out in 'building' anyway.. but just in case.

                vm_states.BUILDING: [vm_states.BUILDING, None]}

        expected_task_state_map = {

                # Always allow updates when task_state doesn't change,

                # but also make sure we don't set resize/rebuild task

                # states for old messages when we've potentially already

                # processed the ACTIVE/None messages.  Ie, these checks

                # will prevent stomping on any ACTIVE/None messages

                # we already processed.

                task_states.REBUILD_BLOCK_DEVICE_MAPPING:

                        [task_states.REBUILD_BLOCK_DEVICE_MAPPING,

                         task_states.REBUILDING],

                task_states.REBUILD_SPAWNING:

                        [task_states.REBUILD_SPAWNING,

                         task_states.REBUILD_BLOCK_DEVICE_MAPPING,

                         task_states.REBUILDING],

                task_states.RESIZE_MIGRATING:

                        [task_states.RESIZE_MIGRATING,

                         task_states.RESIZE_PREP],

                task_states.RESIZE_MIGRATED:

                        [task_states.RESIZE_MIGRATED,

                         task_states.RESIZE_MIGRATING,

                         task_states.RESIZE_PREP],

                task_states.RESIZE_FINISH:

                        [task_states.RESIZE_FINISH,

                         task_states.RESIZE_MIGRATED,

                         task_states.RESIZE_MIGRATING,

                         task_states.RESIZE_PREP]}

        if 'vm_state' in instance_info:

            expected = expected_vm_state_map.get(instance_info['vm_state'])

            if expected is not None:

                instance_info['expected_vm_state'] = expected

        if 'task_state' in instance_info:

            expected = expected_task_state_map.get(instance_info['task_state'])

            if expected is not None:

                instance_info['expected_task_state'] = expected

**** CubicPower OpenStack Study ****

    def instance_update_at_top(self, message, instance, **kwargs):

        """Update an instance in the DB if we're a top level cell."""

        if not self._at_the_top():

            return

        instance_uuid = instance['uuid']

        # Remove things that we can't update in the top level cells.

        # 'metadata' is only updated in the API cell, so don't overwrite

        # it based on what child cells say.  Make sure to update

        # 'cell_name' based on the routing path.

        items_to_remove = ['id', 'security_groups', 'volumes', 'cell_name',

                           'name', 'metadata']

        for key in items_to_remove:

            instance.pop(key, None)

        instance['cell_name'] = _reverse_path(message.routing_path)

        # Fixup info_cache.  We'll have to update this separately if

        # it exists.

        info_cache = instance.pop('info_cache', None)

        if info_cache is not None:

            info_cache.pop('id', None)

            info_cache.pop('instance', None)

        if 'system_metadata' in instance:

            # Make sure we have the dict form that we need for

            # instance_update.

            instance['system_metadata'] = utils.instance_sys_meta(instance)

        LOG.debug(_("Got update for instance: %(instance)s"),

                  {'instance': instance}, instance_uuid=instance_uuid)

        self._apply_expected_states(instance)

        # It's possible due to some weird condition that the instance

        # was already set as deleted... so we'll attempt to update

        # it with permissions that allows us to read deleted.

        with utils.temporary_mutation(message.ctxt, read_deleted="yes"):

            try:

                self.db.instance_update(message.ctxt, instance_uuid,

                        instance, update_cells=False)

            except exception.NotFound:

                # FIXME(comstud): Strange.  Need to handle quotas here,

                # if we actually want this code to remain..

                self.db.instance_create(message.ctxt, instance)

        if info_cache:

            network_info = info_cache.get('network_info')

            if isinstance(network_info, list):

                if not isinstance(network_info, network_model.NetworkInfo):

                    network_info = network_model.NetworkInfo.hydrate(

                            network_info)

                info_cache['network_info'] = network_info.json()

            try:

                self.db.instance_info_cache_update(

                        message.ctxt, instance_uuid, info_cache)

            except exception.InstanceInfoCacheNotFound:

                # Can happen if we try to update a deleted instance's

                # network information.

                pass

**** CubicPower OpenStack Study ****

    def instance_destroy_at_top(self, message, instance, **kwargs):

        """Destroy an instance from the DB if we're a top level cell."""

        if not self._at_the_top():

            return

        instance_uuid = instance['uuid']

        LOG.debug(_("Got update to delete instance"),

                  instance_uuid=instance_uuid)

        try:

            self.db.instance_destroy(message.ctxt, instance_uuid,

                    update_cells=False)

        except exception.InstanceNotFound:

            pass

**** CubicPower OpenStack Study ****

    def instance_delete_everywhere(self, message, instance, delete_type,

                                   **kwargs):

        """Call compute API delete() or soft_delete() in every cell.

        This is used when the API cell doesn't know what cell an instance

        belongs to but the instance was requested to be deleted or

        soft-deleted.  So, we'll run it everywhere.

        """

        LOG.debug(_("Got broadcast to %(delete_type)s delete instance"),

                  {'delete_type': delete_type}, instance=instance)

        if delete_type == 'soft':

            self.compute_api.soft_delete(message.ctxt, instance)

        else:

            self.compute_api.delete(message.ctxt, instance)

**** CubicPower OpenStack Study ****

    def instance_fault_create_at_top(self, message, instance_fault, **kwargs):

        """Destroy an instance from the DB if we're a top level cell."""

        if not self._at_the_top():

            return

        items_to_remove = ['id']

        for key in items_to_remove:

            instance_fault.pop(key, None)

        log_str = _("Got message to create instance fault: "

                    "%(instance_fault)s")

        LOG.debug(log_str, {'instance_fault': instance_fault})

        self.db.instance_fault_create(message.ctxt, instance_fault)

**** CubicPower OpenStack Study ****

    def bw_usage_update_at_top(self, message, bw_update_info, **kwargs):

        """Update Bandwidth usage in the DB if we're a top level cell."""

        if not self._at_the_top():

            return

        self.db.bw_usage_update(message.ctxt, **bw_update_info)

**** CubicPower OpenStack Study ****

    def _sync_instance(self, ctxt, instance):

        if instance['deleted']:

            self.msg_runner.instance_destroy_at_top(ctxt, instance)

        else:

            self.msg_runner.instance_update_at_top(ctxt, instance)

**** CubicPower OpenStack Study ****

    def sync_instances(self, message, project_id, updated_since, deleted,

                       **kwargs):

        projid_str = project_id is None and "" or project_id

        since_str = updated_since is None and "" or updated_since

        LOG.info(_("Forcing a sync of instances, project_id="

                   "%(projid_str)s, updated_since=%(since_str)s"),

                 {'projid_str': projid_str, 'since_str': since_str})

        if updated_since is not None:

            updated_since = timeutils.parse_isotime(updated_since)

        instances = cells_utils.get_instances_to_sync(message.ctxt,

                updated_since=updated_since, project_id=project_id,

                deleted=deleted)

        for instance in instances:

            self._sync_instance(message.ctxt, instance)

**** CubicPower OpenStack Study ****

    def service_get_all(self, message, filters):

        if filters is None:

            filters = {}

        disabled = filters.pop('disabled', None)

        services = self.db.service_get_all(message.ctxt, disabled=disabled)

        ret_services = []

        for service in services:

            service = jsonutils.to_primitive(service)

            for key, val in filters.iteritems():

                if service[key] != val:

                    break

            else:

                ret_services.append(service)

        return ret_services

**** CubicPower OpenStack Study ****

    def compute_node_get_all(self, message, hypervisor_match):

        """Return compute nodes in this cell."""

        if hypervisor_match is not None:

            nodes = self.db.compute_node_search_by_hypervisor(message.ctxt,

                    hypervisor_match)

        else:

            nodes = self.db.compute_node_get_all(message.ctxt)

        return jsonutils.to_primitive(nodes)

**** CubicPower OpenStack Study ****

    def compute_node_stats(self, message):

        """Return compute node stats from this cell."""

        return self.db.compute_node_statistics(message.ctxt)

**** CubicPower OpenStack Study ****

    def consoleauth_delete_tokens(self, message, instance_uuid):

        """Delete consoleauth tokens for an instance in API cells."""

        if not self._at_the_top():

            return

        self.consoleauth_rpcapi.delete_tokens_for_instance(message.ctxt,

                                                           instance_uuid)

**** CubicPower OpenStack Study ****

    def bdm_update_or_create_at_top(self, message, bdm, create):

        """Create or update a block device mapping in API cells.  If

        create is True, only try to create.  If create is None, try to

        update but fall back to create.  If create is False, only attempt

        to update.  This maps to nova-conductor's behavior.

        """

        if not self._at_the_top():

            return

        items_to_remove = ['id']

        for key in items_to_remove:

            bdm.pop(key, None)

        if create is None:

            self.db.block_device_mapping_update_or_create(message.ctxt,

                                                          bdm,

                                                          legacy=False)

            return

        elif create is True:

            self.db.block_device_mapping_create(message.ctxt, bdm,

                                                legacy=False)

            return

        # Unfortunately this update call wants BDM ID... but we don't know

        # what it is in this cell.  Search for it.. try matching either

        # device_name or volume_id.

        dev_name = bdm['device_name']

        vol_id = bdm['volume_id']

        instance_bdms = self.db.block_device_mapping_get_all_by_instance(

                message.ctxt, bdm['instance_uuid'])

        for instance_bdm in instance_bdms:

            if dev_name and instance_bdm['device_name'] == dev_name:

                break

            if vol_id and instance_bdm['volume_id'] == vol_id:

                break

        else:

            LOG.warn(_("No match when trying to update BDM: %(bdm)s"),

                     dict(bdm=bdm))

            return

        self.db.block_device_mapping_update(message.ctxt,

                                            instance_bdm['id'], bdm,

                                            legacy=False)

**** CubicPower OpenStack Study ****

    def bdm_destroy_at_top(self, message, instance_uuid, device_name,

                           volume_id):

        """Destroy a block device mapping in API cells by device name

        or volume_id.  device_name or volume_id can be None, but not both.

        """

        if not self._at_the_top():

            return

        if device_name:

            self.db.block_device_mapping_destroy_by_instance_and_device(

                    message.ctxt, instance_uuid, device_name)

        elif volume_id:

            self.db.block_device_mapping_destroy_by_instance_and_volume(

                    message.ctxt, instance_uuid, volume_id)

**** CubicPower OpenStack Study ****

    def get_migrations(self, message, filters):

        context = message.ctxt

        return self.compute_api.get_migrations(context, filters)

_CELL_MESSAGE_TYPE_TO_MESSAGE_CLS = {'targeted': _TargetedMessage,

                                     'broadcast': _BroadcastMessage,

                                     'response': _ResponseMessage}

_CELL_MESSAGE_TYPE_TO_METHODS_CLS = {'targeted': _TargetedMessageMethods,

                                     'broadcast': _BroadcastMessageMethods,

                                     'response': _ResponseMessageMethods}

#

# Below are the public interfaces into this module.

#

**** CubicPower OpenStack Study ****

class MessageRunner(object):

"""This class is the main interface into creating messages and

processing them.

Public methods in this class are typically called by the CellsManager

to create a new message and process it with the exception of

'message_from_json' which should be used by CellsDrivers to convert

a JSONified message it has received back into the appropriate Message

class.

Private methods are used internally when we need to keep some

'global' state. For instance, eventlet queues used for responses are

held in this class. Also, when a Message is process()ed above and

it's determined we should take action locally,

_process_message_locally() will be called.

When needing to add a new method to call in a Cell2Cell message,

**** CubicPower OpenStack Study ****

    def __init__(self, state_manager):

        self.state_manager = state_manager

        cells_scheduler_cls = importutils.import_class(

                CONF.cells.scheduler)

        self.scheduler = cells_scheduler_cls(self)

        self.response_queues = {}

        self.methods_by_type = {}

        self.our_name = CONF.cells.name

        for msg_type, cls in _CELL_MESSAGE_TYPE_TO_METHODS_CLS.iteritems():

            self.methods_by_type[msg_type] = cls(self)

        self.serializer = objects_base.NovaObjectSerializer()

**** CubicPower OpenStack Study ****

    def _process_message_locally(self, message):

        """Message processing will call this when its determined that

        the message should be processed within this cell.  Find the

        method to call based on the message type, and call it.  The

        caller is responsible for catching exceptions and returning

        results to cells, if needed.

        """

        methods = self.methods_by_type[message.message_type]

        fn = getattr(methods, message.method_name)

        return fn(message, **message.method_kwargs)

**** CubicPower OpenStack Study ****

    def _put_response(self, response_uuid, response):

        """Put a response into a response queue.  This is called when

        a _ResponseMessage is processed in the cell that initiated a

        'call' to another cell.

        """

        resp_queue = self.response_queues.get(response_uuid)

        if not resp_queue:

            # Response queue is gone.  We must have restarted or we

            # received a response after our timeout period.

            return

        resp_queue.put(response)

**** CubicPower OpenStack Study ****

    def _setup_response_queue(self, message):

        """Set up an eventlet queue to use to wait for replies.

        Replies come back from the target cell as a _ResponseMessage

        being sent back to the source.

        """

        resp_queue = queue.Queue()

        self.response_queues[message.uuid] = resp_queue

        return resp_queue

**** CubicPower OpenStack Study ****

    def _cleanup_response_queue(self, message):

        """Stop tracking the response queue either because we're

        done receiving responses, or we've timed out.

        """

        try:

            del self.response_queues[message.uuid]

        except KeyError:

            # Ignore if queue is gone already somehow.

            pass

**** CubicPower OpenStack Study ****

    def _create_response_message(self, ctxt, direction, target_cell,

            response_uuid, response_kwargs, **kwargs):

        """Create a ResponseMessage.  This is used internally within

        the nova.cells.messaging module.

        """

        return _ResponseMessage(self, ctxt, 'parse_responses',

                                response_kwargs, direction, target_cell,

                                response_uuid, **kwargs)

**** CubicPower OpenStack Study ****

    def _get_migrations_for_cell(self, ctxt, cell_name, filters):

        method_kwargs = dict(filters=filters)

        message = _TargetedMessage(self, ctxt, 'get_migrations',

                                   method_kwargs, 'down', cell_name,

                                   need_response=True)

        response = message.process()

        if response.failure and isinstance(response.value[1],

                                           exception.CellRoutingInconsistency):

            return []

        return [response]

**** CubicPower OpenStack Study ****

    def message_from_json(self, json_message):

        """Turns a message in JSON format into an appropriate Message

        instance.  This is called when cells receive a message from

        another cell.

        """

        message_dict = jsonutils.loads(json_message)

        # Need to convert context back.

        ctxt = message_dict['ctxt']

        message_dict['ctxt'] = context.RequestContext.from_dict(ctxt)

        # NOTE(comstud): We also need to re-serialize any objects that

        # exist in 'method_kwargs'.

        method_kwargs = message_dict['method_kwargs']

        for k, v in method_kwargs.items():

            method_kwargs[k] = self.serializer.deserialize_entity(

                    message_dict['ctxt'], v)

        message_type = message_dict.pop('message_type')

        message_cls = _CELL_MESSAGE_TYPE_TO_MESSAGE_CLS[message_type]

        return message_cls(self, **message_dict)

**** CubicPower OpenStack Study ****

    def ask_children_for_capabilities(self, ctxt):

        """Tell child cells to send us capabilities.  This is typically

        called on startup of the nova-cells service.

        """

        child_cells = self.state_manager.get_child_cells()

        for child_cell in child_cells:

            message = _TargetedMessage(self, ctxt,

                                        'announce_capabilities',

                                        dict(), 'down', child_cell)

            message.process()

**** CubicPower OpenStack Study ****

    def ask_children_for_capacities(self, ctxt):

        """Tell child cells to send us capacities.  This is typically

        called on startup of the nova-cells service.

        """

        child_cells = self.state_manager.get_child_cells()

        for child_cell in child_cells:

            message = _TargetedMessage(self, ctxt, 'announce_capacities',

                                        dict(), 'down', child_cell)

            message.process()

**** CubicPower OpenStack Study ****

    def tell_parents_our_capabilities(self, ctxt):

        """Send our capabilities to parent cells."""

        parent_cells = self.state_manager.get_parent_cells()

        if not parent_cells:

            return

        my_cell_info = self.state_manager.get_my_state()

        capabs = self.state_manager.get_our_capabilities()

        LOG.debug(_("Updating parents with our capabilities: %(capabs)s"),

                  {'capabs': capabs})

        # We have to turn the sets into lists so they can potentially

        # be json encoded when the raw message is sent.

        for key, values in capabs.items():

            capabs[key] = list(values)

        method_kwargs = {'cell_name': my_cell_info.name,

                         'capabilities': capabs}

        for cell in parent_cells:

            message = _TargetedMessage(self, ctxt, 'update_capabilities',

                    method_kwargs, 'up', cell, fanout=True)

            message.process()

**** CubicPower OpenStack Study ****

    def tell_parents_our_capacities(self, ctxt):

        """Send our capacities to parent cells."""

        parent_cells = self.state_manager.get_parent_cells()

        if not parent_cells:

            return

        my_cell_info = self.state_manager.get_my_state()

        capacities = self.state_manager.get_our_capacities()

        LOG.debug(_("Updating parents with our capacities: %(capacities)s"),

                  {'capacities': capacities})

        method_kwargs = {'cell_name': my_cell_info.name,

                         'capacities': capacities}

        for cell in parent_cells:

            message = _TargetedMessage(self, ctxt, 'update_capacities',

                    method_kwargs, 'up', cell, fanout=True)

            message.process()

**** CubicPower OpenStack Study ****

    def schedule_run_instance(self, ctxt, target_cell, host_sched_kwargs):

        """Called by the scheduler to tell a child cell to schedule

        a new instance for build.

        """

        method_kwargs = dict(host_sched_kwargs=host_sched_kwargs)

        message = _TargetedMessage(self, ctxt, 'schedule_run_instance',

                                   method_kwargs, 'down', target_cell)

        message.process()

**** CubicPower OpenStack Study ****

    def build_instances(self, ctxt, target_cell, build_inst_kwargs):

        """Called by the cell scheduler to tell a child cell to build

        instance(s).

        """

        method_kwargs = dict(build_inst_kwargs=build_inst_kwargs)

        message = _TargetedMessage(self, ctxt, 'build_instances',

                                   method_kwargs, 'down', target_cell)

        message.process()

**** CubicPower OpenStack Study ****

    def run_compute_api_method(self, ctxt, cell_name, method_info, call):

        """Call a compute API method in a specific cell."""

        message = _TargetedMessage(self, ctxt, 'run_compute_api_method',

                                   dict(method_info=method_info), 'down',

                                   cell_name, need_response=call)

        return message.process()

**** CubicPower OpenStack Study ****

    def instance_update_at_top(self, ctxt, instance):

        """Update an instance at the top level cell."""

        message = _BroadcastMessage(self, ctxt, 'instance_update_at_top',

                                    dict(instance=instance), 'up',

                                    run_locally=False)

        message.process()

**** CubicPower OpenStack Study ****

    def instance_destroy_at_top(self, ctxt, instance):

        """Destroy an instance at the top level cell."""

        message = _BroadcastMessage(self, ctxt, 'instance_destroy_at_top',

                                    dict(instance=instance), 'up',

                                    run_locally=False)

        message.process()

**** CubicPower OpenStack Study ****

    def instance_delete_everywhere(self, ctxt, instance, delete_type):

        """This is used by API cell when it didn't know what cell

        an instance was in, but the instance was requested to be

        deleted or soft_deleted.  So, we'll broadcast this everywhere.

        """

        method_kwargs = dict(instance=instance, delete_type=delete_type)

        message = _BroadcastMessage(self, ctxt,

                                    'instance_delete_everywhere',

                                    method_kwargs, 'down',

                                    run_locally=False)

        message.process()

**** CubicPower OpenStack Study ****

    def instance_fault_create_at_top(self, ctxt, instance_fault):

        """Create an instance fault at the top level cell."""

        message = _BroadcastMessage(self, ctxt,

                                    'instance_fault_create_at_top',

                                    dict(instance_fault=instance_fault),

                                    'up', run_locally=False)

        message.process()

**** CubicPower OpenStack Study ****

    def bw_usage_update_at_top(self, ctxt, bw_update_info):

        """Update bandwidth usage at top level cell."""

        message = _BroadcastMessage(self, ctxt, 'bw_usage_update_at_top',

                                    dict(bw_update_info=bw_update_info),

                                    'up', run_locally=False)

        message.process()

**** CubicPower OpenStack Study ****

    def sync_instances(self, ctxt, project_id, updated_since, deleted):

        """Force a sync of all instances, potentially by project_id,

        and potentially since a certain date/time.

        """

        method_kwargs = dict(project_id=project_id,

                             updated_since=updated_since,

                             deleted=deleted)

        message = _BroadcastMessage(self, ctxt, 'sync_instances',

                                    method_kwargs, 'down',

                                    run_locally=False)

        message.process()

**** CubicPower OpenStack Study ****

    def service_get_all(self, ctxt, filters=None):

        method_kwargs = dict(filters=filters)

        message = _BroadcastMessage(self, ctxt, 'service_get_all',

                                    method_kwargs, 'down',

                                    run_locally=True, need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def service_get_by_compute_host(self, ctxt, cell_name, host_name):

        method_kwargs = dict(host_name=host_name)

        message = _TargetedMessage(self, ctxt,

                                  'service_get_by_compute_host',

                                  method_kwargs, 'down', cell_name,

                                  need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def get_host_uptime(self, ctxt, cell_name, host_name):

        method_kwargs = dict(host_name=host_name)

        message = _TargetedMessage(self, ctxt,

                                   'get_host_uptime',

                                   method_kwargs, 'down', cell_name,

                                   need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def service_update(self, ctxt, cell_name, host_name, binary,

                       params_to_update):

        """Used to enable/disable a service. For compute services, setting to

        disabled stops new builds arriving on that host.

        :param host_name: the name of the host machine that the service is

                          running

        :param binary: The name of the executable that the service runs as

        :param params_to_update: eg. {'disabled': True}

        :returns: the update service object

        """

        method_kwargs = dict(host_name=host_name, binary=binary,

                             params_to_update=params_to_update)

        message = _TargetedMessage(self, ctxt,

                                  'service_update',

                                  method_kwargs, 'down', cell_name,

                                  need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def service_delete(self, ctxt, cell_name, service_id):

        """Deletes the specified service."""

        method_kwargs = {'service_id': service_id}

        message = _TargetedMessage(self, ctxt,

                                   'service_delete',

                                   method_kwargs, 'down', cell_name,

                                   need_response=True)

        message.process()

**** CubicPower OpenStack Study ****

    def proxy_rpc_to_manager(self, ctxt, cell_name, host_name, topic,

                             rpc_message, call, timeout):

        method_kwargs = {'host_name': host_name,

                         'topic': topic,

                         'rpc_message': rpc_message,

                         'timeout': timeout}

        message = _TargetedMessage(self, ctxt,

                                   'proxy_rpc_to_manager',

                                   method_kwargs, 'down', cell_name,

                                   need_response=call)

        return message.process()

**** CubicPower OpenStack Study ****

    def task_log_get_all(self, ctxt, cell_name, task_name,

                         period_beginning, period_ending,

                         host=None, state=None):

        """Get task logs from the DB from all cells or a particular

        cell.

        If 'cell_name' is None or '', get responses from all cells.

        If 'host' is not None, filter by host.

        If 'state' is not None, filter by state.

        Return a list of Response objects.

        """

        method_kwargs = dict(task_name=task_name,

                             period_beginning=period_beginning,

                             period_ending=period_ending,

                             host=host, state=state)

        if cell_name:

            message = _TargetedMessage(self, ctxt, 'task_log_get_all',

                                    method_kwargs, 'down',

                                    cell_name, need_response=True)

            # Caller should get a list of Responses.

            return [message.process()]

        message = _BroadcastMessage(self, ctxt, 'task_log_get_all',

                                    method_kwargs, 'down',

                                    run_locally=True, need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def compute_node_get_all(self, ctxt, hypervisor_match=None):

        """Return list of compute nodes in all child cells."""

        method_kwargs = dict(hypervisor_match=hypervisor_match)

        message = _BroadcastMessage(self, ctxt, 'compute_node_get_all',

                                    method_kwargs, 'down',

                                    run_locally=True, need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def compute_node_stats(self, ctxt):

        """Return compute node stats from all child cells."""

        method_kwargs = dict()

        message = _BroadcastMessage(self, ctxt, 'compute_node_stats',

                                    method_kwargs, 'down',

                                    run_locally=True, need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def compute_node_get(self, ctxt, cell_name, compute_id):

        """Return compute node entry from a specific cell by ID."""

        method_kwargs = dict(compute_id=compute_id)

        message = _TargetedMessage(self, ctxt, 'compute_node_get',

                                    method_kwargs, 'down',

                                    cell_name, need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def actions_get(self, ctxt, cell_name, instance_uuid):

        method_kwargs = dict(instance_uuid=instance_uuid)

        message = _TargetedMessage(self, ctxt, 'actions_get',

                                method_kwargs, 'down',

                                cell_name, need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def action_get_by_request_id(self, ctxt, cell_name, instance_uuid,

                                 request_id):

        method_kwargs = dict(instance_uuid=instance_uuid,

                             request_id=request_id)

        message = _TargetedMessage(self, ctxt, 'action_get_by_request_id',

                                method_kwargs, 'down',

                                cell_name, need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def action_events_get(self, ctxt, cell_name, action_id):

        method_kwargs = dict(action_id=action_id)

        message = _TargetedMessage(self, ctxt, 'action_events_get',

                                method_kwargs, 'down',

                                cell_name, need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def consoleauth_delete_tokens(self, ctxt, instance_uuid):

        """Delete consoleauth tokens for an instance in API cells."""

        message = _BroadcastMessage(self, ctxt, 'consoleauth_delete_tokens',

                                    dict(instance_uuid=instance_uuid),

                                    'up', run_locally=False)

        message.process()

**** CubicPower OpenStack Study ****

    def validate_console_port(self, ctxt, cell_name, instance_uuid,

                              console_port, console_type):

        """Validate console port with child cell compute node."""

        method_kwargs = {'instance_uuid': instance_uuid,

                         'console_port': console_port,

                         'console_type': console_type}

        message = _TargetedMessage(self, ctxt, 'validate_console_port',

                                   method_kwargs, 'down',

                                   cell_name, need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def bdm_update_or_create_at_top(self, ctxt, bdm, create=None):

        """Update/Create a BDM at top level cell."""

        message = _BroadcastMessage(self, ctxt,

                                    'bdm_update_or_create_at_top',

                                    dict(bdm=bdm, create=create),

                                    'up', run_locally=False)

        message.process()

**** CubicPower OpenStack Study ****

    def bdm_destroy_at_top(self, ctxt, instance_uuid, device_name=None,

                           volume_id=None):

        """Destroy a BDM at top level cell."""

        method_kwargs = dict(instance_uuid=instance_uuid,

                             device_name=device_name,

                             volume_id=volume_id)

        message = _BroadcastMessage(self, ctxt, 'bdm_destroy_at_top',

                                    method_kwargs,

                                    'up', run_locally=False)

        message.process()

**** CubicPower OpenStack Study ****

    def get_migrations(self, ctxt, cell_name, run_locally, filters):

        """Fetch all migrations applying the filters for a given cell or all

        cells.

        """

        method_kwargs = dict(filters=filters)

        if cell_name:

            return self._get_migrations_for_cell(ctxt, cell_name, filters)

        message = _BroadcastMessage(self, ctxt, 'get_migrations',

                                    method_kwargs, 'down',

                                    run_locally=run_locally,

                                    need_response=True)

        return message.process()

**** CubicPower OpenStack Study ****

    def _instance_action(self, ctxt, instance, method, extra_kwargs=None,

                         need_response=False):

        """Call instance_ in correct cell for instance."""

        cell_name = instance.cell_name

        if not cell_name:

            LOG.warn(_("No cell_name for %(method)s() from API"),

                     dict(method=method), instance=instance)

            return

        method_kwargs = {'instance': instance}

        if extra_kwargs:

            method_kwargs.update(extra_kwargs)

        message = _TargetedMessage(self, ctxt, method, method_kwargs,

                                   'down', cell_name,

                                   need_response=need_response)

        return message.process()

**** CubicPower OpenStack Study ****

    def instance_update_from_api(self, ctxt, instance,

                                expected_vm_state, expected_task_state,

                                admin_state_reset):

        """Update an instance object in its cell."""

        cell_name = instance.cell_name

        if not cell_name:

            LOG.warn(_("No cell_name for instance update from API"),

                     instance=instance)

            return

        method_kwargs = {'instance': instance,

                         'expected_vm_state': expected_vm_state,

                         'expected_task_state': expected_task_state,

                         'admin_state_reset': admin_state_reset}

        message = _TargetedMessage(self, ctxt, 'instance_update_from_api',

                                   method_kwargs, 'down',

                                   cell_name)

        message.process()

**** CubicPower OpenStack Study ****

    def start_instance(self, ctxt, instance):

        """Start an instance in its cell."""

        self._instance_action(ctxt, instance, 'start_instance')

**** CubicPower OpenStack Study ****

    def stop_instance(self, ctxt, instance, do_cast=True):

        """Stop an instance in its cell."""

        if do_cast:

            self._instance_action(ctxt, instance, 'stop_instance')

        else:

            return self._instance_action(ctxt, instance, 'stop_instance',

                                         need_response=True)

**** CubicPower OpenStack Study ****

    def reboot_instance(self, ctxt, instance, reboot_type):

        """Reboot an instance in its cell."""

        extra_kwargs = dict(reboot_type=reboot_type)

        self._instance_action(ctxt, instance, 'reboot_instance',

                              extra_kwargs=extra_kwargs)

**** CubicPower OpenStack Study ****

    def suspend_instance(self, ctxt, instance):

        """Suspend an instance in its cell."""

        self._instance_action(ctxt, instance, 'suspend_instance')

**** CubicPower OpenStack Study ****

    def resume_instance(self, ctxt, instance):

        """Resume an instance in its cell."""

        self._instance_action(ctxt, instance, 'resume_instance')

**** CubicPower OpenStack Study ****

    def terminate_instance(self, ctxt, instance):

        self._instance_action(ctxt, instance, 'terminate_instance')

**** CubicPower OpenStack Study ****

    def soft_delete_instance(self, ctxt, instance):

        self._instance_action(ctxt, instance, 'soft_delete_instance')

**** CubicPower OpenStack Study ****

    def pause_instance(self, ctxt, instance):

        """Pause an instance in its cell."""

        self._instance_action(ctxt, instance, 'pause_instance')

**** CubicPower OpenStack Study ****

    def unpause_instance(self, ctxt, instance):

        """Unpause an instance in its cell."""

        self._instance_action(ctxt, instance, 'unpause_instance')

**** CubicPower OpenStack Study ****

    def resize_instance(self, ctxt, instance, flavor,

                       extra_instance_updates):

        """Resize an instance in its cell."""

        extra_kwargs = dict(flavor=flavor,

                            extra_instance_updates=extra_instance_updates)

        self._instance_action(ctxt, instance, 'resize_instance',

                              extra_kwargs=extra_kwargs)

**** CubicPower OpenStack Study ****

    def live_migrate_instance(self, ctxt, instance, block_migration,

                              disk_over_commit, host_name):

        """Live migrate an instance in its cell."""

        extra_kwargs = dict(block_migration=block_migration,

                            disk_over_commit=disk_over_commit,

                            host_name=host_name)

        self._instance_action(ctxt, instance, 'live_migrate_instance',

                              extra_kwargs=extra_kwargs)

**** CubicPower OpenStack Study ****

    def revert_resize(self, ctxt, instance):

        """Revert a resize for an instance in its cell."""

        self._instance_action(ctxt, instance, 'revert_resize')

**** CubicPower OpenStack Study ****

    def confirm_resize(self, ctxt, instance):

        """Confirm a resize for an instance in its cell."""

        self._instance_action(ctxt, instance, 'confirm_resize')

**** CubicPower OpenStack Study ****

    def reset_network(self, ctxt, instance):

        """Reset networking for an instance in its cell."""

        self._instance_action(ctxt, instance, 'reset_network')

**** CubicPower OpenStack Study ****

    def inject_network_info(self, ctxt, instance):

        """Inject networking for an instance in its cell."""

        self._instance_action(ctxt, instance, 'inject_network_info')

**** CubicPower OpenStack Study ****

    def snapshot_instance(self, ctxt, instance, image_id):

        """Snapshot an instance in its cell."""

        extra_kwargs = dict(image_id=image_id)

        self._instance_action(ctxt, instance, 'snapshot_instance',

                              extra_kwargs=extra_kwargs)

**** CubicPower OpenStack Study ****

    def backup_instance(self, ctxt, instance, image_id, backup_type,

                        rotation):

        """Backup an instance in its cell."""

        extra_kwargs = dict(image_id=image_id, backup_type=backup_type,

                            rotation=rotation)

        self._instance_action(ctxt, instance, 'backup_instance',

                              extra_kwargs=extra_kwargs)

**** CubicPower OpenStack Study ****

    def rebuild_instance(self, ctxt, instance, image_href, admin_password,

                         files_to_inject, preserve_ephemeral, kwargs):

        extra_kwargs = dict(image_href=image_href,

                            admin_password=admin_password,

                            files_to_inject=files_to_inject,

                            preserve_ephemeral=preserve_ephemeral,

                            kwargs=kwargs)

        self._instance_action(ctxt, instance, 'rebuild_instance',

                              extra_kwargs=extra_kwargs)

    @staticmethod

**** CubicPower OpenStack Study ****

    def get_message_types():

        return _CELL_MESSAGE_TYPE_TO_MESSAGE_CLS.keys()

**** CubicPower OpenStack Study ****

class Response(object):

"""Holds a response from a cell. If there was a failure, 'failure'

will be True and 'response' will contain an encoded Exception.

"""

**** CubicPower OpenStack Study ****

    def __init__(self, cell_name, value, failure):

        self.failure = failure

        self.cell_name = cell_name

        self.value = value

**** CubicPower OpenStack Study ****

    def to_json(self):

        resp_value = self.value

        if self.failure:

            resp_value = serialize_remote_exception(resp_value,

                                                    log_failure=False)

        _dict = {'cell_name': self.cell_name,

                 'value': resp_value,

                 'failure': self.failure}

        return jsonutils.dumps(_dict)

    @classmethod

**** CubicPower OpenStack Study ****

    def from_json(cls, json_message):

        _dict = jsonutils.loads(json_message)

        if _dict['failure']:

            resp_value = deserialize_remote_exception(_dict['value'],

                                                      rpc.get_allowed_exmods())

            _dict['value'] = resp_value

        return cls(**_dict)

**** CubicPower OpenStack Study ****

    def value_or_raise(self):

        if self.failure:

            if isinstance(self.value, (tuple, list)):

                raise self.value[0], self.value[1], self.value[2]

            else:

                raise self.value

        return self.value

_REMOTE_POSTFIX = '_Remote'

def serialize_remote_exception(failure_info, log_failure=True):

    """Prepares exception data to be sent over rpc.

    Failure_info should be a sys.exc_info() tuple.

    """

    tb = traceback.format_exception(*failure_info)

    failure = failure_info[1]

    if log_failure:

        LOG.error(_("Returning exception %s to caller"),

                  six.text_type(failure))

        LOG.error(tb)

    kwargs = {}

    if hasattr(failure, 'kwargs'):

        kwargs = failure.kwargs

    # NOTE(matiu): With cells, it's possible to re-raise remote, remote

    # exceptions. Lets turn it back into the original exception type.

    cls_name = str(failure.__class__.__name__)

    mod_name = str(failure.__class__.__module__)

    if (cls_name.endswith(_REMOTE_POSTFIX) and

            mod_name.endswith(_REMOTE_POSTFIX)):

        cls_name = cls_name[:-len(_REMOTE_POSTFIX)]

        mod_name = mod_name[:-len(_REMOTE_POSTFIX)]

    data = {

        'class': cls_name,

        'module': mod_name,

        'message': six.text_type(failure),

        'tb': tb,

        'args': failure.args,

        'kwargs': kwargs

    }

    json_data = jsonutils.dumps(data)

    return json_data

def deserialize_remote_exception(data, allowed_remote_exmods):

    failure = jsonutils.loads(str(data))

    trace = failure.get('tb', [])

    message = failure.get('message', "") + "\n" + "\n".join(trace)

    name = failure.get('class')

    module = failure.get('module')

    # NOTE(ameade): We DO NOT want to allow just any module to be imported, in

    # order to prevent arbitrary code execution.

    if module != 'exceptions' and module not in allowed_remote_exmods:

        return messaging.RemoteError(name, failure.get('message'), trace)

    try:

        mod = importutils.import_module(module)

        klass = getattr(mod, name)

        if not issubclass(klass, Exception):

            raise TypeError("Can only deserialize Exceptions")

        failure = klass(*failure.get('args', []), **failure.get('kwargs', {}))

    except (AttributeError, TypeError, ImportError):

        return messaging.RemoteError(name, failure.get('message'), trace)

    ex_type = type(failure)

    str_override = lambda self: message

    new_ex_type = type(ex_type.__name__ + _REMOTE_POSTFIX, (ex_type,),

                       {'__str__': str_override, '__unicode__': str_override})

    new_ex_type.__module__ = '%s%s' % (module, _REMOTE_POSTFIX)

    try:

        # NOTE(ameade): Dynamically create a new exception type and swap it in

        # as the new type for the exception. This only works on user defined

        # Exceptions and not core python exceptions. This is important because

        # we cannot necessarily change an exception message so we must override

        # the __str__ method.

        failure.__class__ = new_ex_type

    except TypeError:

        # NOTE(ameade): If a core exception then just add the traceback to the

        # first exception argument.

        failure.args = (message,) + failure.args[1:]

    return failure

**** CubicPower OpenStack Study ****

def serialize_remote_exception(failure_info, log_failure=True):

    """Prepares exception data to be sent over rpc.

    Failure_info should be a sys.exc_info() tuple.

    """

    tb = traceback.format_exception(*failure_info)

    failure = failure_info[1]

    if log_failure:

        LOG.error(_("Returning exception %s to caller"),

                  six.text_type(failure))

        LOG.error(tb)

    kwargs = {}

    if hasattr(failure, 'kwargs'):

        kwargs = failure.kwargs

    # NOTE(matiu): With cells, it's possible to re-raise remote, remote

    # exceptions. Lets turn it back into the original exception type.

    cls_name = str(failure.__class__.__name__)

    mod_name = str(failure.__class__.__module__)

    if (cls_name.endswith(_REMOTE_POSTFIX) and

            mod_name.endswith(_REMOTE_POSTFIX)):

        cls_name = cls_name[:-len(_REMOTE_POSTFIX)]

        mod_name = mod_name[:-len(_REMOTE_POSTFIX)]

    data = {

        'class': cls_name,

        'module': mod_name,

        'message': six.text_type(failure),

        'tb': tb,

        'args': failure.args,

        'kwargs': kwargs

    }

    json_data = jsonutils.dumps(data)

    return json_data

**** CubicPower OpenStack Study ****

def deserialize_remote_exception(data, allowed_remote_exmods):

    failure = jsonutils.loads(str(data))

    trace = failure.get('tb', [])

    message = failure.get('message', "") + "\n" + "\n".join(trace)

    name = failure.get('class')

    module = failure.get('module')

    # NOTE(ameade): We DO NOT want to allow just any module to be imported, in

    # order to prevent arbitrary code execution.

    if module != 'exceptions' and module not in allowed_remote_exmods:

        return messaging.RemoteError(name, failure.get('message'), trace)

    try:

        mod = importutils.import_module(module)

        klass = getattr(mod, name)

        if not issubclass(klass, Exception):

            raise TypeError("Can only deserialize Exceptions")

        failure = klass(*failure.get('args', []), **failure.get('kwargs', {}))

    except (AttributeError, TypeError, ImportError):

        return messaging.RemoteError(name, failure.get('message'), trace)

    ex_type = type(failure)

    str_override = lambda self: message

    new_ex_type = type(ex_type.__name__ + _REMOTE_POSTFIX, (ex_type,),

                       {'__str__': str_override, '__unicode__': str_override})

    new_ex_type.__module__ = '%s%s' % (module, _REMOTE_POSTFIX)

    try:

        # NOTE(ameade): Dynamically create a new exception type and swap it in

        # as the new type for the exception. This only works on user defined

        # Exceptions and not core python exceptions. This is important because

        # we cannot necessarily change an exception message so we must override

        # the __str__ method.

        failure.__class__ = new_ex_type

    except TypeError:

        # NOTE(ameade): If a core exception then just add the traceback to the

        # first exception argument.

        failure.args = (message,) + failure.args[1:]

    return failure