adapt kernel's ipcluster and Launchers to newparallel

IPython/zmq/parallel/ipclusterapp.py

@@ -0,0 +1,502
+#!/usr/bin/env python
+# encoding: utf-8
+"""
+The ipcluster application.
+"""
+
+#-----------------------------------------------------------------------------
+#  Copyright (C) 2008-2009  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 logging
+import os
+import signal
+import logging
+
+from zmq.eventloop import ioloop
+
+from IPython.external.argparse import ArgumentParser, SUPPRESS
+from IPython.utils.importstring import import_item
+from IPython.zmq.parallel.clusterdir import (
+    ApplicationWithClusterDir, ClusterDirConfigLoader,
+    ClusterDirError, PIDFileError
+)
+
+
+#-----------------------------------------------------------------------------
+# Module level variables
+#-----------------------------------------------------------------------------
+
+
+default_config_file_name = u'ipcluster_config.py'
+
+
+_description = """\
+Start an IPython cluster for parallel computing.\n\n
+
+An IPython cluster consists of 1 controller and 1 or more engines.
+This command automates the startup of these processes using a wide
+range of startup methods (SSH, local processes, PBS, mpiexec,
+Windows HPC Server 2008). To start a cluster with 4 engines on your
+local host simply do 'ipcluster start -n 4'. For more complex usage 
+you will typically do 'ipcluster create -p mycluster', then edit
+configuration files, followed by 'ipcluster start -p mycluster -n 4'.
+"""
+
+
+# Exit codes for ipcluster
+
+# This will be the exit code if the ipcluster appears to be running because
+# a .pid file exists
+ALREADY_STARTED = 10
+
+
+# This will be the exit code if ipcluster stop is run, but there is not .pid
+# file to be found.
+ALREADY_STOPPED = 11
+
+
+#-----------------------------------------------------------------------------
+# Command line options
+#-----------------------------------------------------------------------------
+
+
+class IPClusterAppConfigLoader(ClusterDirConfigLoader):
+
+    def _add_arguments(self):
+        # Don't call ClusterDirConfigLoader._add_arguments as we don't want
+        # its defaults on self.parser. Instead, we will put those on
+        # default options on our subparsers.
+        
+        # This has all the common options that all subcommands use
+        parent_parser1 = ArgumentParser(
+            add_help=False,
+            argument_default=SUPPRESS
+        )
+        self._add_ipython_dir(parent_parser1)
+        self._add_log_level(parent_parser1)
+
+        # This has all the common options that other subcommands use
+        parent_parser2 = ArgumentParser(
+            add_help=False,
+            argument_default=SUPPRESS
+        )
+        self._add_cluster_profile(parent_parser2)
+        self._add_cluster_dir(parent_parser2)
+        self._add_work_dir(parent_parser2)
+        paa = parent_parser2.add_argument
+        paa('--log-to-file',
+            action='store_true', dest='Global.log_to_file', 
+            help='Log to a file in the log directory (default is stdout)')
+
+        # Create the object used to create the subparsers.
+        subparsers = self.parser.add_subparsers(
+            dest='Global.subcommand',
+            title='ipcluster subcommands',
+            description=
+            """ipcluster has a variety of subcommands. The general way of 
+            running ipcluster is 'ipcluster <cmd> [options]'. To get help
+            on a particular subcommand do 'ipcluster <cmd> -h'."""
+            # help="For more help, type 'ipcluster <cmd> -h'",
+        )
+
+        # The "list" subcommand parser
+        parser_list = subparsers.add_parser(
+            'list',
+            parents=[parent_parser1],
+            argument_default=SUPPRESS,
+            help="List all clusters in cwd and ipython_dir.",
+            description=
+            """List all available clusters, by cluster directory, that can
+            be found in the current working directly or in the ipython
+            directory. Cluster directories are named using the convention
+            'cluster_<profile>'."""
+        )
+
+        # The "create" subcommand parser
+        parser_create = subparsers.add_parser(
+            'create',
+            parents=[parent_parser1, parent_parser2],
+            argument_default=SUPPRESS,
+            help="Create a new cluster directory.",
+            description=
+            """Create an ipython cluster directory by its profile name or 
+            cluster directory path. Cluster directories contain 
+            configuration, log and security related files and are named 
+            using the convention 'cluster_<profile>'. By default they are
+            located in your ipython directory. Once created, you will
+            probably need to edit the configuration files in the cluster
+            directory to configure your cluster. Most users will create a
+            cluster directory by profile name, 
+            'ipcluster create -p mycluster', which will put the directory
+            in '<ipython_dir>/cluster_mycluster'. 
+            """
+        )
+        paa = parser_create.add_argument
+        paa('--reset-config',
+            dest='Global.reset_config', action='store_true',
+            help=
+            """Recopy the default config files to the cluster directory.
+            You will loose any modifications you have made to these files.""")
+
+        # The "start" subcommand parser
+        parser_start = subparsers.add_parser(
+            'start',
+            parents=[parent_parser1, parent_parser2],
+            argument_default=SUPPRESS,
+            help="Start a cluster.",
+            description=
+            """Start an ipython cluster by its profile name or cluster 
+            directory. Cluster directories contain configuration, log and
+            security related files and are named using the convention
+            'cluster_<profile>' and should be creating using the 'start'
+            subcommand of 'ipcluster'. If your cluster directory is in 
+            the cwd or the ipython directory, you can simply refer to it
+            using its profile name, 'ipcluster start -n 4 -p <profile>`,
+            otherwise use the '--cluster-dir' option.
+            """
+        )
+        paa = parser_start.add_argument
+        paa('-n', '--number',
+            type=int, dest='Global.n',
+            help='The number of engines to start.',
+            metavar='Global.n')
+        paa('--clean-logs',
+            dest='Global.clean_logs', action='store_true',
+            help='Delete old log flies before starting.')
+        paa('--no-clean-logs',
+            dest='Global.clean_logs', action='store_false',
+            help="Don't delete old log flies before starting.")
+        paa('--daemon',
+            dest='Global.daemonize', action='store_true',
+            help='Daemonize the ipcluster program. This implies --log-to-file')
+        paa('--no-daemon',
+            dest='Global.daemonize', action='store_false',
+            help="Dont't daemonize the ipcluster program.")
+
+        # The "stop" subcommand parser
+        parser_stop = subparsers.add_parser(
+            'stop',
+            parents=[parent_parser1, parent_parser2],
+            argument_default=SUPPRESS,
+            help="Stop a running cluster.",
+            description=
+            """Stop a running ipython cluster by its profile name or cluster 
+            directory. Cluster directories are named using the convention
+            'cluster_<profile>'. If your cluster directory is in 
+            the cwd or the ipython directory, you can simply refer to it
+            using its profile name, 'ipcluster stop -p <profile>`, otherwise
+            use the '--cluster-dir' option.
+            """
+        )
+        paa = parser_stop.add_argument
+        paa('--signal',
+            dest='Global.signal', type=int,
+            help="The signal number to use in stopping the cluster (default=2).",
+            metavar="Global.signal")
+
+
+#-----------------------------------------------------------------------------
+# Main application
+#-----------------------------------------------------------------------------
+
+
+class IPClusterApp(ApplicationWithClusterDir):
+
+    name = u'ipclusterz'
+    description = _description
+    usage = None
+    command_line_loader = IPClusterAppConfigLoader
+    default_config_file_name = default_config_file_name
+    default_log_level = logging.INFO
+    auto_create_cluster_dir = False
+
+    def create_default_config(self):
+        super(IPClusterApp, self).create_default_config()
+        self.default_config.Global.controller_launcher = \
+            'IPython.zmq.parallel.launcher.LocalControllerLauncher'
+        self.default_config.Global.engine_launcher = \
+            'IPython.zmq.parallel.launcher.LocalEngineSetLauncher'
+        self.default_config.Global.n = 2
+        self.default_config.Global.reset_config = False
+        self.default_config.Global.clean_logs = True
+        self.default_config.Global.signal = 2
+        self.default_config.Global.daemonize = False
+
+    def find_resources(self):
+        subcommand = self.command_line_config.Global.subcommand
+        if subcommand=='list':
+            self.list_cluster_dirs()
+            # Exit immediately because there is nothing left to do.
+            self.exit()
+        elif subcommand=='create':
+            self.auto_create_cluster_dir = True
+            super(IPClusterApp, self).find_resources()
+        elif subcommand=='start' or subcommand=='stop':
+            self.auto_create_cluster_dir = True
+            try:
+                super(IPClusterApp, self).find_resources()
+            except ClusterDirError:
+                raise ClusterDirError(
+                    "Could not find a cluster directory. A cluster dir must "
+                    "be created before running 'ipcluster start'.  Do "
+                    "'ipcluster create -h' or 'ipcluster list -h' for more "
+                    "information about creating and listing cluster dirs."
+                )
+
+    def list_cluster_dirs(self):
+        # Find the search paths
+        cluster_dir_paths = os.environ.get('IPCLUSTER_DIR_PATH','')
+        if cluster_dir_paths:
+            cluster_dir_paths = cluster_dir_paths.split(':')
+        else:
+            cluster_dir_paths = []
+        try:
+            ipython_dir = self.command_line_config.Global.ipython_dir
+        except AttributeError:
+            ipython_dir = self.default_config.Global.ipython_dir
+        paths = [os.getcwd(), ipython_dir] + \
+            cluster_dir_paths
+        paths = list(set(paths))
+
+        self.log.info('Searching for cluster dirs in paths: %r' % paths)
+        for path in paths:
+            files = os.listdir(path)
+            for f in files:
+                full_path = os.path.join(path, f)
+                if os.path.isdir(full_path) and f.startswith('cluster_'):
+                    profile = full_path.split('_')[-1]
+                    start_cmd = 'ipcluster start -p %s -n 4' % profile
+                    print start_cmd + " ==> " + full_path
+
+    def pre_construct(self):
+        # IPClusterApp.pre_construct() is where we cd to the working directory.
+        super(IPClusterApp, self).pre_construct()
+        config = self.master_config
+        try:
+            daemon = config.Global.daemonize
+            if daemon:
+                config.Global.log_to_file = True
+        except AttributeError:
+            pass
+
+    def construct(self):
+        config = self.master_config
+        subcmd = config.Global.subcommand
+        reset = config.Global.reset_config
+        if subcmd == 'list':
+            return
+        if subcmd == 'create':
+            self.log.info('Copying default config files to cluster directory '
+            '[overwrite=%r]' % (reset,))
+            self.cluster_dir_obj.copy_all_config_files(overwrite=reset)
+        if subcmd =='start':
+            self.cluster_dir_obj.copy_all_config_files(overwrite=False)
+            self.start_logging()
+            self.loop = ioloop.IOLoop.instance()
+            # reactor.callWhenRunning(self.start_launchers)
+            dc = ioloop.DelayedCallback(self.start_launchers, 0, self.loop)
+            dc.start()
+
+    def start_launchers(self):
+        config = self.master_config
+
+        # Create the launchers. In both bases, we set the work_dir of
+        # the launcher to the cluster_dir. This is where the launcher's
+        # subprocesses will be launched. It is not where the controller
+        # and engine will be launched.
+        el_class = import_item(config.Global.engine_launcher)
+        self.engine_launcher = el_class(
+            work_dir=self.cluster_dir, config=config
+        )
+        cl_class = import_item(config.Global.controller_launcher)
+        self.controller_launcher = cl_class(
+            work_dir=self.cluster_dir, config=config
+        )
+
+        # Setup signals
+        signal.signal(signal.SIGINT, self.sigint_handler)
+
+        # Setup the observing of stopping. If the controller dies, shut
+        # everything down as that will be completely fatal for the engines.
+        self.controller_launcher.on_stop(self.stop_launchers)
+        # d1.addCallback(self.stop_launchers)
+        # But, we don't monitor the stopping of engines. An engine dying
+        # is just fine and in principle a user could start a new engine.
+        # Also, if we did monitor engine stopping, it is difficult to
+        # know what to do when only some engines die. Currently, the
+        # observing of engine stopping is inconsistent. Some launchers
+        # might trigger on a single engine stopping, other wait until
+        # all stop.  TODO: think more about how to handle this.
+        
+        # Start the controller and engines
+        self._stopping = False  # Make sure stop_launchers is not called 2x.
+        d = self.start_controller()
+        self.start_engines()
+        self.startup_message()
+        # d.addCallback(self.start_engines)
+        # d.addCallback(self.startup_message)
+        # If the controller or engines fail to start, stop everything
+        # d.addErrback(self.stop_launchers)
+        return d
+
+    def startup_message(self, r=None):
+        logging.info("IPython cluster: started")
+        return r
+        
+    def start_controller(self, r=None):
+        # logging.info("In start_controller")
+        config = self.master_config
+        d = self.controller_launcher.start(
+            cluster_dir=config.Global.cluster_dir
+        )
+        return d
+            
+    def start_engines(self, r=None):
+        # logging.info("In start_engines")
+        config = self.master_config
+        d = self.engine_launcher.start(
+            config.Global.n,
+            cluster_dir=config.Global.cluster_dir
+        )
+        return d
+
+    def stop_controller(self, r=None):
+        # logging.info("In stop_controller")
+        if self.controller_launcher.running:
+            return self.controller_launcher.stop()
+
+    def stop_engines(self, r=None):
+        # logging.info("In stop_engines")
+        if self.engine_launcher.running:
+            d = self.engine_launcher.stop()
+            # d.addErrback(self.log_err)
+            return d
+        else:
+            return None
+
+    def log_err(self, f):
+        logging.error(f.getTraceback())
+        return None
+        
+    def stop_launchers(self, r=None):
+        if not self._stopping:
+            self._stopping = True
+            # if isinstance(r, failure.Failure):
+            #     logging.error('Unexpected error in ipcluster:')
+            #     logging.info(r.getTraceback())
+            logging.error("IPython cluster: stopping")
+            # These return deferreds. We are not doing anything with them
+            # but we are holding refs to them as a reminder that they 
+            # do return deferreds.
+            d1 = self.stop_engines()
+            d2 = self.stop_controller()
+            # Wait a few seconds to let things shut down.
+            dc = ioloop.DelayedCallback(self.loop.stop, 4000, self.loop)
+            dc.start()
+            # reactor.callLater(4.0, reactor.stop)
+
+    def sigint_handler(self, signum, frame):
+        self.stop_launchers()
+        
+    def start_logging(self):
+        # Remove old log files of the controller and engine
+        if self.master_config.Global.clean_logs:
+            log_dir = self.master_config.Global.log_dir
+            for f in os.listdir(log_dir):
+                if f.startswith('ipengine' + '-'):
+                    if f.endswith('.log') or f.endswith('.out') or f.endswith('.err'):
+                        os.remove(os.path.join(log_dir, f))
+                if f.startswith('ipcontroller' + '-'):
+                    if f.endswith('.log') or f.endswith('.out') or f.endswith('.err'):
+                        os.remove(os.path.join(log_dir, f))
+        # This will remote old log files for ipcluster itself
+        super(IPClusterApp, self).start_logging()
+
+    def start_app(self):
+        """Start the application, depending on what subcommand is used."""
+        subcmd = self.master_config.Global.subcommand
+        if subcmd=='create' or subcmd=='list':
+            return
+        elif subcmd=='start':
+            self.start_app_start()
+        elif subcmd=='stop':
+            self.start_app_stop()
+
+    def start_app_start(self):
+        """Start the app for the start subcommand."""
+        config = self.master_config
+        # First see if the cluster is already running
+        try:
+            pid = self.get_pid_from_file()
+        except PIDFileError:
+            pass
+        else:
+            self.log.critical(
+                'Cluster is already running with [pid=%s]. '
+                'use "ipcluster stop" to stop the cluster.' % pid
+            )
+            # Here I exit with a unusual exit status that other processes
+            # can watch for to learn how I existed.
+            self.exit(ALREADY_STARTED)
+
+        # Now log and daemonize
+        self.log.info(
+            'Starting ipcluster with [daemon=%r]' % config.Global.daemonize
+        )
+        # TODO: Get daemonize working on Windows or as a Windows Server.
+        if config.Global.daemonize:
+            if os.name=='posix':
+                from twisted.scripts._twistd_unix import daemonize
+                daemonize()
+
+        # Now write the new pid file AFTER our new forked pid is active.
+        self.write_pid_file()
+        try:
+            self.loop.start()
+        except:
+            logging.info("stopping...")
+        self.remove_pid_file()
+
+    def start_app_stop(self):
+        """Start the app for the stop subcommand."""
+        config = self.master_config
+        try:
+            pid = self.get_pid_from_file()
+        except PIDFileError:
+            self.log.critical(
+                'Problem reading pid file, cluster is probably not running.'
+            )
+            # Here I exit with a unusual exit status that other processes
+            # can watch for to learn how I existed.
+            self.exit(ALREADY_STOPPED)
+        else:
+            if os.name=='posix':
+                sig = config.Global.signal
+                self.log.info(
+                    "Stopping cluster [pid=%r] with [signal=%r]" % (pid, sig)
+                )
+                os.kill(pid, sig)
+            elif os.name=='nt':
+                # As of right now, we don't support daemonize on Windows, so
+                # stop will not do anything. Minimally, it should clean up the
+                # old .pid files.
+                self.remove_pid_file()
+
+
+def launch_new_instance():
+    """Create and run the IPython cluster."""
+    app = IPClusterApp()
+    app.start()
+
+
+if __name__ == '__main__':
+    launch_new_instance()
+

IPython/zmq/parallel/launcher.py

@@ -0,0 +1,824
+#!/usr/bin/env python
+# encoding: utf-8
+"""
+Facilities for launching IPython processes asynchronously.
+"""
+
+#-----------------------------------------------------------------------------
+#  Copyright (C) 2008-2009  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 re
+import sys
+import logging
+
+from signal import SIGINT
+try:
+    from signal import SIGKILL
+except ImportError:
+    SIGKILL=SIGTERM
+
+from subprocess import Popen, PIPE
+
+from zmq.eventloop import ioloop
+
+from IPython.config.configurable import Configurable
+from IPython.utils.traitlets import Str, Int, List, Unicode, Instance
+from IPython.utils.path import get_ipython_module_path
+from IPython.utils.process import find_cmd, pycmd2argv, FindCmdError
+
+# from IPython.kernel.winhpcjob import (
+#     IPControllerTask, IPEngineTask,
+#     IPControllerJob, IPEngineSetJob
+# )
+
+
+#-----------------------------------------------------------------------------
+# Paths to the kernel apps
+#-----------------------------------------------------------------------------
+
+
+ipcluster_cmd_argv = pycmd2argv(get_ipython_module_path(
+    'IPython.zmq.parallel.ipclusterapp'
+))
+
+ipengine_cmd_argv = pycmd2argv(get_ipython_module_path(
+    'IPython.zmq.parallel.ipengineapp'
+))
+
+ipcontroller_cmd_argv = pycmd2argv(get_ipython_module_path(
+    'IPython.zmq.parallel.ipcontrollerapp'
+))
+
+#-----------------------------------------------------------------------------
+# Base launchers and errors
+#-----------------------------------------------------------------------------
+
+
+class LauncherError(Exception):
+    pass
+
+
+class ProcessStateError(LauncherError):
+    pass
+
+
+class UnknownStatus(LauncherError):
+    pass
+
+
+class BaseLauncher(Configurable):
+    """An asbtraction for starting, stopping and signaling a process."""
+
+    # In all of the launchers, the work_dir is where child processes will be
+    # run. This will usually be the cluster_dir, but may not be. any work_dir
+    # passed into the __init__ method will override the config value.
+    # This should not be used to set the work_dir for the actual engine
+    # and controller. Instead, use their own config files or the
+    # controller_args, engine_args attributes of the launchers to add
+    # the --work-dir option.
+    work_dir = Unicode(u'.')
+    loop = Instance('zmq.eventloop.ioloop.IOLoop')
+    def _loop_default(self):
+        return ioloop.IOLoop.instance()
+
+    def __init__(self, work_dir=u'.', config=None):
+        super(BaseLauncher, self).__init__(work_dir=work_dir, config=config)
+        self.state = 'before' # can be before, running, after
+        self.stop_callbacks = []
+        self.start_data = None
+        self.stop_data = None
+
+    @property
+    def args(self):
+        """A list of cmd and args that will be used to start the process.
+
+        This is what is passed to :func:`spawnProcess` and the first element
+        will be the process name.
+        """
+        return self.find_args()
+
+    def find_args(self):
+        """The ``.args`` property calls this to find the args list.
+
+        Subcommand should implement this to construct the cmd and args.
+        """
+        raise NotImplementedError('find_args must be implemented in a subclass')
+
+    @property
+    def arg_str(self):
+        """The string form of the program arguments."""
+        return ' '.join(self.args)
+
+    @property
+    def running(self):
+        """Am I running."""
+        if self.state == 'running':
+            return True
+        else:
+            return False
+
+    def start(self):
+        """Start the process.
+
+        This must return a deferred that fires with information about the
+        process starting (like a pid, job id, etc.).
+        """
+        raise NotImplementedError('start must be implemented in a subclass')
+
+    def stop(self):
+        """Stop the process and notify observers of stopping.
+
+        This must return a deferred that fires with information about the
+        processing stopping, like errors that occur while the process is
+        attempting to be shut down. This deferred won't fire when the process
+        actually stops. To observe the actual process stopping, see
+        :func:`observe_stop`.
+        """
+        raise NotImplementedError('stop must be implemented in a subclass')
+
+    def on_stop(self, f):
+        """Get a deferred that will fire when the process stops.
+
+        The deferred will fire with data that contains information about
+        the exit status of the process.
+        """
+        if self.state=='after':
+            return f(self.stop_data)
+        else:
+            self.stop_callbacks.append(f)
+
+    def notify_start(self, data):
+        """Call this to trigger startup actions.
+
+        This logs the process startup and sets the state to 'running'.  It is
+        a pass-through so it can be used as a callback.
+        """
+
+        logging.info('Process %r started: %r' % (self.args[0], data))
+        self.start_data = data
+        self.state = 'running'
+        return data
+
+    def notify_stop(self, data):
+        """Call this to trigger process stop actions.
+
+        This logs the process stopping and sets the state to 'after'. Call
+        this to trigger all the deferreds from :func:`observe_stop`."""
+
+        logging.info('Process %r stopped: %r' % (self.args[0], data))
+        self.stop_data = data
+        self.state = 'after'
+        for i in range(len(self.stop_callbacks)):
+            d = self.stop_callbacks.pop()
+            d(data)
+        return data
+
+    def signal(self, sig):
+        """Signal the process.
+
+        Return a semi-meaningless deferred after signaling the process.
+
+        Parameters
+        ----------
+        sig : str or int
+            'KILL', 'INT', etc., or any signal number
+        """
+        raise NotImplementedError('signal must be implemented in a subclass')
+
+
+#-----------------------------------------------------------------------------
+# Local process launchers
+#-----------------------------------------------------------------------------
+
+
+class LocalProcessLauncher(BaseLauncher):
+    """Start and stop an external process in an asynchronous manner.
+
+    This will launch the external process with a working directory of
+    ``self.work_dir``.
+    """
+
+    # This is used to to construct self.args, which is passed to 
+    # spawnProcess.
+    cmd_and_args = List([])
+    poll_frequency = Int(100) # in ms
+
+    def __init__(self, work_dir=u'.', config=None):
+        super(LocalProcessLauncher, self).__init__(
+            work_dir=work_dir, config=config
+        )
+        self.process = None
+        self.start_deferred = None
+        self.poller = None
+
+    def find_args(self):
+        return self.cmd_and_args
+
+    def start(self):
+        if self.state == 'before':
+            self.process = Popen(self.args,
+                stdout=PIPE,stderr=PIPE,stdin=PIPE,
+                env=os.environ,
+                cwd=self.work_dir
+            )
+            
+            self.loop.add_handler(self.process.stdout.fileno(), self.handle_stdout, self.loop.READ)
+            self.loop.add_handler(self.process.stderr.fileno(), self.handle_stderr, self.loop.READ)
+            self.poller = ioloop.PeriodicCallback(self.poll, self.poll_frequency, self.loop)
+            self.poller.start()
+            self.notify_start(self.process.pid)
+        else:
+            s = 'The process was already started and has state: %r' % self.state
+            raise ProcessStateError(s)
+
+    def stop(self):
+        return self.interrupt_then_kill()
+
+    def signal(self, sig):
+        if self.state == 'running':
+            self.process.send_signal(sig)
+
+    def interrupt_then_kill(self, delay=2.0):
+        """Send INT, wait a delay and then send KILL."""
+        self.signal(SIGINT)
+        self.killer  = ioloop.DelayedCallback(lambda : self.signal(SIGKILL), delay*1000, self.loop)
+        self.killer.start()
+
+    # callbacks, etc:
+    
+    def handle_stdout(self, fd, events):
+        line = self.process.stdout.readline()
+        # a stopped process will be readable but return empty strings
+        if line:
+            logging.info(line[:-1])
+        else:
+            self.poll()
+    
+    def handle_stderr(self, fd, events):
+        line = self.process.stderr.readline()
+        # a stopped process will be readable but return empty strings
+        if line:
+            logging.error(line[:-1])
+        else:
+            self.poll()
+    
+    def poll(self):
+        status = self.process.poll()
+        if status is not None:
+            self.poller.stop()
+            self.loop.remove_handler(self.process.stdout.fileno())
+            self.loop.remove_handler(self.process.stderr.fileno())
+            self.notify_stop(dict(exit_code=status, pid=self.process.pid))
+        return status
+
+class LocalControllerLauncher(LocalProcessLauncher):
+    """Launch a controller as a regular external process."""
+
+    controller_cmd = List(ipcontroller_cmd_argv, config=True)
+    # Command line arguments to ipcontroller.
+    controller_args = List(['--log-to-file','--log-level', str(logging.ERROR)], config=True)
+
+    def find_args(self):
+        return self.controller_cmd + self.controller_args
+
+    def start(self, cluster_dir):
+        """Start the controller by cluster_dir."""
+        self.controller_args.extend(['--cluster-dir', cluster_dir])
+        self.cluster_dir = unicode(cluster_dir)
+        logging.info("Starting LocalControllerLauncher: %r" % self.args)
+        return super(LocalControllerLauncher, self).start()
+
+
+class LocalEngineLauncher(LocalProcessLauncher):
+    """Launch a single engine as a regular externall process."""
+
+    engine_cmd = List(ipengine_cmd_argv, config=True)
+    # Command line arguments for ipengine.
+    engine_args = List(
+        ['--log-to-file','--log-level', str(logging.ERROR)], config=True
+    )
+
+    def find_args(self):
+        return self.engine_cmd + self.engine_args
+
+    def start(self, cluster_dir):
+        """Start the engine by cluster_dir."""
+        self.engine_args.extend(['--cluster-dir', cluster_dir])
+        self.cluster_dir = unicode(cluster_dir)
+        return super(LocalEngineLauncher, self).start()
+
+
+class LocalEngineSetLauncher(BaseLauncher):
+    """Launch a set of engines as regular external processes."""
+
+    # Command line arguments for ipengine.
+    engine_args = List(
+        ['--log-to-file','--log-level', str(logging.ERROR)], config=True
+    )
+    # launcher class
+    launcher_class = LocalEngineLauncher
+    
+    def __init__(self, work_dir=u'.', config=None):
+        super(LocalEngineSetLauncher, self).__init__(
+            work_dir=work_dir, config=config
+        )
+        self.launchers = {}
+        self.stop_data = {}
+
+    def start(self, n, cluster_dir):
+        """Start n engines by profile or cluster_dir."""
+        self.cluster_dir = unicode(cluster_dir)
+        dlist = []
+        for i in range(n):
+            el = self.launcher_class(work_dir=self.work_dir, config=self.config)
+            # Copy the engine args over to each engine launcher.
+            import copy
+            el.engine_args = copy.deepcopy(self.engine_args)
+            el.on_stop(self._notice_engine_stopped)
+            d = el.start(cluster_dir)
+            if i==0:
+                logging.info("Starting LocalEngineSetLauncher: %r" % el.args)
+            self.launchers[i] = el
+            dlist.append(d)
+        self.notify_start(dlist)
+        # The consumeErrors here could be dangerous
+        # dfinal = gatherBoth(dlist, consumeErrors=True)
+        # dfinal.addCallback(self.notify_start)
+        return dlist
+
+    def find_args(self):
+        return ['engine set']
+
+    def signal(self, sig):
+        dlist = []
+        for el in self.launchers.itervalues():
+            d = el.signal(sig)
+            dlist.append(d)
+        # dfinal = gatherBoth(dlist, consumeErrors=True)
+        return dlist
+
+    def interrupt_then_kill(self, delay=1.0):
+        dlist = []
+        for el in self.launchers.itervalues():
+            d = el.interrupt_then_kill(delay)
+            dlist.append(d)
+        # dfinal = gatherBoth(dlist, consumeErrors=True)
+        return dlist
+
+    def stop(self):
+        return self.interrupt_then_kill()
+    
+    def _notice_engine_stopped(self, data):
+        print "notice", data
+        pid = data['pid']
+        for idx,el in self.launchers.iteritems():
+            if el.process.pid == pid:
+                break
+        self.launchers.pop(idx)
+        self.stop_data[idx] = data
+        if not self.launchers:
+            self.notify_stop(self.stop_data)
+
+
+#-----------------------------------------------------------------------------
+# MPIExec launchers
+#-----------------------------------------------------------------------------
+
+
+class MPIExecLauncher(LocalProcessLauncher):
+    """Launch an external process using mpiexec."""
+
+    # The mpiexec command to use in starting the process.
+    mpi_cmd = List(['mpiexec'], config=True)
+    # The command line arguments to pass to mpiexec.
+    mpi_args = List([], config=True)
+    # The program to start using mpiexec.
+    program = List(['date'], config=True)
+    # The command line argument to the program.
+    program_args = List([], config=True)
+    # The number of instances of the program to start.
+    n = Int(1, config=True)
+
+    def find_args(self):
+        """Build self.args using all the fields."""
+        return self.mpi_cmd + ['-n', self.n] + self.mpi_args + \
+               self.program + self.program_args
+
+    def start(self, n):
+        """Start n instances of the program using mpiexec."""
+        self.n = n
+        return super(MPIExecLauncher, self).start()
+
+
+class MPIExecControllerLauncher(MPIExecLauncher):
+    """Launch a controller using mpiexec."""
+
+    controller_cmd = List(ipcontroller_cmd_argv, config=True)
+    # Command line arguments to ipcontroller.
+    controller_args = List(['--log-to-file','--log-level', str(logging.ERROR)], config=True)
+    n = Int(1, config=False)
+
+    def start(self, cluster_dir):
+        """Start the controller by cluster_dir."""
+        self.controller_args.extend(['--cluster-dir', cluster_dir])
+        self.cluster_dir = unicode(cluster_dir)
+        logging.info("Starting MPIExecControllerLauncher: %r" % self.args)
+        return super(MPIExecControllerLauncher, self).start(1)
+
+    def find_args(self):
+        return self.mpi_cmd + ['-n', self.n] + self.mpi_args + \
+               self.controller_cmd + self.controller_args
+
+
+class MPIExecEngineSetLauncher(MPIExecLauncher):
+
+    engine_cmd = List(ipengine_cmd_argv, config=True)
+    # Command line arguments for ipengine.
+    engine_args = List(
+        ['--log-to-file','--log-level', str(logging.ERROR)], config=True
+    )
+    n = Int(1, config=True)
+
+    def start(self, n, cluster_dir):
+        """Start n engines by profile or cluster_dir."""
+        self.engine_args.extend(['--cluster-dir', cluster_dir])
+        self.cluster_dir = unicode(cluster_dir)
+        self.n = n
+        logging.info('Starting MPIExecEngineSetLauncher: %r' % self.args)
+        return super(MPIExecEngineSetLauncher, self).start(n)
+
+    def find_args(self):
+        return self.mpi_cmd + ['-n', self.n] + self.mpi_args + \
+               self.engine_cmd + self.engine_args
+
+
+#-----------------------------------------------------------------------------
+# SSH launchers
+#-----------------------------------------------------------------------------
+
+# TODO: Get SSH Launcher working again.
+
+class SSHLauncher(LocalProcessLauncher):
+    """A minimal launcher for ssh.
+
+    To be useful this will probably have to be extended to use the ``sshx``
+    idea for environment variables.  There could be other things this needs
+    as well.
+    """
+
+    ssh_cmd = List(['ssh'], config=True)
+    ssh_args = List([], config=True)
+    program = List(['date'], config=True)
+    program_args = List([], config=True)
+    hostname = Str('', config=True)
+    user = Str(os.environ.get('USER','username'), config=True)
+    location = Str('')
+
+    def _hostname_changed(self, name, old, new):
+        self.location = '%s@%s' % (self.user, new)
+
+    def _user_changed(self, name, old, new):
+        self.location = '%s@%s' % (new, self.hostname)
+
+    def find_args(self):
+        return self.ssh_cmd + self.ssh_args + [self.location] + \
+               self.program + self.program_args
+
+    def start(self, cluster_dir, hostname=None, user=None):
+        if hostname is not None:
+            self.hostname = hostname
+        if user is not None:
+            self.user = user
+        return super(SSHLauncher, self).start()
+
+
+class SSHControllerLauncher(SSHLauncher):
+
+    program = List(ipcontroller_cmd_argv, config=True)
+    # Command line arguments to ipcontroller.
+    program_args = List(['--log-to-file','--log-level', str(logging.ERROR)], config=True)
+
+
+class SSHEngineLauncher(SSHLauncher):
+    program = List(ipengine_cmd_argv, config=True)
+    # Command line arguments for ipengine.
+    program_args = List(
+        ['--log-to-file','--log-level', str(logging.ERROR)], config=True
+    )
+    
+class SSHEngineSetLauncher(LocalEngineSetLauncher):
+    launcher_class = SSHEngineLauncher
+
+
+#-----------------------------------------------------------------------------
+# Windows HPC Server 2008 scheduler launchers
+#-----------------------------------------------------------------------------
+
+
+# # This is only used on Windows.
+# def find_job_cmd():
+#     if os.name=='nt':
+#         try:
+#             return find_cmd('job')
+#         except FindCmdError:
+#             return 'job'
+#     else:
+#         return 'job'
+# 
+# 
+# class WindowsHPCLauncher(BaseLauncher):
+# 
+#     # A regular expression used to get the job id from the output of the 
+#     # submit_command.
+#     job_id_regexp = Str(r'\d+', config=True)
+#     # The filename of the instantiated job script.
+#     job_file_name = Unicode(u'ipython_job.xml', config=True)
+#     # The full path to the instantiated job script. This gets made dynamically
+#     # by combining the work_dir with the job_file_name.
+#     job_file = Unicode(u'')
+#     # The hostname of the scheduler to submit the job to
+#     scheduler = Str('', config=True)
+#     job_cmd = Str(find_job_cmd(), config=True)
+# 
+#     def __init__(self, work_dir=u'.', config=None):
+#         super(WindowsHPCLauncher, self).__init__(
+#             work_dir=work_dir, config=config
+#         )
+# 
+#     @property
+#     def job_file(self):
+#         return os.path.join(self.work_dir, self.job_file_name)
+# 
+#     def write_job_file(self, n):
+#         raise NotImplementedError("Implement write_job_file in a subclass.")
+# 
+#     def find_args(self):
+#         return ['job.exe']
+#         
+#     def parse_job_id(self, output):
+#         """Take the output of the submit command and return the job id."""
+#         m = re.search(self.job_id_regexp, output)
+#         if m is not None:
+#             job_id = m.group()
+#         else:
+#             raise LauncherError("Job id couldn't be determined: %s" % output)
+#         self.job_id = job_id
+#         logging.info('Job started with job id: %r' % job_id)
+#         return job_id
+# 
+#     @inlineCallbacks
+#     def start(self, n):
+#         """Start n copies of the process using the Win HPC job scheduler."""
+#         self.write_job_file(n)
+#         args = [
+#             'submit',
+#             '/jobfile:%s' % self.job_file,
+#             '/scheduler:%s' % self.scheduler
+#         ]
+#         logging.info("Starting Win HPC Job: %s" % (self.job_cmd + ' ' + ' '.join(args),))
+#         # Twisted will raise DeprecationWarnings if we try to pass unicode to this
+#         output = yield getProcessOutput(str(self.job_cmd),
+#             [str(a) for a in args],
+#             env=dict((str(k),str(v)) for k,v in os.environ.items()),
+#             path=self.work_dir
+#         )
+#         job_id = self.parse_job_id(output)
+#         self.notify_start(job_id)
+#         defer.returnValue(job_id)
+# 
+#     @inlineCallbacks
+#     def stop(self):
+#         args = [
+#             'cancel',
+#             self.job_id,
+#             '/scheduler:%s' % self.scheduler
+#         ]
+#         logging.info("Stopping Win HPC Job: %s" % (self.job_cmd + ' ' + ' '.join(args),))
+#         try:
+#             # Twisted will raise DeprecationWarnings if we try to pass unicode to this
+#             output = yield getProcessOutput(str(self.job_cmd),
+#                 [str(a) for a in args],
+#                 env=dict((str(k),str(v)) for k,v in os.environ.iteritems()),
+#                 path=self.work_dir
+#             )
+#         except:
+#             output = 'The job already appears to be stoppped: %r' % self.job_id
+#         self.notify_stop(output)  # Pass the output of the kill cmd
+#         defer.returnValue(output)
+# 
+# 
+# class WindowsHPCControllerLauncher(WindowsHPCLauncher):
+# 
+#     job_file_name = Unicode(u'ipcontroller_job.xml', config=True)
+#     extra_args = List([], config=False)
+# 
+#     def write_job_file(self, n):
+#         job = IPControllerJob(config=self.config)
+# 
+#         t = IPControllerTask(config=self.config)
+#         # The tasks work directory is *not* the actual work directory of 
+#         # the controller. It is used as the base path for the stdout/stderr
+#         # files that the scheduler redirects to.
+#         t.work_directory = self.cluster_dir
+#         # Add the --cluster-dir and from self.start().
+#         t.controller_args.extend(self.extra_args)
+#         job.add_task(t)
+# 
+#         logging.info("Writing job description file: %s" % self.job_file)
+#         job.write(self.job_file)
+# 
+#     @property
+#     def job_file(self):
+#         return os.path.join(self.cluster_dir, self.job_file_name)
+# 
+#     def start(self, cluster_dir):
+#         """Start the controller by cluster_dir."""
+#         self.extra_args = ['--cluster-dir', cluster_dir]
+#         self.cluster_dir = unicode(cluster_dir)
+#         return super(WindowsHPCControllerLauncher, self).start(1)
+# 
+# 
+# class WindowsHPCEngineSetLauncher(WindowsHPCLauncher):
+# 
+#     job_file_name = Unicode(u'ipengineset_job.xml', config=True)
+#     extra_args = List([], config=False)
+# 
+#     def write_job_file(self, n):
+#         job = IPEngineSetJob(config=self.config)
+# 
+#         for i in range(n):
+#             t = IPEngineTask(config=self.config)
+#             # The tasks work directory is *not* the actual work directory of 
+#             # the engine. It is used as the base path for the stdout/stderr
+#             # files that the scheduler redirects to.
+#             t.work_directory = self.cluster_dir
+#             # Add the --cluster-dir and from self.start().
+#             t.engine_args.extend(self.extra_args)
+#             job.add_task(t)
+# 
+#         logging.info("Writing job description file: %s" % self.job_file)
+#         job.write(self.job_file)
+# 
+#     @property
+#     def job_file(self):
+#         return os.path.join(self.cluster_dir, self.job_file_name)
+# 
+#     def start(self, n, cluster_dir):
+#         """Start the controller by cluster_dir."""
+#         self.extra_args = ['--cluster-dir', cluster_dir]
+#         self.cluster_dir = unicode(cluster_dir)
+#         return super(WindowsHPCEngineSetLauncher, self).start(n)
+# 
+# 
+# #-----------------------------------------------------------------------------
+# # Batch (PBS) system launchers
+# #-----------------------------------------------------------------------------
+# 
+# # TODO: Get PBS launcher working again.
+# 
+# class BatchSystemLauncher(BaseLauncher):
+#     """Launch an external process using a batch system.
+# 
+#     This class is designed to work with UNIX batch systems like PBS, LSF,
+#     GridEngine, etc.  The overall model is that there are different commands
+#     like qsub, qdel, etc. that handle the starting and stopping of the process.
+# 
+#     This class also has the notion of a batch script. The ``batch_template``
+#     attribute can be set to a string that is a template for the batch script.
+#     This template is instantiated using Itpl. Thus the template can use
+#     ${n} fot the number of instances. Subclasses can add additional variables
+#     to the template dict.
+#     """
+# 
+#     # Subclasses must fill these in.  See PBSEngineSet
+#     # The name of the command line program used to submit jobs.
+#     submit_command = Str('', config=True)
+#     # The name of the command line program used to delete jobs.
+#     delete_command = Str('', config=True)
+#     # A regular expression used to get the job id from the output of the 
+#     # submit_command.
+#     job_id_regexp = Str('', config=True)
+#     # The string that is the batch script template itself.
+#     batch_template = Str('', config=True)
+#     # The filename of the instantiated batch script.
+#     batch_file_name = Unicode(u'batch_script', config=True)
+#     # The full path to the instantiated batch script.
+#     batch_file = Unicode(u'')
+# 
+#     def __init__(self, work_dir=u'.', config=None):
+#         super(BatchSystemLauncher, self).__init__(
+#             work_dir=work_dir, config=config
+#         )
+#         self.batch_file = os.path.join(self.work_dir, self.batch_file_name)
+#         self.context = {}
+# 
+#     def parse_job_id(self, output):
+#         """Take the output of the submit command and return the job id."""
+#         m = re.match(self.job_id_regexp, output)
+#         if m is not None:
+#             job_id = m.group()
+#         else:
+#             raise LauncherError("Job id couldn't be determined: %s" % output)
+#         self.job_id = job_id
+#         logging.info('Job started with job id: %r' % job_id)
+#         return job_id
+# 
+#     def write_batch_script(self, n):
+#         """Instantiate and write the batch script to the work_dir."""
+#         self.context['n'] = n
+#         script_as_string = Itpl.itplns(self.batch_template, self.context)
+#         logging.info('Writing instantiated batch script: %s' % self.batch_file)
+#         f = open(self.batch_file, 'w')
+#         f.write(script_as_string)
+#         f.close()
+# 
+#     @inlineCallbacks
+#     def start(self, n):
+#         """Start n copies of the process using a batch system."""
+#         self.write_batch_script(n)
+#         output = yield getProcessOutput(self.submit_command,
+#             [self.batch_file], env=os.environ)
+#         job_id = self.parse_job_id(output)
+#         self.notify_start(job_id)
+#         defer.returnValue(job_id)
+# 
+#     @inlineCallbacks
+#     def stop(self):
+#         output = yield getProcessOutput(self.delete_command,
+#             [self.job_id], env=os.environ
+#         )
+#         self.notify_stop(output)  # Pass the output of the kill cmd
+#         defer.returnValue(output)
+# 
+# 
+# class PBSLauncher(BatchSystemLauncher):
+#     """A BatchSystemLauncher subclass for PBS."""
+# 
+#     submit_command = Str('qsub', config=True)
+#     delete_command = Str('qdel', config=True)
+#     job_id_regexp = Str(r'\d+', config=True)
+#     batch_template = Str('', config=True)
+#     batch_file_name = Unicode(u'pbs_batch_script', config=True)
+#     batch_file = Unicode(u'')
+# 
+# 
+# class PBSControllerLauncher(PBSLauncher):
+#     """Launch a controller using PBS."""
+# 
+#     batch_file_name = Unicode(u'pbs_batch_script_controller', config=True)
+# 
+#     def start(self, cluster_dir):
+#         """Start the controller by profile or cluster_dir."""
+#         # Here we save profile and cluster_dir in the context so they
+#         # can be used in the batch script template as ${profile} and
+#         # ${cluster_dir}
+#         self.context['cluster_dir'] = cluster_dir
+#         self.cluster_dir = unicode(cluster_dir)
+#         logging.info("Starting PBSControllerLauncher: %r" % self.args)
+#         return super(PBSControllerLauncher, self).start(1)
+# 
+# 
+# class PBSEngineSetLauncher(PBSLauncher):
+# 
+#     batch_file_name = Unicode(u'pbs_batch_script_engines', config=True)
+# 
+#     def start(self, n, cluster_dir):
+#         """Start n engines by profile or cluster_dir."""
+#         self.program_args.extend(['--cluster-dir', cluster_dir])
+#         self.cluster_dir = unicode(cluster_dir)
+#         logging.info('Starting PBSEngineSetLauncher: %r' % self.args)
+#         return super(PBSEngineSetLauncher, self).start(n)
+
+
+#-----------------------------------------------------------------------------
+# A launcher for ipcluster itself!
+#-----------------------------------------------------------------------------
+
+
+class IPClusterLauncher(LocalProcessLauncher):
+    """Launch the ipcluster program in an external process."""
+
+    ipcluster_cmd = List(ipcluster_cmd_argv, config=True)
+    # Command line arguments to pass to ipcluster.
+    ipcluster_args = List(
+        ['--clean-logs', '--log-to-file', '--log-level', str(logging.ERROR)], config=True)
+    ipcluster_subcommand = Str('start')
+    ipcluster_n = Int(2)
+
+    def find_args(self):
+        return self.ipcluster_cmd + [self.ipcluster_subcommand] + \
+            ['-n', repr(self.ipcluster_n)] + self.ipcluster_args
+
+    def start(self):
+        logging.info("Starting ipcluster: %r" % self.args)
+        return super(IPClusterLauncher, self).start()
+

IPython/zmq/parallel/client.py

@@ -1,1172 +1,1164
 """A semi-synchronous Client for the ZMQ controller"""
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2010  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 time
 from getpass import getpass
 from pprint import pprint
 from datetime import datetime
 
 import zmq
 from zmq.eventloop import ioloop, zmqstream
 
 from IPython.external.decorator import decorator
 from IPython.zmq import tunnel
 
 import streamsession as ss
 # from remotenamespace import RemoteNamespace
 from view import DirectView, LoadBalancedView
 from dependency import Dependency, depend, require
 import error
 import map as Map
 from asyncresult import AsyncResult, AsyncMapResult
 from remotefunction import remote,parallel,ParallelFunction,RemoteFunction
 from util import ReverseDict
 
 #--------------------------------------------------------------------------
 # helpers for implementing old MEC API via client.apply
 #--------------------------------------------------------------------------
 
 def _push(ns):
     """helper method for implementing `client.push` via `client.apply`"""
     globals().update(ns)
 
 def _pull(keys):
     """helper method for implementing `client.pull` via `client.apply`"""
     g = globals()
     if isinstance(keys, (list,tuple, set)):
         for key in keys:
             if not g.has_key(key):
                 raise NameError("name '%s' is not defined"%key)
         return map(g.get, keys)
     else:
         if not g.has_key(keys):
             raise NameError("name '%s' is not defined"%keys)
         return g.get(keys)
 
 def _clear():
     """helper method for implementing `client.clear` via `client.apply`"""
     globals().clear()
 
 def _execute(code):
     """helper method for implementing `client.execute` via `client.apply`"""
     exec code in globals()
     
 
 #--------------------------------------------------------------------------
 # Decorators for Client methods
 #--------------------------------------------------------------------------
 
 @decorator
 def spinfirst(f, self, *args, **kwargs):
     """Call spin() to sync state prior to calling the method."""
     self.spin()
     return f(self, *args, **kwargs)
 
 @decorator
 def defaultblock(f, self, *args, **kwargs):
     """Default to self.block; preserve self.block."""
     block = kwargs.get('block',None)
     block = self.block if block is None else block
     saveblock = self.block
     self.block = block
     try:
         ret = f(self, *args, **kwargs)
     finally:
         self.block = saveblock
     return ret
 
 
 #--------------------------------------------------------------------------
 # Classes
 #--------------------------------------------------------------------------
 
-class ResultDict(dict):
-    """A subclass of dict that raises errors if it has them."""
-    def __getitem__(self, key):
-        res = dict.__getitem__(self, key)
-        if isinstance(res, error.KernelError):
-            raise res
-        return res
-
 class Metadata(dict):
     """Subclass of dict for initializing metadata values."""
     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' : '',
             }
         self.update(md)
         self.update(dict(*args, **kwargs))
 
         
 
 class Client(object):
     """A semi-synchronous client to the IPython ZMQ controller
     
     Parameters
     ----------
     
     addr : bytes; zmq url, e.g. 'tcp://127.0.0.1:10101'
         The address of the controller's registration socket.
         [Default: 'tcp://127.0.0.1:10101']
     context : zmq.Context
         Pass an existing zmq.Context instance, otherwise the client will create its own
     username : bytes
         set username to be passed to the Session object
     debug : bool
         flag for lots of message printing for debug purposes
  
     #-------------- 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 public ssh 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.
     
     #------- exec authentication args -------
     # If even localhost is untrusted, you can have some protection against
     # unauthorized execution by using a key.  Messages are still sent
     # as cleartext, so if someone can snoop your loopback traffic this will
     # not help anything.
     
     exec_key : str
         an authentication key or file containing a key
         default: None
     
     
     Attributes
     ----------
     ids : set 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
             
     barrier : wait on one or more msg_ids
     
     execution methods: apply/apply_bound/apply_to/apply_bound
         legacy: execute, run
     
     query methods: queue_status, get_result, purge
     
     control methods: abort, kill
     
     """
     
     
     _connected=False
     _ssh=False
     _engines=None
     _addr='tcp://127.0.0.1:10101'
     _registration_socket=None
     _query_socket=None
     _control_socket=None
     _iopub_socket=None
     _notification_socket=None
     _mux_socket=None
     _task_socket=None
     block = False
     outstanding=None
     results = None
     history = None
     debug = False
     targets = None
     
     def __init__(self, addr='tcp://127.0.0.1:10101', context=None, username=None, debug=False, 
             sshserver=None, sshkey=None, password=None, paramiko=None,
             exec_key=None,):
         if context is None:
             context = zmq.Context()
         self.context = context
         self.targets = 'all'
         self._addr = addr
         self._ssh = bool(sshserver or sshkey or password)
         if self._ssh and sshserver is None:
             # default to the same
             sshserver = addr.split('://')[1].split(':')[0]
         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)
         
         if exec_key is not None and os.path.isfile(exec_key):
             arg = 'keyfile'
         else:
             arg = 'key'
         key_arg = {arg:exec_key}
         if username is None:
             self.session = ss.StreamSession(**key_arg)
         else:
             self.session = ss.StreamSession(username, **key_arg)
         self._registration_socket = self.context.socket(zmq.XREQ)
         self._registration_socket.setsockopt(zmq.IDENTITY, self.session.session)
         if self._ssh:
             tunnel.tunnel_connection(self._registration_socket, addr, sshserver, **ssh_kwargs)
         else:
             self._registration_socket.connect(addr)
         self._engines = ReverseDict()
         self._ids = set()
         self.outstanding=set()
         self.results = {}
         self.metadata = {}
         self.history = []
         self.debug = debug
         self.session.debug = debug
         
         self._notification_handlers = {'registration_notification' : self._register_engine,
                                     'unregistration_notification' : self._unregister_engine,
                                     }
         self._queue_handlers = {'execute_reply' : self._handle_execute_reply,
                                 'apply_reply' : self._handle_apply_reply}
         self._connect(sshserver, ssh_kwargs)
         
     
     @property
     def ids(self):
         """Always up to date ids property."""
         self._flush_notifications()
         return self._ids
     
     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] = bytes(v) # force not unicode
             self._ids.add(eid)
     
     def _build_targets(self, targets):
         """Turn valid target IDs or 'all' into two lists:
         (int_ids, uuids).
         """
         if targets is None:
             targets = self._ids
         elif isinstance(targets, str):
             if targets.lower() == 'all':
                 targets = self._ids
             else:
                 raise TypeError("%r not valid str target, must be 'all'"%(targets))
         elif isinstance(targets, int):
             targets = [targets]
         return [self._engines[t] for t in targets], list(targets)
     
     def _connect(self, sshserver, ssh_kwargs):
         """setup all our socket connections to the controller. This is called from
         __init__."""
         if self._connected:
             return
         self._connected=True
 
         def connect_socket(s, addr):
             if self._ssh:
                 return tunnel.tunnel_connection(s, addr, sshserver, **ssh_kwargs)
             else:
                 return s.connect(addr)
             
         self.session.send(self._registration_socket, 'connection_request')
         idents,msg = self.session.recv(self._registration_socket,mode=0)
         if self.debug:
             pprint(msg)
         msg = ss.Message(msg)
         content = msg.content
         if content.status == 'ok':
             if content.mux:
                 self._mux_socket = self.context.socket(zmq.PAIR)
                 self._mux_socket.setsockopt(zmq.IDENTITY, self.session.session)
                 connect_socket(self._mux_socket, content.mux)
             if content.task:
                 self._task_socket = self.context.socket(zmq.PAIR)
                 self._task_socket.setsockopt(zmq.IDENTITY, self.session.session)
                 connect_socket(self._task_socket, content.task)
             if content.notification:
                 self._notification_socket = self.context.socket(zmq.SUB)
                 connect_socket(self._notification_socket, content.notification)
                 self._notification_socket.setsockopt(zmq.SUBSCRIBE, "")
             if content.query:
                 self._query_socket = self.context.socket(zmq.PAIR)
                 self._query_socket.setsockopt(zmq.IDENTITY, self.session.session)
                 connect_socket(self._query_socket, content.query)
             if content.control:
                 self._control_socket = self.context.socket(zmq.PAIR)
                 self._control_socket.setsockopt(zmq.IDENTITY, self.session.session)
                 connect_socket(self._control_socket, content.control)
             if content.iopub:
                 self._iopub_socket = self.context.socket(zmq.SUB)
                 self._iopub_socket.setsockopt(zmq.SUBSCRIBE, '')
                 self._iopub_socket.setsockopt(zmq.IDENTITY, self.session.session)
                 connect_socket(self._iopub_socket, content.iopub)
             self._update_engines(dict(content.engines))
                 
         else:
             self._connected = False
             raise Exception("Failed to connect!")
     
     #--------------------------------------------------------------------------
     # handlers and callbacks for incoming messages
     #--------------------------------------------------------------------------
     
     def _register_engine(self, msg):
         """Register a new engine, and update our connection info."""
         content = msg['content']
         eid = content['id']
         d = {eid : content['queue']}
         self._update_engines(d)
         self._ids.add(int(eid))
 
     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)
             self._engines.pop(eid)
     
     def _extract_metadata(self, header, parent, content):
         md = {'msg_id' : parent['msg_id'],
               'submitted' : datetime.strptime(parent['date'], ss.ISO8601),
               'started' : datetime.strptime(header['started'], ss.ISO8601),
               'completed' : datetime.strptime(header['date'], ss.ISO8601),
               'received' : datetime.now(),
               'engine_uuid' : header['engine'],
               'engine_id' : self._engines.get(header['engine'], None),
               'follow' : parent['follow'],
               'after' : parent['after'],
               'status' : content['status'],
             }
         return md
     
     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:
             print("got unknown result: %s"%msg_id)
         else:
             self.outstanding.remove(msg_id)
         self.results[msg_id] = ss.unwrap_exception(msg['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:
             print ("got unknown result: %s"%msg_id)
         else:
             self.outstanding.remove(msg_id)
         content = msg['content']
         header = msg['header']
         
         # construct metadata:
         md = self.metadata.setdefault(msg_id, Metadata())
         md.update(self._extract_metadata(header, parent, content))
         self.metadata[msg_id] = md
         
         # construct result:
         if content['status'] == 'ok':
             self.results[msg_id] = ss.unserialize_object(msg['buffers'])[0]
         elif content['status'] == 'aborted':
             self.results[msg_id] = error.AbortedTask(msg_id)
         elif content['status'] == 'resubmitted':
             # TODO: handle resubmission
             pass
         else:
             e = ss.unwrap_exception(content)
             e_uuid = e.engine_info['engineid']
             eid = self._engines[e_uuid]
             e.engine_info['engineid'] = eid
             self.results[msg_id] = e
     
     def _flush_notifications(self):
         """Flush notifications of engine registrations waiting
         in ZMQ queue."""
         msg = self.session.recv(self._notification_socket, mode=zmq.NOBLOCK)
         while msg is not None:
             if self.debug:
                 pprint(msg)
             msg = msg[-1]
             msg_type = msg['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)
             msg = self.session.recv(self._notification_socket, mode=zmq.NOBLOCK)
     
     def _flush_results(self, sock):
         """Flush task or queue results waiting in ZMQ queue."""
         msg = self.session.recv(sock, mode=zmq.NOBLOCK)
         while msg is not None:
             if self.debug:
                 pprint(msg)
             msg = msg[-1]
             msg_type = msg['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)
             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."""
         msg = self.session.recv(sock, mode=zmq.NOBLOCK)
         while msg is not None:
             if self.debug:
                 pprint(msg)
             msg = self.session.recv(sock, mode=zmq.NOBLOCK)
     
     def _flush_iopub(self, sock):
         """Flush replies from the iopub channel waiting
         in the ZMQ queue.
         """
         msg = self.session.recv(sock, mode=zmq.NOBLOCK)
         while msg is not None:
             if self.debug:
                 pprint(msg)
             msg = msg[-1]
             parent = msg['parent_header']
             msg_id = parent['msg_id']
             content = msg['content']
             header = msg['header']
             msg_type = msg['msg_type']
             
             # init metadata:
             md = self.metadata.setdefault(msg_id, Metadata())
             
             if msg_type == 'stream':
                 name = content['name']
                 s = md[name] or ''
                 md[name] = s + content['data']
             elif msg_type == 'pyerr':
                 md.update({'pyerr' : ss.unwrap_exception(content)})
             else:
                 md.update({msg_type : content['data']})
             
             self.metadata[msg_id] = md
             
             msg = self.session.recv(sock, mode=zmq.NOBLOCK)
     
     #--------------------------------------------------------------------------
     # getitem
     #--------------------------------------------------------------------------
     
     def __getitem__(self, key):
         """Dict access returns DirectView multiplexer objects or,
         if key is None, a LoadBalancedView."""
         if key is None:
             return LoadBalancedView(self)
         if isinstance(key, int):
             if key not in self.ids:
                 raise IndexError("No such engine: %i"%key)
             return DirectView(self, key)
         
         if isinstance(key, slice):
             indices = range(len(self.ids))[key]
             ids = sorted(self._ids)
             key = [ ids[i] for i in indices ]
             # newkeys = sorted(self._ids)[thekeys[k]]
         
         if isinstance(key, (tuple, list, xrange)):
             _,targets = self._build_targets(list(key))
             return DirectView(self, targets)
         else:
             raise TypeError("key by int/iterable of ints only, not %s"%(type(key)))
     
     #--------------------------------------------------------------------------
     # Begin public methods
     #--------------------------------------------------------------------------
     
     @property
     def remote(self):
         """property for convenient RemoteFunction generation.
         
         >>> @client.remote
         ... def f():
                 import os
                 print (os.getpid())
         """
         return remote(self, block=self.block)
     
     def spin(self):
         """Flush any registration notifications and execution results
         waiting in the ZMQ queue.
         """
         if self._notification_socket:
             self._flush_notifications()
         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._iopub_socket:
             self._flush_iopub(self._iopub_socket)
     
     def barrier(self, msg_ids=None, timeout=-1):
         """waits on one or more `msg_ids`, for up to `timeout` seconds.
         
         Parameters
         ----------
         msg_ids : 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 msg_ids is None:
             theids = self.outstanding
         else:
             if isinstance(msg_ids, (int, str, AsyncResult)):
                 msg_ids = [msg_ids]
             theids = set()
             for msg_id in msg_ids:
                 if isinstance(msg_id, int):
                     msg_id = self.history[msg_id]
                 elif isinstance(msg_id, AsyncResult):
                     map(theids.add, msg_id.msg_ids)
                     continue
                 theids.add(msg_id)
         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
     #--------------------------------------------------------------------------
     
     @spinfirst
     @defaultblock
     def clear(self, targets=None, block=None):
         """Clear the namespace in target(s)."""
         targets = self._build_targets(targets)[0]
         for t in targets:
             self.session.send(self._control_socket, 'clear_request', content={}, ident=t)
         error = False
         if self.block:
             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 = ss.unwrap_exception(msg['content'])
         if error:
             return error
         
     
     @spinfirst
     @defaultblock
     def abort(self, msg_ids = None, targets=None, block=None):
         """Abort the execution queues of target(s)."""
         targets = self._build_targets(targets)[0]
         if isinstance(msg_ids, basestring):
             msg_ids = [msg_ids]
         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 self.block:
             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 = ss.unwrap_exception(msg['content'])
         if error:
             return error
     
     @spinfirst
     @defaultblock
     def shutdown(self, targets=None, restart=False, controller=False, block=None):
         """Terminates one or more engine processes, optionally including the controller."""
         if controller:
             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 controller:
             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 = ss.unwrap_exception(msg['content'])
         
         if controller:
             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 = ss.unwrap_exception(msg['content'])
         
         if error:
             raise error
     
     #--------------------------------------------------------------------------
     # Execution methods
     #--------------------------------------------------------------------------
     
     @defaultblock
     def execute(self, code, targets='all', block=None):
         """Executes `code` on `targets` in blocking or nonblocking manner.
         
         ``execute`` is always `bound` (affects engine namespace)
         
         Parameters
         ----------
         code : str
                 the code string to be executed
         targets : int/str/list of ints/strs
                 the engines on which to execute
                 default : all
         block : bool
                 whether or not to wait until done to return
                 default: self.block
         """
         result = self.apply(_execute, (code,), targets=targets, block=self.block, bound=True)
         return result
     
     def run(self, code, block=None):
         """Runs `code` on an engine. 
         
         Calls to this are load-balanced.
         
         ``run`` is never `bound` (no effect on engine namespace)
         
         Parameters
         ----------
         code : str
                 the code string to be executed
         block : bool
                 whether or not to wait until done
         
         """
         result = self.apply(_execute, (code,), targets=None, block=block, bound=False)
         return result
     
     def _maybe_raise(self, result):
         """wrapper for maybe raising an exception if apply failed."""
         if isinstance(result, error.RemoteError):
             raise result
         
         return result
             
     def apply(self, f, args=None, kwargs=None, bound=True, block=None, targets=None,
                         after=None, follow=None):
         """Call `f(*args, **kwargs)` on a remote engine(s), returning the result.
         
         This is the central execution command for the client.
         
         Parameters
         ----------
         
         f : function
             The fuction to be called remotely
         args : tuple/list
             The positional arguments passed to `f`
         kwargs : dict
             The keyword arguments passed to `f`
         bound : bool (default: True)
             Whether to execute in the Engine(s) namespace, or in a clean
             namespace not affecting the engine.
         block : bool (default: self.block)
             Whether to wait for the result, or return immediately.
             False:
                 returns msg_id(s)
                 if multiple targets:
                     list of ids
             True:
                 returns actual result(s) of f(*args, **kwargs)
                 if multiple targets:
                     dict of results, by engine ID
         targets : int,list of ints, 'all', None
             Specify the destination of the job.
             if None:
                 Submit via Task queue for load-balancing.
             if 'all':
                 Run on all active engines
             if list:
                 Run on each specified engine
             if int:
                 Run on single engine
             
         after : Dependency or collection of msg_ids
             Only for load-balanced execution (targets=None)
             Specify a list of msg_ids as a time-based dependency.
             This job will only be run *after* the dependencies
             have been met.
             
         follow : Dependency or collection of msg_ids
             Only for load-balanced execution (targets=None)
             Specify a list of msg_ids as a location-based dependency.
             This job will only be run on an engine where this dependency
             is met.
         
         Returns
         -------
         if block is False:
             if single target:
                 return msg_id
             else:
                 return list of msg_ids
                 ? (should this be dict like block=True) ?
         else:
             if single target:
                 return result of f(*args, **kwargs)
             else:
                 return dict of results, keyed by engine
         """
         
         # defaults:
         block = block if block is not None else self.block
         args = args if args is not None else []
         kwargs = kwargs if kwargs is not None else {}
         
         # enforce types of f,args,kwrags
         if not callable(f):
             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 isinstance(after, Dependency):
             after = after.as_dict()
         elif isinstance(after, AsyncResult):
             after=after.msg_ids
         elif after is None:
             after = []
         if isinstance(follow, Dependency):
             follow = follow.as_dict()
         elif isinstance(follow, AsyncResult):
             follow=follow.msg_ids
         elif follow is None:
             follow = []
         options  = dict(bound=bound, block=block, after=after, follow=follow)
             
         if targets is None:
             return self._apply_balanced(f, args, kwargs, **options)
         else:
             return self._apply_direct(f, args, kwargs, targets=targets, **options)
     
     def _apply_balanced(self, f, args, kwargs, bound=True, block=None,
                             after=None, follow=None):
         """The underlying method for applying functions in a load balanced
         manner, via the task queue."""
         
         subheader = dict(after=after, follow=follow)
         bufs = ss.pack_apply_message(f,args,kwargs)
         content = dict(bound=bound)
         
         msg = self.session.send(self._task_socket, "apply_request", 
                 content=content, buffers=bufs, subheader=subheader)
         msg_id = msg['msg_id']
         self.outstanding.add(msg_id)
         self.history.append(msg_id)
         ar = AsyncResult(self, [msg_id], fname=f.__name__)
         if block:
             return ar.get()
         else:
             return ar
     
     def _apply_direct(self, f, args, kwargs, bound=True, block=None, targets=None,
                                 after=None, follow=None):
         """Then underlying method for applying functions to specific engines
         via the MUX queue."""
         
         queues,targets = self._build_targets(targets)
         
         subheader = dict(after=after, follow=follow)
         content = dict(bound=bound)
         bufs = ss.pack_apply_message(f,args,kwargs)
         
         msg_ids = []
         for queue in queues:
             msg = self.session.send(self._mux_socket, "apply_request", 
                     content=content, buffers=bufs,ident=queue, subheader=subheader)
             msg_id = msg['msg_id']
             self.outstanding.add(msg_id)
             self.history.append(msg_id)
             msg_ids.append(msg_id)
         ar = AsyncResult(self, msg_ids, fname=f.__name__)
         if block:
             return ar.get()
         else:
             return ar
     
     #--------------------------------------------------------------------------
     # Map and decorators
     #--------------------------------------------------------------------------
     
     def map(self, f, *sequences):
         """Parallel version of builtin `map`, using all our engines."""
         pf = ParallelFunction(self, f, block=self.block,
                         bound=True, targets='all')
         return pf.map(*sequences)
     
     def parallel(self, bound=True, targets='all', block=True):
         """Decorator for making a ParallelFunction."""
         return parallel(self, bound=bound, targets=targets, block=block)
     
     def remote(self, bound=True, targets='all', block=True):
         """Decorator for making a RemoteFunction."""
         return remote(self, bound=bound, targets=targets, block=block)
     
     #--------------------------------------------------------------------------
     # Data movement
     #--------------------------------------------------------------------------
     
     @defaultblock
     def push(self, ns, targets='all', block=None):
         """Push the contents of `ns` into the namespace on `target`"""
         if not isinstance(ns, dict):
             raise TypeError("Must be a dict, not %s"%type(ns))
         result = self.apply(_push, (ns,), targets=targets, block=block, bound=True)
         return result
     
     @defaultblock
     def pull(self, keys, targets='all', block=None):
         """Pull objects from `target`'s namespace by `keys`"""
         if isinstance(keys, str):
             pass
         elif isinstance(keys, (list,tuple,set)):
             for key in keys:
                 if not isinstance(key, str):
                     raise TypeError
         result = self.apply(_pull, (keys,), targets=targets, block=block, bound=True)
         return result
     
     def scatter(self, key, seq, dist='b', flatten=False, targets='all', block=None):
         """
         Partition a Python sequence and send the partitions to a set of engines.
         """
         block = block if block is not None else self.block
         targets = self._build_targets(targets)[-1]
         mapObject = Map.dists[dist]()
         nparts = len(targets)
         msg_ids = []
         for index, engineid in enumerate(targets):
             partition = mapObject.getPartition(seq, index, nparts)
             if flatten and len(partition) == 1:
                 r = self.push({key: partition[0]}, targets=engineid, block=False)
             else:
                 r = self.push({key: partition}, targets=engineid, block=False)
             msg_ids.extend(r.msg_ids)
         r = AsyncResult(self, msg_ids, fname='scatter')
         if block:
             return r.get()
         else:
             return r
     
     def gather(self, key, dist='b', targets='all', block=None):
         """
         Gather a partitioned sequence on a set of engines as a single local seq.
         """
         block = block if block is not None else self.block
         
         targets = self._build_targets(targets)[-1]
         mapObject = Map.dists[dist]()
         msg_ids = []
         for index, engineid in enumerate(targets):
             msg_ids.extend(self.pull(key, targets=engineid,block=False).msg_ids)
         
         r = AsyncMapResult(self, msg_ids, mapObject, fname='gather')
         if block:
             return r.get()
         else:
             return r
     
     #--------------------------------------------------------------------------
     # Query methods
     #--------------------------------------------------------------------------
     
     @spinfirst
     def get_results(self, msg_ids, status_only=False):
         """Returns the result of the execute or task request with `msg_ids`.
         
         Parameters
         ----------
         msg_ids : list of ints or msg_ids
             if int:
                 Passed as index to self.history for convenience.
         status_only : bool (default: False)
             if False:
                 return the actual results
         
         Returns
         -------
         
         results : dict
             There will always be the keys 'pending' and 'completed', which will
             be lists of msg_ids.
         """
         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, str):
                 raise TypeError("msg_ids must be str, not %r"%msg_id)
             theids.append(msg_id)
         
         completed = []
         local_results = {}
         # temporarily disable local shortcut
         # for msg_id in list(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 ss.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.setdefault(msg_id, Metadata())
                 md.update(self._extract_metadata(header, parent, rcontent))
                 md.update(iodict)
                 
                 if rcontent['status'] == 'ok':
                     res,buffers = ss.unserialize_object(buffers)
                 else:
                     res = ss.unwrap_exception(rcontent)
                     failures.append(res)
                 
                 self.results[msg_id] = res
                 content[msg_id] = res
         
         error.collect_exceptions(failures, "get_results")
         return content
 
     @spinfirst
     def queue_status(self, targets=None, verbose=False):
         """Fetch the status of engine queues.
         
         Parameters
         ----------
         targets : int/str/list of ints/strs
                 the engines on which to execute
                 default : all
         verbose : bool
                 Whether to return lengths only, or lists of ids for each element
         """
         targets = self._build_targets(targets)[1]
         content = dict(targets=targets, 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 ss.unwrap_exception(content)
         return ss.rekey(content)
         
     @spinfirst
     def purge_results(self, msg_ids=[], targets=[]):
         """Tell the controller to forget results.
         
         Individual results can be purged by msg_id, or the entire
         history of specific targets can be purged.
         
         Parameters
         ----------
         msg_ids : str or list of strs
                 the msg_ids whose results should be forgotten.
         targets : int/str/list of ints/strs
                 The targets, by uuid or int_id, whose entire history is to be purged.
                 Use `targets='all'` to scrub everything from the controller's memory.
                 
                 default : None
         """
         if not targets and not msg_ids:
             raise ValueError
         if targets:
             targets = self._build_targets(targets)[1]
         content = dict(targets=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 ss.unwrap_exception(content)
 
     #----------------------------------------
     # activate for %px,%autopx magics
     #----------------------------------------
     def activate(self):
         """Make this `View` active for parallel magic commands.
         
         IPython has a magic command syntax to work with `MultiEngineClient` objects.
         In a given IPython session there is a single active one.  While
         there can be many `Views` created and used by the user, 
         there is only one active one.  The active `View` is used whenever 
         the magic commands %px and %autopx are used.
         
         The activate() method is called on a given `View` to make it 
         active.  Once this has been done, the magic commands can be used.
         """
         
         try:
             # This is injected into __builtins__.
             ip = get_ipython()
         except NameError:
             print "The IPython parallel magics (%result, %px, %autopx) only work within IPython."
         else:
             pmagic = ip.plugin_manager.get_plugin('parallelmagic')
             if pmagic is not None:
                 pmagic.active_multiengine_client = self
             else:
                 print "You must first load the parallelmagic extension " \
                       "by doing '%load_ext parallelmagic'"
 
 class AsynClient(Client):
     """An Asynchronous client, using the Tornado Event Loop.
     !!!unfinished!!!"""
     io_loop = None
     _queue_stream = None
     _notifier_stream = None
     _task_stream = None
     _control_stream = None
     
     def __init__(self, addr, context=None, username=None, debug=False, io_loop=None):
         Client.__init__(self, addr, context, username, debug)
         if io_loop is None:
             io_loop = ioloop.IOLoop.instance()
         self.io_loop = io_loop
         
         self._queue_stream = zmqstream.ZMQStream(self._mux_socket, io_loop)
         self._control_stream = zmqstream.ZMQStream(self._control_socket, io_loop)
         self._task_stream = zmqstream.ZMQStream(self._task_socket, io_loop)
         self._notification_stream = zmqstream.ZMQStream(self._notification_socket, io_loop)
     
     def spin(self):
         for stream in (self.queue_stream, self.notifier_stream, 
                         self.task_stream, self.control_stream):
             stream.flush()
 
 __all__ = [ 'Client', 
             'depend', 
             'require', 
             'remote',
             'parallel',
             'RemoteFunction',
             'ParallelFunction',
             'DirectView',
             'LoadBalancedView',
             'AsyncResult',
             'AsyncMapResult'
             ]

IPython/zmq/parallel/clusterdir.py

@@ -1,549 +1,537
 #!/usr/bin/env python
 # encoding: utf-8
 """
 The IPython cluster directory
 """
 
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2008-2009  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 os
 import shutil
 import sys
 import logging
 import warnings
 
 from IPython.config.loader import PyFileConfigLoader
 from IPython.core.application import Application, BaseAppConfigLoader
 from IPython.config.configurable import Configurable
 from IPython.core.crashhandler import CrashHandler
 from IPython.core import release
 from IPython.utils.path import (
     get_ipython_package_dir,
     expand_path
 )
 from IPython.utils.traitlets import Unicode
 
 #-----------------------------------------------------------------------------
-# Warnings control
-#-----------------------------------------------------------------------------
-# Twisted generates annoying warnings with Python 2.6, as will do other code
-# that imports 'sets' as of today
-warnings.filterwarnings('ignore', 'the sets module is deprecated',
-                        DeprecationWarning )
-
-# This one also comes from Twisted
-warnings.filterwarnings('ignore', 'the sha module is deprecated',
-                        DeprecationWarning)
-
-#-----------------------------------------------------------------------------
 # Module errors
 #-----------------------------------------------------------------------------
 
 class ClusterDirError(Exception):
     pass
 
 
 class PIDFileError(Exception):
     pass
 
 
 #-----------------------------------------------------------------------------
 # Class for managing cluster directories
 #-----------------------------------------------------------------------------
 
 class ClusterDir(Configurable):
     """An object to manage the cluster directory and its resources.
 
     The cluster directory is used by :command:`ipengine`, 
     :command:`ipcontroller` and :command:`ipclsuter` to manage the
     configuration, logging and security of these applications.
 
     This object knows how to find, create and manage these directories. This
     should be used by any code that want's to handle cluster directories.
     """
 
     security_dir_name = Unicode('security')
     log_dir_name = Unicode('log')
     pid_dir_name = Unicode('pid')
     security_dir = Unicode(u'')
     log_dir = Unicode(u'')
     pid_dir = Unicode(u'')
     location = Unicode(u'')
 
     def __init__(self, location=u''):
         super(ClusterDir, self).__init__(location=location)
 
     def _location_changed(self, name, old, new):
         if not os.path.isdir(new):
             os.makedirs(new)
         self.security_dir = os.path.join(new, self.security_dir_name)
         self.log_dir = os.path.join(new, self.log_dir_name)
         self.pid_dir = os.path.join(new, self.pid_dir_name)
         self.check_dirs()
 
     def _log_dir_changed(self, name, old, new):
         self.check_log_dir()
 
     def check_log_dir(self):
         if not os.path.isdir(self.log_dir):
             os.mkdir(self.log_dir)
 
     def _security_dir_changed(self, name, old, new):
         self.check_security_dir()
 
     def check_security_dir(self):
         if not os.path.isdir(self.security_dir):
             os.mkdir(self.security_dir, 0700)
         os.chmod(self.security_dir, 0700)
 
     def _pid_dir_changed(self, name, old, new):
         self.check_pid_dir()
 
     def check_pid_dir(self):
         if not os.path.isdir(self.pid_dir):
             os.mkdir(self.pid_dir, 0700)
         os.chmod(self.pid_dir, 0700)
 
     def check_dirs(self):
         self.check_security_dir()
         self.check_log_dir()
         self.check_pid_dir()
 
     def load_config_file(self, filename):
         """Load a config file from the top level of the cluster dir.
 
         Parameters
         ----------
         filename : unicode or str
             The filename only of the config file that must be located in
             the top-level of the cluster directory.
         """
         loader = PyFileConfigLoader(filename, self.location)
         return loader.load_config()
 
     def copy_config_file(self, config_file, path=None, overwrite=False):
         """Copy a default config file into the active cluster directory.
 
         Default configuration files are kept in :mod:`IPython.config.default`.
         This function moves these from that location to the working cluster
         directory.
         """
         if path is None:
             import IPython.config.default
             path = IPython.config.default.__file__.split(os.path.sep)[:-1]
             path = os.path.sep.join(path)
         src = os.path.join(path, config_file)
         dst = os.path.join(self.location, config_file)
         if not os.path.isfile(dst) or overwrite:
             shutil.copy(src, dst)
 
     def copy_all_config_files(self, path=None, overwrite=False):
         """Copy all config files into the active cluster directory."""
         for f in [u'ipcontroller_config.py', u'ipengine_config.py',
                   u'ipcluster_config.py']:
             self.copy_config_file(f, path=path, overwrite=overwrite)
 
     @classmethod
     def create_cluster_dir(csl, cluster_dir):
         """Create a new cluster directory given a full path.
 
         Parameters
         ----------
         cluster_dir : str
             The full path to the cluster directory.  If it does exist, it will
             be used.  If not, it will be created.
         """
         return ClusterDir(location=cluster_dir)
 
     @classmethod
     def create_cluster_dir_by_profile(cls, path, profile=u'default'):
         """Create a cluster dir by profile name and path.
 
         Parameters
         ----------
         path : str
             The path (directory) to put the cluster directory in.
         profile : str
             The name of the profile.  The name of the cluster directory will
             be "clusterz_<profile>".
         """
         if not os.path.isdir(path):
             raise ClusterDirError('Directory not found: %s' % path)
         cluster_dir = os.path.join(path, u'clusterz_' + profile)
         return ClusterDir(location=cluster_dir)
 
     @classmethod
     def find_cluster_dir_by_profile(cls, ipython_dir, profile=u'default'):
         """Find an existing cluster dir by profile name, return its ClusterDir.
 
         This searches through a sequence of paths for a cluster dir.  If it
         is not found, a :class:`ClusterDirError` exception will be raised.
 
         The search path algorithm is:
         1. ``os.getcwd()``
         2. ``ipython_dir``
         3. The directories found in the ":" separated 
            :env:`IPCLUSTER_DIR_PATH` environment variable.
 
         Parameters
         ----------
         ipython_dir : unicode or str
             The IPython directory to use.
         profile : unicode or str
             The name of the profile.  The name of the cluster directory
             will be "clusterz_<profile>".
         """
         dirname = u'clusterz_' + profile
         cluster_dir_paths = os.environ.get('IPCLUSTER_DIR_PATH','')
         if cluster_dir_paths:
             cluster_dir_paths = cluster_dir_paths.split(':')
         else:
             cluster_dir_paths = []
         paths = [os.getcwd(), ipython_dir] + cluster_dir_paths
         for p in paths:
             cluster_dir = os.path.join(p, dirname)
             if os.path.isdir(cluster_dir):
                 return ClusterDir(location=cluster_dir)
         else:
             raise ClusterDirError('Cluster directory not found in paths: %s' % dirname)
 
     @classmethod
     def find_cluster_dir(cls, cluster_dir):
         """Find/create a cluster dir and return its ClusterDir.
 
         This will create the cluster directory if it doesn't exist.
 
         Parameters
         ----------
         cluster_dir : unicode or str
             The path of the cluster directory.  This is expanded using
             :func:`IPython.utils.genutils.expand_path`.
         """
         cluster_dir = expand_path(cluster_dir)
         if not os.path.isdir(cluster_dir):
             raise ClusterDirError('Cluster directory not found: %s' % cluster_dir)
         return ClusterDir(location=cluster_dir)
 
 
 #-----------------------------------------------------------------------------
 # Command line options
 #-----------------------------------------------------------------------------
 
 class ClusterDirConfigLoader(BaseAppConfigLoader):
 
     def _add_cluster_profile(self, parser):
         paa = parser.add_argument
         paa('-p', '--profile',
             dest='Global.profile',type=unicode,
             help=
             """The string name of the profile to be used. This determines the name
             of the cluster dir as: cluster_<profile>. The default profile is named
             'default'.  The cluster directory is resolve this way if the
             --cluster-dir option is not used.""",
             metavar='Global.profile')
 
     def _add_cluster_dir(self, parser):
         paa = parser.add_argument
         paa('--cluster-dir',
             dest='Global.cluster_dir',type=unicode,
             help="""Set the cluster dir. This overrides the logic used by the
             --profile option.""",
             metavar='Global.cluster_dir')
 
     def _add_work_dir(self, parser):
         paa = parser.add_argument
         paa('--work-dir',
             dest='Global.work_dir',type=unicode,
             help='Set the working dir for the process.',
             metavar='Global.work_dir')
 
     def _add_clean_logs(self, parser):
         paa = parser.add_argument
         paa('--clean-logs',
             dest='Global.clean_logs', action='store_true',
             help='Delete old log flies before starting.')
 
     def _add_no_clean_logs(self, parser):
         paa = parser.add_argument
         paa('--no-clean-logs',
             dest='Global.clean_logs', action='store_false',
             help="Don't Delete old log flies before starting.")
 
     def _add_arguments(self):
         super(ClusterDirConfigLoader, self)._add_arguments()
         self._add_cluster_profile(self.parser)
         self._add_cluster_dir(self.parser)
         self._add_work_dir(self.parser)
         self._add_clean_logs(self.parser)
         self._add_no_clean_logs(self.parser)
 
 
 #-----------------------------------------------------------------------------
 # Crash handler for this application
 #-----------------------------------------------------------------------------
 
 
 _message_template = """\
 Oops, $self.app_name crashed. We do our best to make it stable, but...
 
 A crash report was automatically generated with the following information:
   - A verbatim copy of the crash traceback.
   - Data on your current $self.app_name configuration.
 
 It was left in the file named:
 \t'$self.crash_report_fname'
 If you can email this file to the developers, the information in it will help
 them in understanding and correcting the problem.
 
 You can mail it to: $self.contact_name at $self.contact_email
 with the subject '$self.app_name Crash Report'.
 
 If you want to do it now, the following command will work (under Unix):
 mail -s '$self.app_name Crash Report' $self.contact_email < $self.crash_report_fname
 
 To ensure accurate tracking of this issue, please file a report about it at:
 $self.bug_tracker
 """
 
 class ClusterDirCrashHandler(CrashHandler):
     """sys.excepthook for IPython itself, leaves a detailed report on disk."""
 
     message_template = _message_template
 
     def __init__(self, app):
         contact_name = release.authors['Brian'][0]
         contact_email = release.authors['Brian'][1]
         bug_tracker = 'http://github.com/ipython/ipython/issues'
         super(ClusterDirCrashHandler,self).__init__(
             app, contact_name, contact_email, bug_tracker
         )
 
 
 #-----------------------------------------------------------------------------
 # Main application
 #-----------------------------------------------------------------------------
 
 class ApplicationWithClusterDir(Application):
     """An application that puts everything into a cluster directory.
 
     Instead of looking for things in the ipython_dir, this type of application
     will use its own private directory called the "cluster directory"
     for things like config files, log files, etc.
 
     The cluster directory is resolved as follows:
 
     * If the ``--cluster-dir`` option is given, it is used.
     * If ``--cluster-dir`` is not given, the application directory is 
       resolve using the profile name as ``cluster_<profile>``. The search 
       path for this directory is then i) cwd if it is found there
       and ii) in ipython_dir otherwise.
 
     The config file for the application is to be put in the cluster
     dir and named the value of the ``config_file_name`` class attribute.
     """
 
     command_line_loader = ClusterDirConfigLoader
     crash_handler_class = ClusterDirCrashHandler
     auto_create_cluster_dir = True
     # temporarily override default_log_level to DEBUG
     default_log_level = logging.DEBUG
 
     def create_default_config(self):
         super(ApplicationWithClusterDir, self).create_default_config()
         self.default_config.Global.profile = u'default'
         self.default_config.Global.cluster_dir = u''
         self.default_config.Global.work_dir = os.getcwd()
         self.default_config.Global.log_to_file = False
         self.default_config.Global.log_url = None
         self.default_config.Global.clean_logs = False
 
     def find_resources(self):
         """This resolves the cluster directory.
 
         This tries to find the cluster directory and if successful, it will
         have done:
         * Sets ``self.cluster_dir_obj`` to the :class:`ClusterDir` object for 
           the application.
         * Sets ``self.cluster_dir`` attribute of the application and config
           objects.
 
         The algorithm used for this is as follows:
         1. Try ``Global.cluster_dir``.
         2. Try using ``Global.profile``.
         3. If both of these fail and ``self.auto_create_cluster_dir`` is
            ``True``, then create the new cluster dir in the IPython directory.
         4. If all fails, then raise :class:`ClusterDirError`.
         """
 
         try:
             cluster_dir = self.command_line_config.Global.cluster_dir
         except AttributeError:
             cluster_dir = self.default_config.Global.cluster_dir
         cluster_dir = expand_path(cluster_dir)
         try:
             self.cluster_dir_obj = ClusterDir.find_cluster_dir(cluster_dir)
         except ClusterDirError:
             pass
         else:
             self.log.info('Using existing cluster dir: %s' % \
                 self.cluster_dir_obj.location
             )
             self.finish_cluster_dir()
             return
 
         try:
             self.profile = self.command_line_config.Global.profile
         except AttributeError:
             self.profile = self.default_config.Global.profile
         try:
             self.cluster_dir_obj = ClusterDir.find_cluster_dir_by_profile(
                 self.ipython_dir, self.profile)
         except ClusterDirError:
             pass
         else:
             self.log.info('Using existing cluster dir: %s' % \
                 self.cluster_dir_obj.location
             )
             self.finish_cluster_dir()
             return
 
         if self.auto_create_cluster_dir:
             self.cluster_dir_obj = ClusterDir.create_cluster_dir_by_profile(
                 self.ipython_dir, self.profile
             )
             self.log.info('Creating new cluster dir: %s' % \
                 self.cluster_dir_obj.location
             )
             self.finish_cluster_dir()
         else:
             raise ClusterDirError('Could not find a valid cluster directory.')
 
     def finish_cluster_dir(self):
         # Set the cluster directory
         self.cluster_dir = self.cluster_dir_obj.location
         
         # These have to be set because they could be different from the one
         # that we just computed.  Because command line has the highest
         # priority, this will always end up in the master_config.
         self.default_config.Global.cluster_dir = self.cluster_dir
         self.command_line_config.Global.cluster_dir = self.cluster_dir
 
     def find_config_file_name(self):
         """Find the config file name for this application."""
         # For this type of Application it should be set as a class attribute.
         if not hasattr(self, 'default_config_file_name'):
             self.log.critical("No config filename found")
         else:
             self.config_file_name = self.default_config_file_name
 
     def find_config_file_paths(self):
         # Set the search path to to the cluster directory. We should NOT
         # include IPython.config.default here as the default config files
         # are ALWAYS automatically moved to the cluster directory.
         conf_dir = os.path.join(get_ipython_package_dir(), 'config', 'default')
         self.config_file_paths = (self.cluster_dir,)
 
     def pre_construct(self):
         # The log and security dirs were set earlier, but here we put them
         # into the config and log them.
         config = self.master_config
         sdir = self.cluster_dir_obj.security_dir
         self.security_dir = config.Global.security_dir = sdir
         ldir = self.cluster_dir_obj.log_dir
         self.log_dir = config.Global.log_dir = ldir
         pdir = self.cluster_dir_obj.pid_dir
         self.pid_dir = config.Global.pid_dir = pdir
         self.log.info("Cluster directory set to: %s" % self.cluster_dir)
         config.Global.work_dir = unicode(expand_path(config.Global.work_dir))
         # Change to the working directory. We do this just before construct
         # is called so all the components there have the right working dir.
         self.to_work_dir()
 
     def to_work_dir(self):
         wd = self.master_config.Global.work_dir
         if unicode(wd) != unicode(os.getcwd()):
             os.chdir(wd)
             self.log.info("Changing to working dir: %s" % wd)
 
     def start_logging(self):
         # Remove old log files
         if self.master_config.Global.clean_logs:
             log_dir = self.master_config.Global.log_dir
             for f in os.listdir(log_dir):
                 if f.startswith(self.name + u'-') and f.endswith('.log'):
                     os.remove(os.path.join(log_dir, f))
         # Start logging to the new log file
         if self.master_config.Global.log_to_file:
             log_filename = self.name + u'-' + str(os.getpid()) + u'.log'
             logfile = os.path.join(self.log_dir, log_filename)
             open_log_file = open(logfile, 'w')
         elif self.master_config.Global.log_url:
             open_log_file = None
         else:
             open_log_file = sys.stdout
         logger = logging.getLogger()
         level = self.log_level
         self.log = logger
         # since we've reconnected the logger, we need to reconnect the log-level
         self.log_level = level
         if open_log_file is not None and self._log_handler not in self.log.handlers:
             self.log.addHandler(self._log_handler)
         # log.startLogging(open_log_file)
 
     def write_pid_file(self, overwrite=False):
         """Create a .pid file in the pid_dir with my pid.
 
         This must be called after pre_construct, which sets `self.pid_dir`.
         This raises :exc:`PIDFileError` if the pid file exists already.
         """
         pid_file = os.path.join(self.pid_dir, self.name + u'.pid')
         if os.path.isfile(pid_file):
             pid = self.get_pid_from_file()
             if not overwrite:
                 raise PIDFileError(
                     'The pid file [%s] already exists. \nThis could mean that this '
                     'server is already running with [pid=%s].' % (pid_file, pid)
                 )
         with open(pid_file, 'w') as f:
             self.log.info("Creating pid file: %s" % pid_file)
             f.write(repr(os.getpid())+'\n')
 
     def remove_pid_file(self):
         """Remove the pid file.
 
         This should be called at shutdown by registering a callback with
         :func:`reactor.addSystemEventTrigger`. This needs to return
         ``None``.
         """
         pid_file = os.path.join(self.pid_dir, self.name + u'.pid')
         if os.path.isfile(pid_file):
             try:
                 self.log.info("Removing pid file: %s" % pid_file)
                 os.remove(pid_file)
             except:
                 self.log.warn("Error removing the pid file: %s" % pid_file)
 
     def get_pid_from_file(self):
         """Get the pid from the pid file.
 
         If the  pid file doesn't exist a :exc:`PIDFileError` is raised.
         """
         pid_file = os.path.join(self.pid_dir, self.name + u'.pid')
         if os.path.isfile(pid_file):
             with open(pid_file, 'r') as f:
                 pid = int(f.read().strip())
                 return pid
         else:
             raise PIDFileError('pid file not found: %s' % pid_file)
 

IPython/zmq/parallel/controller.py

@@ -1,345 +1,110
 #!/usr/bin/env python
 """The IPython Controller with 0MQ
-This is the master object that handles connections from engines and clients,
-and monitors traffic through the various queues.
+This is a collection of one Hub and several Schedulers.
 """
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2010  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 os
-import sys
-import time
 import logging
 from multiprocessing import Process
 
 import zmq
-from zmq.eventloop import  ioloop
-from zmq.eventloop.zmqstream import ZMQStream
-# from zmq.devices import ProcessMonitoredQueue
 
 # internal:
 from IPython.utils.importstring import import_item
 from IPython.utils.traitlets import Int, Str, Instance, List, Bool
-from IPython.zmq.entry_point import bind_port
 
-from entry_point import (make_base_argument_parser, select_random_ports, split_ports,
-                        connect_logger, parse_url, signal_children, generate_exec_key,
-                        local_logger)
+from entry_point import signal_children
 
 
-import streamsession as session
-import heartmonitor
 from scheduler import launch_scheduler
 from hub import Hub, HubFactory
 
-from dictdb import DictDB
-try:
-    import pymongo
-except ImportError:
-    MongoDB=None
-else:
-    from mongodb import MongoDB
-    
-#-------------------------------------------------------------------------
-# Entry Point
-#-------------------------------------------------------------------------
+#-----------------------------------------------------------------------------
+# Configurable
+#-----------------------------------------------------------------------------
 
-def make_argument_parser():
-    """Make an argument parser"""
-    parser = make_base_argument_parser()
-    
-    parser.add_argument('--client', type=int, metavar='PORT', default=0,
-                        help='set the XREP port for clients [default: random]')
-    parser.add_argument('--notice', type=int, metavar='PORT', default=0,
-                        help='set the PUB socket for registration notification [default: random]')
-    parser.add_argument('--hb', type=str, metavar='PORTS',
-                        help='set the 2 ports for heartbeats [default: random]')
-    parser.add_argument('--ping', type=int, default=100,
-                        help='set the heartbeat period in ms [default: 100]')
-    parser.add_argument('--monitor', type=int, metavar='PORT', default=0,
-                        help='set the SUB port for queue monitoring [default: random]')
-    parser.add_argument('--mux', type=str, metavar='PORTS',
-                        help='set the XREP ports for the MUX queue [default: random]')
-    parser.add_argument('--task', type=str, metavar='PORTS',
-                        help='set the XREP/XREQ ports for the task queue [default: random]')
-    parser.add_argument('--control', type=str, metavar='PORTS',
-                        help='set the XREP ports for the control queue [default: random]')
-    parser.add_argument('--iopub', type=str, metavar='PORTS',
-                        help='set the PUB/SUB ports for the iopub relay [default: random]')
-    parser.add_argument('--scheduler', type=str, default='lru',
-                        choices = ['pure', 'lru', 'plainrandom', 'weighted', 'twobin','leastload'],
-                        help='select the task scheduler  [default: Python LRU]')
-    parser.add_argument('--mongodb', action='store_true',
-                        help='Use MongoDB task storage [default: in-memory]')
-    parser.add_argument('--session', type=str, default=None,
-                        help='Manually specify the session id.')
-    
-    return parser
 
 class ControllerFactory(HubFactory):
     """Configurable for setting up a Hub and Schedulers."""
     
     scheme = Str('pure', config=True)
     usethreads = Bool(False, config=True)
     
     # internal
     children = List()
     mq_class = Str('zmq.devices.ProcessMonitoredQueue')
     
     def _update_mq(self):
         self.mq_class = 'zmq.devices.%sMonitoredQueue'%('Thread' if self.usethreads else 'Process')
         
     def __init__(self, **kwargs):
         super(ControllerFactory, self).__init__(**kwargs)
         self.subconstructors.append(self.construct_schedulers)
         self._update_mq()
         self.on_trait_change(self._update_mq, 'usethreads')
     
     def start(self):
         super(ControllerFactory, self).start()
         for child in self.children:
             child.start()
         if not self.usethreads:
             signal_children([ getattr(c, 'launcher', c) for c in self.children ])
         
     
     def construct_schedulers(self):
         children = self.children
         mq = import_item(self.mq_class)
         
         # IOPub relay (in a Process)
         q = mq(zmq.PUB, zmq.SUB, zmq.PUB, 'N/A','iopub')
         q.bind_in(self.client_addrs['iopub'])
         q.bind_out(self.engine_addrs['iopub'])
         q.setsockopt_out(zmq.SUBSCRIBE, '')
         q.connect_mon(self.monitor_url)
         q.daemon=True
         children.append(q)
 
         # Multiplexer Queue (in a Process)
         q = mq(zmq.XREP, zmq.XREP, zmq.PUB, 'in', 'out')
         q.bind_in(self.client_addrs['mux'])
         q.bind_out(self.engine_addrs['mux'])
         q.connect_mon(self.monitor_url)
         q.daemon=True
         children.append(q)
 
         # Control Queue (in a Process)
         q = mq(zmq.XREP, zmq.XREP, zmq.PUB, 'incontrol', 'outcontrol')
         q.bind_in(self.client_addrs['control'])
         q.bind_out(self.engine_addrs['control'])
         q.connect_mon(self.monitor_url)
         q.daemon=True
         children.append(q)
         # Task Queue (in a Process)
         if self.scheme == 'pure':
             logging.warn("task::using pure XREQ Task scheduler")
             q = mq(zmq.XREP, zmq.XREQ, zmq.PUB, 'intask', 'outtask')
             q.bind_in(self.client_addrs['task'])
             q.bind_out(self.engine_addrs['task'])
             q.connect_mon(self.monitor_url)
             q.daemon=True
             children.append(q)
         elif self.scheme == 'none':
             logging.warn("task::using no Task scheduler")
             
         else:
             logging.warn("task::using Python %s Task scheduler"%self.scheme)
             sargs = (self.client_addrs['task'], self.engine_addrs['task'], self.monitor_url, self.client_addrs['notification'])
             q = Process(target=launch_scheduler, args=sargs, kwargs = dict(scheme=self.scheme))
             q.daemon=True
             children.append(q)
-        
-    
-def main(argv=None):
-    """DO NOT USE ME ANYMORE"""
-    
-    parser = make_argument_parser()
-    
-    args = parser.parse_args(argv)
-    parse_url(args)
-    
-    iface="%s://%s"%(args.transport,args.ip)+':%i'
-    
-    random_ports = 0
-    if args.hb:
-        hb = split_ports(args.hb, 2)
-    else:
-        hb = select_random_ports(2)
-    if args.mux:
-        mux = split_ports(args.mux, 2)
-    else:
-        mux = None
-        random_ports += 2
-    if args.iopub:
-        iopub = split_ports(args.iopub, 2)
-    else:
-        iopub = None
-        random_ports += 2
-    if args.task:
-        task = split_ports(args.task, 2)
-    else:
-        task = None
-        random_ports += 2
-    if args.control:
-        control = split_ports(args.control, 2)
-    else:
-        control = None
-        random_ports += 2
-    
-    ctx = zmq.Context()
-    loop = ioloop.IOLoop.instance()
-    
-    
-    # Registrar socket
-    reg = ZMQStream(ctx.socket(zmq.XREP), loop)
-    regport = bind_port(reg, args.ip, args.regport)
-    
-    ### Engine connections ###
-    
-    # heartbeat
-    hpub = ctx.socket(zmq.PUB)
-    bind_port(hpub, args.ip, hb[0])
-    hrep = ctx.socket(zmq.XREP)
-    bind_port(hrep, args.ip, hb[1])
-    
-    hmon = heartmonitor.HeartMonitor(loop, ZMQStream(hpub,loop), ZMQStream(hrep,loop),args.ping)
-    hmon.start()
-    
-    ### Client connections ###
-    # Clientele socket
-    c = ZMQStream(ctx.socket(zmq.XREP), loop)
-    cport = bind_port(c, args.ip, args.client)
-    # Notifier socket
-    n = ZMQStream(ctx.socket(zmq.PUB), loop)
-    nport = bind_port(n, args.ip, args.notice)
-    
-    ### Key File ###
-    if args.execkey and not os.path.isfile(args.execkey):
-            generate_exec_key(args.execkey)
-    
-    thesession = session.StreamSession(username=args.ident or "controller", 
-                    keyfile=args.execkey, session=args.session)
-    
-    ### build and launch the queues ###
-    
-    # monitor socket
-    sub = ctx.socket(zmq.SUB)
-    sub.setsockopt(zmq.SUBSCRIBE, "")
-    monport = bind_port(sub, args.ip, args.monitor)
-    sub = ZMQStream(sub, loop)
-    
-    ports = select_random_ports(random_ports)
-    children = []
-    
-    # IOPub relay (in a Process)
-    if not iopub:
-        iopub = (ports.pop(),ports.pop())
-    q = ProcessMonitoredQueue(zmq.SUB, zmq.PUB, zmq.PUB, 'iopub', 'N/A')
-    q.bind_in(iface%iopub[1])
-    q.bind_out(iface%iopub[0])
-    q.setsockopt_in(zmq.SUBSCRIBE, '')
-    q.connect_mon(iface%monport)
-    q.daemon=True
-    q.start()
-    children.append(q.launcher)
-    
-    # Multiplexer Queue (in a Process)
-    if not mux:
-        mux = (ports.pop(),ports.pop())
-    q = ProcessMonitoredQueue(zmq.XREP, zmq.XREP, zmq.PUB, 'in', 'out')
-    q.bind_in(iface%mux[0])
-    q.bind_out(iface%mux[1])
-    q.connect_mon(iface%monport)
-    q.daemon=True
-    q.start()
-    children.append(q.launcher)
-    
-    # Control Queue (in a Process)
-    if not control:
-        control = (ports.pop(),ports.pop())
-    q = ProcessMonitoredQueue(zmq.XREP, zmq.XREP, zmq.PUB, 'incontrol', 'outcontrol')
-    q.bind_in(iface%control[0])
-    q.bind_out(iface%control[1])
-    q.connect_mon(iface%monport)
-    q.daemon=True
-    q.start()
-    children.append(q.launcher)
-    # Task Queue (in a Process)
-    if not task:
-        task = (ports.pop(),ports.pop())
-    if args.scheduler == 'pure':
-        q = ProcessMonitoredQueue(zmq.XREP, zmq.XREQ, zmq.PUB, 'intask', 'outtask')
-        q.bind_in(iface%task[0])
-        q.bind_out(iface%task[1])
-        q.connect_mon(iface%monport)
-        q.daemon=True
-        q.start()
-        children.append(q.launcher)
-    else:
-        log_addr = iface%args.logport if args.logport else None
-        sargs = (iface%task[0], iface%task[1], iface%monport, iface%nport,
-                    log_addr, args.loglevel, args.scheduler)
-        print (sargs)
-        q = Process(target=launch_scheduler, args=sargs)
-        q.daemon=True
-        q.start()
-        children.append(q)
-    
-    if args.mongodb:
-        from mongodb import MongoDB
-        db = MongoDB(thesession.session)
-    else:
-        db = DictDB()
-    time.sleep(.25)
-    
-    # build connection dicts
-    engine_addrs = {
-        'control' : iface%control[1],
-        'mux': iface%mux[1],
-        'heartbeat': (iface%hb[0], iface%hb[1]),
-        'task' : iface%task[1],
-        'iopub' : iface%iopub[1],
-        'monitor' : iface%monport,
-        }
-    
-    client_addrs = {
-        'control' : iface%control[0],
-        'query': iface%cport,
-        'mux': iface%mux[0],
-        'task' : iface%task[0],
-        'iopub' : iface%iopub[0],
-        'notification': iface%nport
-        }
-    
-    # setup logging
-    if args.logport:
-        connect_logger(ctx, iface%args.logport, root="controller", loglevel=args.loglevel)
-    else:
-        local_logger(args.loglevel)
-        
-    # register relay of signals to the children
-    signal_children(children)
-    hub = Hub(loop=loop, session=thesession, monitor=sub, heartmonitor=hmon,
-                registrar=reg, clientele=c, notifier=n, db=db,
-                engine_addrs=engine_addrs, client_addrs=client_addrs)
-    
-    dc = ioloop.DelayedCallback(lambda : print("Controller started..."), 100, loop)
-    dc.start()
-    try:
-        loop.start()
-    except KeyboardInterrupt:
-        print ("interrupted, exiting...", file=sys.__stderr__)
-    
 
-if __name__ == '__main__':
-    main()

IPython/zmq/parallel/engine.py

@@ -1,190 +1,141
 #!/usr/bin/env python
 """A simple engine that talks to a controller over 0MQ.
 it handles registration, etc. and launches a kernel
 connected to the Controller's queue(s).
 """
 from __future__ import print_function
 import sys
 import time
-import traceback
 import uuid
 import logging
 from pprint import pprint
 
 import zmq
 from zmq.eventloop import ioloop, zmqstream
 
 # internal
 from IPython.config.configurable import Configurable
 from IPython.utils.traitlets import Instance, Str, Dict, Int, Type
 # from IPython.utils.localinterfaces import LOCALHOST 
 
 from factory import RegistrationFactory
 
-from streamsession import Message, StreamSession
-from streamkernel import Kernel, make_kernel
+from streamsession import Message
+from streamkernel import Kernel
 import heartmonitor
-from entry_point import (make_base_argument_parser, connect_engine_logger, parse_url,
-                        local_logger)
-# import taskthread
 
 def printer(*msg):
     # print (logging.handlers, file=sys.__stdout__)
     logging.info(str(msg))
 
 class EngineFactory(RegistrationFactory):
     """IPython engine"""
     
     # configurables:
     user_ns=Dict(config=True)
     out_stream_factory=Type('IPython.zmq.iostream.OutStream', config=True)
     display_hook_factory=Type('IPython.zmq.displayhook.DisplayHook', config=True)
     
     # not configurable:
     id=Int(allow_none=True)
     registrar=Instance('zmq.eventloop.zmqstream.ZMQStream')
     kernel=Instance(Kernel)
     
     
     def __init__(self, **kwargs):
         super(EngineFactory, self).__init__(**kwargs)
         ctx = self.context
         
         reg = ctx.socket(zmq.PAIR)
         reg.setsockopt(zmq.IDENTITY, self.ident)
         reg.connect(self.url)
         self.registrar = zmqstream.ZMQStream(reg, self.loop)
         
     def register(self):
         """send the registration_request"""
         
         logging.info("registering")
         content = dict(queue=self.ident, heartbeat=self.ident, control=self.ident)
         self.registrar.on_recv(self.complete_registration)
         # print (self.session.key)
         self.session.send(self.registrar, "registration_request",content=content)
     
     def complete_registration(self, msg):
         # print msg
         ctx = self.context
         loop = self.loop
         identity = self.ident
         print (identity)
         
         idents,msg = self.session.feed_identities(msg)
         msg = Message(self.session.unpack_message(msg))
         
         if msg.content.status == 'ok':
             self.id = int(msg.content.id)
             
             # create Shell Streams (MUX, Task, etc.):
             queue_addr = msg.content.mux
             shell_addrs = [ str(queue_addr) ]
             task_addr = msg.content.task
             if task_addr:
                 shell_addrs.append(str(task_addr))
             shell_streams = []
             for addr in shell_addrs:
                 stream = zmqstream.ZMQStream(ctx.socket(zmq.PAIR), loop)
                 stream.setsockopt(zmq.IDENTITY, identity)
                 stream.connect(addr)
                 shell_streams.append(stream)
             
             # control stream:
             control_addr = str(msg.content.control)
             control_stream = zmqstream.ZMQStream(ctx.socket(zmq.PAIR), loop)
             control_stream.setsockopt(zmq.IDENTITY, identity)
             control_stream.connect(control_addr)
             
             # create iopub stream:
             iopub_addr = msg.content.iopub
             iopub_stream = zmqstream.ZMQStream(ctx.socket(zmq.PUB), loop)
             iopub_stream.setsockopt(zmq.IDENTITY, identity)
             iopub_stream.connect(iopub_addr)
             
             # launch heartbeat
             hb_addrs = msg.content.heartbeat
             # print (hb_addrs)
             
             # # Redirect input streams and set a display hook.
-            # if self.out_stream_factory:
-            #     sys.stdout = self.out_stream_factory(self.session, iopub_stream, u'stdout')
-            #     sys.stdout.topic = 'engine.%i.stdout'%self.id
-            #     sys.stderr = self.out_stream_factory(self.session, iopub_stream, u'stderr')
-            #     sys.stderr.topic = 'engine.%i.stderr'%self.id
-            # if self.display_hook_factory:
-            #     sys.displayhook = self.display_hook_factory(self.session, iopub_stream)
-            #     sys.displayhook.topic = 'engine.%i.pyout'%self.id
+            if self.out_stream_factory:
+                sys.stdout = self.out_stream_factory(self.session, iopub_stream, u'stdout')
+                sys.stdout.topic = 'engine.%i.stdout'%self.id
+                sys.stderr = self.out_stream_factory(self.session, iopub_stream, u'stderr')
+                sys.stderr.topic = 'engine.%i.stderr'%self.id
+            if self.display_hook_factory:
+                sys.displayhook = self.display_hook_factory(self.session, iopub_stream)
+                sys.displayhook.topic = 'engine.%i.pyout'%self.id
             
-            # ioloop.DelayedCallback(self.heart.start, 1000, self.loop).start()
             self.kernel = Kernel(int_id=self.id, ident=self.ident, session=self.session, 
                     control_stream=control_stream,
                     shell_streams=shell_streams, iopub_stream=iopub_stream, loop=loop,
                     user_ns = self.user_ns, config=self.config)
             self.kernel.start()
             
             heart = heartmonitor.Heart(*map(str, hb_addrs), heart_id=identity)
+            # ioloop.DelayedCallback(heart.start, 1000, self.loop).start()
             heart.start()
             
             
         else:
             logging.error("Registration Failed: %s"%msg)
             raise Exception("Registration Failed: %s"%msg)
         
         logging.info("Completed registration with id %i"%self.id)
     
     
     def unregister(self):
         self.session.send(self.registrar, "unregistration_request", content=dict(id=self.id))
         time.sleep(1)
         sys.exit(0)
     
     def start(self):
         dc = ioloop.DelayedCallback(self.register, 0, self.loop)
         dc.start()
 
-
-
-def main(argv=None, user_ns=None):
-    """DO NOT USE ME ANYMORE"""
-    parser = make_base_argument_parser()
-    
-    args = parser.parse_args(argv)
-    
-    parse_url(args)
-    
-    iface="%s://%s"%(args.transport,args.ip)+':%i'
-    
-    loop = ioloop.IOLoop.instance()
-    session = StreamSession(keyfile=args.execkey)
-    # print (session.key)
-    ctx = zmq.Context()
-
-    # setup logging
-    
-    reg_conn = iface % args.regport
-    print (reg_conn, file=sys.__stdout__)
-    print ("Starting the engine...", file=sys.__stderr__)
-    
-    reg = ctx.socket(zmq.PAIR)
-    reg.connect(reg_conn)
-    reg = zmqstream.ZMQStream(reg, loop)
-    
-    e = Engine(context=ctx, loop=loop, session=session, registrar=reg, 
-            ident=args.ident or '', user_ns=user_ns)
-    if args.logport:
-        print ("connecting logger to %s"%(iface%args.logport), file=sys.__stdout__)
-        connect_engine_logger(ctx, iface%args.logport, e, loglevel=args.loglevel)
-    else:
-        local_logger(args.loglevel)
-    
-    dc = ioloop.DelayedCallback(e.start, 0, loop)
-    dc.start()
-    try:
-        loop.start()
-    except KeyboardInterrupt:
-        print ("interrupted, exiting...", file=sys.__stderr__)
-
-# Execution as a script
-if __name__ == '__main__':
-    main()

IPython/zmq/parallel/entry_point.py

@@ -1,167 +1,118
 """ Defines helper functions for creating kernel entry points and process
 launchers.
 
 ************
 NOTE: Most of this module has been deprecated by moving to Configurables
 ************
 """
 
 # Standard library imports.
 import logging
 import atexit
 import sys
 import os
 import stat
 import socket
 from subprocess import Popen, PIPE
 from signal import signal, SIGINT, SIGABRT, SIGTERM
 try:
     from signal import SIGKILL
 except ImportError:
     SIGKILL=None
 
 # System library imports.
 import zmq
 from zmq.log import handlers
 # Local imports.
 from IPython.core.ultratb import FormattedTB
 from IPython.external.argparse import ArgumentParser
 from IPython.zmq.log import EnginePUBHandler
 
-def split_ports(s, n):
-    """Parser helper for multiport strings"""
-    if not s:
-        return tuple([0]*n)
-    ports = map(int, s.split(','))
-    if len(ports) != n:
-        raise ValueError
-    return ports
-
 _random_ports = set()
 
 def select_random_ports(n):
     """Selects and return n random ports that are available."""
     ports = []
     for i in xrange(n):
         sock = socket.socket()
         sock.bind(('', 0))
         while sock.getsockname()[1] in _random_ports:
             sock.close()
             sock = socket.socket()
             sock.bind(('', 0))
         ports.append(sock)
     for i, sock in enumerate(ports):
         port = sock.getsockname()[1]
         sock.close()
         ports[i] = port
         _random_ports.add(port)
     return ports
 
-def parse_url(args):
-    """Ensure args.url contains full transport://interface:port"""
-    if args.url:
-        iface = args.url.split('://',1)
-        if len(args) == 2:
-            args.transport,iface = iface
-        iface = iface.split(':')
-        args.ip = iface[0]
-        if iface[1]:
-            args.regport = iface[1]
-    args.url = "%s://%s:%i"%(args.transport, args.ip,args.regport)
-
 def signal_children(children):
     """Relay interupt/term signals to children, for more solid process cleanup."""
     def terminate_children(sig, frame):
         logging.critical("Got signal %i, terminating children..."%sig)
         for child in children:
             child.terminate()
         
         sys.exit(sig != SIGINT)
         # sys.exit(sig)
     for sig in (SIGINT, SIGABRT, SIGTERM):
         signal(sig, terminate_children)
 
 def generate_exec_key(keyfile):
     import uuid
     newkey = str(uuid.uuid4())
     with open(keyfile, 'w') as f:
         # f.write('ipython-key ')
         f.write(newkey+'\n')
     # set user-only RW permissions (0600)
     # this will have no effect on Windows
     os.chmod(keyfile, stat.S_IRUSR|stat.S_IWUSR)
-        
-
-def make_base_argument_parser():
-    """ Creates an ArgumentParser for the generic arguments supported by all 
-    ipcluster entry points.
-    """
-    
-    parser = ArgumentParser()
-    parser.add_argument('--ip', type=str, default='127.0.0.1',
-                        help='set the controller\'s IP address [default: local]')
-    parser.add_argument('--transport', type=str, default='tcp',
-                        help='set the transport to use [default: tcp]')
-    parser.add_argument('--regport', type=int, metavar='PORT', default=10101,
-                        help='set the XREP port for registration [default: 10101]')
-    parser.add_argument('--logport', type=int, metavar='PORT', default=0,
-                        help='set the PUB port for remote logging [default: log to stdout]')
-    parser.add_argument('--loglevel', type=str, metavar='LEVEL', default=logging.INFO,
-                        help='set the log level [default: INFO]')
-    parser.add_argument('--ident', type=str,
-                        help='set the ZMQ identity [default: random]')
-    parser.add_argument('--packer', type=str, default='json',
-                        choices=['json','pickle'],
-                        help='set the message format method [default: json]')
-    parser.add_argument('--url', type=str,
-                        help='set transport,ip,regport in one arg, e.g. tcp://127.0.0.1:10101')
-    parser.add_argument('--execkey', type=str,
-                        help="File containing key for authenticating requests.")
 
-    return parser
 
 def integer_loglevel(loglevel):
     try:
         loglevel = int(loglevel)
     except ValueError:
         if isinstance(loglevel, str):
             loglevel = getattr(logging, loglevel)
     return loglevel
 
 def connect_logger(context, iface, root="ip", loglevel=logging.DEBUG):
     logger = logging.getLogger()
     if any([isinstance(h, handlers.PUBHandler) for h in logger.handlers]):
         # don't add a second PUBHandler
         return
     loglevel = integer_loglevel(loglevel)
     lsock = context.socket(zmq.PUB)
     lsock.connect(iface)
     handler = handlers.PUBHandler(lsock)
     handler.setLevel(loglevel)
     handler.root_topic = root
     logger.addHandler(handler)
     logger.setLevel(loglevel)
 
 def connect_engine_logger(context, iface, engine, loglevel=logging.DEBUG):
     logger = logging.getLogger()
     if any([isinstance(h, handlers.PUBHandler) for h in logger.handlers]):
         # don't add a second PUBHandler
         return
     loglevel = integer_loglevel(loglevel)
     lsock = context.socket(zmq.PUB)
     lsock.connect(iface)
     handler = EnginePUBHandler(engine, lsock)
     handler.setLevel(loglevel)
     logger.addHandler(handler)
     logger.setLevel(loglevel)
 
 def local_logger(loglevel=logging.DEBUG):
     loglevel = integer_loglevel(loglevel)
     logger = logging.getLogger()
     if any([isinstance(h, logging.StreamHandler) for h in logger.handlers]):
         # don't add a second StreamHandler
         return
     handler = logging.StreamHandler()
     handler.setLevel(loglevel)
     logger.addHandler(handler)
     logger.setLevel(loglevel)

IPython/zmq/parallel/hub.py

@@ -1,1079 +1,1047
 #!/usr/bin/env python
 """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.
 """
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2010  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 sys
 from datetime import datetime
 import time
 import logging
 
 import zmq
 from zmq.eventloop import ioloop
 from zmq.eventloop.zmqstream import ZMQStream
 
 # internal:
 from IPython.config.configurable import Configurable
 from IPython.utils.traitlets import HasTraits, Instance, Int, Str, Dict, Set, List, Bool
 from IPython.utils.importstring import import_item
 
 from entry_point import select_random_ports
 from factory import RegistrationFactory
 
 from streamsession import Message, wrap_exception, ISO8601
 from heartmonitor import HeartMonitor
 from util import validate_url_container
 
 try:
     from pymongo.binary import Binary
 except ImportError:
     MongoDB=None
 else:
     from mongodb import MongoDB
     
 #-----------------------------------------------------------------------------
 # Code
 #-----------------------------------------------------------------------------
 
 def _passer(*args, **kwargs):
     return
 
 def _printer(*args, **kwargs):
     print (args)
     print (kwargs)
 
 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': datetime.strptime(header['date'], ISO8601),
         'client_uuid' : None,
         'engine_uuid' : None,
         'started': None,
         'completed': None,
         'resubmitted': 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 XREQ socket
     registration (str): identity of registration XREQ socket
     heartbeat (str): identity of heartbeat XREQ socket
     """
     id=Int(0)
     queue=Str()
     control=Str()
     registration=Str()
     heartbeat=Str()
     pending=Set()
     
     def __init__(self, **kwargs):
         super(EngineConnector, self).__init__(**kwargs)
         logging.info("engine::Engine Connected: %i"%self.id)
 
 class HubFactory(RegistrationFactory):
     """The Configurable for setting up a Hub."""
     
     # port-pairs for monitoredqueues:
     hb = Instance(list, config=True)
     def _hb_default(self):
         return select_random_ports(2)
     
     mux = Instance(list, config=True)
     def _mux_default(self):
         return select_random_ports(2)
     
     task = Instance(list, config=True)
     def _task_default(self):
         return select_random_ports(2)
     
     control = Instance(list, config=True)
     def _control_default(self):
         return select_random_ports(2)
     
     iopub = Instance(list, config=True)
     def _iopub_default(self):
         return select_random_ports(2)
     
     # single ports:
     mon_port = Instance(int, config=True)
     def _mon_port_default(self):
         return select_random_ports(1)[0]
     
     query_port = Instance(int, config=True)
     def _query_port_default(self):
         return select_random_ports(1)[0]
     
     notifier_port = Instance(int, config=True)
     def _notifier_port_default(self):
         return select_random_ports(1)[0]
     
     ping = Int(1000, config=True) # ping frequency
     
     engine_ip = Str('127.0.0.1', config=True)
     engine_transport = Str('tcp', config=True)
     
     client_ip = Str('127.0.0.1', config=True)
     client_transport = Str('tcp', config=True)
     
     monitor_ip = Str('127.0.0.1', config=True)
     monitor_transport = Str('tcp', config=True)
     
     monitor_url = Str('')
     
     db_class = Str('IPython.zmq.parallel.dictdb.DictDB', config=True)
     
     # not configurable
     db = Instance('IPython.zmq.parallel.dictdb.BaseDB')
     heartmonitor = Instance('IPython.zmq.parallel.heartmonitor.HeartMonitor')
     subconstructors = List()
     _constructed = Bool(False)
     
+    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 _sync_ips(self):
-        self.engine_ip = self.ip
-        self.client_ip = self.ip
-        self.monitor_ip = self.ip
-        self._update_monitor_url()
-    
-    def _sync_transports(self):
-        self.engine_transport = self.transport
-        self.client_transport = self.transport
-        self.monitor_transport = self.transport
+    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()
-        self.on_trait_change(self._sync_ips, 'ip')
-        self.on_trait_change(self._sync_transports, 'transport')
+        # self.on_trait_change(self._sync_ips, 'ip')
+        # self.on_trait_change(self._sync_transports, 'transport')
         self.subconstructors.append(self.construct_hub)
     
     
     def construct(self):
         assert not self._constructed, "already constructed!"
         
         for subc in self.subconstructors:
             subc()
         
         self._constructed = True
         
     
     def start(self):
         assert self._constructed, "must be constructed by self.construct() first!"
         self.heartmonitor.start()
         logging.info("Heartmonitor started")
     
     def construct_hub(self):
         """construct"""
         client_iface = "%s://%s:"%(self.client_transport, self.client_ip) + "%i"
         engine_iface = "%s://%s:"%(self.engine_transport, self.engine_ip) + "%i"
         
         ctx = self.context
         loop = self.loop
         
         # Registrar socket
         reg = ZMQStream(ctx.socket(zmq.XREP), loop)
         reg.bind(client_iface % self.regport)
         logging.info("Hub listening on %s for registration."%(client_iface%self.regport))
         if self.client_ip != self.engine_ip:
             reg.bind(engine_iface % self.regport)
             logging.info("Hub listening on %s for registration."%(engine_iface%self.regport))
         
         ### Engine connections ###
 
         # heartbeat
         hpub = ctx.socket(zmq.PUB)
         hpub.bind(engine_iface % self.hb[0])
         hrep = ctx.socket(zmq.XREP)
         hrep.bind(engine_iface % self.hb[1])
 
         self.heartmonitor = HeartMonitor(loop, ZMQStream(hpub,loop), ZMQStream(hrep,loop), self.ping)
 
         ### Client connections ###
         # Clientele socket
         c = ZMQStream(ctx.socket(zmq.XREP), loop)
         c.bind(client_iface%self.query_port)
         # Notifier socket
         n = ZMQStream(ctx.socket(zmq.PUB), loop)
         n.bind(client_iface%self.notifier_port)
 
         ### build and launch the queues ###
 
         # monitor socket
         sub = ctx.socket(zmq.SUB)
         sub.setsockopt(zmq.SUBSCRIBE, "")
         sub.bind(self.monitor_url)
         sub = ZMQStream(sub, loop)
         
         # connect the db
         self.db = import_item(self.db_class)()
         time.sleep(.25)
 
         # build connection dicts
         self.engine_addrs = {
             'control' : engine_iface%self.control[1],
             'mux': engine_iface%self.mux[1],
             'heartbeat': (engine_iface%self.hb[0], engine_iface%self.hb[1]),
             'task' : engine_iface%self.task[1],
             'iopub' : engine_iface%self.iopub[1],
             # 'monitor' : engine_iface%self.mon_port,
             }
 
         self.client_addrs = {
             'control' : client_iface%self.control[0],
             'query': client_iface%self.query_port,
             'mux': client_iface%self.mux[0],
             'task' : client_iface%self.task[0],
             'iopub' : client_iface%self.iopub[0],
             'notification': client_iface%self.notifier_port
             }
         logging.debug("hub::Hub engine addrs: %s"%self.engine_addrs)
         logging.debug("hub::Hub client addrs: %s"%self.client_addrs)
         self.hub = Hub(loop=loop, session=self.session, monitor=sub, heartmonitor=self.heartmonitor,
                 registrar=reg, clientele=c, notifier=n, db=self.db,
                 engine_addrs=self.engine_addrs, client_addrs=self.client_addrs)
     
 
 class Hub(HasTraits):
     """The IPython Controller Hub with 0MQ connections
     
     Parameters
     ==========
     loop: zmq IOLoop instance
     session: StreamSession object
     <removed> context: zmq context for creating new connections (?)
     queue: ZMQStream for monitoring the command queue (SUB)
     registrar: ZMQStream for engine registration requests (XREP)
     heartbeat: HeartMonitor object checking the pulse of the engines
     clientele: ZMQStream for client connections (XREP)
                 not used for jobs, only query/control commands
     notifier: ZMQStream for broadcasting engine registration changes (PUB)
     db: connection to db for out of memory logging of commands
                 NotImplemented
     engine_addrs: dict of zmq connection information for engines to connect
                 to the queues.
     client_addrs: dict of zmq connection information for engines to connect
                 to the queues.
     """
     # 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
     # mia=None
     incoming_registrations=Dict()
     registration_timeout=Int()
     _idcounter=Int(0)
     
     # objects from constructor:
     loop=Instance(ioloop.IOLoop)
     registrar=Instance(ZMQStream)
     clientele=Instance(ZMQStream)
     monitor=Instance(ZMQStream)
     heartmonitor=Instance(HeartMonitor)
     notifier=Instance(ZMQStream)
     db=Instance(object)
     client_addrs=Dict()
     engine_addrs=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
         registrar: ZMQStream for engine registration
         heartbeat: HeartMonitor object for tracking engines
         # client:
         clientele: ZMQStream for client connections
         # extra:
         db: ZMQStream for db connection (NotImplemented)
         engine_addrs: zmq address/protocol dict for engine connections
         client_addrs: zmq address/protocol dict for client connections
         """
         
         super(Hub, self).__init__(**kwargs)
-        self.ids = set()
-        self.pending = set()
-        # self.keytable={}
-        # self.incoming_registrations={}
-        # self.engines = {}
-        # self.by_ident = {}
-        # self.clients = {}
-        # self.hearts = {}
-        # self.mia = set()
         self.registration_timeout = max(5000, 2*self.heartmonitor.period)
-        # this is the stuff that will move to DB:
-        # self.pending = set() # pending messages, keyed by msg_id
-        # self.queues = {} # pending msg_ids keyed by engine_id
-        # self.tasks = {} # pending msg_ids submitted as tasks, keyed by client_id
-        # self.completed = {} # completed msg_ids keyed by engine_id
-        # self.all_completed = set()
-        # self._idcounter = 0
-        # self.sockets = {}
-        # self.loop = loop
-        # self.session = session
-        # self.registrar = registrar
-        # self.clientele = clientele
-        # self.queue = queue
-        # self.heartmonitor = heartbeat
-        # self.notifier = notifier
-        # self.db = db
         
         # validate connection dicts:
-        # self.client_addrs = client_addrs
         validate_url_container(self.client_addrs)
-        
-        # assert isinstance(self.client_addrs['queue'], str)
-        # assert isinstance(self.client_addrs['control'], str)
-        # self.hb_addrs = hb_addrs
         validate_url_container(self.engine_addrs)
-        # self.engine_addrs = engine_addrs
-        # assert isinstance(self.engine_addrs['queue'], str)
-        # assert isinstance(self.engine_addrs['control'], str)
-        # assert len(engine_addrs['heartbeat']) == 2
         
         # register our callbacks
         self.registrar.on_recv(self.dispatch_register_request)
         self.clientele.on_recv(self.dispatch_client_msg)
         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 = { 'in' : self.save_queue_request,
                                 'out': self.save_queue_result,
                                 'intask': self.save_task_request,
                                 'outtask': self.save_task_result,
                                 'tracktask': self.save_task_destination,
                                 'incontrol': _passer,
                                 'outcontrol': _passer,
                                 'iopub': self.save_iopub_message,
         }
         
         self.client_handlers = {'queue_request': self.queue_status,
                                 'result_request': self.get_results,
                                 'purge_request': self.purge_results,
                                 'load_request': self.check_load,
                                 'resubmit_request': self.resubmit_task,
                                 'shutdown_request': self.shutdown_request,
                                 }
         
         self.registrar_handlers = {'registration_request' : self.register_engine,
                                 'unregistration_request' : self.unregister_engine,
                                 'connection_request': self.connection_request,
         }
         
         logging.info("hub::created hub")
     
     @property
     def _next_id(self):
-        """gemerate a new ID"""
+        """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
             targets = 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(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
     
     def _validate_client_msg(self, msg):
         """validates and unpacks headers of a message. Returns False if invalid,
         (ident, header, parent, content)"""
         client_id = msg[0]
         try:
             msg = self.session.unpack_message(msg[1:], content=True)
         except:
             logging.error("client::Invalid Message %s"%msg, exc_info=True)
             return False
         
         msg_type = msg.get('msg_type', None)
         if msg_type is None:
             return False
         header = msg.get('header')
         # session doesn't handle split content for now:
         return client_id, msg
         
     
     #-----------------------------------------------------------------------------
     # dispatch methods (1 per stream)
     #-----------------------------------------------------------------------------
     
     def dispatch_register_request(self, msg):
         """"""
         logging.debug("registration::dispatch_register_request(%s)"%msg)
         idents,msg = self.session.feed_identities(msg)
         if not idents:
             logging.error("Bad Queue Message: %s"%msg, exc_info=True)
             return
         try:
             msg = self.session.unpack_message(msg,content=True)
         except:
             logging.error("registration::got bad registration message: %s"%msg, exc_info=True)
             return
         
         msg_type = msg['msg_type']
         content = msg['content']
         
         handler = self.registrar_handlers.get(msg_type, None)
         if handler is None:
             logging.error("registration::got bad registration message: %s"%msg)
         else:
             handler(idents, msg)
     
     def dispatch_monitor_traffic(self, msg):
         """all ME and Task queue messages come through here, as well as
         IOPub traffic."""
         logging.debug("monitor traffic: %s"%msg[:2])
         switch = msg[0]
         idents, msg = self.session.feed_identities(msg[1:])
         if not idents:
             logging.error("Bad Monitor Message: %s"%msg)
             return
         handler = self.monitor_handlers.get(switch, None)
         if handler is not None:
             handler(idents, msg)
         else:
             logging.error("Invalid monitor topic: %s"%switch)
         
     
     def dispatch_client_msg(self, msg):
         """Route messages from clients"""
         idents, msg = self.session.feed_identities(msg)
         if not idents:
             logging.error("Bad Client Message: %s"%msg)
             return
         client_id = idents[0]
         try:
             msg = self.session.unpack_message(msg, content=True)
         except:
             content = wrap_exception()
             logging.error("Bad Client Message: %s"%msg, exc_info=True)
             self.session.send(self.clientele, "hub_error", ident=client_id, 
                     content=content)
             return
         
         # print client_id, header, parent, content
         #switch on message type:
         msg_type = msg['msg_type']
         logging.info("client:: client %s requested %s"%(client_id, msg_type))
         handler = self.client_handlers.get(msg_type, None)
         try:
             assert handler is not None, "Bad Message Type: %s"%msg_type
         except:
             content = wrap_exception()
             logging.error("Bad Message Type: %s"%msg_type, exc_info=True)
             self.session.send(self.clientele, "hub_error", ident=client_id, 
                     content=content)
             return
         else:
             handler(client_id, 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."""
         logging.debug("heartbeat::handle_new_heart(%r)"%heart)
         if heart not in self.incoming_registrations:
             logging.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"""
         logging.debug("heartbeat::handle_heart_failure(%r)"%heart)
         eid = self.hearts.get(heart, None)
         queue = self.engines[eid].queue
         if eid is None:
             logging.info("heartbeat::ignoring heart failure %r"%heart)
         else:
             self.unregister_engine(heart, dict(content=dict(id=eid, queue=queue)))
     
     #----------------------- MUX Queue Traffic ------------------------------
     
     def save_queue_request(self, idents, msg):
         if len(idents) < 2:
             logging.error("invalid identity prefix: %s"%idents)
             return
         queue_id, client_id = idents[:2]
         try:
             msg = self.session.unpack_message(msg, content=False)
         except:
             logging.error("queue::client %r sent invalid message to %r: %s"%(client_id, queue_id, msg), exc_info=True)
             return
         
         eid = self.by_ident.get(queue_id, None)
         if eid is None:
             logging.error("queue::target %r not registered"%queue_id)
             logging.debug("queue::    valid are: %s"%(self.by_ident.keys()))
             return
             
         header = msg['header']
         msg_id = header['msg_id']
         record = init_record(msg)
         record['engine_uuid'] = queue_id
         record['client_uuid'] = client_id
         record['queue'] = 'mux'
         if MongoDB is not None and isinstance(self.db, MongoDB):
             record['buffers'] = map(Binary, record['buffers'])
         self.pending.add(msg_id)
         self.queues[eid].append(msg_id)
         self.db.add_record(msg_id, record)
     
     def save_queue_result(self, idents, msg):
         if len(idents) < 2:
             logging.error("invalid identity prefix: %s"%idents)
             return
             
         client_id, queue_id = idents[:2]
         try:
             msg = self.session.unpack_message(msg, content=False)
         except:
             logging.error("queue::engine %r sent invalid message to %r: %s"%(
                     queue_id,client_id, msg), exc_info=True)
             return
         
         eid = self.by_ident.get(queue_id, None)
         if eid is None:
             logging.error("queue::unknown engine %r is sending a reply: "%queue_id)
             logging.debug("queue::       %s"%msg[2:])
             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)
             rheader = msg['header']
             completed = datetime.strptime(rheader['date'], ISO8601)
             started = rheader.get('started', None)
             if started is not None:
                 started = datetime.strptime(started, ISO8601)
             result = {
                 'result_header' : rheader,
                 'result_content': msg['content'],
                 'started' : started,
                 'completed' : completed
             }
             if MongoDB is not None and isinstance(self.db, MongoDB):
                 result['result_buffers'] = map(Binary, msg['buffers'])
             else:
                 result['result_buffers'] = msg['buffers']
             self.db.update_record(msg_id, result)
         else:
             logging.debug("queue:: unknown msg finished %s"%msg_id)
             
     #--------------------- Task Queue Traffic ------------------------------
     
     def save_task_request(self, idents, msg):
         """Save the submission of a task."""
         client_id = idents[0]
         
         try:
             msg = self.session.unpack_message(msg, content=False)
         except:
             logging.error("task::client %r sent invalid task message: %s"%(
                     client_id, msg), exc_info=True)
             return
         record = init_record(msg)
         if MongoDB is not None and isinstance(self.db, MongoDB):
             record['buffers'] = map(Binary, record['buffers'])
         record['client_uuid'] = client_id
         record['queue'] = 'task'
         header = msg['header']
         msg_id = header['msg_id']
         self.pending.add(msg_id)
         self.db.add_record(msg_id, record)
     
     def save_task_result(self, idents, msg):
         """save the result of a completed task."""
         client_id = idents[0]
         try:
             msg = self.session.unpack_message(msg, content=False)
         except:
             logging.error("task::invalid task result message send to %r: %s"%(
                     client_id, msg), exc_info=True)
             raise
             return
         
         parent = msg['parent_header']
         if not parent:
             # print msg
             logging.warn("Task %r had no parent!"%msg)
             return
         msg_id = parent['msg_id']
         
         header = msg['header']
         engine_uuid = header.get('engine', None)
         eid = self.by_ident.get(engine_uuid, None)
         
         if msg_id in self.pending:
             self.pending.remove(msg_id)
             self.all_completed.add(msg_id)
             if eid is not None:
                 self.completed[eid].append(msg_id)
                 if msg_id in self.tasks[eid]:
                     self.tasks[eid].remove(msg_id)
             completed = datetime.strptime(header['date'], ISO8601)
             started = header.get('started', None)
             if started is not None:
                 started = datetime.strptime(started, ISO8601)
             result = {
                 'result_header' : header,
                 'result_content': msg['content'],
                 'started' : started,
                 'completed' : completed,
                 'engine_uuid': engine_uuid
             }
             if MongoDB is not None and isinstance(self.db, MongoDB):
                 result['result_buffers'] = map(Binary, msg['buffers'])
             else:
                 result['result_buffers'] = msg['buffers']
             self.db.update_record(msg_id, result)
             
         else:
             logging.debug("task::unknown task %s finished"%msg_id)
     
     def save_task_destination(self, idents, msg):
         try:
             msg = self.session.unpack_message(msg, content=True)
         except:
             logging.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[engine_uuid]
         
         logging.info("task::task %s arrived on %s"%(msg_id, eid))
         # if msg_id in self.mia:
         #     self.mia.remove(msg_id)
         # else:
         #     logging.debug("task::task %s 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))
         self.db.update_record(msg_id, dict(engine_uuid=engine_uuid))
     
     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.unpack_message(msg, content=True)
         except:
             logging.error("iopub::invalid IOPub message", exc_info=True)
             return
         
         parent = msg['parent_header']
         if not parent:
             logging.error("iopub::invalid IOPub message: %s"%msg)
             return
         msg_id = parent['msg_id']
         msg_type = msg['msg_type']
         content = msg['content']
         
         # ensure msg_id is in db
         try:
             rec = self.db.get_record(msg_id)
         except:
             logging.error("iopub::IOPub message has invalid parent", exc_info=True)
             return
         # 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
         else:
             d[msg_type] = content['data']
         
         self.db.update_record(msg_id, d)
         
     
             
     #-------------------------------------------------------------------------
     # Registration requests
     #-------------------------------------------------------------------------
         
     def connection_request(self, client_id, msg):
         """Reply with connection addresses for clients."""
         logging.info("client::client %s connected"%client_id)
         content = dict(status='ok')
         content.update(self.client_addrs)
         jsonable = {}
         for k,v in self.keytable.iteritems():
             jsonable[str(k)] = v
         content['engines'] = jsonable
         self.session.send(self.registrar, 'connection_reply', content, parent=msg, ident=client_id)
     
     def register_engine(self, reg, msg):
         """Register a new engine."""
         content = msg['content']
         try:
             queue = content['queue']
         except KeyError:
             logging.error("registration::queue not specified", exc_info=True)
             return
         heart = content.get('heartbeat', None)
         """register a new engine, and create the socket(s) necessary"""
         eid = self._next_id
         # print (eid, queue, reg, heart)
         
         logging.debug("registration::register_engine(%i, %r, %r, %r)"%(eid, queue, reg, heart))
         
         content = dict(id=eid,status='ok')
         content.update(self.engine_addrs)
         # check if requesting available IDs:
         if queue in self.by_ident:
             try:
                 raise KeyError("queue_id %r in use"%queue)
             except:
                 content = wrap_exception()
                 logging.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:
                 logging.error("heart_id %r in use"%heart, exc_info=True)
                 content = wrap_exception()
         else:
             for h, pack in self.incoming_registrations.iteritems():
                 if heart == h:
                     try:
                         raise KeyError("heart_id %r in use"%heart)
                     except:
                         logging.error("heart_id %r in use"%heart, exc_info=True)
                         content = wrap_exception()
                     break
                 elif queue == pack[1]:
                     try:
                         raise KeyError("queue_id %r in use"%queue)
                     except:
                         logging.error("queue_id %r in use"%queue, exc_info=True)
                         content = wrap_exception()
                     break
         
         msg = self.session.send(self.registrar, "registration_reply", 
                 content=content, 
                 ident=reg)
         
         if content['status'] == 'ok':
             if heart in self.heartmonitor.hearts:
                 # already beating
                 self.incoming_registrations[heart] = (eid,queue,reg[0],None)
                 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)
         else:
             logging.error("registration::registration %i failed: %s"%(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:
             logging.error("registration::bad engine id for unregistration: %s"%ident, exc_info=True)
             return
         logging.info("registration::unregister_engine(%s)"%eid)
         content=dict(id=eid, queue=self.engines[eid].queue)
         self.ids.remove(eid)
         self.keytable.pop(eid)
         ec = self.engines.pop(eid)
         self.hearts.pop(ec.heartbeat)
         self.by_ident.pop(ec.queue)
         self.completed.pop(eid)
         for msg_id in self.queues.pop(eid):
             msg = self.pending.remove(msg_id)
             ############## TODO: HANDLE IT ################
         
         if self.notifier:
             self.session.send(self.notifier, "unregistration_notification", content=content)
     
     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)
         except KeyError:
             logging.error("registration::tried to finish nonexistant registration", exc_info=True)
             return
         logging.info("registration::finished registering engine %i:%r"%(eid,queue))
         if purge is not None:
             purge.stop()
         control = queue
         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.queues[eid] = list()
         self.tasks[eid] = list()
         self.completed[eid] = list()
         self.hearts[heart] = eid
         content = dict(id=eid, queue=self.engines[eid].queue)
         if self.notifier:
             self.session.send(self.notifier, "registration_notification", content=content)
     
     def _purge_stalled_registration(self, heart):
         if heart in self.incoming_registrations:
             eid = self.incoming_registrations.pop(heart)[0]
             logging.info("registration::purging stalled registration: %i"%eid)
         else:
             pass
             
     #-------------------------------------------------------------------------
     # Client Requests
     #-------------------------------------------------------------------------
     
     def shutdown_request(self, client_id, msg):
         """handle shutdown request."""
         # s = self.context.socket(zmq.XREQ)
         # s.connect(self.client_connections['mux'])
         # time.sleep(0.1)
         # for eid,ec in self.engines.iteritems():
         #     self.session.send(s, 'shutdown_request', content=dict(restart=False), ident=ec.queue)
         # time.sleep(1)
         self.session.send(self.clientele, 'shutdown_reply', content={'status': 'ok'}, ident=client_id)
         dc = ioloop.DelayedCallback(lambda : self._shutdown(), 1000, self.loop)
         dc.start()
     
     def _shutdown(self):
         logging.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 = wrap_exception()
             self.session.send(self.clientele, "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.clientele, "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 = wrap_exception()
             self.session.send(self.clientele, "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[bytes(t)] = {'queue': queue, 'completed': completed , 'tasks': tasks}
             # pending
         self.session.send(self.clientele, "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']
         msg_ids = content.get('msg_ids', [])
         reply = dict(status='ok')
         if msg_ids == 'all':
             self.db.drop_matching_records(dict(completed={'$ne':None}))
         else:
             for msg_id in msg_ids:
                 if msg_id in self.all_completed:
                     self.db.drop_record(msg_id)
                 else:
                     if msg_id in self.pending:
                         try:
                             raise IndexError("msg pending: %r"%msg_id)
                         except:
                             reply = wrap_exception()
                     else:
                         try:
                             raise IndexError("No such msg: %r"%msg_id)
                         except:
                             reply = wrap_exception()
                     break
             eids = content.get('engine_ids', [])
             for eid in eids:
                 if eid not in self.engines:
                     try:
                         raise IndexError("No such engine: %i"%eid)
                     except:
                         reply = wrap_exception()
                     break
                 msg_ids = self.completed.pop(eid)
                 uid = self.engines[eid].queue
                 self.db.drop_matching_records(dict(engine_uuid=uid, completed={'$ne':None}))
         
         self.session.send(self.clientele, 'purge_reply', content=reply, ident=client_id)
     
     def resubmit_task(self, client_id, msg, buffers):
         """Resubmit a task."""
         raise NotImplementedError
     
     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:
             content['results'] = {}
             records = self.db.find_records(dict(msg_id={'$in':msg_ids}))
         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:
                     rec = records[msg_id]
                     io_dict = {}
                     for key in 'pyin pyout pyerr stdout stderr'.split():
                             io_dict[key] = rec[key]
                     content[msg_id] = { 'result_content': rec['result_content'],
                                         'header': rec['header'],
                                         'result_header' : rec['result_header'],
                                         'io' : io_dict,
                                       }
                     buffers.extend(map(str, rec['result_buffers']))
             else:
                 try:
                     raise KeyError('No such message: '+msg_id)
                 except:
                     content = wrap_exception()
                 break
         self.session.send(self.clientele, "result_reply", content=content, 
                                             parent=msg, ident=client_id,
                                             buffers=buffers)
 

IPython/zmq/parallel/ipcontrollerapp.py

@@ -1,322 +1,340
 #!/usr/bin/env python
 # encoding: utf-8
 """
 The IPython controller application.
 """
 
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2008-2009  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 copy
 import sys
 import os
 import logging
 # from twisted.application import service
 # from twisted.internet import reactor
 # from twisted.python import log
 
 import zmq
 from zmq.log.handlers import PUBHandler
 
 from IPython.config.loader import Config
 from IPython.zmq.parallel import factory
 from IPython.zmq.parallel.controller import ControllerFactory
 from IPython.zmq.parallel.clusterdir import (
     ApplicationWithClusterDir,
     ClusterDirConfigLoader
 )
 # from IPython.kernel.fcutil import FCServiceFactory, FURLError
 from IPython.utils.traitlets import Instance, Unicode
 
 from entry_point import generate_exec_key
 
 
 #-----------------------------------------------------------------------------
 # 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 "cluster_<profile>". See the --profile
 and --cluster-dir options for details.
 """
 
 #-----------------------------------------------------------------------------
 # Default interfaces
 #-----------------------------------------------------------------------------
 
 # The default client interfaces for FCClientServiceFactory.interfaces
 default_client_interfaces = Config()
 default_client_interfaces.Default.url_file = 'ipcontroller-client.url'
 
 # Make this a dict we can pass to Config.__init__ for the default
 default_client_interfaces = dict(copy.deepcopy(default_client_interfaces.items()))
 
 
 
 # The default engine interfaces for FCEngineServiceFactory.interfaces
 default_engine_interfaces = Config()
 default_engine_interfaces.Default.url_file = u'ipcontroller-engine.url'
 
 # Make this a dict we can pass to Config.__init__ for the default
 default_engine_interfaces = dict(copy.deepcopy(default_engine_interfaces.items()))
 
 
 #-----------------------------------------------------------------------------
 # Service factories
 #-----------------------------------------------------------------------------
 
 # 
 # class FCClientServiceFactory(FCServiceFactory):
 #     """A Foolscap implementation of the client services."""
 # 
 #     cert_file = Unicode(u'ipcontroller-client.pem', config=True)
 #     interfaces = Instance(klass=Config, kw=default_client_interfaces,
 #                           allow_none=False, config=True)
 # 
 # 
 # class FCEngineServiceFactory(FCServiceFactory):
 #     """A Foolscap implementation of the engine services."""
 # 
 #     cert_file = Unicode(u'ipcontroller-engine.pem', config=True)
 #     interfaces = Instance(klass=dict, kw=default_engine_interfaces,
 #                           allow_none=False, config=True)
 # 
 
 #-----------------------------------------------------------------------------
 # Command line options
 #-----------------------------------------------------------------------------
 
 
 class IPControllerAppConfigLoader(ClusterDirConfigLoader):
 
     def _add_arguments(self):
         super(IPControllerAppConfigLoader, self)._add_arguments()
         paa = self.parser.add_argument
         
         ## Hub Config:
         paa('--mongodb', 
             dest='HubFactory.db_class', action='store_const',
             const='IPython.zmq.parallel.mongodb.MongoDB', 
             help='Use MongoDB task storage [default: in-memory]')
         paa('--hb',
             type=int, dest='HubFactory.hb', nargs=2,
             help='The (2) ports the Hub\'s Heartmonitor will use for the heartbeat '
             'connections [default: random]',
             metavar='Hub.hb_ports')
+        paa('--ping',
+            type=int, dest='HubFactory.ping',
+            help='The frequency at which the Hub pings the engines for heartbeats '
+            ' (in ms) [default: 100]',
+            metavar='Hub.ping')
         
         # Client config
         paa('--client-ip',
             type=str, dest='HubFactory.client_ip', 
             help='The IP address or hostname the Hub will listen on for '
             'client connections. Both engine-ip and client-ip can be set simultaneously '
             'via --ip [default: loopback]',
             metavar='Hub.client_ip')
         paa('--client-transport',
             type=str, dest='HubFactory.client_transport', 
             help='The ZeroMQ transport the Hub will use for '
             'client connections. Both engine-transport and client-transport can be set simultaneously '
             'via --transport [default: tcp]',
             metavar='Hub.client_transport')
         paa('--query',
             type=int, dest='HubFactory.query_port', 
             help='The port on which the Hub XREP socket will listen for result queries from clients [default: random]',
             metavar='Hub.query_port')
         paa('--notifier',
             type=int, dest='HubFactory.notifier_port', 
             help='The port on which the Hub PUB socket will listen for notification connections [default: random]',
             metavar='Hub.notifier_port')
         
         # Engine config
         paa('--engine-ip',
             type=str, dest='HubFactory.engine_ip', 
             help='The IP address or hostname the Hub will listen on for '
             'engine connections. This applies to the Hub and its schedulers'
             'engine-ip and client-ip can be set simultaneously '
             'via --ip [default: loopback]',
             metavar='Hub.engine_ip')
         paa('--engine-transport',
             type=str, dest='HubFactory.engine_transport', 
             help='The ZeroMQ transport the Hub will use for '
             'client connections. Both engine-transport and client-transport can be set simultaneously '
             'via --transport [default: tcp]',
             metavar='Hub.engine_transport')
         
         # Scheduler config
         paa('--mux',
             type=int, dest='ControllerFactory.mux', nargs=2,
             help='The (2) ports the MUX scheduler will listen on for client,engine '
             'connections, respectively [default: random]',
             metavar='Scheduler.mux_ports')
         paa('--task',
             type=int, dest='ControllerFactory.task', nargs=2,
             help='The (2) ports the Task scheduler will listen on for client,engine '
             'connections, respectively [default: random]',
             metavar='Scheduler.task_ports')
         paa('--control',
             type=int, dest='ControllerFactory.control', nargs=2,
             help='The (2) ports the Control scheduler will listen on for client,engine '
             'connections, respectively [default: random]',
             metavar='Scheduler.control_ports')
         paa('--iopub',
             type=int, dest='ControllerFactory.iopub', nargs=2,
             help='The (2) ports the IOPub scheduler will listen on for client,engine '
             'connections, respectively [default: random]',
             metavar='Scheduler.iopub_ports')
         paa('--scheme',
             type=str, dest='ControllerFactory.scheme',
             choices = ['pure', 'lru', 'plainrandom', 'weighted', 'twobin','leastload'],
             help='select the task scheduler scheme  [default: Python LRU]',
             metavar='Scheduler.scheme')
         paa('--usethreads',
             dest='ControllerFactory.usethreads', action="store_true",
             help='Use threads instead of processes for the schedulers',
             )
         
         ## Global config
         paa('--log-to-file',
             action='store_true', dest='Global.log_to_file',
             help='Log to a file in the log directory (default is stdout)')
         paa('--log-url',
             type=str, dest='Global.log_url',
             help='Broadcast logs to an iploggerz process [default: disabled]')
         paa('-r','--reuse-key', 
             action='store_true', dest='Global.reuse_key',
             help='Try to reuse existing execution keys.')
         paa('--no-secure',
             action='store_false', dest='Global.secure',
-            help='Turn off execution keys.')
+            help='Turn off execution keys (default).')
         paa('--secure',
             action='store_true', dest='Global.secure',
-            help='Turn on execution keys (default).')
+            help='Turn on execution keys.')
         paa('--execkey',
             type=str, dest='Global.exec_key',
             help='path to a file containing an execution key.',
             metavar='keyfile')
         factory.add_session_arguments(self.parser)
         factory.add_registration_arguments(self.parser)
 
 
 #-----------------------------------------------------------------------------
 # The main application
 #-----------------------------------------------------------------------------
 
 
 class IPControllerApp(ApplicationWithClusterDir):
 
     name = u'ipcontrollerz'
     description = _description
     command_line_loader = IPControllerAppConfigLoader
     default_config_file_name = default_config_file_name
     auto_create_cluster_dir = True
 
     def create_default_config(self):
         super(IPControllerApp, self).create_default_config()
         # Don't set defaults for Global.secure or Global.reuse_furls
         # as those are set in a component.
         self.default_config.Global.import_statements = []
         self.default_config.Global.clean_logs = True
         self.default_config.Global.secure = False
         self.default_config.Global.reuse_key = False
         self.default_config.Global.exec_key = "exec_key.key"
 
     def pre_construct(self):
         super(IPControllerApp, self).pre_construct()
         c = self.master_config
         # The defaults for these are set in FCClientServiceFactory and
         # FCEngineServiceFactory, so we only set them here if the global
         # options have be set to override the class level defaults.
         
         # if hasattr(c.Global, 'reuse_furls'):
         #     c.FCClientServiceFactory.reuse_furls = c.Global.reuse_furls
         #     c.FCEngineServiceFactory.reuse_furls = c.Global.reuse_furls
         #     del c.Global.reuse_furls
         # if hasattr(c.Global, 'secure'):
         #     c.FCClientServiceFactory.secure = c.Global.secure
         #     c.FCEngineServiceFactory.secure = c.Global.secure
         #     del c.Global.secure
 
     def construct(self):
         # This is the working dir by now.
         sys.path.insert(0, '')
         c = self.master_config
         
         self.import_statements()
 
         if c.Global.secure:
             keyfile = os.path.join(c.Global.security_dir, c.Global.exec_key)
             if not c.Global.reuse_key or not os.path.exists(keyfile):
                 generate_exec_key(keyfile)
             c.SessionFactory.exec_key = keyfile
         else:
             keyfile = os.path.join(c.Global.security_dir, c.Global.exec_key)
             if os.path.exists(keyfile):
                 os.remove(keyfile)
             c.SessionFactory.exec_key = ''
         
         try:
             self.factory = ControllerFactory(config=c)
             self.start_logging()
             self.factory.construct()
         except:
             self.log.error("Couldn't construct the Controller", exc_info=True)
             self.exit(1)
+    
+    def save_urls(self):
+        """save the registration urls to files."""
+        c = self.master_config
+        
+        sec_dir = c.Global.security_dir
+        cf = self.factory
+        
+        with open(os.path.join(sec_dir, 'ipcontroller-engine.url'), 'w') as f:
+            f.write("%s://%s:%s"%(cf.engine_transport, cf.engine_ip, cf.regport))
+        
+        with open(os.path.join(sec_dir, 'ipcontroller-client.url'), 'w') as f:
+            f.write("%s://%s:%s"%(cf.client_transport, cf.client_ip, cf.regport))
         
     
     def import_statements(self):
         statements = self.master_config.Global.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 start_logging(self):
-    #     super(IPControllerApp, self).start_logging()
-    #     if self.master_config.Global.log_url:
-    #         context = self.factory.context
-    #         lsock = context.socket(zmq.PUB)
-    #         lsock.connect(self.master_config.Global.log_url)
-    #         handler = PUBHandler(lsock)
-    #         handler.root_topic = 'controller'
-    #         handler.setLevel(self.log_level)
-    #         self.log.addHandler(handler)
+    def start_logging(self):
+        super(IPControllerApp, self).start_logging()
+        if self.master_config.Global.log_url:
+            context = self.factory.context
+            lsock = context.socket(zmq.PUB)
+            lsock.connect(self.master_config.Global.log_url)
+            handler = PUBHandler(lsock)
+            handler.root_topic = 'controller'
+            handler.setLevel(self.log_level)
+            self.log.addHandler(handler)
     # 
     def start_app(self):
-        # Start the controller service.
+        # Start the subprocesses:
         self.factory.start()
         self.write_pid_file(overwrite=True)
         try:
             self.factory.loop.start()
         except KeyboardInterrupt:
             self.log.critical("Interrupted, Exiting...\n")
 
 
 def launch_new_instance():
     """Create and run the IPython controller"""
     app = IPControllerApp()
     app.start()
 
 
 if __name__ == '__main__':
     launch_new_instance()

setup.py

@@ -1,251 +1,252
 #!/usr/bin/env python
 # -*- coding: utf-8 -*-
 """Setup script for IPython.
 
 Under Posix environments it works like a typical setup.py script. 
 Under Windows, the command sdist is not supported, since IPython 
 requires utilities which are not available under Windows."""
 
 #-----------------------------------------------------------------------------
 #  Copyright (c) 2008-2010, IPython Development Team.
 #  Copyright (c) 2001-2007, Fernando Perez <[e-mail]>
 #  Copyright (c) 2001, Janko Hauser <[e-mail]>
 #  Copyright (c) 2001, Nathaniel Gray <[e-mail]>
 #
 #  Distributed under the terms of the Modified BSD License.
 #
 #  The full license is in the file COPYING.txt, distributed with this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Minimal Python version sanity check
 #-----------------------------------------------------------------------------
 
 import sys
 
 # This check is also made in IPython/__init__, don't forget to update both when
 # changing Python version requirements.
 if sys.version[0:3] < '2.6':
     error = """\
 ERROR: 'IPython requires Python Version 2.6 or above.'
 Exiting."""
     print >> sys.stderr, error
     sys.exit(1)
 
 # At least we're on the python version we need, move on.
 
 #-------------------------------------------------------------------------------
 # Imports
 #-------------------------------------------------------------------------------
 
 # Stdlib imports
 import os
 import shutil
 
 from glob import glob
 
 # BEFORE importing distutils, remove MANIFEST. distutils doesn't properly
 # update it when the contents of directories change.
 if os.path.exists('MANIFEST'): os.remove('MANIFEST')
 
 from distutils.core import setup
 
 # Our own imports
 from IPython.utils.path import target_update
 
 from setupbase import (
     setup_args, 
     find_packages, 
     find_package_data, 
     find_scripts,
     find_data_files,
     check_for_dependencies,
     record_commit_info,
 )
 
 isfile = os.path.isfile
 pjoin = os.path.join
 
 #-----------------------------------------------------------------------------
 # Function definitions
 #-----------------------------------------------------------------------------
 
 def cleanup():
     """Clean up the junk left around by the build process"""
     if "develop" not in sys.argv:
         try:
             shutil.rmtree('ipython.egg-info')
         except:
             try:
                 os.unlink('ipython.egg-info')
             except:
                 pass
 
 #-------------------------------------------------------------------------------
 # Handle OS specific things
 #-------------------------------------------------------------------------------
 
 if os.name == 'posix':
     os_name = 'posix'
 elif os.name in ['nt','dos']:
     os_name = 'windows'
 else:
     print 'Unsupported operating system:',os.name
     sys.exit(1)
 
 # Under Windows, 'sdist' has not been supported.  Now that the docs build with
 # Sphinx it might work, but let's not turn it on until someone confirms that it
 # actually works.
 if os_name == 'windows' and 'sdist' in sys.argv:
     print 'The sdist command is not available under Windows.  Exiting.'
     sys.exit(1)
 
 #-------------------------------------------------------------------------------
 # Things related to the IPython documentation
 #-------------------------------------------------------------------------------
 
 # update the manuals when building a source dist
 if len(sys.argv) >= 2 and sys.argv[1] in ('sdist','bdist_rpm'):
     import textwrap
 
     # List of things to be updated. Each entry is a triplet of args for
     # target_update()
     to_update = [
                   # FIXME - Disabled for now: we need to redo an automatic way
                   # of generating the magic info inside the rst.                
                   #('docs/magic.tex',
                   #['IPython/Magic.py'],
                   #"cd doc && ./update_magic.sh" ),
 
                  ('docs/man/ipcluster.1.gz',
                   ['docs/man/ipcluster.1'],
                   'cd docs/man && gzip -9c ipcluster.1 > ipcluster.1.gz'),
 
                  ('docs/man/ipcontroller.1.gz',
                   ['docs/man/ipcontroller.1'],
                   'cd docs/man && gzip -9c ipcontroller.1 > ipcontroller.1.gz'),
 
                  ('docs/man/ipengine.1.gz',
                   ['docs/man/ipengine.1'],
                   'cd docs/man && gzip -9c ipengine.1 > ipengine.1.gz'),
 
                  ('docs/man/ipython.1.gz',
                   ['docs/man/ipython.1'],
                   'cd docs/man && gzip -9c ipython.1 > ipython.1.gz'),
 
                  ('docs/man/ipython-wx.1.gz',
                   ['docs/man/ipython-wx.1'],
                   'cd docs/man && gzip -9c ipython-wx.1 > ipython-wx.1.gz'),
 
                  ('docs/man/ipythonx.1.gz',
                   ['docs/man/ipythonx.1'],
                   'cd docs/man && gzip -9c ipythonx.1 > ipythonx.1.gz'),
 
                  ('docs/man/irunner.1.gz',
                   ['docs/man/irunner.1'],
                   'cd docs/man && gzip -9c irunner.1 > irunner.1.gz'),
 
                  ('docs/man/pycolor.1.gz',
                   ['docs/man/pycolor.1'],
                   'cd docs/man && gzip -9c pycolor.1 > pycolor.1.gz'),
                  ]
 
     # Only build the docs if sphinx is present
     try:
         import sphinx
     except ImportError:
         pass
     else:
         # The Makefile calls the do_sphinx scripts to build html and pdf, so
         # just one target is enough to cover all manual generation
 
         # First, compute all the dependencies that can force us to rebuild the
         # docs.  Start with the main release file that contains metadata
         docdeps = ['IPython/core/release.py']
         # Inculde all the reST sources
         pjoin = os.path.join
         for dirpath,dirnames,filenames in os.walk('docs/source'):
             if dirpath in ['_static','_templates']:
                 continue
             docdeps += [ pjoin(dirpath,f) for f in filenames
                          if f.endswith('.txt') ]
         # and the examples
         for dirpath,dirnames,filenames in os.walk('docs/example'):
             docdeps += [ pjoin(dirpath,f) for f in filenames
                          if not f.endswith('~') ]
         # then, make them all dependencies for the main PDF (the html will get
         # auto-generated as well).
         to_update.append(
             ('docs/dist/ipython.pdf',
              docdeps,
              "cd docs && make dist")
             )
         
     [ target_update(*t) for t in to_update ]
     
 #---------------------------------------------------------------------------
 # Find all the packages, package data, scripts and data_files
 #---------------------------------------------------------------------------
 
 packages = find_packages()
 package_data = find_package_data()
 scripts = find_scripts()
 data_files = find_data_files()
 
 #---------------------------------------------------------------------------
 # Handle dependencies and setuptools specific things
 #---------------------------------------------------------------------------
 
 # For some commands, use setuptools.  Note that we do NOT list install here!
 # If you want a setuptools-enhanced install, just run 'setupegg.py install'
 if len(set(('develop', 'sdist', 'release', 'bdist_egg', 'bdist_rpm',
            'bdist', 'bdist_dumb', 'bdist_wininst', 'install_egg_info',
            'build_sphinx', 'egg_info', 'easy_install', 'upload',
             )).intersection(sys.argv)) > 0:
     import setuptools
 
 # This dict is used for passing extra arguments that are setuptools 
 # specific to setup
 setuptools_extra_args = {}
 
 if 'setuptools' in sys.modules:
     setuptools_extra_args['zip_safe'] = False
     setuptools_extra_args['entry_points'] = {
         'console_scripts': [
             'ipython = IPython.frontend.terminal.ipapp:launch_new_instance',
             'ipython-qtconsole = IPython.frontend.qt.console.ipythonqt:main',
             'pycolor = IPython.utils.PyColorize:main',
             'ipcontrollerz = IPython.zmq.parallel.ipcontrollerapp:launch_new_instance',
             'ipenginez = IPython.zmq.parallel.ipengineapp:launch_new_instance',
             'iploggerz = IPython.zmq.parallel.iploggerapp:launch_new_instance',
-            'ipclusterz = IPython.zmq.parallel.ipcluster:main',
+            'ipclusterz = IPython.zmq.parallel.ipclusterapp:launch_new_instance',
             'iptest = IPython.testing.iptest:main',
             'irunner = IPython.lib.irunner:main'
         ]
     }
     setup_args['extras_require'] = dict(
+        zmq = 'pyzmq>=2.0.10',
         doc='Sphinx>=0.3',
         test='nose>=0.10.1',
         security='pyOpenSSL>=0.6'
     )
 else:
     # If we are running without setuptools, call this function which will
     # check for dependencies an inform the user what is needed.  This is
     # just to make life easy for users.
     check_for_dependencies()
 
 #---------------------------------------------------------------------------
 # Do the actual setup now
 #---------------------------------------------------------------------------
 
 setup_args['cmdclass'] = {'build_py': record_commit_info('IPython')}
 setup_args['packages'] = packages
 setup_args['package_data'] = package_data
 setup_args['scripts'] = scripts
 setup_args['data_files'] = data_files
 setup_args.update(setuptools_extra_args)
 
 
 if __name__ == '__main__':
     setup(**setup_args)
     cleanup()