add early shutdown detection, and public-ip message to ipcluster/ipengine

When engines fail to connect, the error message will describe the
most likely cause (not instructing the controller to listen on a public interface).  An similar message is included in ipcluster, but due to restrictions, its trigger is purely time-based.

IPython/parallel/apps/ipclusterapp.py

@@ -1,557 +1,591 b''
 #!/usr/bin/env python
 # encoding: utf-8
 """
 The ipcluster application.
 
 Authors:
 
 * Brian Granger
 * MinRK
 
 """
 
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2008-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 import errno
 import logging
 import os
 import re
 import signal
 
 from subprocess import check_call, CalledProcessError, PIPE
 import zmq
 from zmq.eventloop import ioloop
 
 from IPython.config.application import Application, boolean_flag
 from IPython.config.loader import Config
 from IPython.core.application import BaseIPythonApplication
 from IPython.core.profiledir import ProfileDir
 from IPython.utils.daemonize import daemonize
 from IPython.utils.importstring import import_item
 from IPython.utils.sysinfo import num_cpus
 from IPython.utils.traitlets import (Int, Unicode, Bool, CFloat, Dict, List, Any,
                                         DottedObjectName)
 
 from IPython.parallel.apps.baseapp import (
     BaseParallelApplication,
     PIDFileError,
     base_flags, base_aliases
 )
 
 
 #-----------------------------------------------------------------------------
 # Module level variables
 #-----------------------------------------------------------------------------
 
 
 default_config_file_name = u'ipcluster_config.py'
 
 
 _description = """Start an IPython cluster for parallel computing.
 
 An IPython cluster consists of 1 controller and 1 or more engines.
 This command automates the startup of these processes using a wide
 range of startup methods (SSH, local processes, PBS, mpiexec,
 Windows HPC Server 2008). To start a cluster with 4 engines on your
 local host simply do 'ipcluster start --n=4'. For more complex usage
 you will typically do 'ipython profile create mycluster --parallel', then edit
 configuration files, followed by 'ipcluster start --profile=mycluster --n=4'.
 """
 
 _main_examples = """
 ipcluster start --n=4 # start a 4 node cluster on localhost
 ipcluster start -h    # show the help string for the start subcmd
 
 ipcluster stop -h     # show the help string for the stop subcmd
 ipcluster engines -h  # show the help string for the engines subcmd
 """
 
 _start_examples = """
 ipython profile create mycluster --parallel # create mycluster profile
 ipcluster start --profile=mycluster --n=4   # start mycluster with 4 nodes
 """
 
 _stop_examples = """
 ipcluster stop --profile=mycluster  # stop a running cluster by profile name
 """
 
 _engines_examples = """
 ipcluster engines --profile=mycluster --n=4  # start 4 engines only
 """
 
 
 # Exit codes for ipcluster
 
 # This will be the exit code if the ipcluster appears to be running because
 # a .pid file exists
 ALREADY_STARTED = 10
 
 
 # This will be the exit code if ipcluster stop is run, but there is not .pid
 # file to be found.
 ALREADY_STOPPED = 11
 
 # This will be the exit code if ipcluster engines is run, but there is not .pid
 # file to be found.
 NO_CLUSTER = 12
 
 
 #-----------------------------------------------------------------------------
 # Main application
 #-----------------------------------------------------------------------------
 start_help = """Start an IPython cluster for parallel computing
 
 Start an ipython cluster by its profile name or cluster
 directory. Cluster directories contain configuration, log and
 security related files and are named using the convention
 'profile_<name>' and should be creating using the 'start'
 subcommand of 'ipcluster'. If your cluster directory is in
 the cwd or the ipython directory, you can simply refer to it
 using its profile name, 'ipcluster start --n=4 --profile=<profile>`,
 otherwise use the 'profile-dir' option.
 """
 stop_help = """Stop a running IPython cluster
 
 Stop a running ipython cluster by its profile name or cluster
 directory. Cluster directories are named using the convention
 'profile_<name>'. If your cluster directory is in
 the cwd or the ipython directory, you can simply refer to it
 using its profile name, 'ipcluster stop --profile=<profile>`, otherwise
 use the '--profile-dir' option.
 """
 engines_help = """Start engines connected to an existing IPython cluster
 
 Start one or more engines to connect to an existing Cluster
 by profile name or cluster directory.
 Cluster directories contain configuration, log and
 security related files and are named using the convention
 'profile_<name>' and should be creating using the 'start'
 subcommand of 'ipcluster'. If your cluster directory is in
 the cwd or the ipython directory, you can simply refer to it
 using its profile name, 'ipcluster engines --n=4 --profile=<profile>`,
 otherwise use the 'profile-dir' option.
 """
 stop_aliases = dict(
     signal='IPClusterStop.signal',
 )
 stop_aliases.update(base_aliases)
 
 class IPClusterStop(BaseParallelApplication):
     name = u'ipcluster'
     description = stop_help
     examples = _stop_examples
     config_file_name = Unicode(default_config_file_name)
 
     signal = Int(signal.SIGINT, config=True,
         help="signal to use for stopping processes.")
 
     aliases = Dict(stop_aliases)
 
     def start(self):
         """Start the app for the stop subcommand."""
         try:
             pid = self.get_pid_from_file()
         except PIDFileError:
             self.log.critical(
                 'Could not read pid file, cluster is probably not running.'
             )
             # Here I exit with a unusual exit status that other processes
             # can watch for to learn how I existed.
             self.remove_pid_file()
             self.exit(ALREADY_STOPPED)
 
         if not self.check_pid(pid):
             self.log.critical(
                 'Cluster [pid=%r] is not running.' % pid
             )
             self.remove_pid_file()
             # Here I exit with a unusual exit status that other processes
             # can watch for to learn how I existed.
             self.exit(ALREADY_STOPPED)
 
         elif os.name=='posix':
             sig = self.signal
             self.log.info(
                 "Stopping cluster [pid=%r] with [signal=%r]" % (pid, sig)
             )
             try:
                 os.kill(pid, sig)
             except OSError:
                 self.log.error("Stopping cluster failed, assuming already dead.",
                     exc_info=True)
                 self.remove_pid_file()
         elif os.name=='nt':
             try:
                 # kill the whole tree
                 p = check_call(['taskkill', '-pid', str(pid), '-t', '-f'], stdout=PIPE,stderr=PIPE)
             except (CalledProcessError, OSError):
                 self.log.error("Stopping cluster failed, assuming already dead.",
                     exc_info=True)
             self.remove_pid_file()
 
 engine_aliases = {}
 engine_aliases.update(base_aliases)
 engine_aliases.update(dict(
     n='IPClusterEngines.n',
     engines = 'IPClusterEngines.engine_launcher_class',
     daemonize = 'IPClusterEngines.daemonize',
 ))
 engine_flags = {}
 engine_flags.update(base_flags)
 
 engine_flags.update(dict(
     daemonize=(
         {'IPClusterEngines' : {'daemonize' : True}},
         """run the cluster into the background (not available on Windows)""",
     )
 ))
 class IPClusterEngines(BaseParallelApplication):
 
     name = u'ipcluster'
     description = engines_help
     examples = _engines_examples
     usage = None
     config_file_name = Unicode(default_config_file_name)
     default_log_level = logging.INFO
     classes = List()
     def _classes_default(self):
         from IPython.parallel.apps import launcher
         launchers = launcher.all_launchers
         eslaunchers = [ l for l in launchers if 'EngineSet' in l.__name__]
         return [ProfileDir]+eslaunchers
 
     n = Int(num_cpus(), config=True,
         help="""The number of engines to start. The default is to use one for each
         CPU on your machine""")
 
     engine_launcher = Any(config=True, help="Deprecated, use engine_launcher_class")
     def _engine_launcher_changed(self, name, old, new):
         if isinstance(new, basestring):
             self.log.warn("WARNING: %s.engine_launcher is deprecated as of 0.12,"
                     " use engine_launcher_class" % self.__class__.__name__)
             self.engine_launcher_class = new
     engine_launcher_class = DottedObjectName('LocalEngineSetLauncher',
         config=True,
         help="""The class for launching a set of Engines. Change this value
         to use various batch systems to launch your engines, such as PBS,SGE,MPIExec,etc.
         Each launcher class has its own set of configuration options, for making sure
         it will work in your environment.
 
         You can also write your own launcher, and specify it's absolute import path,
         as in 'mymodule.launcher.FTLEnginesLauncher`.
 
         Examples include:
 
             LocalEngineSetLauncher : start engines locally as subprocesses [default]
             MPIExecEngineSetLauncher : use mpiexec to launch in an MPI environment
             PBSEngineSetLauncher : use PBS (qsub) to submit engines to a batch queue
             SGEEngineSetLauncher : use SGE (qsub) to submit engines to a batch queue
             LSFEngineSetLauncher : use LSF (bsub) to submit engines to a batch queue
             SSHEngineSetLauncher : use SSH to start the controller
                                 Note that SSH does *not* move the connection files
                                 around, so you will likely have to do this manually
                                 unless the machines are on a shared file system.
             WindowsHPCEngineSetLauncher : use Windows HPC
 
         If you are using one of IPython's builtin launchers, you can specify just the
         prefix, e.g:
 
             c.IPClusterEngines.engine_launcher_class = 'SSH'
 
         or:
 
             ipcluster start --engines 'MPIExec'
 
         """
         )
     daemonize = Bool(False, config=True,
         help="""Daemonize the ipcluster program. This implies --log-to-file.
         Not available on Windows.
         """)
 
     def _daemonize_changed(self, name, old, new):
         if new:
             self.log_to_file = True
 
+    early_shutdown = Int(30, config=True, help="The timeout (in seconds)")
+    _stopping = False
+    
     aliases = Dict(engine_aliases)
     flags = Dict(engine_flags)
-    _stopping = False
 
     def initialize(self, argv=None):
         super(IPClusterEngines, self).initialize(argv)
         self.init_signal()
         self.init_launchers()
 
     def init_launchers(self):
         self.engine_launcher = self.build_launcher(self.engine_launcher_class, 'EngineSet')
-        self.engine_launcher.on_stop(lambda r: self.loop.stop())
 
     def init_signal(self):
         # Setup signals
         signal.signal(signal.SIGINT, self.sigint_handler)
 
     def build_launcher(self, clsname, kind=None):
         """import and instantiate a Launcher based on importstring"""
         if '.' not in clsname:
             # not a module, presume it's the raw name in apps.launcher
             if kind and kind not in clsname:
                 # doesn't match necessary full class name, assume it's
                 # just 'PBS' or 'MPIExec' prefix:
                 clsname = clsname + kind + 'Launcher'
             clsname = 'IPython.parallel.apps.launcher.'+clsname
         try:
             klass = import_item(clsname)
         except (ImportError, KeyError):
             self.log.fatal("Could not import launcher class: %r"%clsname)
             self.exit(1)
 
         launcher = klass(
             work_dir=u'.', config=self.config, log=self.log,
             profile_dir=self.profile_dir.location, cluster_id=self.cluster_id,
         )
         return launcher
 
+    def engines_started_okay(self):
+        self.log.info("Engines appear to have started successfully")
+        self.early_shutdown = 0
+    
     def start_engines(self):
         self.log.info("Starting %i engines"%self.n)
         self.engine_launcher.start(self.n)
+        self.engine_launcher.on_stop(self.engines_stopped_early)
+        if self.early_shutdown:
+            ioloop.DelayedCallback(self.engines_started_okay, self.early_shutdown*1000, self.loop).start()
+
+    def engines_stopped_early(self, r):
+        if self.early_shutdown and not self._stopping:
+            self.log.error("""
+            Engines shutdown early, they probably failed to connect.
+            
+            Check the engine log files for output.
+            
+            If your controller and engines are not on the same machine, you probably
+            have to instruct the controller to listen on an interface other than localhost.
+            
+            You can set this by adding "--ip='*'" to your ControllerLauncher.controller_args.
+            
+            Be sure to read our security docs before instructing your controller to listen on
+            a public interface.
+            """)
+            self.stop_launchers()
+        
+        return self.engines_stopped(r)
+    
+    def engines_stopped(self, r):
+        return self.loop.stop()
 
     def stop_engines(self):
-        self.log.info("Stopping Engines...")
         if self.engine_launcher.running:
+            self.log.info("Stopping Engines...")
             d = self.engine_launcher.stop()
             return d
         else:
             return None
 
     def stop_launchers(self, r=None):
         if not self._stopping:
             self._stopping = True
             self.log.error("IPython cluster: stopping")
             self.stop_engines()
             # Wait a few seconds to let things shut down.
-            dc = ioloop.DelayedCallback(self.loop.stop, 4000, self.loop)
+            dc = ioloop.DelayedCallback(self.loop.stop, 3000, self.loop)
             dc.start()
 
     def sigint_handler(self, signum, frame):
         self.log.debug("SIGINT received, stopping launchers...")
         self.stop_launchers()
 
     def start_logging(self):
         # Remove old log files of the controller and engine
         if self.clean_logs:
             log_dir = self.profile_dir.log_dir
             for f in os.listdir(log_dir):
                 if re.match(r'ip(engine|controller)z-\d+\.(log|err|out)',f):
                     os.remove(os.path.join(log_dir, f))
         # This will remove old log files for ipcluster itself
         # super(IPBaseParallelApplication, self).start_logging()
 
     def start(self):
         """Start the app for the engines subcommand."""
         self.log.info("IPython cluster: started")
         # First see if the cluster is already running
 
         # Now log and daemonize
         self.log.info(
             'Starting engines with [daemon=%r]' % self.daemonize
         )
         # TODO: Get daemonize working on Windows or as a Windows Server.
         if self.daemonize:
             if os.name=='posix':
                 daemonize()
 
         dc = ioloop.DelayedCallback(self.start_engines, 0, self.loop)
         dc.start()
         # Now write the new pid file AFTER our new forked pid is active.
         # self.write_pid_file()
         try:
             self.loop.start()
         except KeyboardInterrupt:
             pass
         except zmq.ZMQError as e:
             if e.errno == errno.EINTR:
                 pass
             else:
                 raise
 
 start_aliases = {}
 start_aliases.update(engine_aliases)
 start_aliases.update(dict(
     delay='IPClusterStart.delay',
     controller = 'IPClusterStart.controller_launcher_class',
 ))
 start_aliases['clean-logs'] = 'IPClusterStart.clean_logs'
 
 class IPClusterStart(IPClusterEngines):
 
     name = u'ipcluster'
     description = start_help
     examples = _start_examples
     default_log_level = logging.INFO
     auto_create = Bool(True, config=True,
         help="whether to create the profile_dir if it doesn't exist")
     classes = List()
     def _classes_default(self,):
         from IPython.parallel.apps import launcher
         return [ProfileDir] + [IPClusterEngines] + launcher.all_launchers
 
     clean_logs = Bool(True, config=True,
         help="whether to cleanup old logs before starting")
 
     delay = CFloat(1., config=True,
         help="delay (in s) between starting the controller and the engines")
 
     controller_launcher = Any(config=True, help="Deprecated, use controller_launcher_class")
     def _controller_launcher_changed(self, name, old, new):
         if isinstance(new, basestring):
             # old 0.11-style config
             self.log.warn("WARNING: %s.controller_launcher is deprecated as of 0.12,"
                     " use controller_launcher_class" % self.__class__.__name__)
             self.controller_launcher_class = new
     controller_launcher_class = DottedObjectName('LocalControllerLauncher',
         config=True,
         help="""The class for launching a Controller. Change this value if you want
         your controller to also be launched by a batch system, such as PBS,SGE,MPIExec,etc.
 
         Each launcher class has its own set of configuration options, for making sure
         it will work in your environment.
 
         Examples include:
 
             LocalControllerLauncher : start engines locally as subprocesses
             MPIExecControllerLauncher : use mpiexec to launch engines in an MPI universe
             PBSControllerLauncher : use PBS (qsub) to submit engines to a batch queue
             SGEControllerLauncher : use SGE (qsub) to submit engines to a batch queue
             LSFControllerLauncher : use LSF (bsub) to submit engines to a batch queue
             SSHControllerLauncher : use SSH to start the controller
             WindowsHPCControllerLauncher : use Windows HPC
 
         If you are using one of IPython's builtin launchers, you can specify just the
         prefix, e.g:
 
             c.IPClusterStart.controller_launcher_class = 'SSH'
 
         or:
 
             ipcluster start --controller 'MPIExec'
 
         """
         )
     reset = Bool(False, config=True,
         help="Whether to reset config files as part of '--create'."
         )
 
     # flags = Dict(flags)
     aliases = Dict(start_aliases)
 
     def init_launchers(self):
         self.controller_launcher = self.build_launcher(self.controller_launcher_class, 'Controller')
         self.engine_launcher = self.build_launcher(self.engine_launcher_class, 'EngineSet')
         self.controller_launcher.on_stop(self.stop_launchers)
-
+    
+    def engines_stopped(self, r):
+        """prevent parent.engines_stopped from stopping everything on engine shutdown"""
+        pass
+    
     def start_controller(self):
         self.controller_launcher.start()
 
     def stop_controller(self):
         # self.log.info("In stop_controller")
         if self.controller_launcher and self.controller_launcher.running:
             return self.controller_launcher.stop()
 
     def stop_launchers(self, r=None):
         if not self._stopping:
             self.stop_controller()
             super(IPClusterStart, self).stop_launchers()
 
     def start(self):
         """Start the app for the start subcommand."""
         # First see if the cluster is already running
         try:
             pid = self.get_pid_from_file()
         except PIDFileError:
             pass
         else:
             if self.check_pid(pid):
                 self.log.critical(
                     'Cluster is already running with [pid=%s]. '
                     'use "ipcluster stop" to stop the cluster.' % pid
                 )
                 # Here I exit with a unusual exit status that other processes
                 # can watch for to learn how I existed.
                 self.exit(ALREADY_STARTED)
             else:
                 self.remove_pid_file()
 
 
         # Now log and daemonize
         self.log.info(
             'Starting ipcluster with [daemon=%r]' % self.daemonize
         )
         # TODO: Get daemonize working on Windows or as a Windows Server.
         if self.daemonize:
             if os.name=='posix':
                 daemonize()
 
         dc = ioloop.DelayedCallback(self.start_controller, 0, self.loop)
         dc.start()
         dc = ioloop.DelayedCallback(self.start_engines, 1000*self.delay, self.loop)
         dc.start()
         # Now write the new pid file AFTER our new forked pid is active.
         self.write_pid_file()
         try:
             self.loop.start()
         except KeyboardInterrupt:
             pass
         except zmq.ZMQError as e:
             if e.errno == errno.EINTR:
                 pass
             else:
                 raise
         finally:
             self.remove_pid_file()
 
 base='IPython.parallel.apps.ipclusterapp.IPCluster'
 
 class IPClusterApp(Application):
     name = u'ipcluster'
     description = _description
     examples = _main_examples
 
     subcommands = {
                 'start' : (base+'Start', start_help),
                 'stop' : (base+'Stop', stop_help),
                 'engines' : (base+'Engines', engines_help),
     }
 
     # no aliases or flags for parent App
     aliases = Dict()
     flags = Dict()
 
     def start(self):
         if self.subapp is None:
             print "No subcommand specified. Must specify one of: %s"%(self.subcommands.keys())
             print
             self.print_description()
             self.print_subcommands()
             self.exit(1)
         else:
             return self.subapp.start()
 
 def launch_new_instance():
     """Create and run the IPython cluster."""
     app = IPClusterApp.instance()
     app.initialize()
     app.start()
 
 
 if __name__ == '__main__':
     launch_new_instance()
 

IPython/parallel/engine/engine.py

@@ -1,226 +1,234 b''
 """A simple engine that talks to a controller over 0MQ.
 it handles registration, etc. and launches a kernel
 connected to the Controller's Schedulers.
 
 Authors:
 
 * Min RK
 """
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2010-2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 from __future__ import print_function
 
 import sys
 import time
 from getpass import getpass
 
 import zmq
 from zmq.eventloop import ioloop, zmqstream
 
 from IPython.external.ssh import tunnel
 # internal
 from IPython.utils.traitlets import (
     Instance, Dict, Int, Type, CFloat, Unicode, CBytes, Bool
 )
 # from IPython.utils.localinterfaces import LOCALHOST
 
 from IPython.parallel.controller.heartmonitor import Heart
 from IPython.parallel.factory import RegistrationFactory
 from IPython.parallel.util import disambiguate_url, asbytes
 
 from IPython.zmq.session import Message
 
 from .streamkernel import Kernel
 
 class EngineFactory(RegistrationFactory):
     """IPython engine"""
 
     # configurables:
     out_stream_factory=Type('IPython.zmq.iostream.OutStream', config=True,
         help="""The OutStream for handling stdout/err.
         Typically 'IPython.zmq.iostream.OutStream'""")
     display_hook_factory=Type('IPython.zmq.displayhook.ZMQDisplayHook', config=True,
         help="""The class for handling displayhook.
         Typically 'IPython.zmq.displayhook.ZMQDisplayHook'""")
     location=Unicode(config=True,
         help="""The location (an IP address) of the controller.  This is
         used for disambiguating URLs, to determine whether
         loopback should be used to connect or the public address.""")
     timeout=CFloat(2,config=True,
         help="""The time (in seconds) to wait for the Controller to respond
         to registration requests before giving up.""")
     sshserver=Unicode(config=True,
         help="""The SSH server to use for tunneling connections to the Controller.""")
     sshkey=Unicode(config=True,
         help="""The SSH private key file to use when tunneling connections to the Controller.""")
     paramiko=Bool(sys.platform == 'win32', config=True,
         help="""Whether to use paramiko instead of openssh for tunnels.""")
 
     # not configurable:
     user_ns=Dict()
     id=Int(allow_none=True)
     registrar=Instance('zmq.eventloop.zmqstream.ZMQStream')
     kernel=Instance(Kernel)
 
     bident = CBytes()
     ident = Unicode()
     def _ident_changed(self, name, old, new):
         self.bident = asbytes(new)
     using_ssh=Bool(False)
 
 
     def __init__(self, **kwargs):
         super(EngineFactory, self).__init__(**kwargs)
         self.ident = self.session.session
 
     def init_connector(self):
         """construct connection function, which handles tunnels."""
         self.using_ssh = bool(self.sshkey or self.sshserver)
 
         if self.sshkey and not self.sshserver:
             # We are using ssh directly to the controller, tunneling localhost to localhost
             self.sshserver = self.url.split('://')[1].split(':')[0]
 
         if self.using_ssh:
             if tunnel.try_passwordless_ssh(self.sshserver, self.sshkey, self.paramiko):
                 password=False
             else:
                 password = getpass("SSH Password for %s: "%self.sshserver)
         else:
             password = False
 
         def connect(s, url):
             url = disambiguate_url(url, self.location)
             if self.using_ssh:
                 self.log.debug("Tunneling connection to %s via %s"%(url, self.sshserver))
                 return tunnel.tunnel_connection(s, url, self.sshserver,
                             keyfile=self.sshkey, paramiko=self.paramiko,
                             password=password,
                 )
             else:
                 return s.connect(url)
 
         def maybe_tunnel(url):
             """like connect, but don't complete the connection (for use by heartbeat)"""
             url = disambiguate_url(url, self.location)
             if self.using_ssh:
                 self.log.debug("Tunneling connection to %s via %s"%(url, self.sshserver))
                 url,tunnelobj = tunnel.open_tunnel(url, self.sshserver,
                             keyfile=self.sshkey, paramiko=self.paramiko,
                             password=password,
                 )
             return url
         return connect, maybe_tunnel
 
     def register(self):
         """send the registration_request"""
 
         self.log.info("Registering with controller at %s"%self.url)
         ctx = self.context
         connect,maybe_tunnel = self.init_connector()
         reg = ctx.socket(zmq.DEALER)
         reg.setsockopt(zmq.IDENTITY, self.bident)
         connect(reg, self.url)
         self.registrar = zmqstream.ZMQStream(reg, self.loop)
 
 
         content = dict(queue=self.ident, heartbeat=self.ident, control=self.ident)
         self.registrar.on_recv(lambda msg: self.complete_registration(msg, connect, maybe_tunnel))
         # print (self.session.key)
         self.session.send(self.registrar, "registration_request",content=content)
 
     def complete_registration(self, msg, connect, maybe_tunnel):
         # print msg
         self._abort_dc.stop()
         ctx = self.context
         loop = self.loop
         identity = self.bident
         idents,msg = self.session.feed_identities(msg)
         msg = Message(self.session.unserialize(msg))
 
         if msg.content.status == 'ok':
             self.id = int(msg.content.id)
 
             # launch heartbeat
             hb_addrs = msg.content.heartbeat
 
             # possibly forward hb ports with tunnels
             hb_addrs = [ maybe_tunnel(addr) for addr in hb_addrs ]
             heart = Heart(*map(str, hb_addrs), heart_id=identity)
             heart.start()
 
             # create Shell Streams (MUX, Task, etc.):
             queue_addr = msg.content.mux
             shell_addrs = [ str(queue_addr) ]
             task_addr = msg.content.task
             if task_addr:
                 shell_addrs.append(str(task_addr))
 
             # Uncomment this to go back to two-socket model
             # shell_streams = []
             # for addr in shell_addrs:
             #     stream = zmqstream.ZMQStream(ctx.socket(zmq.ROUTER), loop)
             #     stream.setsockopt(zmq.IDENTITY, identity)
             #     stream.connect(disambiguate_url(addr, self.location))
             #     shell_streams.append(stream)
 
             # Now use only one shell stream for mux and tasks
             stream = zmqstream.ZMQStream(ctx.socket(zmq.ROUTER), loop)
             stream.setsockopt(zmq.IDENTITY, identity)
             shell_streams = [stream]
             for addr in shell_addrs:
                 connect(stream, addr)
             # end single stream-socket
 
             # control stream:
             control_addr = str(msg.content.control)
             control_stream = zmqstream.ZMQStream(ctx.socket(zmq.ROUTER), loop)
             control_stream.setsockopt(zmq.IDENTITY, identity)
             connect(control_stream, control_addr)
 
             # create iopub stream:
             iopub_addr = msg.content.iopub
             iopub_stream = zmqstream.ZMQStream(ctx.socket(zmq.PUB), loop)
             iopub_stream.setsockopt(zmq.IDENTITY, identity)
             connect(iopub_stream, iopub_addr)
 
             # # Redirect input streams and set a display hook.
             if self.out_stream_factory:
                 sys.stdout = self.out_stream_factory(self.session, iopub_stream, u'stdout')
                 sys.stdout.topic = 'engine.%i.stdout'%self.id
                 sys.stderr = self.out_stream_factory(self.session, iopub_stream, u'stderr')
                 sys.stderr.topic = 'engine.%i.stderr'%self.id
             if self.display_hook_factory:
                 sys.displayhook = self.display_hook_factory(self.session, iopub_stream)
                 sys.displayhook.topic = 'engine.%i.pyout'%self.id
 
             self.kernel = Kernel(config=self.config, int_id=self.id, ident=self.ident, session=self.session,
                     control_stream=control_stream, shell_streams=shell_streams, iopub_stream=iopub_stream,
                     loop=loop, user_ns = self.user_ns, log=self.log)
             self.kernel.start()
 
 
         else:
             self.log.fatal("Registration Failed: %s"%msg)
             raise Exception("Registration Failed: %s"%msg)
 
         self.log.info("Completed registration with id %i"%self.id)
 
 
     def abort(self):
         self.log.fatal("Registration timed out after %.1f seconds"%self.timeout)
+        if '127' in self.url:
+            self.log.fatal("""
+            If the controller and engines are not on the same machine,
+            you will have to instruct the controller to listen on an external IP (in ipcontroller_config.py):
+                c.HubFactory.ip='*' # for all interfaces, internal and external
+                c.HubFactory.ip='192.168.1.101' # or any interface that the engines can see
+            or tunnel connections via ssh.
+            """)
         self.session.send(self.registrar, "unregistration_request", content=dict(id=self.id))
         time.sleep(1)
         sys.exit(255)
 
     def start(self):
         dc = ioloop.DelayedCallback(self.register, 0, self.loop)
         dc.start()
         self._abort_dc = ioloop.DelayedCallback(self.abort, self.timeout*1000, self.loop)
         self._abort_dc.start()