enables resume of ipcontroller

cleans up EngineConnector and registration messages further, only storing single UUID for each engine (all the stored UUIDs happened to be the same already).

Message spec docs updated to reflect changes to message formats.

IPython/parallel/apps/ipcontrollerapp.py

@@ -1,509 +1,528 b''
 #!/usr/bin/env python
 # encoding: utf-8
 """
 The IPython controller 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
 #-----------------------------------------------------------------------------
 
 from __future__ import with_statement
 
 import json
 import os
 import socket
 import stat
 import sys
 
 from multiprocessing import Process
 from signal import signal, SIGINT, SIGABRT, SIGTERM
 
 import zmq
 from zmq.devices import ProcessMonitoredQueue
 from zmq.log.handlers import PUBHandler
 
 from IPython.core.profiledir import ProfileDir
 
 from IPython.parallel.apps.baseapp import (
     BaseParallelApplication,
     base_aliases,
     base_flags,
     catch_config_error,
 )
 from IPython.utils.importstring import import_item
 from IPython.utils.traitlets import Instance, Unicode, Bool, List, Dict, TraitError
 
 from IPython.zmq.session import (
     Session, session_aliases, session_flags, default_secure
 )
 
 from IPython.parallel.controller.heartmonitor import HeartMonitor
 from IPython.parallel.controller.hub import HubFactory
 from IPython.parallel.controller.scheduler import TaskScheduler,launch_scheduler
 from IPython.parallel.controller.sqlitedb import SQLiteDB
 
 from IPython.parallel.util import split_url, disambiguate_url
 
 # conditional import of MongoDB backend class
 
 try:
     from IPython.parallel.controller.mongodb import MongoDB
 except ImportError:
     maybe_mongo = []
 else:
     maybe_mongo = [MongoDB]
 
 
 #-----------------------------------------------------------------------------
 # Module level variables
 #-----------------------------------------------------------------------------
 
 
 #: The default config file name for this application
 default_config_file_name = u'ipcontroller_config.py'
 
 
 _description = """Start the IPython controller for parallel computing.
 
 The IPython controller provides a gateway between the IPython engines and
 clients. The controller needs to be started before the engines and can be
 configured using command line options or using a cluster directory. Cluster
 directories contain config, log and security files and are usually located in
 your ipython directory and named as "profile_name". See the `profile`
 and `profile-dir` options for details.
 """
 
 _examples = """
 ipcontroller --ip=192.168.0.1 --port=1000  # listen on ip, port for engines
 ipcontroller --scheme=pure  # use the pure zeromq scheduler
 """
 
 
 #-----------------------------------------------------------------------------
 # The main application
 #-----------------------------------------------------------------------------
 flags = {}
 flags.update(base_flags)
 flags.update({
     'usethreads' : ( {'IPControllerApp' : {'use_threads' : True}},
                     'Use threads instead of processes for the schedulers'),
     'sqlitedb' : ({'HubFactory' : {'db_class' : 'IPython.parallel.controller.sqlitedb.SQLiteDB'}},
                     'use the SQLiteDB backend'),
     'mongodb' : ({'HubFactory' : {'db_class' : 'IPython.parallel.controller.mongodb.MongoDB'}},
                     'use the MongoDB backend'),
     'dictdb' : ({'HubFactory' : {'db_class' : 'IPython.parallel.controller.dictdb.DictDB'}},
                     'use the in-memory DictDB backend'),
     'nodb' : ({'HubFactory' : {'db_class' : 'IPython.parallel.controller.dictdb.NoDB'}},
                     """use dummy DB backend, which doesn't store any information.
                     
                     This is the default as of IPython 0.13.
                     
                     To enable delayed or repeated retrieval of results from the Hub,
                     select one of the true db backends.
                     """),
     'reuse' : ({'IPControllerApp' : {'reuse_files' : True}},
-                    'reuse existing json connection files')
+                    'reuse existing json connection files'),
+    'restore' : ({'IPControllerApp' : {'restore_engines' : True, 'reuse_files' : True}},
+                    'Attempt to restore engines from a JSON file.  '
+                    'For use when resuming a crashed controller'),
 })
 
 flags.update(session_flags)
 
 aliases = dict(
     ssh = 'IPControllerApp.ssh_server',
     enginessh = 'IPControllerApp.engine_ssh_server',
     location = 'IPControllerApp.location',
 
     url = 'HubFactory.url',
     ip = 'HubFactory.ip',
     transport = 'HubFactory.transport',
     port = 'HubFactory.regport',
 
     ping = 'HeartMonitor.period',
 
     scheme = 'TaskScheduler.scheme_name',
     hwm = 'TaskScheduler.hwm',
 )
 aliases.update(base_aliases)
 aliases.update(session_aliases)
 
 class IPControllerApp(BaseParallelApplication):
 
     name = u'ipcontroller'
     description = _description
     examples = _examples
     config_file_name = Unicode(default_config_file_name)
     classes = [ProfileDir, Session, HubFactory, TaskScheduler, HeartMonitor, SQLiteDB] + maybe_mongo
     
     # change default to True
     auto_create = Bool(True, config=True,
         help="""Whether to create profile dir if it doesn't exist.""")
     
     reuse_files = Bool(False, config=True,
         help="""Whether to reuse existing json connection files.
         If False, connection files will be removed on a clean exit.
         """
     )
+    restore_engines = Bool(False, config=True,
+        help="""Reload engine state from JSON file
+        """
+    )
     ssh_server = Unicode(u'', config=True,
         help="""ssh url for clients to use when connecting to the Controller
         processes. It should be of the form: [user@]server[:port]. The
         Controller's listening addresses must be accessible from the ssh server""",
     )
     engine_ssh_server = Unicode(u'', config=True,
         help="""ssh url for engines to use when connecting to the Controller
         processes. It should be of the form: [user@]server[:port]. The
         Controller's listening addresses must be accessible from the ssh server""",
     )
     location = Unicode(u'', config=True,
         help="""The external IP or domain name of the Controller, used for disambiguating
         engine and client connections.""",
     )
     import_statements = List([], config=True,
         help="import statements to be run at startup.  Necessary in some environments"
     )
 
     use_threads = Bool(False, config=True,
         help='Use threads instead of processes for the schedulers',
     )
 
     engine_json_file = Unicode('ipcontroller-engine.json', config=True,
         help="JSON filename where engine connection info will be stored.")
     client_json_file = Unicode('ipcontroller-client.json', config=True,
         help="JSON filename where client connection info will be stored.")
     
     def _cluster_id_changed(self, name, old, new):
         super(IPControllerApp, self)._cluster_id_changed(name, old, new)
         self.engine_json_file = "%s-engine.json" % self.name
         self.client_json_file = "%s-client.json" % self.name
 
 
     # internal
     children = List()
     mq_class = Unicode('zmq.devices.ProcessMonitoredQueue')
 
     def _use_threads_changed(self, name, old, new):
         self.mq_class = 'zmq.devices.%sMonitoredQueue'%('Thread' if new else 'Process')
     
     write_connection_files = Bool(True,
         help="""Whether to write connection files to disk.
         True in all cases other than runs with `reuse_files=True` *after the first*
         """
     )
 
     aliases = Dict(aliases)
     flags = Dict(flags)
     
 
     def save_connection_dict(self, fname, cdict):
         """save a connection dict to json file."""
         c = self.config
         url = cdict['registration']
         location = cdict['location']
         
         if not location:
             try:
                 location = socket.gethostbyname_ex(socket.gethostname())[2][-1]
             except (socket.gaierror, IndexError):
                 self.log.warn("Could not identify this machine's IP, assuming 127.0.0.1."
                 " You may need to specify '--location=<external_ip_address>' to help"
                 " IPython decide when to connect via loopback.")
                 location = '127.0.0.1'
             cdict['location'] = location
         fname = os.path.join(self.profile_dir.security_dir, fname)
         self.log.info("writing connection info to %s", fname)
         with open(fname, 'w') as f:
             f.write(json.dumps(cdict, indent=2))
         os.chmod(fname, stat.S_IRUSR|stat.S_IWUSR)
     
     def load_config_from_json(self):
         """load config from existing json connector files."""
         c = self.config
         self.log.debug("loading config from JSON")
         
         # load engine config
         
         fname = os.path.join(self.profile_dir.security_dir, self.engine_json_file)
         self.log.info("loading connection info from %s", fname)
         with open(fname) as f:
             ecfg = json.loads(f.read())
         
         # json gives unicode, Session.key wants bytes
         c.Session.key = ecfg['exec_key'].encode('ascii')
         
         xport,ip = ecfg['interface'].split('://')
         
         c.HubFactory.engine_ip = ip
         c.HubFactory.engine_transport = xport
         
         self.location = ecfg['location']
         if not self.engine_ssh_server:
             self.engine_ssh_server = ecfg['ssh']
         
         # load client config
         
         fname = os.path.join(self.profile_dir.security_dir, self.client_json_file)
         self.log.info("loading connection info from %s", fname)
         with open(fname) as f:
             ccfg = json.loads(f.read())
         
         for key in ('exec_key', 'registration', 'pack', 'unpack'):
             assert ccfg[key] == ecfg[key], "mismatch between engine and client info: %r" % key
         
         xport,addr = ccfg['interface'].split('://')
         
         c.HubFactory.client_transport = xport
         c.HubFactory.client_ip = ip
         if not self.ssh_server:
             self.ssh_server = ccfg['ssh']
         
         # load port config:
         c.HubFactory.regport = ecfg['registration']
         c.HubFactory.hb = (ecfg['hb_ping'], ecfg['hb_pong'])
         c.HubFactory.control = (ccfg['control'], ecfg['control'])
         c.HubFactory.mux = (ccfg['mux'], ecfg['mux'])
         c.HubFactory.task = (ccfg['task'], ecfg['task'])
         c.HubFactory.iopub = (ccfg['iopub'], ecfg['iopub'])
         c.HubFactory.notifier_port = ccfg['notification']
     
     def cleanup_connection_files(self):
         if self.reuse_files:
             self.log.debug("leaving JSON connection files for reuse")
             return
         self.log.debug("cleaning up JSON connection files")
         for f in (self.client_json_file, self.engine_json_file):
             f = os.path.join(self.profile_dir.security_dir, f)
             try:
                 os.remove(f)
             except Exception as e:
                 self.log.error("Failed to cleanup connection file: %s", e)
             else:
                 self.log.debug(u"removed %s", f)
     
     def load_secondary_config(self):
         """secondary config, loading from JSON and setting defaults"""
         if self.reuse_files:
             try:
                 self.load_config_from_json()
             except (AssertionError,IOError) as e:
                 self.log.error("Could not load config from JSON: %s" % e)
             else:
                 # successfully loaded config from JSON, and reuse=True
                 # no need to wite back the same file
                 self.write_connection_files = False
                 
         # switch Session.key default to secure
         default_secure(self.config)
         self.log.debug("Config changed")
         self.log.debug(repr(self.config))
         
     def init_hub(self):
         c = self.config
         
         self.do_import_statements()
         
         try:
             self.factory = HubFactory(config=c, log=self.log)
             # self.start_logging()
             self.factory.init_hub()
         except TraitError:
             raise
         except Exception:
             self.log.error("Couldn't construct the Controller", exc_info=True)
             self.exit(1)
         
         if self.write_connection_files:
             # save to new json config files
             f = self.factory
             base = {
                 'exec_key'  : f.session.key.decode('ascii'),
                 'location'  : self.location,
                 'pack'      : f.session.packer,
                 'unpack'    : f.session.unpacker,
             }
             
             cdict = {'ssh' : self.ssh_server}
             cdict.update(f.client_info)
             cdict.update(base)
             self.save_connection_dict(self.client_json_file, cdict)
             
             edict = {'ssh' : self.engine_ssh_server}
             edict.update(f.engine_info)
             edict.update(base)
             self.save_connection_dict(self.engine_json_file, edict)
 
+        fname = "engines%s.json" % self.cluster_id
+        self.factory.hub.engine_state_file = os.path.join(self.profile_dir.log_dir, fname)
+        if self.restore_engines:
+            self.factory.hub._load_engine_state()
+
     def init_schedulers(self):
         children = self.children
         mq = import_item(str(self.mq_class))
         
         f = self.factory
+        ident = f.session.bsession
         # disambiguate url, in case of *
         monitor_url = disambiguate_url(f.monitor_url)
         # maybe_inproc = 'inproc://monitor' if self.use_threads else monitor_url
         # IOPub relay (in a Process)
         q = mq(zmq.PUB, zmq.SUB, zmq.PUB, b'N/A',b'iopub')
         q.bind_in(f.client_url('iopub'))
+        q.setsockopt_in(zmq.IDENTITY, ident+"_iopub")
         q.bind_out(f.engine_url('iopub'))
         q.setsockopt_out(zmq.SUBSCRIBE, b'')
         q.connect_mon(monitor_url)
         q.daemon=True
         children.append(q)
 
         # Multiplexer Queue (in a Process)
         q = mq(zmq.ROUTER, zmq.ROUTER, zmq.PUB, b'in', b'out')
         q.bind_in(f.client_url('mux'))
-        q.setsockopt_in(zmq.IDENTITY, b'mux')
+        q.setsockopt_in(zmq.IDENTITY, b'mux_in')
         q.bind_out(f.engine_url('mux'))
+        q.setsockopt_out(zmq.IDENTITY, b'mux_out')
         q.connect_mon(monitor_url)
         q.daemon=True
         children.append(q)
 
         # Control Queue (in a Process)
         q = mq(zmq.ROUTER, zmq.ROUTER, zmq.PUB, b'incontrol', b'outcontrol')
         q.bind_in(f.client_url('control'))
-        q.setsockopt_in(zmq.IDENTITY, b'control')
+        q.setsockopt_in(zmq.IDENTITY, b'control_in')
         q.bind_out(f.engine_url('control'))
+        q.setsockopt_out(zmq.IDENTITY, b'control_out')
         q.connect_mon(monitor_url)
         q.daemon=True
         children.append(q)
         try:
             scheme = self.config.TaskScheduler.scheme_name
         except AttributeError:
             scheme = TaskScheduler.scheme_name.get_default_value()
         # Task Queue (in a Process)
         if scheme == 'pure':
             self.log.warn("task::using pure DEALER Task scheduler")
             q = mq(zmq.ROUTER, zmq.DEALER, zmq.PUB, b'intask', b'outtask')
             # q.setsockopt_out(zmq.HWM, hub.hwm)
             q.bind_in(f.client_url('task'))
-            q.setsockopt_in(zmq.IDENTITY, b'task')
+            q.setsockopt_in(zmq.IDENTITY, b'task_in')
             q.bind_out(f.engine_url('task'))
+            q.setsockopt_out(zmq.IDENTITY, b'task_out')
             q.connect_mon(monitor_url)
             q.daemon=True
             children.append(q)
         elif scheme == 'none':
             self.log.warn("task::using no Task scheduler")
 
         else:
             self.log.info("task::using Python %s Task scheduler"%scheme)
             sargs = (f.client_url('task'), f.engine_url('task'),
-                                monitor_url, disambiguate_url(f.client_url('notification')))
+                    monitor_url, disambiguate_url(f.client_url('notification')),
+                    disambiguate_url(f.client_url('registration')),
+            )
             kwargs = dict(logname='scheduler', loglevel=self.log_level,
                             log_url = self.log_url, config=dict(self.config))
             if 'Process' in self.mq_class:
                 # run the Python scheduler in a Process
                 q = Process(target=launch_scheduler, args=sargs, kwargs=kwargs)
                 q.daemon=True
                 children.append(q)
             else:
                 # single-threaded Controller
                 kwargs['in_thread'] = True
                 launch_scheduler(*sargs, **kwargs)
 
     def terminate_children(self):
         child_procs = []
         for child in self.children:
             if isinstance(child, ProcessMonitoredQueue):
                 child_procs.append(child.launcher)
             elif isinstance(child, Process):
                 child_procs.append(child)
         if child_procs:
             self.log.critical("terminating children...")
             for child in child_procs:
                 try:
                     child.terminate()
                 except OSError:
                     # already dead
                     pass
     
     def handle_signal(self, sig, frame):
         self.log.critical("Received signal %i, shutting down", sig)
         self.terminate_children()
         self.loop.stop()
     
     def init_signal(self):
         for sig in (SIGINT, SIGABRT, SIGTERM):
             signal(sig, self.handle_signal)
 
     def do_import_statements(self):
         statements = self.import_statements
         for s in statements:
             try:
                 self.log.msg("Executing statement: '%s'" % s)
                 exec s in globals(), locals()
             except:
                 self.log.msg("Error running statement: %s" % s)
 
     def forward_logging(self):
         if self.log_url:
             self.log.info("Forwarding logging to %s"%self.log_url)
             context = zmq.Context.instance()
             lsock = context.socket(zmq.PUB)
             lsock.connect(self.log_url)
             handler = PUBHandler(lsock)
             handler.root_topic = 'controller'
             handler.setLevel(self.log_level)
             self.log.addHandler(handler)
     
     @catch_config_error
     def initialize(self, argv=None):
         super(IPControllerApp, self).initialize(argv)
         self.forward_logging()
         self.load_secondary_config()
         self.init_hub()
         self.init_schedulers()
     
     def start(self):
         # Start the subprocesses:
         self.factory.start()
         # children must be started before signals are setup,
         # otherwise signal-handling will fire multiple times
         for child in self.children:
             child.start()
         self.init_signal()
 
         self.write_pid_file(overwrite=True)
 
         try:
             self.factory.loop.start()
         except KeyboardInterrupt:
             self.log.critical("Interrupted, Exiting...\n")
         finally:
             self.cleanup_connection_files()
             
 
 
 def launch_new_instance():
     """Create and run the IPython controller"""
     if sys.platform == 'win32':
         # make sure we don't get called from a multiprocessing subprocess
         # this can result in infinite Controllers being started on Windows
         # which doesn't have a proper fork, so multiprocessing is wonky
         
         # this only comes up when IPython has been installed using vanilla
         # setuptools, and *not* distribute.
         import multiprocessing
         p = multiprocessing.current_process()
         # the main process has name 'MainProcess'
         # subprocesses will have names like 'Process-1'
         if p.name != 'MainProcess':
             # we are a subprocess, don't start another Controller!
             return
     app = IPControllerApp.instance()
     app.initialize()
     app.start()
 
 
 if __name__ == '__main__':
     launch_new_instance()

IPython/parallel/client/client.py

@@ -1,1699 +1,1700 b''
 """A semi-synchronous Client for the ZMQ cluster
 
 Authors:
 
 * MinRK
 """
 #-----------------------------------------------------------------------------
 #  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.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 import os
 import json
 import sys
 from threading import Thread, Event
 import time
 import warnings
 from datetime import datetime
 from getpass import getpass
 from pprint import pprint
 
 pjoin = os.path.join
 
 import zmq
 # from zmq.eventloop import ioloop, zmqstream
 
 from IPython.config.configurable import MultipleInstanceError
 from IPython.core.application import BaseIPythonApplication
 from IPython.core.profiledir import ProfileDir, ProfileDirError
 
 from IPython.utils.coloransi import TermColors
 from IPython.utils.jsonutil import rekey
 from IPython.utils.localinterfaces import LOCAL_IPS
 from IPython.utils.path import get_ipython_dir
 from IPython.utils.py3compat import cast_bytes
 from IPython.utils.traitlets import (HasTraits, Integer, Instance, Unicode,
                                     Dict, List, Bool, Set, Any)
 from IPython.external.decorator import decorator
 from IPython.external.ssh import tunnel
 
 from IPython.parallel import Reference
 from IPython.parallel import error
 from IPython.parallel import util
 
 from IPython.zmq.session import Session, Message
 
 from .asyncresult import AsyncResult, AsyncHubResult
 from .view import DirectView, LoadBalancedView
 
 if sys.version_info[0] >= 3:
     # xrange is used in a couple 'isinstance' tests in py2
     # should be just 'range' in 3k
     xrange = range
 
 #--------------------------------------------------------------------------
 # Decorators for Client methods
 #--------------------------------------------------------------------------
 
 @decorator
 def spin_first(f, self, *args, **kwargs):
     """Call spin() to sync state prior to calling the method."""
     self.spin()
     return f(self, *args, **kwargs)
 
 
 #--------------------------------------------------------------------------
 # Classes
 #--------------------------------------------------------------------------
 
 
 class ExecuteReply(object):
     """wrapper for finished Execute results"""
     def __init__(self, msg_id, content, metadata):
         self.msg_id = msg_id
         self._content = content
         self.execution_count = content['execution_count']
         self.metadata = metadata
     
     def __getitem__(self, key):
         return self.metadata[key]
     
     def __getattr__(self, key):
         if key not in self.metadata:
             raise AttributeError(key)
         return self.metadata[key]
     
     def __repr__(self):
         pyout = self.metadata['pyout'] or {'data':{}}
         text_out = pyout['data'].get('text/plain', '')
         if len(text_out) > 32:
             text_out = text_out[:29] + '...'
         
         return "<ExecuteReply[%i]: %s>" % (self.execution_count, text_out)
     
     def _repr_pretty_(self, p, cycle):
         pyout = self.metadata['pyout'] or {'data':{}}
         text_out = pyout['data'].get('text/plain', '')
         
         if not text_out:
             return
         
         try:
             ip = get_ipython()
         except NameError:
             colors = "NoColor"
         else:
             colors = ip.colors
         
         if colors == "NoColor":
             out = normal = ""
         else:
             out = TermColors.Red
             normal = TermColors.Normal
         
         if '\n' in text_out and not text_out.startswith('\n'):
             # add newline for multiline reprs
             text_out = '\n' + text_out
         
         p.text(
             out + u'Out[%i:%i]: ' % (
                 self.metadata['engine_id'], self.execution_count
             ) + normal + text_out
         )
     
     def _repr_html_(self):
         pyout = self.metadata['pyout'] or {'data':{}}
         return pyout['data'].get("text/html")
     
     def _repr_latex_(self):
         pyout = self.metadata['pyout'] or {'data':{}}
         return pyout['data'].get("text/latex")
     
     def _repr_json_(self):
         pyout = self.metadata['pyout'] or {'data':{}}
         return pyout['data'].get("application/json")
     
     def _repr_javascript_(self):
         pyout = self.metadata['pyout'] or {'data':{}}
         return pyout['data'].get("application/javascript")
     
     def _repr_png_(self):
         pyout = self.metadata['pyout'] or {'data':{}}
         return pyout['data'].get("image/png")
     
     def _repr_jpeg_(self):
         pyout = self.metadata['pyout'] or {'data':{}}
         return pyout['data'].get("image/jpeg")
     
     def _repr_svg_(self):
         pyout = self.metadata['pyout'] or {'data':{}}
         return pyout['data'].get("image/svg+xml")
 
 
 class Metadata(dict):
     """Subclass of dict for initializing metadata values.
 
     Attribute access works on keys.
 
     These objects have a strict set of keys - errors will raise if you try
     to add new keys.
     """
     def __init__(self, *args, **kwargs):
         dict.__init__(self)
         md = {'msg_id' : None,
               'submitted' : None,
               'started' : None,
               'completed' : None,
               'received' : None,
               'engine_uuid' : None,
               'engine_id' : None,
               'follow' : None,
               'after' : None,
               'status' : None,
 
               'pyin' : None,
               'pyout' : None,
               'pyerr' : None,
               'stdout' : '',
               'stderr' : '',
               'outputs' : [],
               'outputs_ready' : False,
             }
         self.update(md)
         self.update(dict(*args, **kwargs))
 
     def __getattr__(self, key):
         """getattr aliased to getitem"""
         if key in self.iterkeys():
             return self[key]
         else:
             raise AttributeError(key)
 
     def __setattr__(self, key, value):
         """setattr aliased to setitem, with strict"""
         if key in self.iterkeys():
             self[key] = value
         else:
             raise AttributeError(key)
 
     def __setitem__(self, key, value):
         """strict static key enforcement"""
         if key in self.iterkeys():
             dict.__setitem__(self, key, value)
         else:
             raise KeyError(key)
 
 
 class Client(HasTraits):
     """A semi-synchronous client to the IPython ZMQ cluster
 
     Parameters
     ----------
 
     url_file : str/unicode; path to ipcontroller-client.json
         This JSON file should contain all the information needed to connect to a cluster,
         and is likely the only argument needed.
         Connection information for the Hub's registration.  If a json connector
         file is given, then likely no further configuration is necessary.
         [Default: use profile]
     profile : bytes
         The name of the Cluster profile to be used to find connector information.
         If run from an IPython application, the default profile will be the same
         as the running application, otherwise it will be 'default'.
     context : zmq.Context
         Pass an existing zmq.Context instance, otherwise the client will create its own.
     debug : bool
         flag for lots of message printing for debug purposes
     timeout : int/float
         time (in seconds) to wait for connection replies from the Hub
         [Default: 10]
 
     #-------------- session related args ----------------
 
     config : Config object
         If specified, this will be relayed to the Session for configuration
     username : str
         set username for the session object
 
     #-------------- ssh related args ----------------
     # These are args for configuring the ssh tunnel to be used
     # credentials are used to forward connections over ssh to the Controller
     # Note that the ip given in `addr` needs to be relative to sshserver
     # The most basic case is to leave addr as pointing to localhost (127.0.0.1),
     # and set sshserver as the same machine the Controller is on. However,
     # the only requirement is that sshserver is able to see the Controller
     # (i.e. is within the same trusted network).
 
     sshserver : str
         A string of the form passed to ssh, i.e. 'server.tld' or '[e-mail]:port'
         If keyfile or password is specified, and this is not, it will default to
         the ip given in addr.
     sshkey : str; path to ssh private key file
         This specifies a key to be used in ssh login, default None.
         Regular default ssh keys will be used without specifying this argument.
     password : str
         Your ssh password to sshserver. Note that if this is left None,
         you will be prompted for it if passwordless key based login is unavailable.
     paramiko : bool
         flag for whether to use paramiko instead of shell ssh for tunneling.
         [default: True on win32, False else]
 
 
     Attributes
     ----------
 
     ids : list of int engine IDs
         requesting the ids attribute always synchronizes
         the registration state. To request ids without synchronization,
         use semi-private _ids attributes.
 
     history : list of msg_ids
         a list of msg_ids, keeping track of all the execution
         messages you have submitted in order.
 
     outstanding : set of msg_ids
         a set of msg_ids that have been submitted, but whose
         results have not yet been received.
 
     results : dict
         a dict of all our results, keyed by msg_id
 
     block : bool
         determines default behavior when block not specified
         in execution methods
 
     Methods
     -------
 
     spin
         flushes incoming results and registration state changes
         control methods spin, and requesting `ids` also ensures up to date
 
     wait
         wait on one or more msg_ids
 
     execution methods
         apply
         legacy: execute, run
 
     data movement
         push, pull, scatter, gather
 
     query methods
         queue_status, get_result, purge, result_status
 
     control methods
         abort, shutdown
 
     """
 
 
     block = Bool(False)
     outstanding = Set()
     results = Instance('collections.defaultdict', (dict,))
     metadata = Instance('collections.defaultdict', (Metadata,))
     history = List()
     debug = Bool(False)
     _spin_thread = Any()
     _stop_spinning = Any()
 
     profile=Unicode()
     def _profile_default(self):
         if BaseIPythonApplication.initialized():
             # an IPython app *might* be running, try to get its profile
             try:
                 return BaseIPythonApplication.instance().profile
             except (AttributeError, MultipleInstanceError):
                 # could be a *different* subclass of config.Application,
                 # which would raise one of these two errors.
                 return u'default'
         else:
             return u'default'
 
 
     _outstanding_dict = Instance('collections.defaultdict', (set,))
     _ids = List()
     _connected=Bool(False)
     _ssh=Bool(False)
     _context = Instance('zmq.Context')
     _config = Dict()
     _engines=Instance(util.ReverseDict, (), {})
     # _hub_socket=Instance('zmq.Socket')
     _query_socket=Instance('zmq.Socket')
     _control_socket=Instance('zmq.Socket')
     _iopub_socket=Instance('zmq.Socket')
     _notification_socket=Instance('zmq.Socket')
     _mux_socket=Instance('zmq.Socket')
     _task_socket=Instance('zmq.Socket')
     _task_scheme=Unicode()
     _closed = False
     _ignored_control_replies=Integer(0)
     _ignored_hub_replies=Integer(0)
 
     def __new__(self, *args, **kw):
         # don't raise on positional args
         return HasTraits.__new__(self, **kw)
 
     def __init__(self, url_file=None, profile=None, profile_dir=None, ipython_dir=None,
             context=None, debug=False,
             sshserver=None, sshkey=None, password=None, paramiko=None,
             timeout=10, **extra_args
             ):
         if profile:
             super(Client, self).__init__(debug=debug, profile=profile)
         else:
             super(Client, self).__init__(debug=debug)
         if context is None:
             context = zmq.Context.instance()
         self._context = context
         self._stop_spinning = Event()
         
         if 'url_or_file' in extra_args:
             url_file = extra_args['url_or_file']
             warnings.warn("url_or_file arg no longer supported, use url_file", DeprecationWarning)
         
         if url_file and util.is_url(url_file):
             raise ValueError("single urls cannot be specified, url-files must be used.")
 
         self._setup_profile_dir(self.profile, profile_dir, ipython_dir)
         
         if self._cd is not None:
             if url_file is None:
                 url_file = pjoin(self._cd.security_dir, 'ipcontroller-client.json')
         if url_file is None:
             raise ValueError(
                 "I can't find enough information to connect to a hub!"
                 " Please specify at least one of url_file or profile."
             )
         
         with open(url_file) as f:
             cfg = json.load(f)
         
         self._task_scheme = cfg['task_scheme']
 
         # sync defaults from args, json:
         if sshserver:
             cfg['ssh'] = sshserver
 
         location = cfg.setdefault('location', None)
         
         proto,addr = cfg['interface'].split('://')
         addr = util.disambiguate_ip_address(addr)
         cfg['interface'] = "%s://%s" % (proto, addr)
         
         # turn interface,port into full urls:
         for key in ('control', 'task', 'mux', 'iopub', 'notification', 'registration'):
             cfg[key] = cfg['interface'] + ':%i' % cfg[key]
         
         url = cfg['registration']
         
         if location is not None and addr == '127.0.0.1':
             # location specified, and connection is expected to be local
             if location not in LOCAL_IPS and not sshserver:
                 # load ssh from JSON *only* if the controller is not on
                 # this machine
                 sshserver=cfg['ssh']
             if location not in LOCAL_IPS and not sshserver:
                 # warn if no ssh specified, but SSH is probably needed
                 # This is only a warning, because the most likely cause
                 # is a local Controller on a laptop whose IP is dynamic
                 warnings.warn("""
             Controller appears to be listening on localhost, but not on this machine.
             If this is true, you should specify Client(...,sshserver='you@%s')
             or instruct your controller to listen on an external IP."""%location,
                 RuntimeWarning)
         elif not sshserver:
             # otherwise sync with cfg
             sshserver = cfg['ssh']
 
         self._config = cfg
 
         self._ssh = bool(sshserver or sshkey or password)
         if self._ssh and sshserver is None:
             # default to ssh via localhost
             sshserver = addr
         if self._ssh and password is None:
             if tunnel.try_passwordless_ssh(sshserver, sshkey, paramiko):
                 password=False
             else:
                 password = getpass("SSH Password for %s: "%sshserver)
         ssh_kwargs = dict(keyfile=sshkey, password=password, paramiko=paramiko)
 
         # configure and construct the session
         extra_args['packer'] = cfg['pack']
         extra_args['unpacker'] = cfg['unpack']
         extra_args['key'] = cfg['exec_key']
         
         self.session = Session(**extra_args)
 
         self._query_socket = self._context.socket(zmq.DEALER)
 
         if self._ssh:
             tunnel.tunnel_connection(self._query_socket, cfg['registration'], sshserver, **ssh_kwargs)
         else:
             self._query_socket.connect(cfg['registration'])
 
         self.session.debug = self.debug
 
         self._notification_handlers = {'registration_notification' : self._register_engine,
                                     'unregistration_notification' : self._unregister_engine,
                                     'shutdown_notification' : lambda msg: self.close(),
                                     }
         self._queue_handlers = {'execute_reply' : self._handle_execute_reply,
                                 'apply_reply' : self._handle_apply_reply}
         self._connect(sshserver, ssh_kwargs, timeout)
         
         # last step: setup magics, if we are in IPython:
         
         try:
             ip = get_ipython()
         except NameError:
             return
         else:
             if 'px' not in ip.magics_manager.magics:
                 # in IPython but we are the first Client.
                 # activate a default view for parallel magics.
                 self.activate()
 
     def __del__(self):
         """cleanup sockets, but _not_ context."""
         self.close()
 
     def _setup_profile_dir(self, profile, profile_dir, ipython_dir):
         if ipython_dir is None:
             ipython_dir = get_ipython_dir()
         if profile_dir is not None:
             try:
                 self._cd = ProfileDir.find_profile_dir(profile_dir)
                 return
             except ProfileDirError:
                 pass
         elif profile is not None:
             try:
                 self._cd = ProfileDir.find_profile_dir_by_name(
                     ipython_dir, profile)
                 return
             except ProfileDirError:
                 pass
         self._cd = None
 
     def _update_engines(self, engines):
         """Update our engines dict and _ids from a dict of the form: {id:uuid}."""
         for k,v in engines.iteritems():
             eid = int(k)
-            self._engines[eid] = v
+            if eid not in self._engines:
             self._ids.append(eid)
+            self._engines[eid] = v
         self._ids = sorted(self._ids)
         if sorted(self._engines.keys()) != range(len(self._engines)) and \
                         self._task_scheme == 'pure' and self._task_socket:
             self._stop_scheduling_tasks()
 
     def _stop_scheduling_tasks(self):
         """Stop scheduling tasks because an engine has been unregistered
         from a pure ZMQ scheduler.
         """
         self._task_socket.close()
         self._task_socket = None
         msg = "An engine has been unregistered, and we are using pure " +\
               "ZMQ task scheduling.  Task farming will be disabled."
         if self.outstanding:
             msg += " If you were running tasks when this happened, " +\
                    "some `outstanding` msg_ids may never resolve."
         warnings.warn(msg, RuntimeWarning)
 
     def _build_targets(self, targets):
         """Turn valid target IDs or 'all' into two lists:
         (int_ids, uuids).
         """
         if not self._ids:
             # flush notification socket if no engines yet, just in case
             if not self.ids:
                 raise error.NoEnginesRegistered("Can't build targets without any engines")
 
         if targets is None:
             targets = self._ids
         elif isinstance(targets, basestring):
             if targets.lower() == 'all':
                 targets = self._ids
             else:
                 raise TypeError("%r not valid str target, must be 'all'"%(targets))
         elif isinstance(targets, int):
             if targets < 0:
                 targets = self.ids[targets]
             if targets not in self._ids:
                 raise IndexError("No such engine: %i"%targets)
             targets = [targets]
 
         if isinstance(targets, slice):
             indices = range(len(self._ids))[targets]
             ids = self.ids
             targets = [ ids[i] for i in indices ]
 
         if not isinstance(targets, (tuple, list, xrange)):
             raise TypeError("targets by int/slice/collection of ints only, not %s"%(type(targets)))
 
         return [cast_bytes(self._engines[t]) for t in targets], list(targets)
 
     def _connect(self, sshserver, ssh_kwargs, timeout):
         """setup all our socket connections to the cluster. This is called from
         __init__."""
 
         # Maybe allow reconnecting?
         if self._connected:
             return
         self._connected=True
 
         def connect_socket(s, url):
             # url = util.disambiguate_url(url, self._config['location'])
             if self._ssh:
                 return tunnel.tunnel_connection(s, url, sshserver, **ssh_kwargs)
             else:
                 return s.connect(url)
 
         self.session.send(self._query_socket, 'connection_request')
         # use Poller because zmq.select has wrong units in pyzmq 2.1.7
         poller = zmq.Poller()
         poller.register(self._query_socket, zmq.POLLIN)
         # poll expects milliseconds, timeout is seconds
         evts = poller.poll(timeout*1000)
         if not evts:
             raise error.TimeoutError("Hub connection request timed out")
         idents,msg = self.session.recv(self._query_socket,mode=0)
         if self.debug:
             pprint(msg)
         content = msg['content']
         # self._config['registration'] = dict(content)
         cfg = self._config
         if content['status'] == 'ok':
             self._mux_socket = self._context.socket(zmq.DEALER)
             connect_socket(self._mux_socket, cfg['mux'])
 
             self._task_socket = self._context.socket(zmq.DEALER)
             connect_socket(self._task_socket, cfg['task'])
 
             self._notification_socket = self._context.socket(zmq.SUB)
             self._notification_socket.setsockopt(zmq.SUBSCRIBE, b'')
             connect_socket(self._notification_socket, cfg['notification'])
 
             self._control_socket = self._context.socket(zmq.DEALER)
             connect_socket(self._control_socket, cfg['control'])
 
             self._iopub_socket = self._context.socket(zmq.SUB)
             self._iopub_socket.setsockopt(zmq.SUBSCRIBE, b'')
             connect_socket(self._iopub_socket, cfg['iopub'])
 
             self._update_engines(dict(content['engines']))
         else:
             self._connected = False
             raise Exception("Failed to connect!")
 
     #--------------------------------------------------------------------------
     # handlers and callbacks for incoming messages
     #--------------------------------------------------------------------------
 
     def _unwrap_exception(self, content):
         """unwrap exception, and remap engine_id to int."""
         e = error.unwrap_exception(content)
         # print e.traceback
         if e.engine_info:
             e_uuid = e.engine_info['engine_uuid']
             eid = self._engines[e_uuid]
             e.engine_info['engine_id'] = eid
         return e
 
     def _extract_metadata(self, header, parent, content):
         md = {'msg_id' : parent['msg_id'],
               'received' : datetime.now(),
               'engine_uuid' : header.get('engine', None),
               'follow' : parent.get('follow', []),
               'after' : parent.get('after', []),
               'status' : content['status'],
             }
 
         if md['engine_uuid'] is not None:
             md['engine_id'] = self._engines.get(md['engine_uuid'], None)
 
         if 'date' in parent:
             md['submitted'] = parent['date']
         if 'started' in header:
             md['started'] = header['started']
         if 'date' in header:
             md['completed'] = header['date']
         return md
 
     def _register_engine(self, msg):
         """Register a new engine, and update our connection info."""
         content = msg['content']
         eid = content['id']
-        d = {eid : content['queue']}
+        d = {eid : content['uuid']}
         self._update_engines(d)
 
     def _unregister_engine(self, msg):
         """Unregister an engine that has died."""
         content = msg['content']
         eid = int(content['id'])
         if eid in self._ids:
             self._ids.remove(eid)
             uuid = self._engines.pop(eid)
 
             self._handle_stranded_msgs(eid, uuid)
 
         if self._task_socket and self._task_scheme == 'pure':
             self._stop_scheduling_tasks()
 
     def _handle_stranded_msgs(self, eid, uuid):
         """Handle messages known to be on an engine when the engine unregisters.
 
         It is possible that this will fire prematurely - that is, an engine will
         go down after completing a result, and the client will be notified
         of the unregistration and later receive the successful result.
         """
 
         outstanding = self._outstanding_dict[uuid]
 
         for msg_id in list(outstanding):
             if msg_id in self.results:
                 # we already
                 continue
             try:
                 raise error.EngineError("Engine %r died while running task %r"%(eid, msg_id))
             except:
                 content = error.wrap_exception()
             # build a fake message:
             parent = {}
             header = {}
             parent['msg_id'] = msg_id
             header['engine'] = uuid
             header['date'] = datetime.now()
             msg = dict(parent_header=parent, header=header, content=content)
             self._handle_apply_reply(msg)
 
     def _handle_execute_reply(self, msg):
         """Save the reply to an execute_request into our results.
 
         execute messages are never actually used. apply is used instead.
         """
 
         parent = msg['parent_header']
         msg_id = parent['msg_id']
         if msg_id not in self.outstanding:
             if msg_id in self.history:
                 print ("got stale result: %s"%msg_id)
             else:
                 print ("got unknown result: %s"%msg_id)
         else:
             self.outstanding.remove(msg_id)
 
         content = msg['content']
         header = msg['header']
 
         # construct metadata:
         md = self.metadata[msg_id]
         md.update(self._extract_metadata(header, parent, content))
         # is this redundant?
         self.metadata[msg_id] = md
         
         e_outstanding = self._outstanding_dict[md['engine_uuid']]
         if msg_id in e_outstanding:
             e_outstanding.remove(msg_id)
 
         # construct result:
         if content['status'] == 'ok':
             self.results[msg_id] = ExecuteReply(msg_id, content, md)
         elif content['status'] == 'aborted':
             self.results[msg_id] = error.TaskAborted(msg_id)
         elif content['status'] == 'resubmitted':
             # TODO: handle resubmission
             pass
         else:
             self.results[msg_id] = self._unwrap_exception(content)
 
     def _handle_apply_reply(self, msg):
         """Save the reply to an apply_request into our results."""
         parent = msg['parent_header']
         msg_id = parent['msg_id']
         if msg_id not in self.outstanding:
             if msg_id in self.history:
                 print ("got stale result: %s"%msg_id)
                 print self.results[msg_id]
                 print msg
             else:
                 print ("got unknown result: %s"%msg_id)
         else:
             self.outstanding.remove(msg_id)
         content = msg['content']
         header = msg['header']
 
         # construct metadata:
         md = self.metadata[msg_id]
         md.update(self._extract_metadata(header, parent, content))
         # is this redundant?
         self.metadata[msg_id] = md
 
         e_outstanding = self._outstanding_dict[md['engine_uuid']]
         if msg_id in e_outstanding:
             e_outstanding.remove(msg_id)
 
         # construct result:
         if content['status'] == 'ok':
             self.results[msg_id] = util.unserialize_object(msg['buffers'])[0]
         elif content['status'] == 'aborted':
             self.results[msg_id] = error.TaskAborted(msg_id)
         elif content['status'] == 'resubmitted':
             # TODO: handle resubmission
             pass
         else:
             self.results[msg_id] = self._unwrap_exception(content)
 
     def _flush_notifications(self):
         """Flush notifications of engine registrations waiting
         in ZMQ queue."""
         idents,msg = self.session.recv(self._notification_socket, mode=zmq.NOBLOCK)
         while msg is not None:
             if self.debug:
                 pprint(msg)
             msg_type = msg['header']['msg_type']
             handler = self._notification_handlers.get(msg_type, None)
             if handler is None:
                 raise Exception("Unhandled message type: %s"%msg.msg_type)
             else:
                 handler(msg)
             idents,msg = self.session.recv(self._notification_socket, mode=zmq.NOBLOCK)
 
     def _flush_results(self, sock):
         """Flush task or queue results waiting in ZMQ queue."""
         idents,msg = self.session.recv(sock, mode=zmq.NOBLOCK)
         while msg is not None:
             if self.debug:
                 pprint(msg)
             msg_type = msg['header']['msg_type']
             handler = self._queue_handlers.get(msg_type, None)
             if handler is None:
                 raise Exception("Unhandled message type: %s"%msg.msg_type)
             else:
                 handler(msg)
             idents,msg = self.session.recv(sock, mode=zmq.NOBLOCK)
 
     def _flush_control(self, sock):
         """Flush replies from the control channel waiting
         in the ZMQ queue.
 
         Currently: ignore them."""
         if self._ignored_control_replies <= 0:
             return
         idents,msg = self.session.recv(sock, mode=zmq.NOBLOCK)
         while msg is not None:
             self._ignored_control_replies -= 1
             if self.debug:
                 pprint(msg)
             idents,msg = self.session.recv(sock, mode=zmq.NOBLOCK)
 
     def _flush_ignored_control(self):
         """flush ignored control replies"""
         while self._ignored_control_replies > 0:
             self.session.recv(self._control_socket)
             self._ignored_control_replies -= 1
 
     def _flush_ignored_hub_replies(self):
         ident,msg = self.session.recv(self._query_socket, mode=zmq.NOBLOCK)
         while msg is not None:
             ident,msg = self.session.recv(self._query_socket, mode=zmq.NOBLOCK)
 
     def _flush_iopub(self, sock):
         """Flush replies from the iopub channel waiting
         in the ZMQ queue.
         """
         idents,msg = self.session.recv(sock, mode=zmq.NOBLOCK)
         while msg is not None:
             if self.debug:
                 pprint(msg)
             parent = msg['parent_header']
             # ignore IOPub messages with no parent.
             # Caused by print statements or warnings from before the first execution.
             if not parent:
                 continue
             msg_id = parent['msg_id']
             content = msg['content']
             header = msg['header']
             msg_type = msg['header']['msg_type']
 
             # init metadata:
             md = self.metadata[msg_id]
 
             if msg_type == 'stream':
                 name = content['name']
                 s = md[name] or ''
                 md[name] = s + content['data']
             elif msg_type == 'pyerr':
                 md.update({'pyerr' : self._unwrap_exception(content)})
             elif msg_type == 'pyin':
                 md.update({'pyin' : content['code']})
             elif msg_type == 'display_data':
                 md['outputs'].append(content)
             elif msg_type == 'pyout':
                 md['pyout'] = content
             elif msg_type == 'status':
                 # idle message comes after all outputs
                 if content['execution_state'] == 'idle':
                     md['outputs_ready'] = True
             else:
                 # unhandled msg_type (status, etc.)
                 pass
 
             # reduntant?
             self.metadata[msg_id] = md
 
             idents,msg = self.session.recv(sock, mode=zmq.NOBLOCK)
 
     #--------------------------------------------------------------------------
     # len, getitem
     #--------------------------------------------------------------------------
 
     def __len__(self):
         """len(client) returns # of engines."""
         return len(self.ids)
 
     def __getitem__(self, key):
         """index access returns DirectView multiplexer objects
 
         Must be int, slice, or list/tuple/xrange of ints"""
         if not isinstance(key, (int, slice, tuple, list, xrange)):
             raise TypeError("key by int/slice/iterable of ints only, not %s"%(type(key)))
         else:
             return self.direct_view(key)
 
     #--------------------------------------------------------------------------
     # Begin public methods
     #--------------------------------------------------------------------------
 
     @property
     def ids(self):
         """Always up-to-date ids property."""
         self._flush_notifications()
         # always copy:
         return list(self._ids)
 
     def activate(self, targets='all', suffix=''):
         """Create a DirectView and register it with IPython magics
         
         Defines the magics `%px, %autopx, %pxresult, %%px`
         
         Parameters
         ----------
         
         targets: int, list of ints, or 'all'
             The engines on which the view's magics will run
         suffix: str [default: '']
             The suffix, if any, for the magics.  This allows you to have
             multiple views associated with parallel magics at the same time.
             
             e.g. ``rc.activate(targets=0, suffix='0')`` will give you
             the magics ``%px0``, ``%pxresult0``, etc. for running magics just
             on engine 0.
         """
         view = self.direct_view(targets)
         view.block = True
         view.activate(suffix)
         return view
 
     def close(self):
         if self._closed:
             return
         self.stop_spin_thread()
         snames = filter(lambda n: n.endswith('socket'), dir(self))
         for socket in map(lambda name: getattr(self, name), snames):
             if isinstance(socket, zmq.Socket) and not socket.closed:
                 socket.close()
         self._closed = True
 
     def _spin_every(self, interval=1):
         """target func for use in spin_thread"""
         while True:
             if self._stop_spinning.is_set():
                 return
             time.sleep(interval)
             self.spin()
 
     def spin_thread(self, interval=1):
         """call Client.spin() in a background thread on some regular interval
         
         This helps ensure that messages don't pile up too much in the zmq queue
         while you are working on other things, or just leaving an idle terminal.
         
         It also helps limit potential padding of the `received` timestamp
         on AsyncResult objects, used for timings.
         
         Parameters
         ----------
         
         interval : float, optional
             The interval on which to spin the client in the background thread
             (simply passed to time.sleep).
         
         Notes
         -----
         
         For precision timing, you may want to use this method to put a bound
         on the jitter (in seconds) in `received` timestamps used
         in AsyncResult.wall_time.
         
         """
         if self._spin_thread is not None:
             self.stop_spin_thread()
         self._stop_spinning.clear()
         self._spin_thread = Thread(target=self._spin_every, args=(interval,))
         self._spin_thread.daemon = True
         self._spin_thread.start()
     
     def stop_spin_thread(self):
         """stop background spin_thread, if any"""
         if self._spin_thread is not None:
             self._stop_spinning.set()
             self._spin_thread.join()
             self._spin_thread = None
 
     def spin(self):
         """Flush any registration notifications and execution results
         waiting in the ZMQ queue.
         """
         if self._notification_socket:
             self._flush_notifications()
         if self._iopub_socket:
             self._flush_iopub(self._iopub_socket)
         if self._mux_socket:
             self._flush_results(self._mux_socket)
         if self._task_socket:
             self._flush_results(self._task_socket)
         if self._control_socket:
             self._flush_control(self._control_socket)
         if self._query_socket:
             self._flush_ignored_hub_replies()
 
     def wait(self, jobs=None, timeout=-1):
         """waits on one or more `jobs`, for up to `timeout` seconds.
 
         Parameters
         ----------
 
         jobs : int, str, or list of ints and/or strs, or one or more AsyncResult objects
                 ints are indices to self.history
                 strs are msg_ids
                 default: wait on all outstanding messages
         timeout : float
                 a time in seconds, after which to give up.
                 default is -1, which means no timeout
 
         Returns
         -------
 
         True : when all msg_ids are done
         False : timeout reached, some msg_ids still outstanding
         """
         tic = time.time()
         if jobs is None:
             theids = self.outstanding
         else:
             if isinstance(jobs, (int, basestring, AsyncResult)):
                 jobs = [jobs]
             theids = set()
             for job in jobs:
                 if isinstance(job, int):
                     # index access
                     job = self.history[job]
                 elif isinstance(job, AsyncResult):
                     map(theids.add, job.msg_ids)
                     continue
                 theids.add(job)
         if not theids.intersection(self.outstanding):
             return True
         self.spin()
         while theids.intersection(self.outstanding):
             if timeout >= 0 and ( time.time()-tic ) > timeout:
                 break
             time.sleep(1e-3)
             self.spin()
         return len(theids.intersection(self.outstanding)) == 0
 
     #--------------------------------------------------------------------------
     # Control methods
     #--------------------------------------------------------------------------
 
     @spin_first
     def clear(self, targets=None, block=None):
         """Clear the namespace in target(s)."""
         block = self.block if block is None else block
         targets = self._build_targets(targets)[0]
         for t in targets:
             self.session.send(self._control_socket, 'clear_request', content={}, ident=t)
         error = False
         if block:
             self._flush_ignored_control()
             for i in range(len(targets)):
                 idents,msg = self.session.recv(self._control_socket,0)
                 if self.debug:
                     pprint(msg)
                 if msg['content']['status'] != 'ok':
                     error = self._unwrap_exception(msg['content'])
         else:
             self._ignored_control_replies += len(targets)
         if error:
             raise error
 
 
     @spin_first
     def abort(self, jobs=None, targets=None, block=None):
         """Abort specific jobs from the execution queues of target(s).
 
         This is a mechanism to prevent jobs that have already been submitted
         from executing.
 
         Parameters
         ----------
 
         jobs : msg_id, list of msg_ids, or AsyncResult
             The jobs to be aborted
             
             If unspecified/None: abort all outstanding jobs.
 
         """
         block = self.block if block is None else block
         jobs = jobs if jobs is not None else list(self.outstanding)
         targets = self._build_targets(targets)[0]
         
         msg_ids = []
         if isinstance(jobs, (basestring,AsyncResult)):
             jobs = [jobs]
         bad_ids = filter(lambda obj: not isinstance(obj, (basestring, AsyncResult)), jobs)
         if bad_ids:
             raise TypeError("Invalid msg_id type %r, expected str or AsyncResult"%bad_ids[0])
         for j in jobs:
             if isinstance(j, AsyncResult):
                 msg_ids.extend(j.msg_ids)
             else:
                 msg_ids.append(j)
         content = dict(msg_ids=msg_ids)
         for t in targets:
             self.session.send(self._control_socket, 'abort_request',
                     content=content, ident=t)
         error = False
         if block:
             self._flush_ignored_control()
             for i in range(len(targets)):
                 idents,msg = self.session.recv(self._control_socket,0)
                 if self.debug:
                     pprint(msg)
                 if msg['content']['status'] != 'ok':
                     error = self._unwrap_exception(msg['content'])
         else:
             self._ignored_control_replies += len(targets)
         if error:
             raise error
 
     @spin_first
     def shutdown(self, targets='all', restart=False, hub=False, block=None):
         """Terminates one or more engine processes, optionally including the hub.
         
         Parameters
         ----------
         
         targets: list of ints or 'all' [default: all]
             Which engines to shutdown.
         hub: bool [default: False]
             Whether to include the Hub.  hub=True implies targets='all'.
         block: bool [default: self.block]
             Whether to wait for clean shutdown replies or not.
         restart: bool [default: False]
             NOT IMPLEMENTED
             whether to restart engines after shutting them down.
         """
         
         if restart:
             raise NotImplementedError("Engine restart is not yet implemented")
         
         block = self.block if block is None else block
         if hub:
             targets = 'all'
         targets = self._build_targets(targets)[0]
         for t in targets:
             self.session.send(self._control_socket, 'shutdown_request',
                         content={'restart':restart},ident=t)
         error = False
         if block or hub:
             self._flush_ignored_control()
             for i in range(len(targets)):
                 idents,msg = self.session.recv(self._control_socket, 0)
                 if self.debug:
                     pprint(msg)
                 if msg['content']['status'] != 'ok':
                     error = self._unwrap_exception(msg['content'])
         else:
             self._ignored_control_replies += len(targets)
 
         if hub:
             time.sleep(0.25)
             self.session.send(self._query_socket, 'shutdown_request')
             idents,msg = self.session.recv(self._query_socket, 0)
             if self.debug:
                 pprint(msg)
             if msg['content']['status'] != 'ok':
                 error = self._unwrap_exception(msg['content'])
 
         if error:
             raise error
 
     #--------------------------------------------------------------------------
     # Execution related methods
     #--------------------------------------------------------------------------
 
     def _maybe_raise(self, result):
         """wrapper for maybe raising an exception if apply failed."""
         if isinstance(result, error.RemoteError):
             raise result
 
         return result
 
     def send_apply_request(self, socket, f, args=None, kwargs=None, subheader=None, track=False,
                             ident=None):
         """construct and send an apply message via a socket.
 
         This is the principal method with which all engine execution is performed by views.
         """
 
         if self._closed:
             raise RuntimeError("Client cannot be used after its sockets have been closed")
         
         # defaults:
         args = args if args is not None else []
         kwargs = kwargs if kwargs is not None else {}
         subheader = subheader if subheader is not None else {}
 
         # validate arguments
         if not callable(f) and not isinstance(f, Reference):
             raise TypeError("f must be callable, not %s"%type(f))
         if not isinstance(args, (tuple, list)):
             raise TypeError("args must be tuple or list, not %s"%type(args))
         if not isinstance(kwargs, dict):
             raise TypeError("kwargs must be dict, not %s"%type(kwargs))
         if not isinstance(subheader, dict):
             raise TypeError("subheader must be dict, not %s"%type(subheader))
 
         bufs = util.pack_apply_message(f,args,kwargs)
 
         msg = self.session.send(socket, "apply_request", buffers=bufs, ident=ident,
                             subheader=subheader, track=track)
 
         msg_id = msg['header']['msg_id']
         self.outstanding.add(msg_id)
         if ident:
             # possibly routed to a specific engine
             if isinstance(ident, list):
                 ident = ident[-1]
             if ident in self._engines.values():
                 # save for later, in case of engine death
                 self._outstanding_dict[ident].add(msg_id)
         self.history.append(msg_id)
         self.metadata[msg_id]['submitted'] = datetime.now()
 
         return msg
 
     def send_execute_request(self, socket, code, silent=True, subheader=None, ident=None):
         """construct and send an execute request via a socket.
 
         """
 
         if self._closed:
             raise RuntimeError("Client cannot be used after its sockets have been closed")
         
         # defaults:
         subheader = subheader if subheader is not None else {}
 
         # validate arguments
         if not isinstance(code, basestring):
             raise TypeError("code must be text, not %s" % type(code))
         if not isinstance(subheader, dict):
             raise TypeError("subheader must be dict, not %s" % type(subheader))
         
         content = dict(code=code, silent=bool(silent), user_variables=[], user_expressions={})
 
 
         msg = self.session.send(socket, "execute_request", content=content, ident=ident,
                             subheader=subheader)
 
         msg_id = msg['header']['msg_id']
         self.outstanding.add(msg_id)
         if ident:
             # possibly routed to a specific engine
             if isinstance(ident, list):
                 ident = ident[-1]
             if ident in self._engines.values():
                 # save for later, in case of engine death
                 self._outstanding_dict[ident].add(msg_id)
         self.history.append(msg_id)
         self.metadata[msg_id]['submitted'] = datetime.now()
 
         return msg
 
     #--------------------------------------------------------------------------
     # construct a View object
     #--------------------------------------------------------------------------
 
     def load_balanced_view(self, targets=None):
         """construct a DirectView object.
 
         If no arguments are specified, create a LoadBalancedView
         using all engines.
 
         Parameters
         ----------
 
         targets: list,slice,int,etc. [default: use all engines]
             The subset of engines across which to load-balance
         """
         if targets == 'all':
             targets = None
         if targets is not None:
             targets = self._build_targets(targets)[1]
         return LoadBalancedView(client=self, socket=self._task_socket, targets=targets)
 
     def direct_view(self, targets='all'):
         """construct a DirectView object.
 
         If no targets are specified, create a DirectView using all engines.
         
         rc.direct_view('all') is distinguished from rc[:] in that 'all' will
         evaluate the target engines at each execution, whereas rc[:] will connect to
         all *current* engines, and that list will not change.
         
         That is, 'all' will always use all engines, whereas rc[:] will not use
         engines added after the DirectView is constructed.
 
         Parameters
         ----------
 
         targets: list,slice,int,etc. [default: use all engines]
             The engines to use for the View
         """
         single = isinstance(targets, int)
         # allow 'all' to be lazily evaluated at each execution
         if targets != 'all':
             targets = self._build_targets(targets)[1]
         if single:
             targets = targets[0]
         return DirectView(client=self, socket=self._mux_socket, targets=targets)
 
     #--------------------------------------------------------------------------
     # Query methods
     #--------------------------------------------------------------------------
 
     @spin_first
     def get_result(self, indices_or_msg_ids=None, block=None):
         """Retrieve a result by msg_id or history index, wrapped in an AsyncResult object.
 
         If the client already has the results, no request to the Hub will be made.
 
         This is a convenient way to construct AsyncResult objects, which are wrappers
         that include metadata about execution, and allow for awaiting results that
         were not submitted by this Client.
 
         It can also be a convenient way to retrieve the metadata associated with
         blocking execution, since it always retrieves
 
         Examples
         --------
         ::
 
             In [10]: r = client.apply()
 
         Parameters
         ----------
 
         indices_or_msg_ids : integer history index, str msg_id, or list of either
             The indices or msg_ids of indices to be retrieved
 
         block : bool
             Whether to wait for the result to be done
 
         Returns
         -------
 
         AsyncResult
             A single AsyncResult object will always be returned.
 
         AsyncHubResult
             A subclass of AsyncResult that retrieves results from the Hub
 
         """
         block = self.block if block is None else block
         if indices_or_msg_ids is None:
             indices_or_msg_ids = -1
 
         if not isinstance(indices_or_msg_ids, (list,tuple)):
             indices_or_msg_ids = [indices_or_msg_ids]
 
         theids = []
         for id in indices_or_msg_ids:
             if isinstance(id, int):
                 id = self.history[id]
             if not isinstance(id, basestring):
                 raise TypeError("indices must be str or int, not %r"%id)
             theids.append(id)
 
         local_ids = filter(lambda msg_id: msg_id in self.history or msg_id in self.results, theids)
         remote_ids = filter(lambda msg_id: msg_id not in local_ids, theids)
 
         if remote_ids:
             ar = AsyncHubResult(self, msg_ids=theids)
         else:
             ar = AsyncResult(self, msg_ids=theids)
 
         if block:
             ar.wait()
 
         return ar
 
     @spin_first
     def resubmit(self, indices_or_msg_ids=None, subheader=None, block=None):
         """Resubmit one or more tasks.
 
         in-flight tasks may not be resubmitted.
 
         Parameters
         ----------
 
         indices_or_msg_ids : integer history index, str msg_id, or list of either
             The indices or msg_ids of indices to be retrieved
 
         block : bool
             Whether to wait for the result to be done
 
         Returns
         -------
 
         AsyncHubResult
             A subclass of AsyncResult that retrieves results from the Hub
 
         """
         block = self.block if block is None else block
         if indices_or_msg_ids is None:
             indices_or_msg_ids = -1
 
         if not isinstance(indices_or_msg_ids, (list,tuple)):
             indices_or_msg_ids = [indices_or_msg_ids]
 
         theids = []
         for id in indices_or_msg_ids:
             if isinstance(id, int):
                 id = self.history[id]
             if not isinstance(id, basestring):
                 raise TypeError("indices must be str or int, not %r"%id)
             theids.append(id)
 
         content = dict(msg_ids = theids)
 
         self.session.send(self._query_socket, 'resubmit_request', content)
 
         zmq.select([self._query_socket], [], [])
         idents,msg = self.session.recv(self._query_socket, zmq.NOBLOCK)
         if self.debug:
             pprint(msg)
         content = msg['content']
         if content['status'] != 'ok':
             raise self._unwrap_exception(content)
         mapping = content['resubmitted']
         new_ids = [ mapping[msg_id] for msg_id in theids ]
 
         ar = AsyncHubResult(self, msg_ids=new_ids)
 
         if block:
             ar.wait()
 
         return ar
 
     @spin_first
     def result_status(self, msg_ids, status_only=True):
         """Check on the status of the result(s) of the apply request with `msg_ids`.
 
         If status_only is False, then the actual results will be retrieved, else
         only the status of the results will be checked.
 
         Parameters
         ----------
 
         msg_ids : list of msg_ids
             if int:
                 Passed as index to self.history for convenience.
         status_only : bool (default: True)
             if False:
                 Retrieve the actual results of completed tasks.
 
         Returns
         -------
 
         results : dict
             There will always be the keys 'pending' and 'completed', which will
             be lists of msg_ids that are incomplete or complete. If `status_only`
             is False, then completed results will be keyed by their `msg_id`.
         """
         if not isinstance(msg_ids, (list,tuple)):
             msg_ids = [msg_ids]
 
         theids = []
         for msg_id in msg_ids:
             if isinstance(msg_id, int):
                 msg_id = self.history[msg_id]
             if not isinstance(msg_id, basestring):
                 raise TypeError("msg_ids must be str, not %r"%msg_id)
             theids.append(msg_id)
 
         completed = []
         local_results = {}
 
         # comment this block out to temporarily disable local shortcut:
         for msg_id in theids:
             if msg_id in self.results:
                 completed.append(msg_id)
                 local_results[msg_id] = self.results[msg_id]
                 theids.remove(msg_id)
 
         if theids: # some not locally cached
             content = dict(msg_ids=theids, status_only=status_only)
             msg = self.session.send(self._query_socket, "result_request", content=content)
             zmq.select([self._query_socket], [], [])
             idents,msg = self.session.recv(self._query_socket, zmq.NOBLOCK)
             if self.debug:
                 pprint(msg)
             content = msg['content']
             if content['status'] != 'ok':
                 raise self._unwrap_exception(content)
             buffers = msg['buffers']
         else:
             content = dict(completed=[],pending=[])
 
         content['completed'].extend(completed)
 
         if status_only:
             return content
 
         failures = []
         # load cached results into result:
         content.update(local_results)
 
         # update cache with results:
         for msg_id in sorted(theids):
             if msg_id in content['completed']:
                 rec = content[msg_id]
                 parent = rec['header']
                 header = rec['result_header']
                 rcontent = rec['result_content']
                 iodict = rec['io']
                 if isinstance(rcontent, str):
                     rcontent = self.session.unpack(rcontent)
 
                 md = self.metadata[msg_id]
                 md.update(self._extract_metadata(header, parent, rcontent))
                 if rec.get('received'):
                     md['received'] = rec['received']
                 md.update(iodict)
                 
                 if rcontent['status'] == 'ok':
                     if header['msg_type'] == 'apply_reply':
                         res,buffers = util.unserialize_object(buffers)
                     elif header['msg_type'] == 'execute_reply':
                         res = ExecuteReply(msg_id, rcontent, md)
                     else:
                         raise KeyError("unhandled msg type: %r" % header[msg_type])
                 else:
                     res = self._unwrap_exception(rcontent)
                     failures.append(res)
 
                 self.results[msg_id] = res
                 content[msg_id] = res
 
         if len(theids) == 1 and failures:
             raise failures[0]
 
         error.collect_exceptions(failures, "result_status")
         return content
 
     @spin_first
     def queue_status(self, targets='all', verbose=False):
         """Fetch the status of engine queues.
 
         Parameters
         ----------
 
         targets : int/str/list of ints/strs
                 the engines whose states are to be queried.
                 default : all
         verbose : bool
                 Whether to return lengths only, or lists of ids for each element
         """
         if targets == 'all':
             # allow 'all' to be evaluated on the engine
             engine_ids = None
         else:
             engine_ids = self._build_targets(targets)[1]
         content = dict(targets=engine_ids, verbose=verbose)
         self.session.send(self._query_socket, "queue_request", content=content)
         idents,msg = self.session.recv(self._query_socket, 0)
         if self.debug:
             pprint(msg)
         content = msg['content']
         status = content.pop('status')
         if status != 'ok':
             raise self._unwrap_exception(content)
         content = rekey(content)
         if isinstance(targets, int):
             return content[targets]
         else:
             return content
 
     @spin_first
     def purge_results(self, jobs=[], targets=[]):
         """Tell the Hub to forget results.
 
         Individual results can be purged by msg_id, or the entire
         history of specific targets can be purged.
 
         Use `purge_results('all')` to scrub everything from the Hub's db.
 
         Parameters
         ----------
 
         jobs : str or list of str or AsyncResult objects
                 the msg_ids whose results should be forgotten.
         targets : int/str/list of ints/strs
                 The targets, by int_id, whose entire history is to be purged.
 
                 default : None
         """
         if not targets and not jobs:
             raise ValueError("Must specify at least one of `targets` and `jobs`")
         if targets:
             targets = self._build_targets(targets)[1]
 
         # construct msg_ids from jobs
         if jobs == 'all':
             msg_ids = jobs
         else:
             msg_ids = []
             if isinstance(jobs, (basestring,AsyncResult)):
                 jobs = [jobs]
             bad_ids = filter(lambda obj: not isinstance(obj, (basestring, AsyncResult)), jobs)
             if bad_ids:
                 raise TypeError("Invalid msg_id type %r, expected str or AsyncResult"%bad_ids[0])
             for j in jobs:
                 if isinstance(j, AsyncResult):
                     msg_ids.extend(j.msg_ids)
                 else:
                     msg_ids.append(j)
 
         content = dict(engine_ids=targets, msg_ids=msg_ids)
         self.session.send(self._query_socket, "purge_request", content=content)
         idents, msg = self.session.recv(self._query_socket, 0)
         if self.debug:
             pprint(msg)
         content = msg['content']
         if content['status'] != 'ok':
             raise self._unwrap_exception(content)
 
     @spin_first
     def hub_history(self):
         """Get the Hub's history
 
         Just like the Client, the Hub has a history, which is a list of msg_ids.
         This will contain the history of all clients, and, depending on configuration,
         may contain history across multiple cluster sessions.
 
         Any msg_id returned here is a valid argument to `get_result`.
 
         Returns
         -------
 
         msg_ids : list of strs
                 list of all msg_ids, ordered by task submission time.
         """
 
         self.session.send(self._query_socket, "history_request", content={})
         idents, msg = self.session.recv(self._query_socket, 0)
 
         if self.debug:
             pprint(msg)
         content = msg['content']
         if content['status'] != 'ok':
             raise self._unwrap_exception(content)
         else:
             return content['history']
 
     @spin_first
     def db_query(self, query, keys=None):
         """Query the Hub's TaskRecord database
 
         This will return a list of task record dicts that match `query`
 
         Parameters
         ----------
 
         query : mongodb query dict
             The search dict. See mongodb query docs for details.
         keys : list of strs [optional]
             The subset of keys to be returned.  The default is to fetch everything but buffers.
             'msg_id' will *always* be included.
         """
         if isinstance(keys, basestring):
             keys = [keys]
         content = dict(query=query, keys=keys)
         self.session.send(self._query_socket, "db_request", content=content)
         idents, msg = self.session.recv(self._query_socket, 0)
         if self.debug:
             pprint(msg)
         content = msg['content']
         if content['status'] != 'ok':
             raise self._unwrap_exception(content)
 
         records = content['records']
 
         buffer_lens = content['buffer_lens']
         result_buffer_lens = content['result_buffer_lens']
         buffers = msg['buffers']
         has_bufs = buffer_lens is not None
         has_rbufs = result_buffer_lens is not None
         for i,rec in enumerate(records):
             # relink buffers
             if has_bufs:
                 blen = buffer_lens[i]
                 rec['buffers'], buffers = buffers[:blen],buffers[blen:]
             if has_rbufs:
                 blen = result_buffer_lens[i]
                 rec['result_buffers'], buffers = buffers[:blen],buffers[blen:]
 
         return records
 
 __all__ = [ 'Client' ]

IPython/parallel/controller/hub.py

@@ -1,1341 +1,1401 b''
 """The IPython Controller Hub with 0MQ
 This is the master object that handles connections from engines and clients,
 and monitors traffic through the various queues.
 
 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.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 from __future__ import print_function
 
+import json
+import os
 import sys
 import time
 from datetime import datetime
 
 import zmq
 from zmq.eventloop import ioloop
 from zmq.eventloop.zmqstream import ZMQStream
 
 # internal:
 from IPython.utils.importstring import import_item
 from IPython.utils.py3compat import cast_bytes
 from IPython.utils.traitlets import (
         HasTraits, Instance, Integer, Unicode, Dict, Set, Tuple, CBytes, DottedObjectName
         )
 
 from IPython.parallel import error, util
 from IPython.parallel.factory import RegistrationFactory
 
 from IPython.zmq.session import SessionFactory
 
 from .heartmonitor import HeartMonitor
 
 #-----------------------------------------------------------------------------
 # Code
 #-----------------------------------------------------------------------------
 
 def _passer(*args, **kwargs):
     return
 
 def _printer(*args, **kwargs):
     print (args)
     print (kwargs)
 
 def empty_record():
     """Return an empty dict with all record keys."""
     return {
         'msg_id' : None,
         'header' : None,
         'content': None,
         'buffers': None,
         'submitted': None,
         'client_uuid' : None,
         'engine_uuid' : None,
         'started': None,
         'completed': None,
         'resubmitted': None,
         'received': None,
         'result_header' : None,
         'result_content' : None,
         'result_buffers' : None,
         'queue' : None,
         'pyin' : None,
         'pyout': None,
         'pyerr': None,
         'stdout': '',
         'stderr': '',
     }
 
 def init_record(msg):
     """Initialize a TaskRecord based on a request."""
     header = msg['header']
     return {
         'msg_id' : header['msg_id'],
         'header' : header,
         'content': msg['content'],
         'buffers': msg['buffers'],
         'submitted': header['date'],
         'client_uuid' : None,
         'engine_uuid' : None,
         'started': None,
         'completed': None,
         'resubmitted': None,
         'received': None,
         'result_header' : None,
         'result_content' : None,
         'result_buffers' : None,
         'queue' : None,
         'pyin' : None,
         'pyout': None,
         'pyerr': None,
         'stdout': '',
         'stderr': '',
     }
 
 
 class EngineConnector(HasTraits):
     """A simple object for accessing the various zmq connections of an object.
     Attributes are:
     id (int): engine ID
-    uuid (str): uuid (unused?)
-    queue (str): identity of queue's DEALER socket
-    registration (str): identity of registration DEALER socket
-    heartbeat (str): identity of heartbeat DEALER socket
+    uuid (unicode): engine UUID
+    pending: set of msg_ids
+    stallback: DelayedCallback for stalled registration
     """
+    
     id=Integer(0)
-    queue=CBytes()
-    control=CBytes()
-    registration=CBytes()
-    heartbeat=CBytes()
+    uuid = Unicode()
     pending=Set()
+    stallback = Instance(ioloop.DelayedCallback)
+
 
 _db_shortcuts = {
     'sqlitedb' : 'IPython.parallel.controller.sqlitedb.SQLiteDB',
     'mongodb'  : 'IPython.parallel.controller.mongodb.MongoDB',
     'dictdb'   : 'IPython.parallel.controller.dictdb.DictDB',
     'nodb'     : 'IPython.parallel.controller.dictdb.NoDB',
 }
 
 class HubFactory(RegistrationFactory):
     """The Configurable for setting up a Hub."""
 
     # port-pairs for monitoredqueues:
     hb = Tuple(Integer,Integer,config=True,
         help="""PUB/ROUTER Port pair for Engine heartbeats""")
     def _hb_default(self):
         return tuple(util.select_random_ports(2))
 
     mux = Tuple(Integer,Integer,config=True,
         help="""Client/Engine Port pair for MUX queue""")
 
     def _mux_default(self):
         return tuple(util.select_random_ports(2))
 
     task = Tuple(Integer,Integer,config=True,
         help="""Client/Engine Port pair for Task queue""")
     def _task_default(self):
         return tuple(util.select_random_ports(2))
 
     control = Tuple(Integer,Integer,config=True,
         help="""Client/Engine Port pair for Control queue""")
 
     def _control_default(self):
         return tuple(util.select_random_ports(2))
 
     iopub = Tuple(Integer,Integer,config=True,
         help="""Client/Engine Port pair for IOPub relay""")
 
     def _iopub_default(self):
         return tuple(util.select_random_ports(2))
 
     # single ports:
     mon_port = Integer(config=True,
         help="""Monitor (SUB) port for queue traffic""")
 
     def _mon_port_default(self):
         return util.select_random_ports(1)[0]
 
     notifier_port = Integer(config=True,
         help="""PUB port for sending engine status notifications""")
 
     def _notifier_port_default(self):
         return util.select_random_ports(1)[0]
 
     engine_ip = Unicode('127.0.0.1', config=True,
         help="IP on which to listen for engine connections. [default: loopback]")
     engine_transport = Unicode('tcp', config=True,
         help="0MQ transport for engine connections. [default: tcp]")
 
     client_ip = Unicode('127.0.0.1', config=True,
         help="IP on which to listen for client connections. [default: loopback]")
     client_transport = Unicode('tcp', config=True,
         help="0MQ transport for client connections. [default : tcp]")
 
     monitor_ip = Unicode('127.0.0.1', config=True,
         help="IP on which to listen for monitor messages. [default: loopback]")
     monitor_transport = Unicode('tcp', config=True,
         help="0MQ transport for monitor messages. [default : tcp]")
 
     monitor_url = Unicode('')
 
     db_class = DottedObjectName('NoDB',
         config=True, help="""The class to use for the DB backend
         
         Options include:
         
         SQLiteDB: SQLite
         MongoDB : use MongoDB
         DictDB  : in-memory storage (fastest, but be mindful of memory growth of the Hub)
         NoDB    : disable database altogether (default)
         
         """)
 
     # not configurable
     db = Instance('IPython.parallel.controller.dictdb.BaseDB')
     heartmonitor = Instance('IPython.parallel.controller.heartmonitor.HeartMonitor')
 
     def _ip_changed(self, name, old, new):
         self.engine_ip = new
         self.client_ip = new
         self.monitor_ip = new
         self._update_monitor_url()
 
     def _update_monitor_url(self):
         self.monitor_url = "%s://%s:%i" % (self.monitor_transport, self.monitor_ip, self.mon_port)
 
     def _transport_changed(self, name, old, new):
         self.engine_transport = new
         self.client_transport = new
         self.monitor_transport = new
         self._update_monitor_url()
 
     def __init__(self, **kwargs):
         super(HubFactory, self).__init__(**kwargs)
         self._update_monitor_url()
 
 
     def construct(self):
         self.init_hub()
 
     def start(self):
         self.heartmonitor.start()
         self.log.info("Heartmonitor started")
 
     def client_url(self, channel):
         """return full zmq url for a named client channel"""
         return "%s://%s:%i" % (self.client_transport, self.client_ip, self.client_info[channel])
     
     def engine_url(self, channel):
         """return full zmq url for a named engine channel"""
         return "%s://%s:%i" % (self.engine_transport, self.engine_ip, self.engine_info[channel])
     
     def init_hub(self):
         """construct Hub object"""
 
         ctx = self.context
         loop = self.loop
 
         try:
             scheme = self.config.TaskScheduler.scheme_name
         except AttributeError:
             from .scheduler import TaskScheduler
             scheme = TaskScheduler.scheme_name.get_default_value()
         
         # build connection dicts
         engine = self.engine_info = {
             'interface'     : "%s://%s" % (self.engine_transport, self.engine_ip),
             'registration'  : self.regport,
             'control'       : self.control[1],
             'mux'           : self.mux[1],
             'hb_ping'       : self.hb[0],
             'hb_pong'       : self.hb[1],
             'task'          : self.task[1],
             'iopub'         : self.iopub[1],
             }
 
         client = self.client_info = {
             'interface'     : "%s://%s" % (self.client_transport, self.client_ip),
             'registration'  : self.regport,
             'control'       : self.control[0],
             'mux'           : self.mux[0],
             'task'          : self.task[0],
             'task_scheme'   : scheme,
             'iopub'         : self.iopub[0],
             'notification'  : self.notifier_port,
             }
         
         self.log.debug("Hub engine addrs: %s", self.engine_info)
         self.log.debug("Hub client addrs: %s", self.client_info)
         
         # Registrar socket
         q = ZMQStream(ctx.socket(zmq.ROUTER), loop)
         q.bind(self.client_url('registration'))
         self.log.info("Hub listening on %s for registration.", self.client_url('registration'))
         if self.client_ip != self.engine_ip:
             q.bind(self.engine_url('registration'))
             self.log.info("Hub listening on %s for registration.", self.engine_url('registration'))
 
         ### Engine connections ###
 
         # heartbeat
         hpub = ctx.socket(zmq.PUB)
         hpub.bind(self.engine_url('hb_ping'))
         hrep = ctx.socket(zmq.ROUTER)
         hrep.bind(self.engine_url('hb_pong'))
         self.heartmonitor = HeartMonitor(loop=loop, config=self.config, log=self.log,
                                 pingstream=ZMQStream(hpub,loop),
                                 pongstream=ZMQStream(hrep,loop)
                             )
 
         ### Client connections ###
         
         # Notifier socket
         n = ZMQStream(ctx.socket(zmq.PUB), loop)
         n.bind(self.client_url('notification'))
 
         ### build and launch the queues ###
 
         # monitor socket
         sub = ctx.socket(zmq.SUB)
         sub.setsockopt(zmq.SUBSCRIBE, b"")
         sub.bind(self.monitor_url)
         sub.bind('inproc://monitor')
         sub = ZMQStream(sub, loop)
 
         # connect the db
         db_class = _db_shortcuts.get(self.db_class.lower(), self.db_class)
         self.log.info('Hub using DB backend: %r', (db_class.split('.')[-1]))
         self.db = import_item(str(db_class))(session=self.session.session,
                                             config=self.config, log=self.log)
         time.sleep(.25)
 
         # resubmit stream
         r = ZMQStream(ctx.socket(zmq.DEALER), loop)
         url = util.disambiguate_url(self.client_url('task'))
         r.connect(url)
 
         self.hub = Hub(loop=loop, session=self.session, monitor=sub, heartmonitor=self.heartmonitor,
                 query=q, notifier=n, resubmit=r, db=self.db,
                 engine_info=self.engine_info, client_info=self.client_info,
                 log=self.log)
 
 
 class Hub(SessionFactory):
     """The IPython Controller Hub with 0MQ connections
 
     Parameters
     ==========
     loop: zmq IOLoop instance
     session: Session object
     <removed> context: zmq context for creating new connections (?)
     queue: ZMQStream for monitoring the command queue (SUB)
     query: ZMQStream for engine registration and client queries requests (ROUTER)
     heartbeat: HeartMonitor object checking the pulse of the engines
     notifier: ZMQStream for broadcasting engine registration changes (PUB)
     db: connection to db for out of memory logging of commands
                 NotImplemented
     engine_info: dict of zmq connection information for engines to connect
                 to the queues.
     client_info: dict of zmq connection information for engines to connect
                 to the queues.
     """
+    
+    engine_state_file = Unicode()
+    
     # internal data structures:
     ids=Set() # engine IDs
     keytable=Dict()
     by_ident=Dict()
     engines=Dict()
     clients=Dict()
     hearts=Dict()
     pending=Set()
     queues=Dict()  # pending msg_ids keyed by engine_id
     tasks=Dict() # pending msg_ids submitted as tasks, keyed by client_id
     completed=Dict() # completed msg_ids keyed by engine_id
     all_completed=Set() # completed msg_ids keyed by engine_id
     dead_engines=Set() # completed msg_ids keyed by engine_id
     unassigned=Set() # set of task msg_ds not yet assigned a destination
     incoming_registrations=Dict()
     registration_timeout=Integer()
     _idcounter=Integer(0)
 
     # objects from constructor:
     query=Instance(ZMQStream)
     monitor=Instance(ZMQStream)
     notifier=Instance(ZMQStream)
     resubmit=Instance(ZMQStream)
     heartmonitor=Instance(HeartMonitor)
     db=Instance(object)
     client_info=Dict()
     engine_info=Dict()
 
 
     def __init__(self, **kwargs):
         """
         # universal:
         loop: IOLoop for creating future connections
         session: streamsession for sending serialized data
         # engine:
         queue: ZMQStream for monitoring queue messages
         query: ZMQStream for engine+client registration and client requests
         heartbeat: HeartMonitor object for tracking engines
         # extra:
         db: ZMQStream for db connection (NotImplemented)
         engine_info: zmq address/protocol dict for engine connections
         client_info: zmq address/protocol dict for client connections
         """
 
         super(Hub, self).__init__(**kwargs)
         self.registration_timeout = max(5000, 2*self.heartmonitor.period)
 
         # register our callbacks
         self.query.on_recv(self.dispatch_query)
         self.monitor.on_recv(self.dispatch_monitor_traffic)
 
         self.heartmonitor.add_heart_failure_handler(self.handle_heart_failure)
         self.heartmonitor.add_new_heart_handler(self.handle_new_heart)
 
         self.monitor_handlers = {b'in' : self.save_queue_request,
                                 b'out': self.save_queue_result,
                                 b'intask': self.save_task_request,
                                 b'outtask': self.save_task_result,
                                 b'tracktask': self.save_task_destination,
                                 b'incontrol': _passer,
                                 b'outcontrol': _passer,
                                 b'iopub': self.save_iopub_message,
         }
 
         self.query_handlers = {'queue_request': self.queue_status,
                                 'result_request': self.get_results,
                                 'history_request': self.get_history,
                                 'db_request': self.db_query,
                                 'purge_request': self.purge_results,
                                 'load_request': self.check_load,
                                 'resubmit_request': self.resubmit_task,
                                 'shutdown_request': self.shutdown_request,
                                 'registration_request' : self.register_engine,
                                 'unregistration_request' : self.unregister_engine,
                                 'connection_request': self.connection_request,
         }
 
         # ignore resubmit replies
         self.resubmit.on_recv(lambda msg: None, copy=False)
 
         self.log.info("hub::created hub")
 
     @property
     def _next_id(self):
         """gemerate a new ID.
 
         No longer reuse old ids, just count from 0."""
         newid = self._idcounter
         self._idcounter += 1
         return newid
         # newid = 0
         # incoming = [id[0] for id in self.incoming_registrations.itervalues()]
         # # print newid, self.ids, self.incoming_registrations
         # while newid in self.ids or newid in incoming:
         #     newid += 1
         # return newid
 
     #-----------------------------------------------------------------------------
     # message validation
     #-----------------------------------------------------------------------------
 
     def _validate_targets(self, targets):
         """turn any valid targets argument into a list of integer ids"""
         if targets is None:
             # default to all
             return self.ids
 
         if isinstance(targets, (int,str,unicode)):
             # only one target specified
             targets = [targets]
         _targets = []
         for t in targets:
             # map raw identities to ids
             if isinstance(t, (str,unicode)):
                 t = self.by_ident.get(cast_bytes(t), t)
             _targets.append(t)
         targets = _targets
         bad_targets = [ t for t in targets if t not in self.ids ]
         if bad_targets:
             raise IndexError("No Such Engine: %r" % bad_targets)
         if not targets:
             raise IndexError("No Engines Registered")
         return targets
 
     #-----------------------------------------------------------------------------
     # dispatch methods (1 per stream)
     #-----------------------------------------------------------------------------
 
 
     @util.log_errors
     def dispatch_monitor_traffic(self, msg):
         """all ME and Task queue messages come through here, as well as
         IOPub traffic."""
         self.log.debug("monitor traffic: %r", msg[0])
         switch = msg[0]
         try:
             idents, msg = self.session.feed_identities(msg[1:])
         except ValueError:
             idents=[]
         if not idents:
             self.log.error("Monitor message without topic: %r", msg)
             return
         handler = self.monitor_handlers.get(switch, None)
         if handler is not None:
             handler(idents, msg)
         else:
             self.log.error("Unrecognized monitor topic: %r", switch)
 
 
     @util.log_errors
     def dispatch_query(self, msg):
         """Route registration requests and queries from clients."""
         try:
             idents, msg = self.session.feed_identities(msg)
         except ValueError:
             idents = []
         if not idents:
             self.log.error("Bad Query Message: %r", msg)
             return
         client_id = idents[0]
         try:
             msg = self.session.unserialize(msg, content=True)
         except Exception:
             content = error.wrap_exception()
             self.log.error("Bad Query Message: %r", msg, exc_info=True)
             self.session.send(self.query, "hub_error", ident=client_id,
                     content=content)
             return
         # print client_id, header, parent, content
         #switch on message type:
         msg_type = msg['header']['msg_type']
         self.log.info("client::client %r requested %r", client_id, msg_type)
         handler = self.query_handlers.get(msg_type, None)
         try:
             assert handler is not None, "Bad Message Type: %r" % msg_type
         except:
             content = error.wrap_exception()
             self.log.error("Bad Message Type: %r", msg_type, exc_info=True)
             self.session.send(self.query, "hub_error", ident=client_id,
                     content=content)
             return
 
         else:
             handler(idents, msg)
 
     def dispatch_db(self, msg):
         """"""
         raise NotImplementedError
 
     #---------------------------------------------------------------------------
     # handler methods (1 per event)
     #---------------------------------------------------------------------------
 
     #----------------------- Heartbeat --------------------------------------
 
     def handle_new_heart(self, heart):
         """handler to attach to heartbeater.
         Called when a new heart starts to beat.
         Triggers completion of registration."""
         self.log.debug("heartbeat::handle_new_heart(%r)", heart)
         if heart not in self.incoming_registrations:
             self.log.info("heartbeat::ignoring new heart: %r", heart)
         else:
             self.finish_registration(heart)
 
 
     def handle_heart_failure(self, heart):
         """handler to attach to heartbeater.
         called when a previously registered heart fails to respond to beat request.
         triggers unregistration"""
         self.log.debug("heartbeat::handle_heart_failure(%r)", heart)
         eid = self.hearts.get(heart, None)
-        queue = self.engines[eid].queue
+        uuid = self.engines[eid].uuid
         if eid is None or self.keytable[eid] in self.dead_engines:
             self.log.info("heartbeat::ignoring heart failure %r (not an engine or already dead)", heart)
         else:
-            self.unregister_engine(heart, dict(content=dict(id=eid, queue=queue)))
+            self.unregister_engine(heart, dict(content=dict(id=eid, queue=uuid)))
 
     #----------------------- MUX Queue Traffic ------------------------------
 
     def save_queue_request(self, idents, msg):
         if len(idents) < 2:
             self.log.error("invalid identity prefix: %r", idents)
             return
         queue_id, client_id = idents[:2]
         try:
             msg = self.session.unserialize(msg)
         except Exception:
             self.log.error("queue::client %r sent invalid message to %r: %r", client_id, queue_id, msg, exc_info=True)
             return
 
         eid = self.by_ident.get(queue_id, None)
         if eid is None:
             self.log.error("queue::target %r not registered", queue_id)
             self.log.debug("queue::    valid are: %r", self.by_ident.keys())
             return
         record = init_record(msg)
         msg_id = record['msg_id']
         self.log.info("queue::client %r submitted request %r to %s", client_id, msg_id, eid)
         # Unicode in records
         record['engine_uuid'] = queue_id.decode('ascii')
         record['client_uuid'] = msg['header']['session']
         record['queue'] = 'mux'
 
         try:
             # it's posible iopub arrived first:
             existing = self.db.get_record(msg_id)
             for key,evalue in existing.iteritems():
                 rvalue = record.get(key, None)
                 if evalue and rvalue and evalue != rvalue:
                     self.log.warn("conflicting initial state for record: %r:%r <%r> %r", msg_id, rvalue, key, evalue)
                 elif evalue and not rvalue:
                     record[key] = evalue
             try:
                 self.db.update_record(msg_id, record)
             except Exception:
                 self.log.error("DB Error updating record %r", msg_id, exc_info=True)
         except KeyError:
             try:
                 self.db.add_record(msg_id, record)
             except Exception:
                 self.log.error("DB Error adding record %r", msg_id, exc_info=True)
 
 
         self.pending.add(msg_id)
         self.queues[eid].append(msg_id)
 
     def save_queue_result(self, idents, msg):
         if len(idents) < 2:
             self.log.error("invalid identity prefix: %r", idents)
             return
 
         client_id, queue_id = idents[:2]
         try:
             msg = self.session.unserialize(msg)
         except Exception:
             self.log.error("queue::engine %r sent invalid message to %r: %r",
                     queue_id, client_id, msg, exc_info=True)
             return
 
         eid = self.by_ident.get(queue_id, None)
         if eid is None:
             self.log.error("queue::unknown engine %r is sending a reply: ", queue_id)
             return
 
         parent = msg['parent_header']
         if not parent:
             return
         msg_id = parent['msg_id']
         if msg_id in self.pending:
             self.pending.remove(msg_id)
             self.all_completed.add(msg_id)
             self.queues[eid].remove(msg_id)
             self.completed[eid].append(msg_id)
             self.log.info("queue::request %r completed on %s", msg_id, eid)
         elif msg_id not in self.all_completed:
             # it could be a result from a dead engine that died before delivering the
             # result
             self.log.warn("queue:: unknown msg finished %r", msg_id)
             return
         # update record anyway, because the unregistration could have been premature
         rheader = msg['header']
         completed = rheader['date']
         started = rheader.get('started', None)
         result = {
             'result_header' : rheader,
             'result_content': msg['content'],
             'received': datetime.now(),
             'started' : started,
             'completed' : completed
         }
 
         result['result_buffers'] = msg['buffers']
         try:
             self.db.update_record(msg_id, result)
         except Exception:
             self.log.error("DB Error updating record %r", msg_id, exc_info=True)
 
 
     #--------------------- Task Queue Traffic ------------------------------
 
     def save_task_request(self, idents, msg):
         """Save the submission of a task."""
         client_id = idents[0]
 
         try:
             msg = self.session.unserialize(msg)
         except Exception:
             self.log.error("task::client %r sent invalid task message: %r",
                     client_id, msg, exc_info=True)
             return
         record = init_record(msg)
 
         record['client_uuid'] = msg['header']['session']
         record['queue'] = 'task'
         header = msg['header']
         msg_id = header['msg_id']
         self.pending.add(msg_id)
         self.unassigned.add(msg_id)
         try:
             # it's posible iopub arrived first:
             existing = self.db.get_record(msg_id)
             if existing['resubmitted']:
                 for key in ('submitted', 'client_uuid', 'buffers'):
                     # don't clobber these keys on resubmit
                     # submitted and client_uuid should be different
                     # and buffers might be big, and shouldn't have changed
                     record.pop(key)
                     # still check content,header which should not change
                     # but are not expensive to compare as buffers
 
             for key,evalue in existing.iteritems():
                 if key.endswith('buffers'):
                     # don't compare buffers
                     continue
                 rvalue = record.get(key, None)
                 if evalue and rvalue and evalue != rvalue:
                     self.log.warn("conflicting initial state for record: %r:%r <%r> %r", msg_id, rvalue, key, evalue)
                 elif evalue and not rvalue:
                     record[key] = evalue
             try:
                 self.db.update_record(msg_id, record)
             except Exception:
                 self.log.error("DB Error updating record %r", msg_id, exc_info=True)
         except KeyError:
             try:
                 self.db.add_record(msg_id, record)
             except Exception:
                 self.log.error("DB Error adding record %r", msg_id, exc_info=True)
         except Exception:
             self.log.error("DB Error saving task request %r", msg_id, exc_info=True)
 
     def save_task_result(self, idents, msg):
         """save the result of a completed task."""
         client_id = idents[0]
         try:
             msg = self.session.unserialize(msg)
         except Exception:
             self.log.error("task::invalid task result message send to %r: %r",
                     client_id, msg, exc_info=True)
             return
 
         parent = msg['parent_header']
         if not parent:
             # print msg
             self.log.warn("Task %r had no parent!", msg)
             return
         msg_id = parent['msg_id']
         if msg_id in self.unassigned:
             self.unassigned.remove(msg_id)
 
         header = msg['header']
         engine_uuid = header.get('engine', u'')
         eid = self.by_ident.get(cast_bytes(engine_uuid), None)
         
         status = header.get('status', None)
 
         if msg_id in self.pending:
             self.log.info("task::task %r finished on %s", msg_id, eid)
             self.pending.remove(msg_id)
             self.all_completed.add(msg_id)
             if eid is not None:
                 if status != 'aborted':
                     self.completed[eid].append(msg_id)
                 if msg_id in self.tasks[eid]:
                     self.tasks[eid].remove(msg_id)
             completed = header['date']
             started = header.get('started', None)
             result = {
                 'result_header' : header,
                 'result_content': msg['content'],
                 'started' : started,
                 'completed' : completed,
                 'received' : datetime.now(),
                 'engine_uuid': engine_uuid,
             }
 
             result['result_buffers'] = msg['buffers']
             try:
                 self.db.update_record(msg_id, result)
             except Exception:
                 self.log.error("DB Error saving task request %r", msg_id, exc_info=True)
 
         else:
             self.log.debug("task::unknown task %r finished", msg_id)
 
     def save_task_destination(self, idents, msg):
         try:
             msg = self.session.unserialize(msg, content=True)
         except Exception:
             self.log.error("task::invalid task tracking message", exc_info=True)
             return
         content = msg['content']
         # print (content)
         msg_id = content['msg_id']
         engine_uuid = content['engine_id']
         eid = self.by_ident[cast_bytes(engine_uuid)]
 
         self.log.info("task::task %r arrived on %r", msg_id, eid)
         if msg_id in self.unassigned:
             self.unassigned.remove(msg_id)
         # else:
         #     self.log.debug("task::task %r not listed as MIA?!"%(msg_id))
 
         self.tasks[eid].append(msg_id)
         # self.pending[msg_id][1].update(received=datetime.now(),engine=(eid,engine_uuid))
         try:
             self.db.update_record(msg_id, dict(engine_uuid=engine_uuid))
         except Exception:
             self.log.error("DB Error saving task destination %r", msg_id, exc_info=True)
 
 
     def mia_task_request(self, idents, msg):
         raise NotImplementedError
         client_id = idents[0]
         # content = dict(mia=self.mia,status='ok')
         # self.session.send('mia_reply', content=content, idents=client_id)
 
 
     #--------------------- IOPub Traffic ------------------------------
 
     def save_iopub_message(self, topics, msg):
         """save an iopub message into the db"""
         # print (topics)
         try:
             msg = self.session.unserialize(msg, content=True)
         except Exception:
             self.log.error("iopub::invalid IOPub message", exc_info=True)
             return
 
         parent = msg['parent_header']
         if not parent:
             self.log.warn("iopub::IOPub message lacks parent: %r", msg)
             return
         msg_id = parent['msg_id']
         msg_type = msg['header']['msg_type']
         content = msg['content']
 
         # ensure msg_id is in db
         try:
             rec = self.db.get_record(msg_id)
         except KeyError:
             rec = empty_record()
             rec['msg_id'] = msg_id
             self.db.add_record(msg_id, rec)
         # stream
         d = {}
         if msg_type == 'stream':
             name = content['name']
             s = rec[name] or ''
             d[name] = s + content['data']
 
         elif msg_type == 'pyerr':
             d['pyerr'] = content
         elif msg_type == 'pyin':
             d['pyin'] = content['code']
         elif msg_type in ('display_data', 'pyout'):
             d[msg_type] = content
         elif msg_type == 'status':
             pass
         else:
             self.log.warn("unhandled iopub msg_type: %r", msg_type)
 
         if not d:
             return
 
         try:
             self.db.update_record(msg_id, d)
         except Exception:
             self.log.error("DB Error saving iopub message %r", msg_id, exc_info=True)
 
 
 
     #-------------------------------------------------------------------------
     # Registration requests
     #-------------------------------------------------------------------------
 
     def connection_request(self, client_id, msg):
         """Reply with connection addresses for clients."""
         self.log.info("client::client %r connected", client_id)
         content = dict(status='ok')
         jsonable = {}
         for k,v in self.keytable.iteritems():
             if v not in self.dead_engines:
-                jsonable[str(k)] = v.decode('ascii')
+                jsonable[str(k)] = v
         content['engines'] = jsonable
         self.session.send(self.query, 'connection_reply', content, parent=msg, ident=client_id)
 
     def register_engine(self, reg, msg):
         """Register a new engine."""
         content = msg['content']
         try:
-            queue = cast_bytes(content['queue'])
+            uuid = content['uuid']
         except KeyError:
             self.log.error("registration::queue not specified", exc_info=True)
             return
-        heart = content.get('heartbeat', None)
-        if heart:
-            heart = cast_bytes(heart)
-        """register a new engine, and create the socket(s) necessary"""
+
         eid = self._next_id
-        # print (eid, queue, reg, heart)
 
-        self.log.debug("registration::register_engine(%i, %r, %r, %r)", eid, queue, reg, heart)
+        self.log.debug("registration::register_engine(%i, %r)", eid, uuid)
 
         content = dict(id=eid,status='ok')
         # check if requesting available IDs:
-        if queue in self.by_ident:
+        if uuid in self.by_ident:
             try:
-                raise KeyError("queue_id %r in use" % queue)
+                raise KeyError("uuid %r in use" % uuid)
             except:
                 content = error.wrap_exception()
-                self.log.error("queue_id %r in use", queue, exc_info=True)
-        elif heart in self.hearts: # need to check unique hearts?
-            try:
-                raise KeyError("heart_id %r in use" % heart)
-            except:
-                self.log.error("heart_id %r in use", heart, exc_info=True)
-                content = error.wrap_exception()
+                self.log.error("uuid %r in use", uuid, exc_info=True)
         else:
-            for h, pack in self.incoming_registrations.iteritems():
-                if heart == h:
+            for h, ec in self.incoming_registrations.iteritems():
+                if uuid == h:
                     try:
-                        raise KeyError("heart_id %r in use" % heart)
+                        raise KeyError("heart_id %r in use" % uuid)
                     except:
-                        self.log.error("heart_id %r in use", heart, exc_info=True)
+                        self.log.error("heart_id %r in use", uuid, exc_info=True)
                         content = error.wrap_exception()
                     break
-                elif queue == pack[1]:
+                elif uuid == ec.uuid:
                     try:
-                        raise KeyError("queue_id %r in use" % queue)
+                        raise KeyError("uuid %r in use" % uuid)
                     except:
-                        self.log.error("queue_id %r in use", queue, exc_info=True)
+                        self.log.error("uuid %r in use", uuid, exc_info=True)
                         content = error.wrap_exception()
                     break
 
         msg = self.session.send(self.query, "registration_reply",
                 content=content,
                 ident=reg)
 
+        heart = util.asbytes(uuid)
+
         if content['status'] == 'ok':
             if heart in self.heartmonitor.hearts:
                 # already beating
-                self.incoming_registrations[heart] = (eid,queue,reg[0],None)
+                self.incoming_registrations[heart] = EngineConnector(id=eid,uuid=uuid)
                 self.finish_registration(heart)
             else:
                 purge = lambda : self._purge_stalled_registration(heart)
                 dc = ioloop.DelayedCallback(purge, self.registration_timeout, self.loop)
                 dc.start()
-                self.incoming_registrations[heart] = (eid,queue,reg[0],dc)
+                self.incoming_registrations[heart] = EngineConnector(id=eid,uuid=uuid,stallback=dc)
         else:
             self.log.error("registration::registration %i failed: %r", eid, content['evalue'])
+        
         return eid
 
     def unregister_engine(self, ident, msg):
         """Unregister an engine that explicitly requested to leave."""
         try:
             eid = msg['content']['id']
         except:
             self.log.error("registration::bad engine id for unregistration: %r", ident, exc_info=True)
             return
         self.log.info("registration::unregister_engine(%r)", eid)
         # print (eid)
         uuid = self.keytable[eid]
-        content=dict(id=eid, queue=uuid.decode('ascii'))
+        content=dict(id=eid, uuid=uuid)
         self.dead_engines.add(uuid)
         # self.ids.remove(eid)
         # uuid = self.keytable.pop(eid)
         #
         # ec = self.engines.pop(eid)
         # self.hearts.pop(ec.heartbeat)
         # self.by_ident.pop(ec.queue)
         # self.completed.pop(eid)
         handleit = lambda : self._handle_stranded_msgs(eid, uuid)
         dc = ioloop.DelayedCallback(handleit, self.registration_timeout, self.loop)
         dc.start()
         ############## TODO: HANDLE IT ################
 
+        self._save_engine_state()
+
         if self.notifier:
             self.session.send(self.notifier, "unregistration_notification", content=content)
 
     def _handle_stranded_msgs(self, eid, uuid):
         """Handle messages known to be on an engine when the engine unregisters.
 
         It is possible that this will fire prematurely - that is, an engine will
         go down after completing a result, and the client will be notified
         that the result failed and later receive the actual result.
         """
 
         outstanding = self.queues[eid]
 
         for msg_id in outstanding:
             self.pending.remove(msg_id)
             self.all_completed.add(msg_id)
             try:
                 raise error.EngineError("Engine %r died while running task %r" % (eid, msg_id))
             except:
                 content = error.wrap_exception()
             # build a fake header:
             header = {}
             header['engine'] = uuid
             header['date'] = datetime.now()
             rec = dict(result_content=content, result_header=header, result_buffers=[])
             rec['completed'] = header['date']
             rec['engine_uuid'] = uuid
             try:
                 self.db.update_record(msg_id, rec)
             except Exception:
                 self.log.error("DB Error handling stranded msg %r", msg_id, exc_info=True)
 
 
     def finish_registration(self, heart):
         """Second half of engine registration, called after our HeartMonitor
         has received a beat from the Engine's Heart."""
         try:
-            (eid,queue,reg,purge) = self.incoming_registrations.pop(heart)
+            ec = self.incoming_registrations.pop(heart)
         except KeyError:
             self.log.error("registration::tried to finish nonexistant registration", exc_info=True)
             return
-        self.log.info("registration::finished registering engine %i:%r", eid, queue)
-        if purge is not None:
-            purge.stop()
-        control = queue
+        self.log.info("registration::finished registering engine %i:%s", ec.id, ec.uuid)
+        if ec.stallback is not None:
+            ec.stallback.stop()
+        eid = ec.id
         self.ids.add(eid)
-        self.keytable[eid] = queue
-        self.engines[eid] = EngineConnector(id=eid, queue=queue, registration=reg,
-                                    control=control, heartbeat=heart)
-        self.by_ident[queue] = eid
+        self.keytable[eid] = ec.uuid
+        self.engines[eid] = ec
+        self.by_ident[ec.uuid] = ec.id
         self.queues[eid] = list()
         self.tasks[eid] = list()
         self.completed[eid] = list()
         self.hearts[heart] = eid
-        content = dict(id=eid, queue=self.engines[eid].queue.decode('ascii'))
+        content = dict(id=eid, uuid=self.engines[eid].uuid)
         if self.notifier:
             self.session.send(self.notifier, "registration_notification", content=content)
         self.log.info("engine::Engine Connected: %i", eid)
 
+        self._save_engine_state()
+
     def _purge_stalled_registration(self, heart):
         if heart in self.incoming_registrations:
-            eid = self.incoming_registrations.pop(heart)[0]
-            self.log.info("registration::purging stalled registration: %i", eid)
+            ec = self.incoming_registrations.pop(heart)
+            self.log.info("registration::purging stalled registration: %i", ec.id)
         else:
             pass
 
     #-------------------------------------------------------------------------
+    # Engine State
+    #-------------------------------------------------------------------------
+
+
+    def _cleanup_engine_state_file(self):
+        """cleanup engine state mapping"""
+        
+        if os.path.exists(self.engine_state_file):
+            self.log.debug("cleaning up engine state: %s", self.engine_state_file)
+            try:
+                os.remove(self.engine_state_file)
+            except IOError:
+                self.log.error("Couldn't cleanup file: %s", self.engine_state_file, exc_info=True)
+
+
+    def _save_engine_state(self):
+        """save engine mapping to JSON file"""
+        if not self.engine_state_file:
+            return
+        self.log.debug("save engine state to %s" % self.engine_state_file)
+        state = {}
+        engines = {}
+        for eid, ec in self.engines.iteritems():
+            if ec.uuid not in self.dead_engines:
+                engines[eid] = ec.uuid
+        
+        state['engines'] = engines
+        
+        state['next_id'] = self._idcounter
+        
+        with open(self.engine_state_file, 'w') as f:
+            json.dump(state, f)
+
+
+    def _load_engine_state(self):
+        """load engine mapping from JSON file"""
+        if not os.path.exists(self.engine_state_file):
+            return
+        
+        self.log.info("loading engine state from %s" % self.engine_state_file)
+        
+        with open(self.engine_state_file) as f:
+            state = json.load(f)
+        
+        save_notifier = self.notifier
+        self.notifier = None
+        for eid, uuid in state['engines'].iteritems():
+            heart = uuid.encode('ascii')
+            # start with this heart as current and beating:
+            self.heartmonitor.responses.add(heart)
+            self.heartmonitor.hearts.add(heart)
+            
+            self.incoming_registrations[heart] = EngineConnector(id=int(eid), uuid=uuid)
+            self.finish_registration(heart)
+        
+        self.notifier = save_notifier
+        
+        self._idcounter = state['next_id']
+
+    #-------------------------------------------------------------------------
     # Client Requests
     #-------------------------------------------------------------------------
 
     def shutdown_request(self, client_id, msg):
         """handle shutdown request."""
         self.session.send(self.query, 'shutdown_reply', content={'status': 'ok'}, ident=client_id)
         # also notify other clients of shutdown
         self.session.send(self.notifier, 'shutdown_notice', content={'status': 'ok'})
         dc = ioloop.DelayedCallback(lambda : self._shutdown(), 1000, self.loop)
         dc.start()
 
     def _shutdown(self):
         self.log.info("hub::hub shutting down.")
         time.sleep(0.1)
         sys.exit(0)
 
 
     def check_load(self, client_id, msg):
         content = msg['content']
         try:
             targets = content['targets']
             targets = self._validate_targets(targets)
         except:
             content = error.wrap_exception()
             self.session.send(self.query, "hub_error",
                     content=content, ident=client_id)
             return
 
         content = dict(status='ok')
         # loads = {}
         for t in targets:
             content[bytes(t)] = len(self.queues[t])+len(self.tasks[t])
         self.session.send(self.query, "load_reply", content=content, ident=client_id)
 
 
     def queue_status(self, client_id, msg):
         """Return the Queue status of one or more targets.
         if verbose: return the msg_ids
         else: return len of each type.
         keys: queue (pending MUX jobs)
             tasks (pending Task jobs)
             completed (finished jobs from both queues)"""
         content = msg['content']
         targets = content['targets']
         try:
             targets = self._validate_targets(targets)
         except:
             content = error.wrap_exception()
             self.session.send(self.query, "hub_error",
                     content=content, ident=client_id)
             return
         verbose = content.get('verbose', False)
         content = dict(status='ok')
         for t in targets:
             queue = self.queues[t]
             completed = self.completed[t]
             tasks = self.tasks[t]
             if not verbose:
                 queue = len(queue)
                 completed = len(completed)
                 tasks = len(tasks)
             content[str(t)] = {'queue': queue, 'completed': completed , 'tasks': tasks}
         content['unassigned'] = list(self.unassigned) if verbose else len(self.unassigned)
         # print (content)
         self.session.send(self.query, "queue_reply", content=content, ident=client_id)
 
     def purge_results(self, client_id, msg):
         """Purge results from memory. This method is more valuable before we move
         to a DB based message storage mechanism."""
         content = msg['content']
         self.log.info("Dropping records with %s", content)
         msg_ids = content.get('msg_ids', [])
         reply = dict(status='ok')
         if msg_ids == 'all':
             try:
                 self.db.drop_matching_records(dict(completed={'$ne':None}))
             except Exception:
                 reply = error.wrap_exception()
         else:
             pending = filter(lambda m: m in self.pending, msg_ids)
             if pending:
                 try:
                     raise IndexError("msg pending: %r" % pending[0])
                 except:
                     reply = error.wrap_exception()
             else:
                 try:
                     self.db.drop_matching_records(dict(msg_id={'$in':msg_ids}))
                 except Exception:
                     reply = error.wrap_exception()
 
             if reply['status'] == 'ok':
                 eids = content.get('engine_ids', [])
                 for eid in eids:
                     if eid not in self.engines:
                         try:
                             raise IndexError("No such engine: %i" % eid)
                         except:
                             reply = error.wrap_exception()
                         break
-                    uid = self.engines[eid].queue
+                    uid = self.engines[eid].uuid
                     try:
                         self.db.drop_matching_records(dict(engine_uuid=uid, completed={'$ne':None}))
                     except Exception:
                         reply = error.wrap_exception()
                         break
 
         self.session.send(self.query, 'purge_reply', content=reply, ident=client_id)
 
     def resubmit_task(self, client_id, msg):
         """Resubmit one or more tasks."""
         def finish(reply):
             self.session.send(self.query, 'resubmit_reply', content=reply, ident=client_id)
 
         content = msg['content']
         msg_ids = content['msg_ids']
         reply = dict(status='ok')
         try:
             records = self.db.find_records({'msg_id' : {'$in' : msg_ids}}, keys=[
                 'header', 'content', 'buffers'])
         except Exception:
             self.log.error('db::db error finding tasks to resubmit', exc_info=True)
             return finish(error.wrap_exception())
 
         # validate msg_ids
         found_ids = [ rec['msg_id'] for rec in records ]
         pending_ids = [ msg_id for msg_id in found_ids if msg_id in self.pending ]
         if len(records) > len(msg_ids):
             try:
                 raise RuntimeError("DB appears to be in an inconsistent state."
                     "More matching records were found than should exist")
             except Exception:
                 return finish(error.wrap_exception())
         elif len(records) < len(msg_ids):
             missing = [ m for m in msg_ids if m not in found_ids ]
             try:
                 raise KeyError("No such msg(s): %r" % missing)
             except KeyError:
                 return finish(error.wrap_exception())
         elif pending_ids:
             pass
             # no need to raise on resubmit of pending task, now that we
             # resubmit under new ID, but do we want to raise anyway?
             # msg_id = invalid_ids[0]
             # try:
             #     raise ValueError("Task(s) %r appears to be inflight" % )
             # except Exception:
             #     return finish(error.wrap_exception())
 
         # mapping of original IDs to resubmitted IDs
         resubmitted = {}
 
         # send the messages
         for rec in records:
             header = rec['header']
             msg = self.session.msg(header['msg_type'], parent=header)
             msg_id = msg['msg_id']
             msg['content'] = rec['content']
             
             # use the old header, but update msg_id and timestamp
             fresh = msg['header']
             header['msg_id'] = fresh['msg_id']
             header['date'] = fresh['date']
             msg['header'] = header
 
             self.session.send(self.resubmit, msg, buffers=rec['buffers'])
 
             resubmitted[rec['msg_id']] = msg_id
             self.pending.add(msg_id)
             msg['buffers'] = rec['buffers']
             try:
                 self.db.add_record(msg_id, init_record(msg))
             except Exception:
                 self.log.error("db::DB Error updating record: %s", msg_id, exc_info=True)
                 return finish(error.wrap_exception())
 
         finish(dict(status='ok', resubmitted=resubmitted))
         
         # store the new IDs in the Task DB
         for msg_id, resubmit_id in resubmitted.iteritems():
             try:
                 self.db.update_record(msg_id, {'resubmitted' : resubmit_id})
             except Exception:
                 self.log.error("db::DB Error updating record: %s", msg_id, exc_info=True)
 
 
     def _extract_record(self, rec):
         """decompose a TaskRecord dict into subsection of reply for get_result"""
         io_dict = {}
         for key in ('pyin', 'pyout', 'pyerr', 'stdout', 'stderr'):
                 io_dict[key] = rec[key]
         content = { 'result_content': rec['result_content'],
                             'header': rec['header'],
                             'result_header' : rec['result_header'],
                             'received' : rec['received'],
                             'io' : io_dict,
                           }
         if rec['result_buffers']:
             buffers = map(bytes, rec['result_buffers'])
         else:
             buffers = []
 
         return content, buffers
 
     def get_results(self, client_id, msg):
         """Get the result of 1 or more messages."""
         content = msg['content']
         msg_ids = sorted(set(content['msg_ids']))
         statusonly = content.get('status_only', False)
         pending = []
         completed = []
         content = dict(status='ok')
         content['pending'] = pending
         content['completed'] = completed
         buffers = []
         if not statusonly:
             try:
                 matches = self.db.find_records(dict(msg_id={'$in':msg_ids}))
                 # turn match list into dict, for faster lookup
                 records = {}
                 for rec in matches:
                     records[rec['msg_id']] = rec
             except Exception:
                 content = error.wrap_exception()
                 self.session.send(self.query, "result_reply", content=content,
                                                     parent=msg, ident=client_id)
                 return
         else:
             records = {}
         for msg_id in msg_ids:
             if msg_id in self.pending:
                 pending.append(msg_id)
             elif msg_id in self.all_completed:
                 completed.append(msg_id)
                 if not statusonly:
                     c,bufs = self._extract_record(records[msg_id])
                     content[msg_id] = c
                     buffers.extend(bufs)
             elif msg_id in records:
                 if rec['completed']:
                     completed.append(msg_id)
                     c,bufs = self._extract_record(records[msg_id])
                     content[msg_id] = c
                     buffers.extend(bufs)
                 else:
                     pending.append(msg_id)
             else:
                 try:
                     raise KeyError('No such message: '+msg_id)
                 except:
                     content = error.wrap_exception()
                 break
         self.session.send(self.query, "result_reply", content=content,
                                             parent=msg, ident=client_id,
                                             buffers=buffers)
 
     def get_history(self, client_id, msg):
         """Get a list of all msg_ids in our DB records"""
         try:
             msg_ids = self.db.get_history()
         except Exception as e:
             content = error.wrap_exception()
         else:
             content = dict(status='ok', history=msg_ids)
 
         self.session.send(self.query, "history_reply", content=content,
                                             parent=msg, ident=client_id)
 
     def db_query(self, client_id, msg):
         """Perform a raw query on the task record database."""
         content = msg['content']
         query = content.get('query', {})
         keys = content.get('keys', None)
         buffers = []
         empty = list()
         try:
             records = self.db.find_records(query, keys)
         except Exception as e:
             content = error.wrap_exception()
         else:
             # extract buffers from reply content:
             if keys is not None:
                 buffer_lens = [] if 'buffers' in keys else None
                 result_buffer_lens = [] if 'result_buffers' in keys else None
             else:
                 buffer_lens = None
                 result_buffer_lens = None
 
             for rec in records:
                 # buffers may be None, so double check
                 b = rec.pop('buffers', empty) or empty
                 if buffer_lens is not None:
                     buffer_lens.append(len(b))
                     buffers.extend(b)
                 rb = rec.pop('result_buffers', empty) or empty
                 if result_buffer_lens is not None:
                     result_buffer_lens.append(len(rb))
                     buffers.extend(rb)
             content = dict(status='ok', records=records, buffer_lens=buffer_lens,
                                     result_buffer_lens=result_buffer_lens)
         # self.log.debug (content)
         self.session.send(self.query, "db_reply", content=content,
                                             parent=msg, ident=client_id,
                                             buffers=buffers)
 

IPython/parallel/controller/scheduler.py

@@ -1,768 +1,794 b''
 """The Python scheduler for rich scheduling.
 
 The Pure ZMQ scheduler does not allow routing schemes other than LRU,
 nor does it check msg_id DAG dependencies. For those, a slightly slower
 Python Scheduler exists.
 
 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.
 #-----------------------------------------------------------------------------
 
 #----------------------------------------------------------------------
 # Imports
 #----------------------------------------------------------------------
 
 from __future__ import print_function
 
 import logging
 import sys
 import time
 
 from datetime import datetime, timedelta
 from random import randint, random
 from types import FunctionType
 
 try:
     import numpy
 except ImportError:
     numpy = None
 
 import zmq
 from zmq.eventloop import ioloop, zmqstream
 
 # local imports
 from IPython.external.decorator import decorator
 from IPython.config.application import Application
 from IPython.config.loader import Config
 from IPython.utils.traitlets import Instance, Dict, List, Set, Integer, Enum, CBytes
 from IPython.utils.py3compat import cast_bytes
 
 from IPython.parallel import error, util
 from IPython.parallel.factory import SessionFactory
 from IPython.parallel.util import connect_logger, local_logger
 
 from .dependency import Dependency
 
 @decorator
 def logged(f,self,*args,**kwargs):
     # print ("#--------------------")
     self.log.debug("scheduler::%s(*%s,**%s)", f.func_name, args, kwargs)
     # print ("#--")
     return f(self,*args, **kwargs)
 
 #----------------------------------------------------------------------
 # Chooser functions
 #----------------------------------------------------------------------
 
 def plainrandom(loads):
     """Plain random pick."""
     n = len(loads)
     return randint(0,n-1)
 
 def lru(loads):
     """Always pick the front of the line.
 
     The content of `loads` is ignored.
 
     Assumes LRU ordering of loads, with oldest first.
     """
     return 0
 
 def twobin(loads):
     """Pick two at random, use the LRU of the two.
 
     The content of loads is ignored.
 
     Assumes LRU ordering of loads, with oldest first.
     """
     n = len(loads)
     a = randint(0,n-1)
     b = randint(0,n-1)
     return min(a,b)
 
 def weighted(loads):
     """Pick two at random using inverse load as weight.
 
     Return the less loaded of the two.
     """
     # weight 0 a million times more than 1:
     weights = 1./(1e-6+numpy.array(loads))
     sums = weights.cumsum()
     t = sums[-1]
     x = random()*t
     y = random()*t
     idx = 0
     idy = 0
     while sums[idx] < x:
         idx += 1
     while sums[idy] < y:
         idy += 1
     if weights[idy] > weights[idx]:
         return idy
     else:
         return idx
 
 def leastload(loads):
     """Always choose the lowest load.
 
     If the lowest load occurs more than once, the first
     occurance will be used.  If loads has LRU ordering, this means
     the LRU of those with the lowest load is chosen.
     """
     return loads.index(min(loads))
 
 #---------------------------------------------------------------------
 # Classes
 #---------------------------------------------------------------------
 
 
 # store empty default dependency:
 MET = Dependency([])
 
 
 class Job(object):
     """Simple container for a job"""
     def __init__(self, msg_id, raw_msg, idents, msg, header, targets, after, follow, timeout):
         self.msg_id = msg_id
         self.raw_msg = raw_msg
         self.idents = idents
         self.msg = msg
         self.header = header
         self.targets = targets
         self.after = after
         self.follow = follow
         self.timeout = timeout
         
         
         self.timestamp = time.time()
         self.blacklist = set()
 
     @property
     def dependents(self):
         return self.follow.union(self.after)
 
 class TaskScheduler(SessionFactory):
     """Python TaskScheduler object.
 
     This is the simplest object that supports msg_id based
     DAG dependencies. *Only* task msg_ids are checked, not
     msg_ids of jobs submitted via the MUX queue.
 
     """
 
     hwm = Integer(1, config=True,
         help="""specify the High Water Mark (HWM) for the downstream
         socket in the Task scheduler. This is the maximum number
         of allowed outstanding tasks on each engine.
         
         The default (1) means that only one task can be outstanding on each
         engine.  Setting TaskScheduler.hwm=0 means there is no limit, and the
         engines continue to be assigned tasks while they are working,
         effectively hiding network latency behind computation, but can result
         in an imbalance of work when submitting many heterogenous tasks all at
         once.  Any positive value greater than one is a compromise between the
         two.
 
         """
     )
     scheme_name = Enum(('leastload', 'pure', 'lru', 'plainrandom', 'weighted', 'twobin'),
         'leastload', config=True, allow_none=False,
         help="""select the task scheduler scheme  [default: Python LRU]
         Options are: 'pure', 'lru', 'plainrandom', 'weighted', 'twobin','leastload'"""
     )
     def _scheme_name_changed(self, old, new):
         self.log.debug("Using scheme %r"%new)
         self.scheme = globals()[new]
 
     # input arguments:
     scheme = Instance(FunctionType) # function for determining the destination
     def _scheme_default(self):
         return leastload
     client_stream = Instance(zmqstream.ZMQStream) # client-facing stream
     engine_stream = Instance(zmqstream.ZMQStream) # engine-facing stream
     notifier_stream = Instance(zmqstream.ZMQStream) # hub-facing sub stream
     mon_stream = Instance(zmqstream.ZMQStream) # hub-facing pub stream
+    query_stream = Instance(zmqstream.ZMQStream) # hub-facing DEALER stream
 
     # internals:
     graph = Dict() # dict by msg_id of [ msg_ids that depend on key ]
     retries = Dict() # dict by msg_id of retries remaining (non-neg ints)
     # waiting = List() # list of msg_ids ready to run, but haven't due to HWM
     depending = Dict() # dict by msg_id of Jobs
     pending = Dict() # dict by engine_uuid of submitted tasks
     completed = Dict() # dict by engine_uuid of completed tasks
     failed = Dict() # dict by engine_uuid of failed tasks
     destinations = Dict() # dict by msg_id of engine_uuids where jobs ran (reverse of completed+failed)
     clients = Dict() # dict by msg_id for who submitted the task
     targets = List() # list of target IDENTs
     loads = List() # list of engine loads
     # full = Set() # set of IDENTs that have HWM outstanding tasks
     all_completed = Set() # set of all completed tasks
     all_failed = Set() # set of all failed tasks
     all_done = Set() # set of all finished tasks=union(completed,failed)
     all_ids = Set() # set of all submitted task IDs
 
     auditor = Instance('zmq.eventloop.ioloop.PeriodicCallback')
 
     ident = CBytes() # ZMQ identity. This should just be self.session.session
                      # but ensure Bytes
     def _ident_default(self):
         return self.session.bsession
 
     def start(self):
+        self.query_stream.on_recv(self.dispatch_query_reply)
+        self.session.send(self.query_stream, "connection_request", {})
+        
         self.engine_stream.on_recv(self.dispatch_result, copy=False)
         self.client_stream.on_recv(self.dispatch_submission, copy=False)
 
         self._notification_handlers = dict(
             registration_notification = self._register_engine,
             unregistration_notification = self._unregister_engine
         )
         self.notifier_stream.on_recv(self.dispatch_notification)
         self.auditor = ioloop.PeriodicCallback(self.audit_timeouts, 2e3, self.loop) # 1 Hz
         self.auditor.start()
         self.log.info("Scheduler started [%s]"%self.scheme_name)
 
     def resume_receiving(self):
         """Resume accepting jobs."""
         self.client_stream.on_recv(self.dispatch_submission, copy=False)
 
     def stop_receiving(self):
         """Stop accepting jobs while there are no engines.
         Leave them in the ZMQ queue."""
         self.client_stream.on_recv(None)
 
     #-----------------------------------------------------------------------
     # [Un]Registration Handling
     #-----------------------------------------------------------------------
 
     
+    def dispatch_query_reply(self, msg):
+        """handle reply to our initial connection request"""
+        try:
+            idents,msg = self.session.feed_identities(msg)
+        except ValueError:
+            self.log.warn("task::Invalid Message: %r",msg)
+            return
+        try:
+            msg = self.session.unserialize(msg)
+        except ValueError:
+            self.log.warn("task::Unauthorized message from: %r"%idents)
+            return
+        
+        content = msg['content']
+        for uuid in content.get('engines', {}).values():
+            self._register_engine(asbytes(uuid))
+
+    
     @util.log_errors
     def dispatch_notification(self, msg):
         """dispatch register/unregister events."""
         try:
             idents,msg = self.session.feed_identities(msg)
         except ValueError:
             self.log.warn("task::Invalid Message: %r",msg)
             return
         try:
             msg = self.session.unserialize(msg)
         except ValueError:
             self.log.warn("task::Unauthorized message from: %r"%idents)
             return
 
         msg_type = msg['header']['msg_type']
 
         handler = self._notification_handlers.get(msg_type, None)
         if handler is None:
             self.log.error("Unhandled message type: %r"%msg_type)
         else:
             try:
-                handler(cast_bytes(msg['content']['queue']))
+                handler(cast_bytes(msg['content']['uuid']))
             except Exception:
                 self.log.error("task::Invalid notification msg: %r", msg, exc_info=True)
 
     def _register_engine(self, uid):
         """New engine with ident `uid` became available."""
         # head of the line:
         self.targets.insert(0,uid)
         self.loads.insert(0,0)
 
         # initialize sets
         self.completed[uid] = set()
         self.failed[uid] = set()
         self.pending[uid] = {}
 
         # rescan the graph:
         self.update_graph(None)
 
     def _unregister_engine(self, uid):
         """Existing engine with ident `uid` became unavailable."""
         if len(self.targets) == 1:
             # this was our only engine
             pass
 
         # handle any potentially finished tasks:
         self.engine_stream.flush()
 
         # don't pop destinations, because they might be used later
         # map(self.destinations.pop, self.completed.pop(uid))
         # map(self.destinations.pop, self.failed.pop(uid))
 
         # prevent this engine from receiving work
         idx = self.targets.index(uid)
         self.targets.pop(idx)
         self.loads.pop(idx)
 
         # wait 5 seconds before cleaning up pending jobs, since the results might
         # still be incoming
         if self.pending[uid]:
             dc = ioloop.DelayedCallback(lambda : self.handle_stranded_tasks(uid), 5000, self.loop)
             dc.start()
         else:
             self.completed.pop(uid)
             self.failed.pop(uid)
 
 
     def handle_stranded_tasks(self, engine):
         """Deal with jobs resident in an engine that died."""
         lost = self.pending[engine]
         for msg_id in lost.keys():
             if msg_id not in self.pending[engine]:
                 # prevent double-handling of messages
                 continue
 
             raw_msg = lost[msg_id].raw_msg
             idents,msg = self.session.feed_identities(raw_msg, copy=False)
             parent = self.session.unpack(msg[1].bytes)
             idents = [engine, idents[0]]
 
             # build fake error reply
             try:
                 raise error.EngineError("Engine %r died while running task %r"%(engine, msg_id))
             except:
                 content = error.wrap_exception()
             # build fake header
             header = dict(
                 status='error',
                 engine=engine,
                 date=datetime.now(),
             )
             msg = self.session.msg('apply_reply', content, parent=parent, subheader=header)
             raw_reply = map(zmq.Message, self.session.serialize(msg, ident=idents))
             # and dispatch it
             self.dispatch_result(raw_reply)
 
         # finally scrub completed/failed lists
         self.completed.pop(engine)
         self.failed.pop(engine)
 
 
     #-----------------------------------------------------------------------
     # Job Submission
     #-----------------------------------------------------------------------
     
     
     @util.log_errors
     def dispatch_submission(self, raw_msg):
         """Dispatch job submission to appropriate handlers."""
         # ensure targets up to date:
         self.notifier_stream.flush()
         try:
             idents, msg = self.session.feed_identities(raw_msg, copy=False)
             msg = self.session.unserialize(msg, content=False, copy=False)
         except Exception:
             self.log.error("task::Invaid task msg: %r"%raw_msg, exc_info=True)
             return
 
 
         # send to monitor
         self.mon_stream.send_multipart([b'intask']+raw_msg, copy=False)
 
         header = msg['header']
         msg_id = header['msg_id']
         self.all_ids.add(msg_id)
 
         # get targets as a set of bytes objects
         # from a list of unicode objects
         targets = header.get('targets', [])
         targets = map(cast_bytes, targets)
         targets = set(targets)
 
         retries = header.get('retries', 0)
         self.retries[msg_id] = retries
 
         # time dependencies
         after = header.get('after', None)
         if after:
             after = Dependency(after)
             if after.all:
                 if after.success:
                     after = Dependency(after.difference(self.all_completed),
                                 success=after.success,
                                 failure=after.failure,
                                 all=after.all,
                     )
                 if after.failure:
                     after = Dependency(after.difference(self.all_failed),
                                 success=after.success,
                                 failure=after.failure,
                                 all=after.all,
                     )
             if after.check(self.all_completed, self.all_failed):
                 # recast as empty set, if `after` already met,
                 # to prevent unnecessary set comparisons
                 after = MET
         else:
             after = MET
 
         # location dependencies
         follow = Dependency(header.get('follow', []))
 
         # turn timeouts into datetime objects:
         timeout = header.get('timeout', None)
         if timeout:
             # cast to float, because jsonlib returns floats as decimal.Decimal,
             # which timedelta does not accept
             timeout = datetime.now() + timedelta(0,float(timeout),0)
 
         job = Job(msg_id=msg_id, raw_msg=raw_msg, idents=idents, msg=msg,
                  header=header, targets=targets, after=after, follow=follow,
                  timeout=timeout,
         )
 
         # validate and reduce dependencies:
         for dep in after,follow:
             if not dep: # empty dependency
                 continue
             # check valid:
             if msg_id in dep or dep.difference(self.all_ids):
                 self.depending[msg_id] = job
                 return self.fail_unreachable(msg_id, error.InvalidDependency)
             # check if unreachable:
             if dep.unreachable(self.all_completed, self.all_failed):
                 self.depending[msg_id] = job
                 return self.fail_unreachable(msg_id)
 
         if after.check(self.all_completed, self.all_failed):
             # time deps already met, try to run
             if not self.maybe_run(job):
                 # can't run yet
                 if msg_id not in self.all_failed:
                     # could have failed as unreachable
                     self.save_unmet(job)
         else:
             self.save_unmet(job)
 
     def audit_timeouts(self):
         """Audit all waiting tasks for expired timeouts."""
         now = datetime.now()
         for msg_id in self.depending.keys():
             # must recheck, in case one failure cascaded to another:
             if msg_id in self.depending:
                 job = self.depending[msg_id]
                 if job.timeout and job.timeout < now:
                     self.fail_unreachable(msg_id, error.TaskTimeout)
 
     def fail_unreachable(self, msg_id, why=error.ImpossibleDependency):
         """a task has become unreachable, send a reply with an ImpossibleDependency
         error."""
         if msg_id not in self.depending:
             self.log.error("msg %r already failed!", msg_id)
             return
         job = self.depending.pop(msg_id)
         for mid in job.dependents:
             if mid in self.graph:
                 self.graph[mid].remove(msg_id)
 
         try:
             raise why()
         except:
             content = error.wrap_exception()
 
         self.all_done.add(msg_id)
         self.all_failed.add(msg_id)
 
         msg = self.session.send(self.client_stream, 'apply_reply', content,
                                                 parent=job.header, ident=job.idents)
         self.session.send(self.mon_stream, msg, ident=[b'outtask']+job.idents)
 
         self.update_graph(msg_id, success=False)
 
     def maybe_run(self, job):
         """check location dependencies, and run if they are met."""
         msg_id = job.msg_id
         self.log.debug("Attempting to assign task %s", msg_id)
         if not self.targets:
             # no engines, definitely can't run
             return False
         
         if job.follow or job.targets or job.blacklist or self.hwm:
             # we need a can_run filter
             def can_run(idx):
                 # check hwm
                 if self.hwm and self.loads[idx] == self.hwm:
                     return False
                 target = self.targets[idx]
                 # check blacklist
                 if target in job.blacklist:
                     return False
                 # check targets
                 if job.targets and target not in job.targets:
                     return False
                 # check follow
                 return job.follow.check(self.completed[target], self.failed[target])
 
             indices = filter(can_run, range(len(self.targets)))
 
             if not indices:
                 # couldn't run
                 if job.follow.all:
                     # check follow for impossibility
                     dests = set()
                     relevant = set()
                     if job.follow.success:
                         relevant = self.all_completed
                     if job.follow.failure:
                         relevant = relevant.union(self.all_failed)
                     for m in job.follow.intersection(relevant):
                         dests.add(self.destinations[m])
                     if len(dests) > 1:
                         self.depending[msg_id] = job
                         self.fail_unreachable(msg_id)
                         return False
                 if job.targets:
                     # check blacklist+targets for impossibility
                     job.targets.difference_update(job.blacklist)
                     if not job.targets or not job.targets.intersection(self.targets):
                         self.depending[msg_id] = job
                         self.fail_unreachable(msg_id)
                         return False
                 return False
         else:
             indices = None
 
         self.submit_task(job, indices)
         return True
 
     def save_unmet(self, job):
         """Save a message for later submission when its dependencies are met."""
         msg_id = job.msg_id
         self.depending[msg_id] = job
         # track the ids in follow or after, but not those already finished
         for dep_id in job.after.union(job.follow).difference(self.all_done):
             if dep_id not in self.graph:
                 self.graph[dep_id] = set()
             self.graph[dep_id].add(msg_id)
 
     def submit_task(self, job, indices=None):
         """Submit a task to any of a subset of our targets."""
         if indices:
             loads = [self.loads[i] for i in indices]
         else:
             loads = self.loads
         idx = self.scheme(loads)
         if indices:
             idx = indices[idx]
         target = self.targets[idx]
         # print (target, map(str, msg[:3]))
         # send job to the engine
         self.engine_stream.send(target, flags=zmq.SNDMORE, copy=False)
         self.engine_stream.send_multipart(job.raw_msg, copy=False)
         # update load
         self.add_job(idx)
         self.pending[target][job.msg_id] = job
         # notify Hub
         content = dict(msg_id=job.msg_id, engine_id=target.decode('ascii'))
         self.session.send(self.mon_stream, 'task_destination', content=content,
                         ident=[b'tracktask',self.ident])
 
 
     #-----------------------------------------------------------------------
     # Result Handling
     #-----------------------------------------------------------------------
     
     
     @util.log_errors
     def dispatch_result(self, raw_msg):
         """dispatch method for result replies"""
         try:
             idents,msg = self.session.feed_identities(raw_msg, copy=False)
             msg = self.session.unserialize(msg, content=False, copy=False)
             engine = idents[0]
             try:
                 idx = self.targets.index(engine)
             except ValueError:
                 pass # skip load-update for dead engines
             else:
                 self.finish_job(idx)
         except Exception:
             self.log.error("task::Invaid result: %r", raw_msg, exc_info=True)
             return
 
         header = msg['header']
         parent = msg['parent_header']
         if header.get('dependencies_met', True):
             success = (header['status'] == 'ok')
             msg_id = parent['msg_id']
             retries = self.retries[msg_id]
             if not success and retries > 0:
                 # failed
                 self.retries[msg_id] = retries - 1
                 self.handle_unmet_dependency(idents, parent)
             else:
                 del self.retries[msg_id]
                 # relay to client and update graph
                 self.handle_result(idents, parent, raw_msg, success)
                 # send to Hub monitor
                 self.mon_stream.send_multipart([b'outtask']+raw_msg, copy=False)
         else:
             self.handle_unmet_dependency(idents, parent)
 
     def handle_result(self, idents, parent, raw_msg, success=True):
         """handle a real task result, either success or failure"""
         # first, relay result to client
         engine = idents[0]
         client = idents[1]
         # swap_ids for ROUTER-ROUTER mirror
         raw_msg[:2] = [client,engine]
         # print (map(str, raw_msg[:4]))
         self.client_stream.send_multipart(raw_msg, copy=False)
         # now, update our data structures
         msg_id = parent['msg_id']
         self.pending[engine].pop(msg_id)
         if success:
             self.completed[engine].add(msg_id)
             self.all_completed.add(msg_id)
         else:
             self.failed[engine].add(msg_id)
             self.all_failed.add(msg_id)
         self.all_done.add(msg_id)
         self.destinations[msg_id] = engine
 
         self.update_graph(msg_id, success)
 
     def handle_unmet_dependency(self, idents, parent):
         """handle an unmet dependency"""
         engine = idents[0]
         msg_id = parent['msg_id']
 
         job = self.pending[engine].pop(msg_id)
         job.blacklist.add(engine)
 
         if job.blacklist == job.targets:
             self.depending[msg_id] = job
             self.fail_unreachable(msg_id)
         elif not self.maybe_run(job):
             # resubmit failed
             if msg_id not in self.all_failed:
                 # put it back in our dependency tree
                 self.save_unmet(job)
 
         if self.hwm:
             try:
                 idx = self.targets.index(engine)
             except ValueError:
                 pass # skip load-update for dead engines
             else:
                 if self.loads[idx] == self.hwm-1:
                     self.update_graph(None)
 
 
 
     def update_graph(self, dep_id=None, success=True):
         """dep_id just finished. Update our dependency
         graph and submit any jobs that just became runable.
 
         Called with dep_id=None to update entire graph for hwm, but without finishing
         a task.
         """
         # print ("\n\n***********")
         # pprint (dep_id)
         # pprint (self.graph)
         # pprint (self.depending)
         # pprint (self.all_completed)
         # pprint (self.all_failed)
         # print ("\n\n***********\n\n")
         # update any jobs that depended on the dependency
         jobs = self.graph.pop(dep_id, [])
 
         # recheck *all* jobs if
         # a) we have HWM and an engine just become no longer full
         # or b) dep_id was given as None
         
         if dep_id is None or self.hwm and any( [ load==self.hwm-1 for load in self.loads ]):
             jobs = self.depending.keys()
         
         for msg_id in sorted(jobs, key=lambda msg_id: self.depending[msg_id].timestamp):
             job = self.depending[msg_id]
 
             if job.after.unreachable(self.all_completed, self.all_failed)\
                     or job.follow.unreachable(self.all_completed, self.all_failed):
                 self.fail_unreachable(msg_id)
 
             elif job.after.check(self.all_completed, self.all_failed): # time deps met, maybe run
                 if self.maybe_run(job):
 
                     self.depending.pop(msg_id)
                     for mid in job.dependents:
                         if mid in self.graph:
                             self.graph[mid].remove(msg_id)
 
     #----------------------------------------------------------------------
     # methods to be overridden by subclasses
     #----------------------------------------------------------------------
 
     def add_job(self, idx):
         """Called after self.targets[idx] just got the job with header.
         Override with subclasses.  The default ordering is simple LRU.
         The default loads are the number of outstanding jobs."""
         self.loads[idx] += 1
         for lis in (self.targets, self.loads):
             lis.append(lis.pop(idx))
 
 
     def finish_job(self, idx):
         """Called after self.targets[idx] just finished a job.
         Override with subclasses."""
         self.loads[idx] -= 1
 
 
 
-def launch_scheduler(in_addr, out_addr, mon_addr, not_addr, config=None,
+def launch_scheduler(in_addr, out_addr, mon_addr, not_addr, reg_addr, config=None,
                         logname='root', log_url=None, loglevel=logging.DEBUG,
                         identity=b'task', in_thread=False):
 
     ZMQStream = zmqstream.ZMQStream
 
     if config:
         # unwrap dict back into Config
         config = Config(config)
 
     if in_thread:
         # use instance() to get the same Context/Loop as our parent
         ctx = zmq.Context.instance()
         loop = ioloop.IOLoop.instance()
     else:
         # in a process, don't use instance()
         # for safety with multiprocessing
         ctx = zmq.Context()
         loop = ioloop.IOLoop()
     ins = ZMQStream(ctx.socket(zmq.ROUTER),loop)
-    ins.setsockopt(zmq.IDENTITY, identity)
+    ins.setsockopt(zmq.IDENTITY, identity+'_in')
     ins.bind(in_addr)
 
     outs = ZMQStream(ctx.socket(zmq.ROUTER),loop)
-    outs.setsockopt(zmq.IDENTITY, identity)
+    outs.setsockopt(zmq.IDENTITY, identity+'_out')
     outs.bind(out_addr)
     mons = zmqstream.ZMQStream(ctx.socket(zmq.PUB),loop)
     mons.connect(mon_addr)
     nots = zmqstream.ZMQStream(ctx.socket(zmq.SUB),loop)
     nots.setsockopt(zmq.SUBSCRIBE, b'')
     nots.connect(not_addr)
 
+    querys = ZMQStream(ctx.socket(zmq.DEALER),loop)
+    querys.connect(reg_addr)
+    
     # setup logging.
     if in_thread:
         log = Application.instance().log
     else:
         if log_url:
             log = connect_logger(logname, ctx, log_url, root="scheduler", loglevel=loglevel)
         else:
             log = local_logger(logname, loglevel)
 
     scheduler = TaskScheduler(client_stream=ins, engine_stream=outs,
                             mon_stream=mons, notifier_stream=nots,
+                            query_stream=querys,
                             loop=loop, log=log,
                             config=config)
     scheduler.start()
     if not in_thread:
         try:
             loop.start()
         except KeyboardInterrupt:
             scheduler.log.critical("Interrupted, exiting...")
 

IPython/parallel/engine/engine.py

@@ -1,231 +1,231 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, Integer, Type, CFloat, Unicode, CBytes, Bool
 )
 from IPython.utils.py3compat import cast_bytes
 
 from IPython.parallel.controller.heartmonitor import Heart
 from IPython.parallel.factory import RegistrationFactory
 from IPython.parallel.util import disambiguate_url
 
 from IPython.zmq.session import Message
 from IPython.zmq.ipkernel 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(5, 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:
     connection_info = Dict()
     user_ns = Dict()
     id = Integer(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 = cast_bytes(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 str(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)
+        content = dict(uuid=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 = self.session.unserialize(msg)
         content = msg['content']
         info = self.connection_info
         
         def url(key):
             """get zmq url for given channel"""
             return str(info["interface"] + ":%i" % info[key])
         
         if content['status'] == 'ok':
             self.id = int(content['id'])
 
             # launch heartbeat
             # possibly forward hb ports with tunnels
             hb_ping = maybe_tunnel(url('hb_ping'))
             hb_pong = maybe_tunnel(url('hb_pong'))
 
             heart = Heart(hb_ping, hb_pong, heart_id=identity)
             heart.start()
 
             # create Shell Connections (MUX, Task, etc.):
             shell_addrs = url('mux'), url('task')
 
             # 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)
 
             # control stream:
             control_addr = url('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 = url('iopub')
             iopub_socket = ctx.socket(zmq.PUB)
             iopub_socket.setsockopt(zmq.IDENTITY, identity)
             connect(iopub_socket, iopub_addr)
 
             # disable history:
             self.config.HistoryManager.hist_file = ':memory:'
             
             # Redirect input streams and set a display hook.
             if self.out_stream_factory:
                 sys.stdout = self.out_stream_factory(self.session, iopub_socket, u'stdout')
                 sys.stdout.topic = cast_bytes('engine.%i.stdout' % self.id)
                 sys.stderr = self.out_stream_factory(self.session, iopub_socket, u'stderr')
                 sys.stderr.topic = cast_bytes('engine.%i.stderr' % self.id)
             if self.display_hook_factory:
                 sys.displayhook = self.display_hook_factory(self.session, iopub_socket)
                 sys.displayhook.topic = cast_bytes('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_socket=iopub_socket,
                     loop=loop, user_ns=self.user_ns, log=self.log)
             self.kernel.shell.display_pub.topic = cast_bytes('engine.%i.displaypub' % self.id)
             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 self.url.startswith('127.'):
             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()
 

docs/source/development/parallel_messages.txt

@@ -1,378 +1,368 b''
 .. _parallel_messages:
 
 Messaging for Parallel Computing
 ================================
 
 This is an extension of the :ref:`messaging <messaging>` doc. Diagrams of the connections
 can be found in the :ref:`parallel connections <parallel_connections>` doc.
 
 
 ZMQ messaging is also used in the parallel computing IPython system. All messages to/from
 kernels remain the same as the single kernel model, and are forwarded through a ZMQ Queue
 device. The controller receives all messages and replies in these channels, and saves
 results for future use.
 
 The Controller
 --------------
 
 The controller is the central collection of processes in the IPython parallel computing
 model. It has two major components:
 
     * The Hub
     * A collection of Schedulers
 
 The Hub
 -------
 
 The Hub is the central process for monitoring the state of the engines, and all task
 requests and results.  It has no role in execution and does no relay of messages, so
 large blocking requests or database actions in the Hub do not have the ability to impede
 job submission and results.
 
 Registration (``ROUTER``)
 ***********************
 
 The first function of the Hub is to facilitate and monitor connections of clients
 and engines. Both client and engine registration are handled by the same socket, so only
 one ip/port pair is needed to connect any number of connections and clients.
 
 Engines register with the ``zmq.IDENTITY`` of their two ``DEALER`` sockets, one for the
 queue, which receives execute requests, and one for the heartbeat, which is used to
 monitor the survival of the Engine process.
 
 Message type: ``registration_request``::
 
     content = {
-        'queue'   : 'abcd-1234-...', # the MUX queue zmq.IDENTITY
-        'control'   : 'abcd-1234-...', # the control queue zmq.IDENTITY
-        'heartbeat' : 'abcd-1234-...' # the heartbeat zmq.IDENTITY
+        'uuid'   : 'abcd-1234-...', # the zmq.IDENTITY of the engine's sockets
     }
 
 .. note::
 
     these are always the same, at least for now.
 
 The Controller replies to an Engine's registration request with the engine's integer ID,
 and all the remaining connection information for connecting the heartbeat process, and
 kernel queue socket(s). The message status will be an error if the Engine requests IDs that
 already in use.
 
 Message type: ``registration_reply``::
 
     content = {
         'status' : 'ok', # or 'error'
         # if ok:
         'id' : 0, # int, the engine id
-        'queue' : 'tcp://127.0.0.1:12345', # connection for engine side of the queue
-        'control' : 'tcp://...', # addr for control queue
-        'heartbeat' : ('tcp://...','tcp://...'), # tuple containing two interfaces needed for heartbeat
-        'task' : 'tcp://...', # addr for task queue, or None if no task queue running
     }
 
 Clients use the same socket as engines to start their connections. Connection requests
 from clients need no information:
 
 Message type: ``connection_request``::
     
     content = {}
 
 The reply to a Client registration request contains the connection information for the
 multiplexer and load balanced queues, as well as the address for direct hub
 queries. If any of these addresses is `None`, that functionality is not available.
 
 Message type: ``connection_reply``::
 
     content = {
         'status' : 'ok', # or 'error'
-        # if ok:
-        'queue' : 'tcp://127.0.0.1:12345', # connection for client side of the MUX queue
-        'task' : ('lru','tcp...'), # routing scheme and addr for task queue (len 2 tuple)
-        'query' : 'tcp...', # addr for methods to query the hub, like queue_request, etc.
-        'control' : 'tcp...', # addr for control methods, like abort, etc.
     }
 
 Heartbeat
 *********
 
 The hub uses a heartbeat system to monitor engines, and track when they become
 unresponsive. As described in :ref:`messaging <messaging>`, and shown in :ref:`connections
 <parallel_connections>`.
 
 Notification (``PUB``)
 **********************
 
 The hub publishes all engine registration/unregistration events on a ``PUB`` socket.
 This allows clients to have up-to-date engine ID sets without polling. Registration
 notifications contain both the integer engine ID and the queue ID, which is necessary for
 sending messages via the Multiplexer Queue and Control Queues.
 
 Message type: ``registration_notification``::
 
     content = {
         'id' : 0, # engine ID that has been registered
-        'queue' : 'engine_id' # the IDENT for the engine's queue
+        'uuid' : 'engine_id' # the IDENT for the engine's sockets
     }
 
 Message type : ``unregistration_notification``::
 
     content = {
         'id' : 0 # engine ID that has been unregistered
+        'uuid' : 'engine_id' # the IDENT for the engine's sockets
     }
 
 
 Client Queries (``ROUTER``)
 *************************
 
 The hub monitors and logs all queue traffic, so that clients can retrieve past
 results or monitor pending tasks. This information may reside in-memory on the Hub, or
 on disk in a database (SQLite and MongoDB are currently supported).  These requests are
 handled by the same socket as registration.
 
 
 :func:`queue_request` requests can specify multiple engines to query via the `targets`
 element. A verbose flag can be passed, to determine whether the result should be the list
 of `msg_ids` in the queue or simply the length of each list.
 
 Message type: ``queue_request``::
 
     content = {
         'verbose' : True, # whether return should be lists themselves or just lens
         'targets' : [0,3,1] # list of ints
     }
 
 The content of a reply to a :func:`queue_request` request is a dict, keyed by the engine
 IDs. Note that they will be the string representation of the integer keys, since JSON
 cannot handle number keys.  The three keys of each dict are::
 
     'completed' :  messages submitted via any queue that ran on the engine
     'queue' : jobs submitted via MUX queue, whose results have not been received
     'tasks' : tasks that are known to have been submitted to the engine, but 
                 have not completed.  Note that with the pure zmq scheduler, this will
                 always be 0/[].
 
 Message type: ``queue_reply``::
 
     content = {
         'status' : 'ok', # or 'error'
         # if verbose=False:
         '0' : {'completed' : 1, 'queue' : 7, 'tasks' : 0},
         # if verbose=True:
         '1' : {'completed' : ['abcd-...','1234-...'], 'queue' : ['58008-'], 'tasks' : []},
     }
 
 Clients can request individual results directly from the hub. This is primarily for
 gathering results of executions not submitted by the requesting client, as the client
 will have all its own results already. Requests are made by msg_id, and can contain one or
 more msg_id. An additional boolean key 'statusonly' can be used to not request the
 results, but simply poll the status of the jobs.
 
 Message type: ``result_request``::
 
     content = {
         'msg_ids' : ['uuid','...'], # list of strs
         'targets' : [1,2,3], # list of int ids or uuids
         'statusonly' : False, # bool
     }
 
 The :func:`result_request` reply contains the content objects of the actual execution
 reply messages. If `statusonly=True`, then there will be only the 'pending' and
 'completed' lists.  
 
 
 Message type: ``result_reply``::
 
     content = {
         'status' : 'ok', # else error
         # if ok:
         'acbd-...' : msg, # the content dict is keyed by msg_ids,
                          # values are the result messages
                         # there will be none of these if `statusonly=True`
         'pending' : ['msg_id','...'], # msg_ids still pending
         'completed' : ['msg_id','...'], # list of completed msg_ids
     }
     buffers = ['bufs','...'] # the buffers that contained the results of the objects.
                             # this will be empty if no messages are complete, or if 
                             # statusonly is True.
 
 For memory management purposes, Clients can also instruct the hub to forget the
 results of messages. This can be done by message ID or engine ID. Individual messages are
 dropped by msg_id, and all messages completed on an engine are dropped by engine ID. This
 may no longer be necessary with the mongodb-based message logging backend.
 
 If the msg_ids element is the string ``'all'`` instead of a list, then all completed
 results are forgotten.
 
 Message type: ``purge_request``::
 
     content = {
         'msg_ids' : ['id1', 'id2',...], # list of msg_ids or 'all'
         'engine_ids' : [0,2,4] # list of engine IDs
     }
 
 The reply to a purge request is simply the status 'ok' if the request succeeded, or an
 explanation of why it failed, such as requesting the purge of a nonexistent or pending
 message.
 
 Message type: ``purge_reply``::
 
     content = {
         'status' : 'ok', # or 'error'
     }
 
 
 Schedulers
 ----------
 
 There are three basic schedulers:
 
   * Task Scheduler
   * MUX Scheduler
   * Control Scheduler
 
 The MUX and Control schedulers are simple MonitoredQueue ØMQ devices, with ``ROUTER``
 sockets on either side. This allows the queue to relay individual messages to particular
 targets via ``zmq.IDENTITY`` routing. The Task scheduler may be a MonitoredQueue ØMQ
 device, in which case the client-facing socket is ``ROUTER``, and the engine-facing socket
 is ``DEALER``.  The result of this is that client-submitted messages are load-balanced via 
 the ``DEALER`` socket, but the engine's replies to each message go to the requesting client.
 
 Raw ``DEALER`` scheduling is quite primitive, and doesn't allow message introspection, so
 there are also Python Schedulers that can be used. These Schedulers behave in much the
 same way as a MonitoredQueue does from the outside, but have rich internal logic to
 determine destinations, as well as handle dependency graphs Their sockets are always
 ``ROUTER`` on both sides.
 
 The Python task schedulers have an additional message type, which informs the Hub of
 the destination of a task as soon as that destination is known.
 
 Message type: ``task_destination``::
 
     content = {
         'msg_id' : 'abcd-1234-...', # the msg's uuid
         'engine_id' : '1234-abcd-...', # the destination engine's zmq.IDENTITY
     }
 
 :func:`apply` and :func:`apply_bound`
 *************************************
 
 In terms of message classes, the MUX scheduler and Task scheduler relay the exact same
 message types.  Their only difference lies in how the destination is selected.
 
 The `Namespace <http://gist.github.com/483294>`_ model suggests that execution be able to
 use the model::
 
     ns.apply(f, *args, **kwargs)
     
 which takes `f`, a function in the user's namespace, and executes ``f(*args, **kwargs)``
 on a remote engine, returning the result (or, for non-blocking, information facilitating
 later retrieval of the result). This model, unlike the execute message which just uses a
 code string, must be able to send arbitrary (pickleable) Python objects. And ideally, copy
 as little data as we can. The `buffers` property of a Message was introduced for this
 purpose.
 
 Utility method :func:`build_apply_message` in :mod:`IPython.zmq.streamsession` wraps a
 function signature and builds a sendable buffer format for minimal data copying (exactly
 zero copies of numpy array data or buffers or large strings).
 
 Message type: ``apply_request``::
 
     content = {
         'bound' : True, # whether to execute in the engine's namespace or unbound
         'after' : ['msg_id',...], # list of msg_ids or output of Dependency.as_dict()
         'follow' : ['msg_id',...], # list of msg_ids or output of Dependency.as_dict()
         
     }
     buffers = ['...'] # at least 3 in length
                     # as built by build_apply_message(f,args,kwargs)
 
 after/follow represent task dependencies. 'after' corresponds to a time dependency. The
 request will not arrive at an engine until the 'after' dependency tasks have completed.
 'follow' corresponds to a location dependency. The task will be submitted to the same
 engine as these msg_ids (see :class:`Dependency` docs for details).
 
 Message type: ``apply_reply``::
 
     content = {
         'status' : 'ok' # 'ok' or 'error'
         # other error info here, as in other messages
     }
     buffers = ['...'] # either 1 or 2 in length
                     # a serialization of the return value of f(*args,**kwargs)
                     # only populated if status is 'ok'
 
 All engine execution and data movement is performed via apply messages.
 
 Control Messages
 ----------------
 
 Messages that interact with the engines, but are not meant to execute code, are submitted
 via the Control queue. These messages have high priority, and are thus received and
 handled before any execution requests.
 
 Clients may want to clear the namespace on the engine. There are no arguments nor
 information involved in this request, so the content is empty.
 
 Message type: ``clear_request``::
 
     content = {}
 
 Message type: ``clear_reply``::
 
     content = {
         'status' : 'ok' # 'ok' or 'error'
         # other error info here, as in other messages
     }
 
 Clients may want to abort tasks that have not yet run. This can by done by message id, or
 all enqueued messages can be aborted if None is specified.
 
 Message type: ``abort_request``::
 
     content = {
         'msg_ids' : ['1234-...', '...'] # list of msg_ids or None
     }
 
 Message type: ``abort_reply``::
 
     content = {
         'status' : 'ok' # 'ok' or 'error'
         # other error info here, as in other messages
     }
 
 The last action a client may want to do is shutdown the kernel. If a kernel receives a
 shutdown request, then it aborts all queued messages, replies to the request, and exits.
 
 Message type: ``shutdown_request``::
 
     content = {}
 
 Message type: ``shutdown_reply``::
 
     content = {
         'status' : 'ok' # 'ok' or 'error'
         # other error info here, as in other messages
     }
 
 
 Implementation
 --------------
 
 There are a few differences in implementation between the `StreamSession` object used in
 the newparallel branch and the `Session` object, the main one being that messages are
 sent in parts, rather than as a single serialized object. `StreamSession` objects also
 take pack/unpack functions, which are to be used when serializing/deserializing objects.
 These can be any functions that translate to/from formats that ZMQ sockets can send
 (buffers,bytes, etc.).
 
 Split Sends
 ***********
 
 Previously, messages were bundled as a single json object and one call to
 :func:`socket.send_json`. Since the hub inspects all messages, and doesn't need to
 see the content of the messages, which can be large, messages are now serialized and sent in
 pieces. All messages are sent in at least 3 parts: the header, the parent header, and the
 content. This allows the controller to unpack and inspect the (always small) header,
 without spending time unpacking the content unless the message is bound for the
 controller. Buffers are added on to the end of the message, and can be any objects that
 present the buffer interface.