launcher updates for PBS

IPython/config/default/ipclusterz_config.py

@@ -1,184 +1,192
 import os
 
 c = get_config()
 
 #-----------------------------------------------------------------------------
 # Select which launchers to use
 #-----------------------------------------------------------------------------
 
 # This allows you to control what method is used to start the controller
 # and engines.  The following methods are currently supported:
 # - Start as a regular process on localhost.
 # - Start using mpiexec.
 # - Start using the Windows HPC Server 2008 scheduler
 # - Start using PBS
 # - Start using SSH
 
 
 # The selected launchers can be configured below.
 
 # Options are:
 # - LocalControllerLauncher
 # - MPIExecControllerLauncher
 # - PBSControllerLauncher
 # - WindowsHPCControllerLauncher
 # c.Global.controller_launcher = 'IPython.zmq.parallel.launcher.LocalControllerLauncher'
 
 # Options are:
 # - LocalEngineSetLauncher
 # - MPIExecEngineSetLauncher
 # - PBSEngineSetLauncher
 # - WindowsHPCEngineSetLauncher
 # c.Global.engine_launcher = 'IPython.zmq.parallel.launcher.LocalEngineSetLauncher'
 
 #-----------------------------------------------------------------------------
 # Global configuration
 #-----------------------------------------------------------------------------
 
 # The default number of engines that will be started. This is overridden by
 # the -n command line option: "ipcluster start -n 4"
 # c.Global.n = 2
 
 # Log to a file in cluster_dir/log, otherwise just log to sys.stdout.
 # c.Global.log_to_file = False
 
 # Remove old logs from cluster_dir/log before starting.
 # c.Global.clean_logs = True
 
 # The working directory for the process. The application will use os.chdir
 # to change to this directory before starting.
 # c.Global.work_dir = os.getcwd()
 
 
 #-----------------------------------------------------------------------------
 # Local process launchers
 #-----------------------------------------------------------------------------
 
 # The command line arguments to call the controller with.
 # c.LocalControllerLauncher.controller_args = \
 #    ['--log-to-file','--log-level', '40']
 
 # The working directory for the controller
 # c.LocalEngineSetLauncher.work_dir = u''
 
 # Command line argument passed to the engines.
 # c.LocalEngineSetLauncher.engine_args = ['--log-to-file','--log-level', '40']
 
 #-----------------------------------------------------------------------------
 # MPIExec launchers
 #-----------------------------------------------------------------------------
 
 # The mpiexec/mpirun command to use in started the controller.
 # c.MPIExecControllerLauncher.mpi_cmd = ['mpiexec']
 
 # Additional arguments to pass to the actual mpiexec command.
 # c.MPIExecControllerLauncher.mpi_args = []
 
 # The command line argument to call the controller with.
 # c.MPIExecControllerLauncher.controller_args = \
 #     ['--log-to-file','--log-level', '40']
 
 
 # The mpiexec/mpirun command to use in started the controller.
 # c.MPIExecEngineSetLauncher.mpi_cmd = ['mpiexec']
 
 # Additional arguments to pass to the actual mpiexec command.
 # c.MPIExecEngineSetLauncher.mpi_args = []
 
 # Command line argument passed to the engines.
 # c.MPIExecEngineSetLauncher.engine_args = ['--log-to-file','--log-level', '40']
 
 # The default number of engines to start if not given elsewhere.
 # c.MPIExecEngineSetLauncher.n = 1
 
 #-----------------------------------------------------------------------------
 # SSH launchers
 #-----------------------------------------------------------------------------
 
 # Todo
 
 
 #-----------------------------------------------------------------------------
 # Unix batch (PBS) schedulers launchers
 #-----------------------------------------------------------------------------
 
 # The command line program to use to submit a PBS job.
 # c.PBSControllerLauncher.submit_command = 'qsub'
 
 # The command line program to use to delete a PBS job.
 # c.PBSControllerLauncher.delete_command = 'qdel'
 
 # A regular expression that takes the output of qsub and find the job id.
 # c.PBSControllerLauncher.job_id_regexp = r'\d+'
 
 # The batch submission script used to start the controller. This is where
 # environment variables would be setup, etc. This string is interpolated using
 # the Itpl module in IPython.external. Basically, you can use ${n} for the
 # number of engine and ${cluster_dir} for the cluster_dir.
-# c.PBSControllerLauncher.batch_template = """"""
+# c.PBSControllerLauncher.batch_template = """
+# #PBS -l nprocs=$n
+# 
+# ipcontrollerz --cluster-dir $cluster_dir
+# """
 
 # The name of the instantiated batch script that will actually be used to
 # submit the job. This will be written to the cluster directory.
 # c.PBSControllerLauncher.batch_file_name = u'pbs_batch_script_controller'
 
 
 # The command line program to use to submit a PBS job.
 # c.PBSEngineSetLauncher.submit_command = 'qsub'
 
 # The command line program to use to delete a PBS job.
 # c.PBSEngineSetLauncher.delete_command = 'qdel'
 
 # A regular expression that takes the output of qsub and find the job id.
 # c.PBSEngineSetLauncher.job_id_regexp = r'\d+'
 
 # The batch submission script used to start the engines. This is where
 # environment variables would be setup, etc. This string is interpolated using
 # the Itpl module in IPython.external. Basically, you can use ${n} for the
 # number of engine and ${cluster_dir} for the cluster_dir.
-# c.PBSEngineSetLauncher.batch_template = """"""
+# c.PBSEngineSetLauncher.batch_template = """
+# #PBS -l nprocs=$n
+# 
+# ipenginez --cluster-dir $cluster_dir$s
+# """
 
 # The name of the instantiated batch script that will actually be used to
 # submit the job. This will be written to the cluster directory.
 # c.PBSEngineSetLauncher.batch_file_name = u'pbs_batch_script_engines'
 
 #-----------------------------------------------------------------------------
 # Windows HPC Server 2008 launcher configuration
 #-----------------------------------------------------------------------------
 
 # c.IPControllerJob.job_name = 'IPController'
 # c.IPControllerJob.is_exclusive = False
 # c.IPControllerJob.username = r'USERDOMAIN\USERNAME'
 # c.IPControllerJob.priority = 'Highest'
 # c.IPControllerJob.requested_nodes = ''
 # c.IPControllerJob.project = 'MyProject'
 
 # c.IPControllerTask.task_name = 'IPController'
 # c.IPControllerTask.controller_cmd = [u'ipcontroller.exe']
 # c.IPControllerTask.controller_args = ['--log-to-file', '--log-level', '40']
 # c.IPControllerTask.environment_variables = {}
 
 # c.WindowsHPCControllerLauncher.scheduler = 'HEADNODE'
 # c.WindowsHPCControllerLauncher.job_file_name = u'ipcontroller_job.xml'
 
 
 # c.IPEngineSetJob.job_name = 'IPEngineSet'
 # c.IPEngineSetJob.is_exclusive = False
 # c.IPEngineSetJob.username = r'USERDOMAIN\USERNAME'
 # c.IPEngineSetJob.priority = 'Highest'
 # c.IPEngineSetJob.requested_nodes = ''
 # c.IPEngineSetJob.project = 'MyProject'
 
 # c.IPEngineTask.task_name = 'IPEngine'
 # c.IPEngineTask.engine_cmd = [u'ipengine.exe']
 # c.IPEngineTask.engine_args = ['--log-to-file', '--log-level', '40']
 # c.IPEngineTask.environment_variables = {}
 
 # c.WindowsHPCEngineSetLauncher.scheduler = 'HEADNODE'
 # c.WindowsHPCEngineSetLauncher.job_file_name = u'ipengineset_job.xml'
 
 
 
 
 
 
 

IPython/zmq/parallel/ipclusterapp.py

@@ -1,592 +1,592
 #!/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 errno
 import logging
 import os
 import re
 import signal
 
 import zmq
 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'ipclusterz_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 'ipclusterz start -n 4'. For more complex usage 
 you will typically do 'ipclusterz create -p mycluster', then edit
 configuration files, followed by 'ipclusterz 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
 
 # This will be the exit code if ipcluster engines is run, but there is not .pid
 # file to be found.
 NO_CLUSTER = 12
 
 
 #-----------------------------------------------------------------------------
 # 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 'ipclusterz <cmd> [options]'. To get help
             on a particular subcommand do 'ipclusterz <cmd> -h'."""
             # help="For more help, type 'ipclusterz <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>'."""
+            'clusterz_<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
+            using the convention 'clusterz_<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, 
             'ipclusterz create -p mycluster', which will put the directory
-            in '<ipython_dir>/cluster_mycluster'. 
+            in '<ipython_dir>/clusterz_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'
+            'clusterz_<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, 'ipclusterz 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.")
         paa('--delay',
             type=float, dest='Global.delay',
             help="Specify the delay (in seconds) between starting the controller and starting the engine(s).")
 
         # 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 
+            'clusterz_<profile>'. If your cluster directory is in 
             the cwd or the ipython directory, you can simply refer to it
             using its profile name, 'ipclusterz 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")
         
         # the "engines" subcommand parser
         parser_engines = subparsers.add_parser(
             'engines',
             parents=[parent_parser1, parent_parser2],
             argument_default=SUPPRESS,
             help="Attach some engines to an existing controller or cluster.",
             description=
             """Start one or more engines to connect to an existing Cluster
             by profile name or cluster directory.
             Cluster directories contain configuration, log and
             security related files and are named using the convention
-            'cluster_<profile>' and should be creating using the 'start'
+            'clusterz_<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, 'ipclusterz engines -n 4 -p <profile>`,
             otherwise use the '--cluster-dir' option.
             """
         )
         paa = parser_engines.add_argument
         paa('-n', '--number',
             type=int, dest='Global.n',
             help='The number of engines to start.',
             metavar='Global.n')
         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.")
 
 #-----------------------------------------------------------------------------
 # 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.delay = 2
         self.default_config.Global.reset_config = False
         self.default_config.Global.clean_logs = True
         self.default_config.Global.signal = signal.SIGINT
         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 'ipclusterz start'.  Do "
                     "'ipclusterz create -h' or 'ipclusterz list -h' for more "
                     "information about creating and listing cluster dirs."
                 )
         elif subcommand=='engines':
             self.auto_create_cluster_dir = False
             try:
                 super(IPClusterApp, self).find_resources()
             except ClusterDirError:
                 raise ClusterDirError(
                     "Could not find a cluster directory. A cluster dir must "
                     "be created before running 'ipclusterz start'.  Do "
                     "'ipclusterz create -h' or 'ipclusterz 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_'):
+                if os.path.isdir(full_path) and f.startswith('clusterz_'):
                     profile = full_path.split('_')[-1]
                     start_cmd = 'ipclusterz 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()
         if subcmd == 'engines':
             self.start_logging()
             self.loop = ioloop.IOLoop.instance()
             # reactor.callWhenRunning(self.start_launchers)
             engine_only = lambda : self.start_launchers(controller=False)
             dc = ioloop.DelayedCallback(engine_only, 0, self.loop)
             dc.start()
 
     def start_launchers(self, controller=True):
         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.
         if controller:
             cl_class = import_item(config.Global.controller_launcher)
             self.controller_launcher = cl_class(
                 work_dir=self.cluster_dir, config=config,
                 logname=self.log.name
             )
             # 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)
             # 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.
         else:
             self.controller_launcher = None
         
         el_class = import_item(config.Global.engine_launcher)
         self.engine_launcher = el_class(
             work_dir=self.cluster_dir, config=config, logname=self.log.name
         )
 
         # Setup signals
         signal.signal(signal.SIGINT, self.sigint_handler)
 
         # Start the controller and engines
         self._stopping = False  # Make sure stop_launchers is not called 2x.
         if controller:
             self.start_controller()
         dc = ioloop.DelayedCallback(self.start_engines, 1000*config.Global.delay*controller, self.loop)
         dc.start()
         self.startup_message()
 
     def startup_message(self, r=None):
         self.log.info("IPython cluster: started")
         return r
         
     def start_controller(self, r=None):
         # self.log.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):
         # self.log.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):
         # self.log.info("In stop_controller")
         if self.controller_launcher and self.controller_launcher.running:
             return self.controller_launcher.stop()
 
     def stop_engines(self, r=None):
         # self.log.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):
         self.log.error(f.getTraceback())
         return None
         
     def stop_launchers(self, r=None):
         if not self._stopping:
             self._stopping = True
             # if isinstance(r, failure.Failure):
             #     self.log.error('Unexpected error in ipcluster:')
             #     self.log.info(r.getTraceback())
             self.log.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 re.match(r'ip(engine|controller)z-\d+\.(log|err|out)',f):
                     os.remove(os.path.join(log_dir, f))
         # This will remove old log files for ipcluster itself
         super(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()
         elif subcmd=='engines':
             self.start_app_engines()
 
     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 "ipclusterz 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 ipclusterz 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 KeyboardInterrupt:
             pass
         except zmq.ZMQError as e:
             if e.errno == errno.EINTR:
                 pass
             else:
                 raise
         self.remove_pid_file()
 
     def start_app_engines(self):
         """Start the app for the start subcommand."""
         config = self.master_config
         # First see if the cluster is already running
         
         # Now log and daemonize
         self.log.info(
             'Starting engines 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 KeyboardInterrupt:
             pass
         except zmq.ZMQError as e:
             if e.errno == errno.EINTR:
                 pass
             else:
                 raise
         # 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/ipcontrollerapp.py

@@ -1,427 +1,427
 #!/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 os
 import logging
 import socket
 import stat
 import sys
 import uuid
 
 import zmq
 from zmq.log.handlers import PUBHandler
 from zmq.utils import jsonapi as json
 
 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.zmq.parallel.util import disambiguate_ip_address, split_url
 # from IPython.kernel.fcutil import FCServiceFactory, FURLError
 from IPython.utils.traitlets import Instance, Unicode
 
 
 
 #-----------------------------------------------------------------------------
 # Module level variables
 #-----------------------------------------------------------------------------
 
 
 #: The default config file name for this application
 default_config_file_name = u'ipcontrollerz_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
+your ipython directory and named as "clusterz_<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='HubFactory.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',
             )
         paa('--hwm',
             dest='ControllerFactory.hwm', type=int,
             help='specify the High Water Mark (HWM) for the downstream '
             'socket in the pure ZMQ scheduler. This is the maximum number '
             'of allowed outstanding tasks on each engine.',
             )
         
         ## 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-files', 
             action='store_true', dest='Global.reuse_files',
             help='Try to reuse existing json connection files.')
         paa('--no-secure',
             action='store_false', dest='Global.secure',
             help='Turn off execution keys (default).')
         paa('--secure',
             action='store_true', dest='Global.secure',
             help='Turn on execution keys.')
         paa('--execkey',
             type=str, dest='Global.exec_key',
             help='path to a file containing an execution key.',
             metavar='keyfile')
         paa('--ssh',
             type=str, dest='Global.sshserver',
             help='ssh url for clients to use when connecting to the Controller '
             'processes. It should be of the form: [user@]server[:port]. The '
             'Controller\'s listening addresses must be accessible from the ssh server',
             metavar='Global.sshserver')
         paa('--location',
             type=str, dest='Global.location',
             help="The external IP or domain name of this machine, used for disambiguating "
             "engine and client connections.",
             metavar='Global.location')
         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 = True
         self.default_config.Global.reuse_files = False
         self.default_config.Global.exec_key = "exec_key.key"
         self.default_config.Global.sshserver = None
         self.default_config.Global.location = None
 
     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 save_connection_dict(self, fname, cdict):
         """save a connection dict to json file."""
         c = self.master_config
         url = cdict['url']
         location = cdict['location']
         if not location:
             try:
                 proto,ip,port = split_url(url)
             except AssertionError:
                 pass
             else:
                 location = socket.gethostbyname_ex(socket.gethostname())[2][-1]
             cdict['location'] = location
         fname = os.path.join(c.Global.security_dir, fname)
         with open(fname, 'w') as f:
             f.write(json.dumps(cdict, indent=2))
         os.chmod(fname, stat.S_IRUSR|stat.S_IWUSR)
     
     def load_config_from_json(self):
         """load config from existing json connector files."""
         c = self.master_config
         # load from engine config
         with open(os.path.join(c.Global.security_dir, 'ipcontroller-engine.json')) as f:
             cfg = json.loads(f.read())
         key = c.SessionFactory.exec_key = cfg['exec_key']
         xport,addr = cfg['url'].split('://')
         c.HubFactory.engine_transport = xport
         ip,ports = addr.split(':')
         c.HubFactory.engine_ip = ip
         c.HubFactory.regport = int(ports)
         c.Global.location = cfg['location']
         
         # load client config
         with open(os.path.join(c.Global.security_dir, 'ipcontroller-client.json')) as f:
             cfg = json.loads(f.read())
         assert key == cfg['exec_key'], "exec_key mismatch between engine and client keys"
         xport,addr = cfg['url'].split('://')
         c.HubFactory.client_transport = xport
         ip,ports = addr.split(':')
         c.HubFactory.client_ip = ip
         c.Global.sshserver = cfg['ssh']
         assert int(ports) == c.HubFactory.regport, "regport mismatch"
     
     def construct(self):
         # This is the working dir by now.
         sys.path.insert(0, '')
         c = self.master_config
         
         self.import_statements()
         reusing = c.Global.reuse_files
         if reusing:
             try:
                 self.load_config_from_json()
             except (AssertionError,IOError):
                 reusing=False
         # check again, because reusing may have failed:
         if reusing:
             pass
         elif c.Global.secure:
             keyfile = os.path.join(c.Global.security_dir, c.Global.exec_key)
             key = str(uuid.uuid4())
             with open(keyfile, 'w') as f:
                 f.write(key)
             os.chmod(keyfile, stat.S_IRUSR|stat.S_IWUSR)
             c.SessionFactory.exec_key = key
         else:
             c.SessionFactory.exec_key = ''
             key = None
         
         try:
             self.factory = ControllerFactory(config=c, logname=self.log.name)
             self.start_logging()
             self.factory.construct()
         except:
             self.log.error("Couldn't construct the Controller", exc_info=True)
             self.exit(1)
         
         if not reusing:
             # save to new json config files
             f = self.factory
             cdict = {'exec_key' : key,
                     'ssh' : c.Global.sshserver,
                     'url' : "%s://%s:%s"%(f.client_transport, f.client_ip, f.regport),
                     'location' : c.Global.location
                     }
             self.save_connection_dict('ipcontroller-client.json', cdict)
             edict = cdict
             edict['url']="%s://%s:%s"%((f.client_transport, f.client_ip, f.regport))
             self.save_connection_dict('ipcontroller-engine.json', edict)
         
     
     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_app(self):
         # 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()

IPython/zmq/parallel/ipengineapp.py

@@ -1,294 +1,294
 #!/usr/bin/env python
 # encoding: utf-8
 """
 The IPython engine 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 json
 import os
 import sys
 
 import zmq
 from zmq.eventloop import ioloop
 
 from IPython.zmq.parallel.clusterdir import (
     ApplicationWithClusterDir,
     ClusterDirConfigLoader
 )
 from IPython.zmq.log import EnginePUBHandler
 
 from IPython.zmq.parallel import factory
 from IPython.zmq.parallel.engine import EngineFactory
 from IPython.zmq.parallel.streamkernel import Kernel
 from IPython.zmq.parallel.util import disambiguate_url
 from IPython.utils.importstring import import_item
 
 
 #-----------------------------------------------------------------------------
 # Module level variables
 #-----------------------------------------------------------------------------
 
 #: The default config file name for this application
 default_config_file_name = u'ipenginez_config.py'
 
 
 mpi4py_init = """from mpi4py import MPI as mpi
 mpi.size = mpi.COMM_WORLD.Get_size()
 mpi.rank = mpi.COMM_WORLD.Get_rank()
 """
 
 
 pytrilinos_init = """from PyTrilinos import Epetra
 class SimpleStruct:
 pass
 mpi = SimpleStruct()
 mpi.rank = 0
 mpi.size = 0
 """
 
 
 _description = """Start an IPython engine for parallel computing.\n\n
 
 IPython engines run in parallel and perform computations on behalf of a client
 and controller. A controller needs to be started before the engines. The
 engine 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>".
+usually located in your ipython directory and named as "clusterz_<profile>".
 See the --profile and --cluster-dir options for details.
 """
 
 #-----------------------------------------------------------------------------
 # Command line options
 #-----------------------------------------------------------------------------
 
 
 class IPEngineAppConfigLoader(ClusterDirConfigLoader):
 
     def _add_arguments(self):
         super(IPEngineAppConfigLoader, self)._add_arguments()
         paa = self.parser.add_argument
         # Controller config
         paa('--file', '-f',
             type=unicode, dest='Global.url_file',
             help='The full location of the file containing the connection information fo '
             'controller. If this is not given, the file must be in the '
             'security directory of the cluster directory.  This location is '
             'resolved using the --profile and --app-dir options.',
             metavar='Global.url_file')
         # MPI
         paa('--mpi',
             type=str, dest='MPI.use',
             help='How to enable MPI (mpi4py, pytrilinos, or empty string to disable).',
             metavar='MPI.use')
         # 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',
             dest='Global.log_url',
             help="url of ZMQ logger, as started with iploggerz")
         # paa('--execkey',
         #     type=str, dest='Global.exec_key',
         #     help='path to a file containing an execution key.',
         #     metavar='keyfile')
         # paa('--no-secure',
         #     action='store_false', dest='Global.secure',
         #     help='Turn off execution keys.')
         # paa('--secure',
         #     action='store_true', dest='Global.secure',
         #     help='Turn on execution keys (default).')
         # init command
         paa('-c',
             type=str, dest='Global.extra_exec_lines',
             help='specify a command to be run at startup')
         
         factory.add_session_arguments(self.parser)
         factory.add_registration_arguments(self.parser)
 
 
 #-----------------------------------------------------------------------------
 # Main application
 #-----------------------------------------------------------------------------
 
 
 class IPEngineApp(ApplicationWithClusterDir):
 
     name = u'ipenginez'
     description = _description
     command_line_loader = IPEngineAppConfigLoader
     default_config_file_name = default_config_file_name
     auto_create_cluster_dir = True
 
     def create_default_config(self):
         super(IPEngineApp, self).create_default_config()
 
         # The engine should not clean logs as we don't want to remove the
         # active log files of other running engines.
         self.default_config.Global.clean_logs = False
         self.default_config.Global.secure = True
 
         # Global config attributes
         self.default_config.Global.exec_lines = []
         self.default_config.Global.extra_exec_lines = ''
 
         # Configuration related to the controller
         # This must match the filename (path not included) that the controller
         # used for the FURL file.
         self.default_config.Global.url_file = u''
         self.default_config.Global.url_file_name = u'ipcontroller-engine.json'
         # If given, this is the actual location of the controller's FURL file.
         # If not, this is computed using the profile, app_dir and furl_file_name
         # self.default_config.Global.key_file_name = u'exec_key.key'
         # self.default_config.Global.key_file = u''
 
         # MPI related config attributes
         self.default_config.MPI.use = ''
         self.default_config.MPI.mpi4py = mpi4py_init
         self.default_config.MPI.pytrilinos = pytrilinos_init
 
     def post_load_command_line_config(self):
         pass
 
     def pre_construct(self):
         super(IPEngineApp, self).pre_construct()
         # self.find_cont_url_file()
         self.find_url_file()
         if self.master_config.Global.extra_exec_lines:
             self.master_config.Global.exec_lines.append(self.master_config.Global.extra_exec_lines)
 
     # def find_key_file(self):
     #     """Set the key file.
     # 
     #     Here we don't try to actually see if it exists for is valid as that
     #     is hadled by the connection logic.
     #     """
     #     config = self.master_config
     #     # Find the actual controller key file
     #     if not config.Global.key_file:
     #         try_this = os.path.join(
     #             config.Global.cluster_dir, 
     #             config.Global.security_dir,
     #             config.Global.key_file_name
     #         )
     #         config.Global.key_file = try_this
         
     def find_url_file(self):
         """Set the key file.
 
         Here we don't try to actually see if it exists for is valid as that
         is hadled by the connection logic.
         """
         config = self.master_config
         # Find the actual controller key file
         if not config.Global.url_file:
             try_this = os.path.join(
                 config.Global.cluster_dir, 
                 config.Global.security_dir,
                 config.Global.url_file_name
             )
             config.Global.url_file = try_this
         
     def construct(self):
         # This is the working dir by now.
         sys.path.insert(0, '')
         config = self.master_config
         # if os.path.exists(config.Global.key_file) and config.Global.secure:
         #     config.SessionFactory.exec_key = config.Global.key_file
         if os.path.exists(config.Global.url_file):
             with open(config.Global.url_file) as f:
                 d = json.loads(f.read())
                 for k,v in d.iteritems():
                     if isinstance(v, unicode):
                         d[k] = v.encode()
             if d['exec_key']:
                 config.SessionFactory.exec_key = d['exec_key']
             d['url'] = disambiguate_url(d['url'], d['location'])
             config.RegistrationFactory.url=d['url']
             config.EngineFactory.location = d['location']
         
         
         
         config.Kernel.exec_lines = config.Global.exec_lines
 
         self.start_mpi()
 
         # Create the underlying shell class and EngineService
         # shell_class = import_item(self.master_config.Global.shell_class)
         try:
             self.engine = EngineFactory(config=config, logname=self.log.name)
         except:
             self.log.error("Couldn't start the Engine", exc_info=True)
             self.exit(1)
         
         self.start_logging()
 
         # Create the service hierarchy
         # self.main_service = service.MultiService()
         # self.engine_service.setServiceParent(self.main_service)
         # self.tub_service = Tub()
         # self.tub_service.setServiceParent(self.main_service)
         # # This needs to be called before the connection is initiated
         # self.main_service.startService()
 
         # This initiates the connection to the controller and calls
         # register_engine to tell the controller we are ready to do work
         # self.engine_connector = EngineConnector(self.tub_service)
 
         # self.log.info("Using furl file: %s" % self.master_config.Global.furl_file)
 
         # reactor.callWhenRunning(self.call_connect)
 
 
     def start_logging(self):
         super(IPEngineApp, self).start_logging()
         if self.master_config.Global.log_url:
             context = self.engine.context
             lsock = context.socket(zmq.PUB)
             lsock.connect(self.master_config.Global.log_url)
             handler = EnginePUBHandler(self.engine, lsock)
             handler.setLevel(self.log_level)
             self.log.addHandler(handler)
     
     def start_mpi(self):
         global mpi
         mpikey = self.master_config.MPI.use
         mpi_import_statement = self.master_config.MPI.get(mpikey, None)
         if mpi_import_statement is not None:
             try:
                 self.log.info("Initializing MPI:")
                 self.log.info(mpi_import_statement)
                 exec mpi_import_statement in globals()
             except:
                 mpi = None
         else:
             mpi = None
 
 
     def start_app(self):
         self.engine.start()
         try:
             self.engine.loop.start()
         except KeyboardInterrupt:
             self.log.critical("Engine Interrupted, shutting down...\n")
 
 
 def launch_new_instance():
     """Create and run the IPython controller"""
     app = IPEngineApp()
     app.start()
 
 
 if __name__ == '__main__':
     launch_new_instance()
 

IPython/zmq/parallel/iploggerapp.py

@@ -1,132 +1,132
 #!/usr/bin/env python
 # encoding: utf-8
 """
 A simple IPython logger application
 """
 
 #-----------------------------------------------------------------------------
 #  Copyright (C) 2011  The IPython Development Team
 #
 #  Distributed under the terms of the BSD License.  The full license is in
 #  the file COPYING, distributed as part of this software.
 #-----------------------------------------------------------------------------
 
 #-----------------------------------------------------------------------------
 # Imports
 #-----------------------------------------------------------------------------
 
 import os
 import sys
 
 import zmq
 
 from IPython.zmq.parallel.clusterdir import (
     ApplicationWithClusterDir,
     ClusterDirConfigLoader
 )
 from .logwatcher import LogWatcher
 
 #-----------------------------------------------------------------------------
 # Module level variables
 #-----------------------------------------------------------------------------
 
 #: The default config file name for this application
 default_config_file_name = u'iplogger_config.py'
 
 _description = """Start an IPython logger for parallel computing.\n\n
 
 IPython controllers and engines (and your own processes) can broadcast log messages
 by registering a `zmq.log.handlers.PUBHandler` with the `logging` module. The
 logger 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>".
+usually located in your ipython directory and named as "clusterz_<profile>".
 See the --profile and --cluster-dir options for details.
 """
 
 #-----------------------------------------------------------------------------
 # Command line options
 #-----------------------------------------------------------------------------
 
 
 class IPLoggerAppConfigLoader(ClusterDirConfigLoader):
 
     def _add_arguments(self):
         super(IPLoggerAppConfigLoader, self)._add_arguments()
         paa = self.parser.add_argument
         # Controller config
         paa('--url',
             type=str, dest='LogWatcher.url',
             help='The url the LogWatcher will listen on',
             )
         # MPI
         paa('--topics',
             type=str, dest='LogWatcher.topics', nargs='+',
             help='What topics to subscribe to',
             metavar='topics')
         # 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)')
         
 
 #-----------------------------------------------------------------------------
 # Main application
 #-----------------------------------------------------------------------------
 
 
 class IPLoggerApp(ApplicationWithClusterDir):
 
     name = u'iploggerz'
     description = _description
     command_line_loader = IPLoggerAppConfigLoader
     default_config_file_name = default_config_file_name
     auto_create_cluster_dir = True
 
     def create_default_config(self):
         super(IPLoggerApp, self).create_default_config()
 
         # The engine should not clean logs as we don't want to remove the
         # active log files of other running engines.
         self.default_config.Global.clean_logs = False
 
         # If given, this is the actual location of the logger's URL file.
         # If not, this is computed using the profile, app_dir and furl_file_name
         self.default_config.Global.url_file_name = u'iplogger.url'
         self.default_config.Global.url_file = u''
 
     def post_load_command_line_config(self):
         pass
 
     def pre_construct(self):
         super(IPLoggerApp, self).pre_construct()
 
     def construct(self):
         # This is the working dir by now.
         sys.path.insert(0, '')
 
         self.start_logging()
 
         try:
             self.watcher = LogWatcher(config=self.master_config, logname=self.log.name)
         except:
             self.log.error("Couldn't start the LogWatcher", exc_info=True)
             self.exit(1)
         
 
     def start_app(self):
         try:
             self.watcher.start()
             self.watcher.loop.start()
         except KeyboardInterrupt:
             self.log.critical("Logging Interrupted, shutting down...\n")
 
 
 def launch_new_instance():
     """Create and run the IPython LogWatcher"""
     app = IPLoggerApp()
     app.start()
 
 
 if __name__ == '__main__':
     launch_new_instance()
 

IPython/zmq/parallel/launcher.py

@@ -1,844 +1,847
 #!/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 logging
 import os
 import re
 import sys
 
 from signal import SIGINT, SIGTERM
 try:
     from signal import SIGKILL
 except ImportError:
     SIGKILL=SIGTERM
 
 from subprocess import Popen, PIPE, STDOUT
 try:
     from subprocess import check_output
 except ImportError:
     # pre-2.7:
     from StringIO import StringIO
     
     def check_output(*args, **kwargs):
         sio = StringIO()
         kwargs.update(dict(stdout=PIPE, stderr=STDOUT))
         p = Popen(*args, **kwargs)
         out,err = p.communicate()
         return out
 
 from zmq.eventloop import ioloop
 
 from IPython.external import Itpl
 # from IPython.config.configurable import Configurable
-from IPython.utils.traitlets import Str, Int, List, Unicode, Instance
+from IPython.utils.traitlets import Str, Int, List, Unicode, Dict, Instance
 from IPython.utils.path import get_ipython_module_path
 from IPython.utils.process import find_cmd, pycmd2argv, FindCmdError
 
 from .factory import LoggingFactory
 
 # load winhpcjob from IPython.kernel
 try:
     from IPython.kernel.winhpcjob import (
         IPControllerTask, IPEngineTask,
         IPControllerJob, IPEngineSetJob
     )
 except ImportError:
     pass
 
 
 #-----------------------------------------------------------------------------
 # 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(LoggingFactory):
     """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, **kwargs):
         super(BaseLauncher, self).__init__(work_dir=work_dir, config=config, **kwargs)
         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.
         """
 
         self.log.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`."""
 
         self.log.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, **kwargs):
         super(LocalProcessLauncher, self).__init__(
             work_dir=work_dir, config=config, **kwargs
         )
         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:
             self.log.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:
             self.log.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.INFO)], 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)
         self.log.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.INFO)], 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.INFO)], config=True
     )
     # launcher class
     launcher_class = LocalEngineLauncher
     
     def __init__(self, work_dir=u'.', config=None, **kwargs):
         super(LocalEngineSetLauncher, self).__init__(
             work_dir=work_dir, config=config, **kwargs
         )
         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, logname=self.log.name)
             # 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:
                 self.log.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.INFO)], 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)
         self.log.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.INFO)], 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
         self.log.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(['-tt'], 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):
         print self.config
         if hostname is not None:
             self.hostname = hostname
         if user is not None:
             self.user = user
         print (self.location, hostname, user)
         return super(SSHLauncher, self).start()
     
     def signal(self, sig):
         if self.state == 'running':
             # send escaped ssh connection-closer
             self.process.stdin.write('~.')
             self.process.stdin.flush()
         
 
 
 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.INFO)], 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.INFO)], 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, **kwargs):
         super(WindowsHPCLauncher, self).__init__(
             work_dir=work_dir, config=config, **kwargs
         )
 
     @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
         self.log.info('Job started with job id: %r' % job_id)
         return job_id
 
     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
         ]
         self.log.info("Starting Win HPC Job: %s" % (self.job_cmd + ' ' + ' '.join(args),))
         # Twisted will raise DeprecationWarnings if we try to pass unicode to this
         output = check_output([self.job_cmd]+args,
             env=os.environ,
             cwd=self.work_dir,
             stderr=STDOUT
         )
         job_id = self.parse_job_id(output)
         # self.notify_start(job_id)
         return job_id
 
     def stop(self):
         args = [
             'cancel',
             self.job_id,
             '/scheduler:%s' % self.scheduler
         ]
         self.log.info("Stopping Win HPC Job: %s" % (self.job_cmd + ' ' + ' '.join(args),))
         try:
             output = check_output([self.job_cmd]+args,
                 env=os.environ,
                 cwd=self.work_dir,
                 stderr=STDOUT
             )
         except:
             output = 'The job already appears to be stoppped: %r' % self.job_id
         self.notify_stop(output)  # Pass the output of the kill cmd
         return 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)
 
         self.log.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)
 
         self.log.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'')
+    # the format dict used with batch_template:
+    context = Dict()
 
+    
+    def find_args(self):
+        return [self.submit_command]
+    
     def __init__(self, work_dir=u'.', config=None, **kwargs):
         super(BatchSystemLauncher, self).__init__(
             work_dir=work_dir, config=config, **kwargs
         )
         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
         self.log.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)
         self.log.info('Writing instantiated batch script: %s' % self.batch_file)
         f = open(self.batch_file, 'w')
         f.write(script_as_string)
         f.close()
 
-    def start(self, n):
+    def start(self, n, cluster_dir):
         """Start n copies of the process using a batch system."""
+        # 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)
         self.write_batch_script(n)
         output = check_output([self.submit_command, self.batch_file], env=os.environ, stdout=STDOUT)
         job_id = self.parse_job_id(output)
         # self.notify_start(job_id)
         return job_id
 
     def stop(self):
         output = check_output([self.delete_command, self.job_id], env=os.environ, stderr=STDOUT)
         self.notify_stop(output)  # Pass the output of the kill cmd
         return 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)
         self.log.info("Starting PBSControllerLauncher: %r" % self.args)
-        return super(PBSControllerLauncher, self).start(1)
+        return super(PBSControllerLauncher, self).start(1, cluster_dir)
 
 
 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)
         self.log.info('Starting PBSEngineSetLauncher: %r' % self.args)
-        return super(PBSEngineSetLauncher, self).start(n)
+        return super(PBSEngineSetLauncher, self).start(n, cluster_dir)
 
 
 #-----------------------------------------------------------------------------
 # 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.INFO)], 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):
         self.log.info("Starting ipcluster: %r" % self.args)
         return super(IPClusterLauncher, self).start()