def wait(self):
        """Loop waiting on children to die and respawning as necessary."""

        LOG.debug(_('Full set of CONF:'))
        CONF.log_opt_values(LOG, std_logging.DEBUG)

        while self.running:
            wrap = self._wait_child()
            if not wrap:
                # Yield to other threads if no children have exited
                # Sleep for a short time to avoid excessive CPU usage
                # (see bug #1095346)
                eventlet.greenthread.sleep(.01)
                continue

            while self.running and len(wrap.children) < wrap.workers:
                self._start_child(wrap)

        if self.sigcaught:
            signame = {signal.SIGTERM: 'SIGTERM',
                       signal.SIGINT: 'SIGINT'}[self.sigcaught]
            LOG.info(_('Caught %s, stopping children'), signame)

        for pid in self.children:
            try:
                os.kill(pid, signal.SIGTERM)
            except OSError as exc:
                if exc.errno != errno.ESRCH:
                    raise

        # Wait for children to die
        if self.children:
            LOG.info(_('Waiting on %d children to exit'), len(self.children))
            while self.children:
                self._wait_child()

def load_xcat_resp(message):
    """Abstract information from xCat REST response body.

    As default, xCat response will in format of JSON and can be
    converted to Python dictionary, would looks like:
    {"data": [{"info": [info,]}, {"data": [data,]}, ..., {"error": [error,]}]}

    Returns a Python dictionary, looks like:
    {'info': [info,],
     'data': [data,],
     'error': [error,]}
    """
    resp_list = jsonloads(message)['data']
    keys = constants.XCAT_RESPONSE_KEYS

    resp = {}
    try:
        for k in keys:
            resp[k] = []

        for d in resp_list:
            for k in keys:
                if d.get(k) is not None:
                    resp[k].append(d.get(k))
    except Exception:
        LOG.error(_("Invalid data returned from xCat: %s") % message)
        raise exception.zVMInvalidxCatResponseDataError(msg=message)

    if not verify_xcat_resp(resp):
        LOG.error(_("Error returned from xCAT: %s") % message)
        raise exception.zVMInvalidxCatResponseDataError(msg=message)
    else:
        return resp

    def register(self, proxy, in_addr, zmq_type_in, out_addr=None,
                 zmq_type_out=None, in_bind=True, out_bind=True,
                 subscribe=None):

        LOG.info(_("Registering reactor"))

        if zmq_type_in not in (zmq.PULL, zmq.SUB):
            raise RPCException("Bad input socktype")

        # Items push in.
        inq = ZmqSocket(in_addr, zmq_type_in, bind=in_bind,
                        subscribe=subscribe)

        self.proxies[inq] = proxy
        self.sockets.append(inq)

        LOG.info(_("In reactor registered"))

        if not out_addr:
            return

        if zmq_type_out not in (zmq.PUSH, zmq.PUB):
            raise RPCException("Bad output socktype")

        # Items push out.
        outq = ZmqSocket(out_addr, zmq_type_out, bind=out_bind)

        self.mapping[inq] = outq
        self.mapping[outq] = inq
        self.sockets.append(outq)

        LOG.info(_("Out reactor registered"))

    def wait(self):
        """Loop waiting on children to die and respawning as necessary."""

        LOG.debug(_('Full set of CONF:'))
        CONF.log_opt_values(LOG, std_logging.DEBUG)

        while True:
            self.handle_signal()
            self._respawn_children()
            if self.sigcaught:
                signame = _signo_to_signame(self.sigcaught)
                LOG.info(_('Caught %s, stopping children'), signame)
            if not _is_sighup_and_daemon(self.sigcaught):
                break

            for pid in self.children:
                os.kill(pid, signal.SIGHUP)
            self.running = True
            self.sigcaught = None

        for pid in self.children:
            try:
                os.kill(pid, signal.SIGTERM)
            except OSError as exc:
                if exc.errno != errno.ESRCH:
                    raise

        # Wait for children to die
        if self.children:
            LOG.info(_('Waiting on %d children to exit'), len(self.children))
            while self.children:
                self._wait_child()

    def _start_child(self, wrap):
        if len(wrap.forktimes) > wrap.workers:
            # Limit ourselves to one process a second (over the period of
            # number of workers * 1 second). This will allow workers to
            # start up quickly but ensure we don't fork off children that
            # die instantly too quickly.
            if time.time() - wrap.forktimes[0] < wrap.workers:
                LOG.info(_('Forking too fast, sleeping'))
                time.sleep(1)

            wrap.forktimes.pop(0)

        wrap.forktimes.append(time.time())

        pid = os.fork()
        if pid == 0:
            launcher = self._child_process(wrap.service)
            while True:
                self._child_process_handle_signal()
                status, signo = self._child_wait_for_exit_or_signal(launcher)
                if not _is_sighup_and_daemon(signo):
                    break
                launcher.restart()

            os._exit(status)

        LOG.info(_('Started child %d'), pid)

        wrap.children.add(pid)
        self.children[pid] = wrap

        return pid

    def _wait_for_exit_or_signal(self, ready_callback=None):
        status = None
        signo = 0

        LOG.debug(_('Full set of CONF:'))
        CONF.log_opt_values(LOG, std_logging.DEBUG)

        try:
            if ready_callback:
                ready_callback()
            super(ServiceLauncher, self).wait()
        except SignalExit as exc:
            signame = _signo_to_signame(exc.signo)
            LOG.info(_('Caught %s, exiting'), signame)
            status = exc.code
            signo = exc.signo
        except SystemExit as exc:
            status = exc.code
        finally:
            self.stop()
            if rpc:
                try:
                    rpc.cleanup()
                except Exception:
                    # We're shutting down, so it doesn't matter at this point.
                    LOG.exception(_('Exception during rpc cleanup.'))

        return status, signo

def ssh_execute(ssh, cmd, process_input=None,
                addl_env=None, check_exit_code=True):
    LOG.debug(_('Running cmd (SSH): %s'), cmd)
    if addl_env:
        raise InvalidArgumentError(_('Environment not supported over SSH'))

    if process_input:
        # This is (probably) fixable if we need it...
        raise InvalidArgumentError(_('process_input not supported over SSH'))

    stdin_stream, stdout_stream, stderr_stream = ssh.exec_command(cmd)
    channel = stdout_stream.channel

    # NOTE(justinsb): This seems suspicious...
    # ...other SSH clients have buffering issues with this approach
    stdout = stdout_stream.read()
    stderr = stderr_stream.read()
    stdin_stream.close()

    exit_status = channel.recv_exit_status()

    # exit_status == -1 if no exit code was returned
    if exit_status != -1:
        LOG.debug(_('Result was %s') % exit_status)
        if check_exit_code and exit_status != 0:
            raise ProcessExecutionError(exit_code=exit_status,
                                        stdout=stdout,
                                        stderr=stderr,
                                        cmd=cmd)

    return (stdout, stderr)

def _multi_send(method, context, topic, msg, timeout=None,
                envelope=False, _msg_id=None):
    """Wraps the sending of messages.

    Dispatches to the matchmaker and sends message to all relevant hosts.
    """
    conf = CONF
    LOG.debug(_("%(msg)s") % {'msg': ' '.join(map(pformat, (topic, msg)))})

    queues = _get_matchmaker().queues(topic)
    LOG.debug(_("Sending message(s) to: %s"), queues)

    # Don't stack if we have no matchmaker results
    if not queues:
        LOG.warn(_("No matchmaker results. Not casting."))
        # While not strictly a timeout, callers know how to handle
        # this exception and a timeout isn't too big a lie.
        raise rpc_common.Timeout(_("No match from matchmaker."))

    # This supports brokerless fanout (addresses > 1)
    for queue in queues:
        (_topic, ip_addr) = queue
        _addr = "tcp://%s:%s" % (ip_addr, conf.rpc_zmq_port)

        if method.__name__ == '_cast':
            eventlet.spawn_n(method, _addr, context,
                             _topic, msg, timeout, envelope,
                             _msg_id)
            return
        return method(_addr, context, _topic, msg, timeout,
                      envelope)

            def publisher(waiter):
                LOG.info(_("Creating proxy for topic: %s"), topic)

                try:
                    # The topic is received over the network,
                    # don't trust this input.
                    if self.badchars.search(topic) is not None:
                        emsg = _("Topic contained dangerous characters.")
                        LOG.warn(emsg)
                        raise RPCException(emsg)

                    out_sock = ZmqSocket("ipc://%s/zmq_topic_%s" %
                                         (ipc_dir, topic),
                                         sock_type, bind=True)
                except RPCException:
                    waiter.send_exception(*sys.exc_info())
                    return

                self.topic_proxy[topic] = eventlet.queue.LightQueue(
                    CONF.rpc_zmq_topic_backlog)
                self.sockets.append(out_sock)

                # It takes some time for a pub socket to open,
                # before we can have any faith in doing a send() to it.
                if sock_type == zmq.PUB:
                    eventlet.sleep(.5)

                waiter.send(True)

                while(True):
                    data = self.topic_proxy[topic].get()
                    out_sock.send(data, copy=False)

    def __init__(self, addr, zmq_type, bind=True, subscribe=None):
        self.sock = _get_ctxt().socket(zmq_type)
        self.addr = addr
        self.type = zmq_type
        self.subscriptions = []

        # Support failures on sending/receiving on wrong socket type.
        self.can_recv = zmq_type in (zmq.PULL, zmq.SUB)
        self.can_send = zmq_type in (zmq.PUSH, zmq.PUB)
        self.can_sub = zmq_type in (zmq.SUB, )

        # Support list, str, & None for subscribe arg (cast to list)
        do_sub = {
            list: subscribe,
            str: [subscribe],
            type(None): []
        }[type(subscribe)]

        for f in do_sub:
            self.subscribe(f)

        str_data = {'addr': addr, 'type': self.socket_s(),
                    'subscribe': subscribe, 'bind': bind}

        LOG.debug(_("Connecting to %(addr)s with %(type)s"), str_data)
        LOG.debug(_("-> Subscribed to %(subscribe)s"), str_data)
        LOG.debug(_("-> bind: %(bind)s"), str_data)

        try:
            if bind:
                self.sock.bind(addr)
            else:
                self.sock.connect(addr)
        except Exception:
            raise RPCException(_("Could not open socket."))

    def run_periodic_tasks(self, context, raise_on_error=False):
        """Tasks to be run at a periodic interval."""
        idle_for = DEFAULT_INTERVAL
        for task_name, task in self._periodic_tasks:
            full_task_name = '.'.join([self.__class__.__name__, task_name])

            now = timeutils.utcnow()
            spacing = self._periodic_spacing[task_name]
            last_run = self._periodic_last_run[task_name]

            # If a periodic task is _nearly_ due, then we'll run it early
            if spacing is not None and last_run is not None:
                due = last_run + datetime.timedelta(seconds=spacing)
                if not timeutils.is_soon(due, 0.2):
                    idle_for = min(idle_for, timeutils.delta_seconds(now, due))
                    continue

            if spacing is not None:
                idle_for = min(idle_for, spacing)

            LOG.debug(_("Running periodic task %(full_task_name)s"),
                      {"full_task_name": full_task_name})
            self._periodic_last_run[task_name] = timeutils.utcnow()

            try:
                task(self, context)
            except Exception as e:
                if raise_on_error:
                    raise
                LOG.exception(_("Error during %(full_task_name)s: %(e)s"),
                              {"full_task_name": full_task_name, "e": e})
            time.sleep(0)

        return idle_for

    def __call__(self, message_data):
        """Consumer callback to call a method on a proxy object.

        Parses the message for validity and fires off a thread to call the
        proxy object method.

        Message data should be a dictionary with two keys:
            method: string representing the method to call
            args: dictionary of arg: value

        Example: {'method': 'echo', 'args': {'value': 42}}

        """
        # It is important to clear the context here, because at this point
        # the previous context is stored in local.store.context
        if hasattr(local.store, "context"):
            del local.store.context
        rpc_common._safe_log(LOG.debug, _("received %s"), message_data)
        self.msg_id_cache.check_duplicate_message(message_data)
        ctxt = unpack_context(self.conf, message_data)
        method = message_data.get("method")
        args = message_data.get("args", {})
        version = message_data.get("version")
        namespace = message_data.get("namespace")
        if not method:
            LOG.warn(_("no method for message: %s") % message_data)
            ctxt.reply(_("No method for message: %s") % message_data, connection_pool=self.connection_pool)
            return
        self.pool.spawn_n(self._process_data, ctxt, version, method, namespace, args)

    def _process_data(self, ctxt, version, method, namespace, args):
        """Process a message in a new thread.

        If the proxy object we have has a dispatch method
        (see rpc.dispatcher.RpcDispatcher), pass it the version,
        method, and args and let it dispatch as appropriate.  If not, use
        the old behavior of magically calling the specified method on the
        proxy we have here.
        """
        ctxt.update_store()
        try:
            rval = self.proxy.dispatch(ctxt, version, method, namespace, **args)
            # Check if the result was a generator
            if inspect.isgenerator(rval):
                for x in rval:
                    ctxt.reply(x, None, connection_pool=self.connection_pool)
            else:
                ctxt.reply(rval, None, connection_pool=self.connection_pool)
            # This final None tells multicall that it is done.
            ctxt.reply(ending=True, connection_pool=self.connection_pool)
        except rpc_common.ClientException as e:
            LOG.debug(_("Expected exception during message handling (%s)") % e._exc_info[1])
            ctxt.reply(None, e._exc_info, connection_pool=self.connection_pool, log_failure=False)
        except Exception:
            # sys.exc_info() is deleted by LOG.exception().
            exc_info = sys.exc_info()
            LOG.error(_("Exception during message handling"), exc_info=exc_info)
            ctxt.reply(None, exc_info, connection_pool=self.connection_pool)

        def _inner():
            if initial_delay:
                greenthread.sleep(initial_delay)

            try:
                while self._running:
                    start = timeutils.utcnow()
                    self.f(*self.args, **self.kw)
                    end = timeutils.utcnow()
                    if not self._running:
                        break
                    delay = interval - timeutils.delta_seconds(start, end)
                    if delay <= 0:
                        LOG.warn(_('task run outlasted interval by %s sec') %
                                 -delay)
                    greenthread.sleep(delay if delay > 0 else 0)
            except LoopingCallDone as e:
                self.stop()
                done.send(e.retvalue)
            except Exception:
                LOG.exception(_('in fixed duration looping call'))
                done.send_exception(*sys.exc_info())
                return
            else:
                done.send(True)

    def consume(self, sock):
        #TODO(ewindisch): use zero-copy (i.e. references, not copying)
        data = sock.recv()
        LOG.debug(_("CONSUMER RECEIVED DATA: %s"), data)

        proxy = self.proxies[sock]

        if data[2] == 'cast':  # Legacy protocol
            packenv = data[3]

            ctx, msg = _deserialize(packenv)
            request = rpc_common.deserialize_msg(msg)
            ctx = RpcContext.unmarshal(ctx)
        elif data[2] == 'impl_zmq_v2':
            packenv = data[4:]

            msg = unflatten_envelope(packenv)
            request = rpc_common.deserialize_msg(msg)

            # Unmarshal only after verifying the message.
            ctx = RpcContext.unmarshal(data[3])
        else:
            LOG.error(_("ZMQ Envelope version unsupported or unknown."))
            return

        self.pool.spawn_n(self.process, proxy, ctx, request)

    def create_consumer(self, topic, proxy, fanout=False):
        # Register with matchmaker.
        _get_matchmaker().register(topic, CONF.rpc_zmq_host)

        # Subscription scenarios
        if fanout:
            sock_type = zmq.SUB
            subscribe = ('', fanout)[type(fanout) == str]
            topic = 'fanout~' + topic.split('.', 1)[0]
        else:
            sock_type = zmq.PULL
            subscribe = None
            topic = '.'.join((topic.split('.', 1)[0], CONF.rpc_zmq_host))

        if topic in self.topics:
            LOG.info(_("Skipping topic registration. Already registered."))
            return

        # Receive messages from (local) proxy
        inaddr = "ipc://%s/zmq_topic_%s" % \
            (CONF.rpc_zmq_ipc_dir, topic)

        LOG.debug(_("Consumer is a zmq.%s"),
                  ['PULL', 'SUB'][sock_type == zmq.SUB])

        self.reactor.register(proxy, inaddr, sock_type,
                              subscribe=subscribe, in_bind=False)
        self.topics.append(topic)

    def _wait_child(self):
        try:
            # Don't block if no child processes have exited
            pid, status = os.waitpid(0, os.WNOHANG)
            if not pid:
                return None
        except OSError as exc:
            if exc.errno not in (errno.EINTR, errno.ECHILD):
                raise
            return None

        if os.WIFSIGNALED(status):
            sig = os.WTERMSIG(status)
            LOG.info(_('Child %(pid)d killed by signal %(sig)d'),
                     dict(pid=pid, sig=sig))
        else:
            code = os.WEXITSTATUS(status)
            LOG.info(_('Child %(pid)s exited with status %(code)d'),
                     dict(pid=pid, code=code))

        if pid not in self.children:
            LOG.warning(_('pid %d not in child list'), pid)
            return None

        wrap = self.children.pop(pid)
        wrap.children.remove(pid)
        return wrap

def _parse_check(rule):
    """
    Parse a single base check rule into an appropriate Check object.
    """

    # Handle the special checks
    if rule == '!':
        return FalseCheck()
    elif rule == '@':
        return TrueCheck()

    try:
        kind, match = rule.split(':', 1)
    except Exception:
        LOG.exception(_("Failed to understand rule %(rule)s") % locals())
        # If the rule is invalid, we'll fail closed
        return FalseCheck()

    # Find what implements the check
    if kind in _checks:
        return _checks[kind](kind, match)
    elif None in _checks:
        return _checks[None](kind, match)
    else:
        LOG.error(_("No handler for matches of kind %s") % kind)
        return FalseCheck()

    def consume_in_thread(self):
        """Runs the ZmqProxy service."""
        ipc_dir = CONF.rpc_zmq_ipc_dir
        consume_in = "tcp://%s:%s" % \
            (CONF.rpc_zmq_bind_address,
             CONF.rpc_zmq_port)
        consumption_proxy = InternalContext(None)

        try:
            os.makedirs(ipc_dir)
        except os.error:
            if not os.path.isdir(ipc_dir):
                with excutils.save_and_reraise_exception():
                    LOG.error(_("Required IPC directory does not exist at"
                                " %s") % (ipc_dir, ))
        try:
            self.register(consumption_proxy,
                          consume_in,
                          zmq.PULL)
        except zmq.ZMQError:
            if os.access(ipc_dir, os.X_OK):
                with excutils.save_and_reraise_exception():
                    LOG.error(_("Permission denied to IPC directory at"
                                " %s") % (ipc_dir, ))
            with excutils.save_and_reraise_exception():
                LOG.error(_("Could not create ZeroMQ receiver daemon. "
                            "Socket may already be in use."))

        super(ZmqProxy, self).consume_in_thread()

    def create_xcat_mgt_network(self, zhcp, mgt_ip, mgt_mask, mgt_vswitch):
        url = self._xcat_url.xdsh("/%s" % zhcp)
        xdsh_commands = ('command=smcli Virtual_Network_Adapter_Query'
                  ' -T %s -v 0800') % self._xcat_node_name
        body = [xdsh_commands]
        result = xcatutils.xcat_request("PUT", url, body)['data'][0][0]
        code = result.split("\n")
        # return code 212: Adapter does not exist
        new_nic = ''
        if len(code) == 4 and code[1].split(': ')[2] == '212':
            new_nic = ('vmcp define nic 0800 type qdio\n' +
                    'vmcp couple 0800 system %s\n' % (mgt_vswitch))
        elif len(code) == 7:
            status = code[4].split(': ')[2]
            if status == 'Coupled and active':
                # we just assign the IP/mask,
                # no matter if it is assigned or not
                LOG.info(_("Assign IP for NIC 800."))
            else:
                LOG.error(_("NIC 800 staus is unknown."))
                return
        else:
            raise exception.zvmException(
                    msg="Unknown information from SMAPI")

        url = self._xcat_url.xdsh("/%s") % self._xcat_node_name
        cmd = new_nic + ('/usr/bin/perl /usr/sbin/sspqeth2.pl ' +
              '-a %s -d 0800 0801 0802 -e eth2 -m %s -g %s'
              % (mgt_ip, mgt_mask, mgt_ip))
        xdsh_commands = 'command=%s' % cmd
        body = [xdsh_commands]
        xcatutils.xcat_request("PUT", url, body)