Show More
@@ -2,66 +2,154 b' import os' | |||
|
2 | 2 | |
|
3 | 3 | c = get_config() |
|
4 | 4 | |
|
5 | # Options are: | |
|
6 | # * LocalControllerLauncher | |
|
7 | # * PBSControllerLauncher | |
|
5 | #----------------------------------------------------------------------------- | |
|
6 | # Select which launchers to use | |
|
7 | #----------------------------------------------------------------------------- | |
|
8 | ||
|
9 | # This allows you to control what method is used to start the controller | |
|
10 | # and engines. The following methods are currently supported: | |
|
11 | # * Start as a regular process on localhost. | |
|
12 | # * Start using mpiexec. | |
|
13 | # * Start using PBS | |
|
14 | # * Start using SSH (currently broken) | |
|
15 | ||
|
16 | # The selected launchers can be configured below. | |
|
17 | ||
|
18 | # Options are (LocalControllerLauncher, MPIExecControllerLauncher, | |
|
19 | # PBSControllerLauncher) | |
|
8 | 20 | # c.Global.controller_launcher = 'IPython.kernel.launcher.LocalControllerLauncher' |
|
9 | 21 | |
|
10 | # Options are: | |
|
11 |
# |
|
|
12 | # * MPIExecEngineSetLauncher | |
|
13 | # * PBSEngineSetLauncher | |
|
22 | # Options are (LocalEngineSetLauncher, MPIExecEngineSetLauncher, | |
|
23 | # PBSEngineSetLauncher) | |
|
14 | 24 | # c.Global.engine_launcher = 'IPython.kernel.launcher.LocalEngineSetLauncher' |
|
15 | 25 | |
|
16 | # c.Global.log_to_file = False | |
|
26 | #----------------------------------------------------------------------------- | |
|
27 | # Global configuration | |
|
28 | #----------------------------------------------------------------------------- | |
|
29 | ||
|
30 | # The default number of engine that will be started. This is overridden by | |
|
31 | # the -n command line option: "ipcluster start -n 4" | |
|
17 | 32 | # c.Global.n = 2 |
|
18 | # c.Global.reset_config = False | |
|
19 | # c.Global.clean_logs = True | |
|
20 | 33 | |
|
21 | # c.MPIExecLauncher.mpi_cmd = ['mpiexec'] | |
|
22 | # c.MPIExecLauncher.mpi_args = [] | |
|
23 | # c.MPIExecLauncher.program = [] | |
|
24 | # c.MPIExecLauncher.program_args = [] | |
|
25 | # c.MPIExecLauncher.n = 1 | |
|
34 | # Log to a file in cluster_dir/log, otherwise just log to sys.stdout. | |
|
35 | # c.Global.log_to_file = False | |
|
26 | 36 | |
|
27 | # c.SSHLauncher.ssh_cmd = ['ssh'] | |
|
28 | # c.SSHLauncher.ssh_args = [] | |
|
29 | # c.SSHLauncher.program = [] | |
|
30 | # s.SSHLauncher.program_args = [] | |
|
31 | # c.SSHLauncher.hostname = '' | |
|
32 | # c.SSHLauncher.user = os.environ['USER'] | |
|
37 | # Remove old logs from cluster_dir/log before starting. | |
|
38 | # c.Global.clean_logs = True | |
|
33 | 39 | |
|
34 | # c.PBSLauncher.submit_command = 'qsub' | |
|
35 | # c.PBSLauncher.delete_command = 'qdel' | |
|
36 | # c.PBSLauncher.job_id_regexp = '\d+' | |
|
37 | # c.PBSLauncher.batch_template = """""" | |
|
38 | # c.PBSLauncher.batch_file_name = u'pbs_batch_script' | |
|
40 | #----------------------------------------------------------------------------- | |
|
41 | # Controller launcher configuration | |
|
42 | #----------------------------------------------------------------------------- | |
|
39 | 43 | |
|
40 | # c.LocalControllerLauncher.controller_args = [] | |
|
44 | # Configure how the controller is started. The configuration of the controller | |
|
45 | # can also bet setup by editing the controller config file: | |
|
46 | # ipcontroller_config.py | |
|
41 | 47 | |
|
48 | # The command line arguments to call the controller with. | |
|
49 | # c.LocalControllerLauncher.controller_args = \ | |
|
50 | # ['--log-to-file','--log-level', '40'] | |
|
51 | ||
|
52 | # The mpiexec/mpirun command to use in started the controller. | |
|
42 | 53 | # c.MPIExecControllerLauncher.mpi_cmd = ['mpiexec'] |
|
54 | ||
|
55 | # Additional arguments to pass to the actual mpiexec command. | |
|
43 | 56 | # c.MPIExecControllerLauncher.mpi_args = [] |
|
44 | # c.MPIExecControllerLauncher.controller_args = [] | |
|
45 | # c.MPIExecControllerLauncher.n = 1 | |
|
46 | 57 | |
|
58 | # The command line argument to call the controller with. | |
|
59 | # c.MPIExecControllerLauncher.controller_args = \ | |
|
60 | # ['--log-to-file','--log-level', '40'] | |
|
61 | ||
|
62 | # The command line program to use to submit a PBS job. | |
|
47 | 63 | # c.PBSControllerLauncher.submit_command = 'qsub' |
|
64 | ||
|
65 | # The command line program to use to delete a PBS job. | |
|
48 | 66 | # c.PBSControllerLauncher.delete_command = 'qdel' |
|
67 | ||
|
68 | # A regular expression that takes the output of qsub and find the job id. | |
|
49 | 69 | # c.PBSControllerLauncher.job_id_regexp = '\d+' |
|
70 | ||
|
71 | # The batch submission script used to start the controller. This is where | |
|
72 | # environment variables would be setup, etc. This string is interpolated using | |
|
73 | # the Itpl module in IPython.external. Basically, you can use ${profile} for | |
|
74 | # the controller profile or ${cluster_dir} for the cluster_dir. | |
|
50 | 75 | # c.PBSControllerLauncher.batch_template = """""" |
|
51 | # c.PBSLauncher.batch_file_name = u'pbs_batch_script' | |
|
52 | 76 | |
|
53 | # c.LocalEngineLauncher.engine_args = [] | |
|
77 | # The name of the instantiated batch script that will actually be used to | |
|
78 | # submit the job. This will be written to the cluster directory. | |
|
79 | # c.PBSControllerLauncher.batch_file_name = u'pbs_batch_script_controller' | |
|
80 | ||
|
81 | #----------------------------------------------------------------------------- | |
|
82 | # Engine launcher configuration | |
|
83 | #----------------------------------------------------------------------------- | |
|
54 | 84 | |
|
55 | # c.LocalEngineSetLauncher.engine_args = [] | |
|
85 | # Command line argument passed to the engines. | |
|
86 | # c.LocalEngineSetLauncher.engine_args = ['--log-to-file','--log-level', '40'] | |
|
56 | 87 | |
|
88 | # The mpiexec/mpirun command to use in started the controller. | |
|
57 | 89 | # c.MPIExecEngineSetLauncher.mpi_cmd = ['mpiexec'] |
|
90 | ||
|
91 | # Additional arguments to pass to the actual mpiexec command. | |
|
58 | 92 | # c.MPIExecEngineSetLauncher.mpi_args = [] |
|
59 | # c.MPIExecEngineSetLauncher.controller_args = [] | |
|
93 | ||
|
94 | # Command line argument passed to the engines. | |
|
95 | # c.MPIExecEngineSetLauncher.engine_args = ['--log-to-file','--log-level', '40'] | |
|
96 | ||
|
97 | # The default number of engines to start if not given elsewhere. | |
|
60 | 98 | # c.MPIExecEngineSetLauncher.n = 1 |
|
61 | 99 | |
|
100 | # The command line program to use to submit a PBS job. | |
|
62 | 101 | # c.PBSEngineSetLauncher.submit_command = 'qsub' |
|
102 | ||
|
103 | # The command line program to use to delete a PBS job. | |
|
63 | 104 | # c.PBSEngineSetLauncher.delete_command = 'qdel' |
|
105 | ||
|
106 | # A regular expression that takes the output of qsub and find the job id. | |
|
64 | 107 | # c.PBSEngineSetLauncher.job_id_regexp = '\d+' |
|
108 | ||
|
109 | # The batch submission script used to start the engines. This is where | |
|
110 | # environment variables would be setup, etc. This string is interpolated using | |
|
111 | # the Itpl module in IPython.external. Basically, you can use ${n} for the | |
|
112 | # number of engine, ${profile} or the engine profile and ${cluster_dir} | |
|
113 | # for the cluster_dir. | |
|
65 | 114 | # c.PBSEngineSetLauncher.batch_template = """""" |
|
66 | # c.PBSEngineSetLauncher.batch_file_name = u'pbs_batch_script' | |
|
115 | ||
|
116 | # The name of the instantiated batch script that will actually be used to | |
|
117 | # submit the job. This will be written to the cluster directory. | |
|
118 | # c.PBSEngineSetLauncher.batch_file_name = u'pbs_batch_script_engines' | |
|
119 | ||
|
120 | #----------------------------------------------------------------------------- | |
|
121 | # Base launcher configuration | |
|
122 | #----------------------------------------------------------------------------- | |
|
123 | ||
|
124 | # The various launchers are organized into an inheritance hierarchy. | |
|
125 | # The configurations can also be iherited and the following attributes | |
|
126 | # allow you to configure the base classes. | |
|
127 | ||
|
128 | # c.MPIExecLauncher.mpi_cmd = ['mpiexec'] | |
|
129 | # c.MPIExecLauncher.mpi_args = [] | |
|
130 | # c.MPIExecLauncher.program = [] | |
|
131 | # c.MPIExecLauncher.program_args = [] | |
|
132 | # c.MPIExecLauncher.n = 1 | |
|
133 | ||
|
134 | # c.SSHLauncher.ssh_cmd = ['ssh'] | |
|
135 | # c.SSHLauncher.ssh_args = [] | |
|
136 | # c.SSHLauncher.program = [] | |
|
137 | # s.SSHLauncher.program_args = [] | |
|
138 | # c.SSHLauncher.hostname = '' | |
|
139 | # c.SSHLauncher.user = os.environ['USER'] | |
|
140 | ||
|
141 | # c.BatchSystemLauncher.submit_command | |
|
142 | # c.BatchSystemLauncher.delete_command | |
|
143 | # c.BatchSystemLauncher.job_id_regexp | |
|
144 | # c.BatchSystemLauncher.batch_template | |
|
145 | # c.BatchSystemLauncher.batch_file_name | |
|
146 | ||
|
147 | # c.PBSLauncher.submit_command = 'qsub' | |
|
148 | # c.PBSLauncher.delete_command = 'qdel' | |
|
149 | # c.PBSLauncher.job_id_regexp = '\d+' | |
|
150 | # c.PBSLauncher.batch_template = """""" | |
|
151 | # c.PBSLauncher.batch_file_name = u'pbs_batch_script' | |
|
152 | ||
|
153 | ||
|
154 | ||
|
67 | 155 |
@@ -7,10 +7,32 b' c = get_config()' | |||
|
7 | 7 | #----------------------------------------------------------------------------- |
|
8 | 8 | |
|
9 | 9 | # Basic Global config attributes |
|
10 | ||
|
11 | # Start up messages are logged to stdout using the logging module. | |
|
12 | # These all happen before the twisted reactor is started and are | |
|
13 | # useful for debugging purposes. Can be (10=DEBUG,20=INFO,30=WARN,40=CRITICAL) | |
|
14 | # and smaller is more verbose. | |
|
15 | # c.Global.log_level = 20 | |
|
16 | ||
|
17 | # Log to a file in cluster_dir/log, otherwise just log to sys.stdout. | |
|
10 | 18 | # c.Global.log_to_file = False |
|
19 | ||
|
20 | # Remove old logs from cluster_dir/log before starting. | |
|
11 | 21 | # c.Global.clean_logs = True |
|
22 | ||
|
23 | # A list of Python statements that will be run before starting the | |
|
24 | # controller. This is provided because occasionally certain things need to | |
|
25 | # be imported in the controller for pickling to work. | |
|
12 | 26 | # c.Global.import_statements = ['import math'] |
|
27 | ||
|
28 | # Reuse the controller's FURL files. If False, FURL files are regenerated | |
|
29 | # each time the controller is run. If True, they will be reused, *but*, you | |
|
30 | # also must set the network ports by hand. If set, this will override the | |
|
31 | # values set for the client and engine connections below. | |
|
13 | 32 | # c.Global.reuse_furls = True |
|
33 | ||
|
34 | # Enable SSL encryption on all connections to the controller. If set, this | |
|
35 | # will override the values set for the client and engine connections below. | |
|
14 | 36 | # c.Global.secure = True |
|
15 | 37 | |
|
16 | 38 | #----------------------------------------------------------------------------- |
@@ -18,13 +40,67 b' c = get_config()' | |||
|
18 | 40 | #----------------------------------------------------------------------------- |
|
19 | 41 | |
|
20 | 42 | # Basic client service config attributes |
|
43 | ||
|
44 | # The network interface the controller will listen on for client connections. | |
|
45 | # This should be an IP address or hostname of the controller's host. The empty | |
|
46 | # string means listen on all interfaces. | |
|
21 | 47 | # c.FCClientServiceFactory.ip = '' |
|
48 | ||
|
49 | # The TCP/IP port the controller will listen on for client connections. If 0 | |
|
50 | # a random port will be used. If the controller's host has a firewall running | |
|
51 | # it must allow incoming traffic on this port. | |
|
22 | 52 | # c.FCClientServiceFactory.port = 0 |
|
53 | ||
|
54 | # The client learns how to connect to the controller by looking at the | |
|
55 | # location field embedded in the FURL. If this field is empty, all network | |
|
56 | # interfaces that the controller is listening on will be listed. To have the | |
|
57 | # client connect on a particular interface, list it here. | |
|
23 | 58 | # c.FCClientServiceFactory.location = '' |
|
59 | ||
|
60 | # Use SSL encryption for the client connection. | |
|
24 | 61 | # c.FCClientServiceFactory.secure = True |
|
62 | ||
|
63 | # Reuse the client FURL each time the controller is started. If set, you must | |
|
64 | # also pick a specific network port above (FCClientServiceFactory.port). | |
|
25 | 65 | # c.FCClientServiceFactory.reuse_furls = False |
|
26 | 66 | |
|
27 | # You shouldn't have to modify the rest of this section | |
|
67 | #----------------------------------------------------------------------------- | |
|
68 | # Configure the engine services | |
|
69 | #----------------------------------------------------------------------------- | |
|
70 | ||
|
71 | # Basic config attributes for the engine services. | |
|
72 | ||
|
73 | # The network interface the controller will listen on for engine connections. | |
|
74 | # This should be an IP address or hostname of the controller's host. The empty | |
|
75 | # string means listen on all interfaces. | |
|
76 | # c.FCEngineServiceFactory.ip = '' | |
|
77 | ||
|
78 | # The TCP/IP port the controller will listen on for engine connections. If 0 | |
|
79 | # a random port will be used. If the controller's host has a firewall running | |
|
80 | # it must allow incoming traffic on this port. | |
|
81 | # c.FCEngineServiceFactory.port = 0 | |
|
82 | ||
|
83 | # The engine learns how to connect to the controller by looking at the | |
|
84 | # location field embedded in the FURL. If this field is empty, all network | |
|
85 | # interfaces that the controller is listening on will be listed. To have the | |
|
86 | # client connect on a particular interface, list it here. | |
|
87 | # c.FCEngineServiceFactory.location = '' | |
|
88 | ||
|
89 | # Use SSL encryption for the engine connection. | |
|
90 | # c.FCEngineServiceFactory.secure = True | |
|
91 | ||
|
92 | # Reuse the client FURL each time the controller is started. If set, you must | |
|
93 | # also pick a specific network port above (FCClientServiceFactory.port). | |
|
94 | # c.FCEngineServiceFactory.reuse_furls = False | |
|
95 | ||
|
96 | #----------------------------------------------------------------------------- | |
|
97 | # Developer level configuration attributes | |
|
98 | #----------------------------------------------------------------------------- | |
|
99 | ||
|
100 | # You shouldn't have to modify anything in this section. These attributes | |
|
101 | # are more for developers who want to change the behavior of the controller | |
|
102 | # at a fundamental level. | |
|
103 | ||
|
28 | 104 | # c.FCClientServiceFactory.cert_file = 'ipcontroller-client.pem' |
|
29 | 105 | |
|
30 | 106 | # default_client_interfaces = Config() |
@@ -44,18 +120,6 b' c = get_config()' | |||
|
44 | 120 | # |
|
45 | 121 | # c.FCEngineServiceFactory.interfaces = default_client_interfaces |
|
46 | 122 | |
|
47 | #----------------------------------------------------------------------------- | |
|
48 | # Configure the engine services | |
|
49 | #----------------------------------------------------------------------------- | |
|
50 | ||
|
51 | # Basic config attributes for the engine services | |
|
52 | # c.FCEngineServiceFactory.ip = '' | |
|
53 | # c.FCEngineServiceFactory.port = 0 | |
|
54 | # c.FCEngineServiceFactory.location = '' | |
|
55 | # c.FCEngineServiceFactory.secure = True | |
|
56 | # c.FCEngineServiceFactory.reuse_furls = False | |
|
57 | ||
|
58 | # You shouldn't have to modify the rest of this section | |
|
59 | 123 | # c.FCEngineServiceFactory.cert_file = 'ipcontroller-engine.pem' |
|
60 | 124 | |
|
61 | 125 | # default_engine_interfaces = Config() |
@@ -1,23 +1,61 b'' | |||
|
1 | 1 | c = get_config() |
|
2 | 2 | |
|
3 | #----------------------------------------------------------------------------- | |
|
4 | # Global configuration | |
|
5 | #----------------------------------------------------------------------------- | |
|
6 | ||
|
7 | # Start up messages are logged to stdout using the logging module. | |
|
8 | # These all happen before the twisted reactor is started and are | |
|
9 | # useful for debugging purposes. Can be (10=DEBUG,20=INFO,30=WARN,40=CRITICAL) | |
|
10 | # and smaller is more verbose. | |
|
11 | # c.Global.log_level = 20 | |
|
12 | ||
|
13 | # Log to a file in cluster_dir/log, otherwise just log to sys.stdout. | |
|
3 | 14 | # c.Global.log_to_file = False |
|
4 | # c.Global.clean_logs = False | |
|
15 | ||
|
16 | # Remove old logs from cluster_dir/log before starting. | |
|
17 | # c.Global.clean_logs = True | |
|
18 | ||
|
19 | # A list of strings that will be executed in the users namespace on the engine | |
|
20 | # before it connects to the controller. | |
|
5 | 21 | # c.Global.exec_lines = ['import numpy'] |
|
6 | # c.Global.log_level = 10 | |
|
7 | # c.Global.shell_class = 'IPython.kernel.core.interpreter.Interpreter' | |
|
8 | # c.Global.furl_file_name = 'ipcontroller-engine.furl' | |
|
9 | # c.Global.furl_file = '' | |
|
10 | # The max number of connection attemps and the initial delay between | |
|
22 | ||
|
23 | # The engine will try to connect to the controller multiple times, to allow | |
|
24 | # the controller time to startup and write its FURL file. These parameters | |
|
25 | # control the number of retries (connect_max_tries) and the initial delay | |
|
26 | # (connect_delay) between attemps. The actual delay between attempts gets | |
|
27 | # longer each time by a factor of 1.5 (delay[i] = 1.5*delay[i-1]) | |
|
11 | 28 | # those attemps. |
|
12 | 29 | # c.Global.connect_delay = 0.1 |
|
13 | 30 | # c.Global.connect_max_tries = 15 |
|
14 | 31 | |
|
32 | # By default, the engine will look for the controller's FURL file in its own | |
|
33 | # cluster directory. Sometimes, the FURL file will be elsewhere and this | |
|
34 | # attribute can be set to the full path of the FURL file. | |
|
35 | # c.Global.furl_file = '' | |
|
36 | ||
|
37 | #----------------------------------------------------------------------------- | |
|
38 | # MPI configuration | |
|
39 | #----------------------------------------------------------------------------- | |
|
15 | 40 | |
|
41 | # Upon starting the engine can be configured to call MPI_Init. This section | |
|
42 | # configures that. | |
|
43 | ||
|
44 | # Select which MPI section to execute to setup MPI. The value of this | |
|
45 | # attribute must match the name of another attribute in the MPI config | |
|
46 | # section (mpi4py, pytrilinos, etc.). This can also be set by the --mpi | |
|
47 | # command line option. | |
|
16 | 48 | # c.MPI.use = '' |
|
49 | ||
|
50 | # Initialize MPI using mpi4py. To use this, set c.MPI.use = 'mpi4py' to use | |
|
51 | # --mpi=mpi4py at the command line. | |
|
17 | 52 | # c.MPI.mpi4py = """from mpi4py import MPI as mpi |
|
18 | 53 | # mpi.size = mpi.COMM_WORLD.Get_size() |
|
19 | 54 | # mpi.rank = mpi.COMM_WORLD.Get_rank() |
|
20 | 55 | # """ |
|
56 | ||
|
57 | # Initialize MPI using pytrilinos. To use this, set c.MPI.use = 'pytrilinos' | |
|
58 | # to use --mpi=pytrilinos at the command line. | |
|
21 | 59 | # c.MPI.pytrilinos = """from PyTrilinos import Epetra |
|
22 | 60 | # class SimpleStruct: |
|
23 | 61 | # pass |
@@ -26,3 +64,23 b' c = get_config()' | |||
|
26 | 64 | # mpi.size = 0 |
|
27 | 65 | # """ |
|
28 | 66 | |
|
67 | #----------------------------------------------------------------------------- | |
|
68 | # Developer level configuration attributes | |
|
69 | #----------------------------------------------------------------------------- | |
|
70 | ||
|
71 | # You shouldn't have to modify anything in this section. These attributes | |
|
72 | # are more for developers who want to change the behavior of the controller | |
|
73 | # at a fundamental level. | |
|
74 | ||
|
75 | # You should not have to change these attributes. | |
|
76 | ||
|
77 | # c.Global.shell_class = 'IPython.kernel.core.interpreter.Interpreter' | |
|
78 | ||
|
79 | # c.Global.furl_file_name = 'ipcontroller-engine.furl' | |
|
80 | ||
|
81 | ||
|
82 | ||
|
83 | ||
|
84 | ||
|
85 | ||
|
86 |
@@ -156,6 +156,11 b' class AsyncClientConnector(object):' | |||
|
156 | 156 | ipythondir : str |
|
157 | 157 | The location of the ipythondir if different from the default. |
|
158 | 158 | This is used if the cluster directory is being found by profile. |
|
159 | delay : float | |
|
160 | The initial delay between re-connection attempts. Susequent delays | |
|
161 | get longer according to ``delay[i] = 1.5*delay[i-1]``. | |
|
162 | max_tries : int | |
|
163 | The max number of re-connection attempts. | |
|
159 | 164 | |
|
160 | 165 | Returns |
|
161 | 166 | ------- |
@@ -193,7 +198,12 b' class AsyncClientConnector(object):' | |||
|
193 | 198 | ipythondir : str |
|
194 | 199 | The location of the ipythondir if different from the default. |
|
195 | 200 | This is used if the cluster directory is being found by profile. |
|
196 | ||
|
201 | delay : float | |
|
202 | The initial delay between re-connection attempts. Susequent delays | |
|
203 | get longer according to ``delay[i] = 1.5*delay[i-1]``. | |
|
204 | max_tries : int | |
|
205 | The max number of re-connection attempts. | |
|
206 | ||
|
197 | 207 | Returns |
|
198 | 208 | ------- |
|
199 | 209 | A deferred to the actual client class. |
@@ -233,6 +243,11 b' class AsyncClientConnector(object):' | |||
|
233 | 243 | ipythondir : str |
|
234 | 244 | The location of the ipythondir if different from the default. |
|
235 | 245 | This is used if the cluster directory is being found by profile. |
|
246 | delay : float | |
|
247 | The initial delay between re-connection attempts. Susequent delays | |
|
248 | get longer according to ``delay[i] = 1.5*delay[i-1]``. | |
|
249 | max_tries : int | |
|
250 | The max number of re-connection attempts. | |
|
236 | 251 | |
|
237 | 252 | Returns |
|
238 | 253 | ------- |
@@ -332,6 +347,11 b' class ClientConnector(object):' | |||
|
332 | 347 | ipythondir : str |
|
333 | 348 | The location of the ipythondir if different from the default. |
|
334 | 349 | This is used if the cluster directory is being found by profile. |
|
350 | delay : float | |
|
351 | The initial delay between re-connection attempts. Susequent delays | |
|
352 | get longer according to ``delay[i] = 1.5*delay[i-1]``. | |
|
353 | max_tries : int | |
|
354 | The max number of re-connection attempts. | |
|
335 | 355 | |
|
336 | 356 | Returns |
|
337 | 357 | ------- |
@@ -368,6 +388,11 b' class ClientConnector(object):' | |||
|
368 | 388 | ipythondir : str |
|
369 | 389 | The location of the ipythondir if different from the default. |
|
370 | 390 | This is used if the cluster directory is being found by profile. |
|
391 | delay : float | |
|
392 | The initial delay between re-connection attempts. Susequent delays | |
|
393 | get longer according to ``delay[i] = 1.5*delay[i-1]``. | |
|
394 | max_tries : int | |
|
395 | The max number of re-connection attempts. | |
|
371 | 396 | |
|
372 | 397 | Returns |
|
373 | 398 | ------- |
@@ -640,8 +665,22 b' class Cluster(object):' | |||
|
640 | 665 | def get_multiengine_client(self, delay=DELAY, max_tries=MAX_TRIES): |
|
641 | 666 | """Get the multiengine client for the running cluster. |
|
642 | 667 | |
|
643 | If this fails, it means that the cluster has not finished starting. | |
|
644 | Usually waiting a few seconds are re-trying will solve this. | |
|
668 | This will try to attempt to the controller multiple times. If this | |
|
669 | fails altogether, try looking at the following: | |
|
670 | * Make sure the controller is starting properly by looking at its | |
|
671 | log files. | |
|
672 | * Make sure the controller is writing its FURL file in the location | |
|
673 | expected by the client. | |
|
674 | * Make sure a firewall on the controller's host is not blocking the | |
|
675 | client from connecting. | |
|
676 | ||
|
677 | Parameters | |
|
678 | ---------- | |
|
679 | delay : float | |
|
680 | The initial delay between re-connection attempts. Susequent delays | |
|
681 | get longer according to ``delay[i] = 1.5*delay[i-1]``. | |
|
682 | max_tries : int | |
|
683 | The max number of re-connection attempts. | |
|
645 | 684 | """ |
|
646 | 685 | if self.client_connector is None: |
|
647 | 686 | self.client_connector = ClientConnector() |
@@ -653,8 +692,22 b' class Cluster(object):' | |||
|
653 | 692 | def get_task_client(self, delay=DELAY, max_tries=MAX_TRIES): |
|
654 | 693 | """Get the task client for the running cluster. |
|
655 | 694 | |
|
656 | If this fails, it means that the cluster has not finished starting. | |
|
657 | Usually waiting a few seconds are re-trying will solve this. | |
|
695 | This will try to attempt to the controller multiple times. If this | |
|
696 | fails altogether, try looking at the following: | |
|
697 | * Make sure the controller is starting properly by looking at its | |
|
698 | log files. | |
|
699 | * Make sure the controller is writing its FURL file in the location | |
|
700 | expected by the client. | |
|
701 | * Make sure a firewall on the controller's host is not blocking the | |
|
702 | client from connecting. | |
|
703 | ||
|
704 | Parameters | |
|
705 | ---------- | |
|
706 | delay : float | |
|
707 | The initial delay between re-connection attempts. Susequent delays | |
|
708 | get longer according to ``delay[i] = 1.5*delay[i-1]``. | |
|
709 | max_tries : int | |
|
710 | The max number of re-connection attempts. | |
|
658 | 711 | """ |
|
659 | 712 | if self.client_connector is None: |
|
660 | 713 | self.client_connector = ClientConnector() |
@@ -52,6 +52,8 b' class UnknownStatus(LauncherError):' | |||
|
52 | 52 | class BaseLauncher(Component): |
|
53 | 53 | """An asbtraction for starting, stopping and signaling a process.""" |
|
54 | 54 | |
|
55 | # A directory for files related to the process. But, we don't cd to | |
|
56 | # this directory, | |
|
55 | 57 | working_dir = Unicode(u'') |
|
56 | 58 | |
|
57 | 59 | def __init__(self, working_dir, parent=None, name=None, config=None): |
@@ -64,11 +66,18 b' class BaseLauncher(Component):' | |||
|
64 | 66 | |
|
65 | 67 | @property |
|
66 | 68 | def args(self): |
|
67 |
"""A list of cmd and args that will be used to start the process. |
|
|
69 | """A list of cmd and args that will be used to start the process. | |
|
70 | ||
|
71 | This is what is passed to :func:`spawnProcess` and the first element | |
|
72 | will be the process name. | |
|
73 | """ | |
|
68 | 74 | return self.find_args() |
|
69 | 75 | |
|
70 | 76 | def find_args(self): |
|
71 |
"""The ``.args`` property calls this to find the args list. |
|
|
77 | """The ``.args`` property calls this to find the args list. | |
|
78 | ||
|
79 | Subcommand should implement this to construct the cmd and args. | |
|
80 | """ | |
|
72 | 81 | raise NotImplementedError('find_args must be implemented in a subclass') |
|
73 | 82 | |
|
74 | 83 | @property |
@@ -78,6 +87,7 b' class BaseLauncher(Component):' | |||
|
78 | 87 | |
|
79 | 88 | @property |
|
80 | 89 | def running(self): |
|
90 | """Am I running.""" | |
|
81 | 91 | if self.state == 'running': |
|
82 | 92 | return True |
|
83 | 93 | else: |
@@ -87,7 +97,7 b' class BaseLauncher(Component):' | |||
|
87 | 97 | """Start the process. |
|
88 | 98 | |
|
89 | 99 | This must return a deferred that fires with information about the |
|
90 | process starting (like a pid, job id, etc.) | |
|
100 | process starting (like a pid, job id, etc.). | |
|
91 | 101 | """ |
|
92 | 102 | return defer.fail( |
|
93 | 103 | Failure(NotImplementedError( |
@@ -96,12 +106,13 b' class BaseLauncher(Component):' | |||
|
96 | 106 | ) |
|
97 | 107 | |
|
98 | 108 | def stop(self): |
|
99 |
"""Stop the process and notify observers of |
|
|
109 | """Stop the process and notify observers of stopping. | |
|
100 | 110 | |
|
101 |
This must return a deferred that fires with |
|
|
102 | while the process is attempting to be shut down. This deferred | |
|
103 | won't fire when the process actually stops. These events are | |
|
104 | handled by calling :func:`observe_stop`. | |
|
111 | This must return a deferred that fires with information about the | |
|
112 | processing stopping, like errors that occur while the process is | |
|
113 | attempting to be shut down. This deferred won't fire when the process | |
|
114 | actually stops. To observe the actual process stopping, see | |
|
115 | :func:`observe_stop`. | |
|
105 | 116 | """ |
|
106 | 117 | return defer.fail( |
|
107 | 118 | Failure(NotImplementedError( |
@@ -123,9 +134,9 b' class BaseLauncher(Component):' | |||
|
123 | 134 | return d |
|
124 | 135 | |
|
125 | 136 | def notify_start(self, data): |
|
126 | """Call this to tigger startup actions. | |
|
137 | """Call this to trigger startup actions. | |
|
127 | 138 | |
|
128 | This logs the process startup and sets the state to running. It is | |
|
139 | This logs the process startup and sets the state to 'running'. It is | |
|
129 | 140 | a pass-through so it can be used as a callback. |
|
130 | 141 | """ |
|
131 | 142 | |
@@ -135,7 +146,10 b' class BaseLauncher(Component):' | |||
|
135 | 146 | return data |
|
136 | 147 | |
|
137 | 148 | def notify_stop(self, data): |
|
138 |
"""Call this to trigger |
|
|
149 | """Call this to trigger process stop actions. | |
|
150 | ||
|
151 | This logs the process stopping and sets the state to 'after'. Call | |
|
152 | this to trigger all the deferreds from :func:`observe_stop`.""" | |
|
139 | 153 | |
|
140 | 154 | log.msg('Process %r stopped: %r' % (self.args[0], data)) |
|
141 | 155 | self.stop_data = data |
@@ -205,6 +219,8 b' class LocalProcessLauncherProtocol(ProcessProtocol):' | |||
|
205 | 219 | class LocalProcessLauncher(BaseLauncher): |
|
206 | 220 | """Start and stop an external process in an asynchronous manner.""" |
|
207 | 221 | |
|
222 | # This is used to to construct self.args, which is passed to | |
|
223 | # spawnProcess. | |
|
208 | 224 | cmd_and_args = List([]) |
|
209 | 225 | |
|
210 | 226 | def __init__(self, working_dir, parent=None, name=None, config=None): |
@@ -246,24 +262,33 b' class LocalProcessLauncher(BaseLauncher):' | |||
|
246 | 262 | |
|
247 | 263 | @inlineCallbacks |
|
248 | 264 | def interrupt_then_kill(self, delay=2.0): |
|
265 | """Send INT, wait a delay and then send KILL.""" | |
|
249 | 266 | yield self.signal('INT') |
|
250 | 267 | yield sleep_deferred(delay) |
|
251 | 268 | yield self.signal('KILL') |
|
252 | 269 | |
|
253 | 270 | |
|
254 | 271 | class MPIExecLauncher(LocalProcessLauncher): |
|
272 | """Launch an external process using mpiexec.""" | |
|
255 | 273 | |
|
274 | # The mpiexec command to use in starting the process. | |
|
256 | 275 | mpi_cmd = List(['mpiexec'], config=True) |
|
276 | # The command line arguments to pass to mpiexec. | |
|
257 | 277 | mpi_args = List([], config=True) |
|
278 | # The program to start using mpiexec. | |
|
258 | 279 | program = List(['date'], config=True) |
|
280 | # The command line argument to the program. | |
|
259 | 281 | program_args = List([], config=True) |
|
282 | # The number of instances of the program to start. | |
|
260 | 283 | n = Int(1, config=True) |
|
261 | 284 | |
|
262 | 285 | def find_args(self): |
|
286 | """Build self.args using all the fields.""" | |
|
263 | 287 | return self.mpi_cmd + ['-n', self.n] + self.mpi_args + \ |
|
264 | 288 | self.program + self.program_args |
|
265 | 289 | |
|
266 | 290 | def start(self, n): |
|
291 | """Start n instances of the program using mpiexec.""" | |
|
267 | 292 | self.n = n |
|
268 | 293 | return super(MPIExecLauncher, self).start() |
|
269 | 294 | |
@@ -307,13 +332,32 b' class WindowsHPCLauncher(BaseLauncher):' | |||
|
307 | 332 | |
|
308 | 333 | |
|
309 | 334 | class BatchSystemLauncher(BaseLauncher): |
|
335 | """Launch an external process using a batch system. | |
|
336 | ||
|
337 | This class is designed to work with UNIX batch systems like PBS, LSF, | |
|
338 | GridEngine, etc. The overall model is that there are different commands | |
|
339 | like qsub, qdel, etc. that handle the starting and stopping of the process. | |
|
340 | ||
|
341 | This class also has the notion of a batch script. The ``batch_template`` | |
|
342 | attribute can be set to a string that is a template for the batch script. | |
|
343 | This template is instantiated using Itpl. Thus the template can use | |
|
344 | ${n} fot the number of instances. Subclasses can add additional variables | |
|
345 | to the template dict. | |
|
346 | """ | |
|
310 | 347 | |
|
311 | 348 | # Subclasses must fill these in. See PBSEngineSet |
|
349 | # The name of the command line program used to submit jobs. | |
|
312 | 350 | submit_command = Str('', config=True) |
|
351 | # The name of the command line program used to delete jobs. | |
|
313 | 352 | delete_command = Str('', config=True) |
|
353 | # A regular expression used to get the job id from the output of the | |
|
354 | # submit_command. | |
|
314 | 355 | job_id_regexp = Str('', config=True) |
|
356 | # The string that is the batch script template itself. | |
|
315 | 357 | batch_template = Str('', config=True) |
|
358 | # The filename of the instantiated batch script. | |
|
316 | 359 | batch_file_name = Unicode(u'batch_script', config=True) |
|
360 | # The full path to the instantiated batch script. | |
|
317 | 361 | batch_file = Unicode(u'') |
|
318 | 362 | |
|
319 | 363 | def __init__(self, working_dir, parent=None, name=None, config=None): |
@@ -324,6 +368,7 b' class BatchSystemLauncher(BaseLauncher):' | |||
|
324 | 368 | self.context = {} |
|
325 | 369 | |
|
326 | 370 | def parse_job_id(self, output): |
|
371 | """Take the output of the submit command and return the job id.""" | |
|
327 | 372 | m = re.match(self.job_id_regexp, output) |
|
328 | 373 | if m is not None: |
|
329 | 374 | job_id = m.group() |
@@ -334,6 +379,7 b' class BatchSystemLauncher(BaseLauncher):' | |||
|
334 | 379 | return job_id |
|
335 | 380 | |
|
336 | 381 | def write_batch_script(self, n): |
|
382 | """Instantiate and write the batch script to the working_dir.""" | |
|
337 | 383 | self.context['n'] = n |
|
338 | 384 | script_as_string = Itpl.itplns(self.batch_template, self.context) |
|
339 | 385 | log.msg('Writing instantiated batch script: %s' % self.batch_file) |
@@ -361,6 +407,7 b' class BatchSystemLauncher(BaseLauncher):' | |||
|
361 | 407 | |
|
362 | 408 | |
|
363 | 409 | class PBSLauncher(BatchSystemLauncher): |
|
410 | """A BatchSystemLauncher subclass for PBS.""" | |
|
364 | 411 | |
|
365 | 412 | submit_command = Str('qsub', config=True) |
|
366 | 413 | delete_command = Str('qdel', config=True) |
@@ -375,6 +422,7 b' class PBSLauncher(BatchSystemLauncher):' | |||
|
375 | 422 | #----------------------------------------------------------------------------- |
|
376 | 423 | |
|
377 | 424 | def find_controller_cmd(): |
|
425 | """Find the command line ipcontroller program in a cross platform way.""" | |
|
378 | 426 | if sys.platform == 'win32': |
|
379 | 427 | # This logic is needed because the ipcontroller script doesn't |
|
380 | 428 | # always get installed in the same way or in the same location. |
@@ -392,14 +440,17 b' def find_controller_cmd():' | |||
|
392 | 440 | |
|
393 | 441 | |
|
394 | 442 | class LocalControllerLauncher(LocalProcessLauncher): |
|
443 | """Launch a controller as a regular external process.""" | |
|
395 | 444 | |
|
396 | 445 | controller_cmd = List(find_controller_cmd()) |
|
446 | # Command line arguments to ipcontroller. | |
|
397 | 447 | controller_args = List(['--log-to-file','--log-level', '40'], config=True) |
|
398 | 448 | |
|
399 | 449 | def find_args(self): |
|
400 | 450 | return self.controller_cmd + self.controller_args |
|
401 | 451 | |
|
402 | 452 | def start(self, profile=None, cluster_dir=None): |
|
453 | """Start the controller by profile or cluster_dir.""" | |
|
403 | 454 | if cluster_dir is not None: |
|
404 | 455 | self.controller_args.extend(['--cluster-dir', cluster_dir]) |
|
405 | 456 | if profile is not None: |
@@ -413,12 +464,15 b' class WindowsHPCControllerLauncher(WindowsHPCLauncher):' | |||
|
413 | 464 | |
|
414 | 465 | |
|
415 | 466 | class MPIExecControllerLauncher(MPIExecLauncher): |
|
467 | """Launch a controller using mpiexec.""" | |
|
416 | 468 | |
|
417 | 469 | controller_cmd = List(find_controller_cmd(), config=False) |
|
470 | # Command line arguments to ipcontroller. | |
|
418 | 471 | controller_args = List(['--log-to-file','--log-level', '40'], config=True) |
|
419 | 472 | n = Int(1, config=False) |
|
420 | 473 | |
|
421 | 474 | def start(self, profile=None, cluster_dir=None): |
|
475 | """Start the controller by profile or cluster_dir.""" | |
|
422 | 476 | if cluster_dir is not None: |
|
423 | 477 | self.controller_args.extend(['--cluster-dir', cluster_dir]) |
|
424 | 478 | if profile is not None: |
@@ -426,15 +480,18 b' class MPIExecControllerLauncher(MPIExecLauncher):' | |||
|
426 | 480 | log.msg("Starting MPIExecControllerLauncher: %r" % self.args) |
|
427 | 481 | return super(MPIExecControllerLauncher, self).start(1) |
|
428 | 482 | |
|
429 | ||
|
430 | 483 | def find_args(self): |
|
431 | 484 | return self.mpi_cmd + ['-n', self.n] + self.mpi_args + \ |
|
432 | 485 | self.controller_cmd + self.controller_args |
|
433 | 486 | |
|
434 | 487 | |
|
435 | 488 | class PBSControllerLauncher(PBSLauncher): |
|
489 | """Launch a controller using PBS.""" | |
|
490 | ||
|
491 | batch_file_name = Unicode(u'pbs_batch_script_controller', config=True) | |
|
436 | 492 | |
|
437 | 493 | def start(self, profile=None, cluster_dir=None): |
|
494 | """Start the controller by profile or cluster_dir.""" | |
|
438 | 495 | # Here we save profile and cluster_dir in the context so they |
|
439 | 496 | # can be used in the batch script template as ${profile} and |
|
440 | 497 | # ${cluster_dir} |
@@ -456,6 +513,7 b' class SSHControllerLauncher(SSHLauncher):' | |||
|
456 | 513 | |
|
457 | 514 | |
|
458 | 515 | def find_engine_cmd(): |
|
516 | """Find the command line ipengine program in a cross platform way.""" | |
|
459 | 517 | if sys.platform == 'win32': |
|
460 | 518 | # This logic is needed because the ipengine script doesn't |
|
461 | 519 | # always get installed in the same way or in the same location. |
@@ -473,8 +531,10 b' def find_engine_cmd():' | |||
|
473 | 531 | |
|
474 | 532 | |
|
475 | 533 | class LocalEngineLauncher(LocalProcessLauncher): |
|
534 | """Launch a single engine as a regular externall process.""" | |
|
476 | 535 | |
|
477 | 536 | engine_cmd = List(find_engine_cmd()) |
|
537 | # Command line arguments for ipengine. | |
|
478 | 538 | engine_args = List( |
|
479 | 539 | ['--log-to-file','--log-level', '40'], config=True |
|
480 | 540 | ) |
@@ -483,6 +543,7 b' class LocalEngineLauncher(LocalProcessLauncher):' | |||
|
483 | 543 | return self.engine_cmd + self.engine_args |
|
484 | 544 | |
|
485 | 545 | def start(self, profile=None, cluster_dir=None): |
|
546 | """Start the engine by profile or cluster_dir.""" | |
|
486 | 547 | if cluster_dir is not None: |
|
487 | 548 | self.engine_args.extend(['--cluster-dir', cluster_dir]) |
|
488 | 549 | if profile is not None: |
@@ -491,7 +552,9 b' class LocalEngineLauncher(LocalProcessLauncher):' | |||
|
491 | 552 | |
|
492 | 553 | |
|
493 | 554 | class LocalEngineSetLauncher(BaseLauncher): |
|
555 | """Launch a set of engines as regular external processes.""" | |
|
494 | 556 | |
|
557 | # Command line arguments for ipengine. | |
|
495 | 558 | engine_args = List( |
|
496 | 559 | ['--log-to-file','--log-level', '40'], config=True |
|
497 | 560 | ) |
@@ -503,6 +566,7 b' class LocalEngineSetLauncher(BaseLauncher):' | |||
|
503 | 566 | self.launchers = [] |
|
504 | 567 | |
|
505 | 568 | def start(self, n, profile=None, cluster_dir=None): |
|
569 | """Start n engines by profile or cluster_dir.""" | |
|
506 | 570 | dlist = [] |
|
507 | 571 | for i in range(n): |
|
508 | 572 | el = LocalEngineLauncher(self.working_dir, self) |
@@ -551,12 +615,14 b' class LocalEngineSetLauncher(BaseLauncher):' | |||
|
551 | 615 | class MPIExecEngineSetLauncher(MPIExecLauncher): |
|
552 | 616 | |
|
553 | 617 | engine_cmd = List(find_engine_cmd(), config=False) |
|
618 | # Command line arguments for ipengine. | |
|
554 | 619 | engine_args = List( |
|
555 | 620 | ['--log-to-file','--log-level', '40'], config=True |
|
556 | 621 | ) |
|
557 | 622 | n = Int(1, config=True) |
|
558 | 623 | |
|
559 | 624 | def start(self, n, profile=None, cluster_dir=None): |
|
625 | """Start n engines by profile or cluster_dir.""" | |
|
560 | 626 | if cluster_dir is not None: |
|
561 | 627 | self.engine_args.extend(['--cluster-dir', cluster_dir]) |
|
562 | 628 | if profile is not None: |
@@ -575,7 +641,10 b' class WindowsHPCEngineSetLauncher(WindowsHPCLauncher):' | |||
|
575 | 641 | |
|
576 | 642 | class PBSEngineSetLauncher(PBSLauncher): |
|
577 | 643 | |
|
644 | batch_file_name = Unicode(u'pbs_batch_script_engines', config=True) | |
|
645 | ||
|
578 | 646 | def start(self, n, profile=None, cluster_dir=None): |
|
647 | """Start n engines by profile or cluster_dir.""" | |
|
579 | 648 | if cluster_dir is not None: |
|
580 | 649 | self.program_args.extend(['--cluster-dir', cluster_dir]) |
|
581 | 650 | if profile is not None: |
@@ -594,6 +663,7 b' class SSHEngineSetLauncher(BaseLauncher):' | |||
|
594 | 663 | |
|
595 | 664 | |
|
596 | 665 | def find_ipcluster_cmd(): |
|
666 | """Find the command line ipcluster program in a cross platform way.""" | |
|
597 | 667 | if sys.platform == 'win32': |
|
598 | 668 | # This logic is needed because the ipcluster script doesn't |
|
599 | 669 | # always get installed in the same way or in the same location. |
@@ -611,8 +681,10 b' def find_ipcluster_cmd():' | |||
|
611 | 681 | |
|
612 | 682 | |
|
613 | 683 | class IPClusterLauncher(LocalProcessLauncher): |
|
684 | """Launch the ipcluster program in an external process.""" | |
|
614 | 685 | |
|
615 | 686 | ipcluster_cmd = List(find_ipcluster_cmd()) |
|
687 | # Command line arguments to pass to ipcluster. | |
|
616 | 688 | ipcluster_args = List( |
|
617 | 689 | ['--clean-logs', '--log-to-file', '--log-level', '40'], config=True) |
|
618 | 690 | ipcluster_subcommand = Str('start') |
General Comments 0
You need to be logged in to leave comments.
Login now