globus-job-submit-ws

Home -> Toolkit -> Docs -> 4.0 -> Execution -> Wsgram	Website Email Lists Search:

Name

globus-job-submit-ws — Batch job submission script for WS GRAM (requires globus_wsrf_gram_client_tools update package from http://www.globus.org/toolkit/downloads/development/)

Synopsis

globus-job-submit-ws GLOBAL OPTIONS

GLOBAL OPTIONS:

[-help|-usage] [-version] [-verbose] [-dumpcmd] [-dumpjdd] -o|-job-epr-output-file filename [-contact contact string] [-factory-type ResourceID] [-z|-authorization authz] [-delegation-epr-file EPR file] [-gridftp URL] [-jobid job ID] [-np|-count N] [-host-count N] [-max-time minutes] [-max-memory MB] [-min-memory MB] [-project project ID] [-queue queue ID] [-dir directory] [-job-type job type] [-stdin file] [-stdout file] [-stderr file] [-library-path path...] [-env name=value[ name=value]...] [-args arg[ arg]...] {[HOST CLAUSE] | [-: OPTIONS HOST CLAUSE[ -: OPTIONS HOST CLAUSE]......]}

OPTIONS:

[-np|-count N] [-host-count N] [-max-time minutes] [-max-memory MB] [-min-memory MB] [-project project ID] [-queue queue ID] [-dir directory] [-job-type job type] [-stdin file] [-stdout file] [-stderr file] [-env name=value[ name=value]...]

HOST CLAUSE:

contact string [[-l] | [-s]] executable arg[ arg]...

Description

	Important
	This package requires the globus_wsrf_gram_client_tools update package from http://www.globus.org/toolkit/downloads/development/.

globus-job-submit-ws is a wrapper script around globusrun-ws... It makes it easier to submit batch jobs. Batch jobs are defined as those that are submitted but not actively monitored for state changes. One must use globusrun-ws to check the job's state. Also, when the job is finished one can use globus-job-get-output-ws to fetch the output of the job and globus-job-cliean-ws to destroy the job and clean up the job's scratch space.

Command options

Global Options

The global options are used to 1) affect the script execution, 2) control the job submission, and 3) provide defaults for subjobs when using the multijob syntax.

-help|-usage: Prints the command usage and exits.
-version: Prints the version of the client program and exits.
-verbose: Prints extra script execution debug messages.
-dumpcmd: Prints the actual globusrun-ws command that would have been executed and exits.
-dumpjdd: Prints the actual Job Description Document that would have been passed to globusrun-ws and exits.
-o|-job-epr-output-file: Causes the job's EPR to be written to the specified file. This argument is not optional.
-contact: Specifies the factory contact string for where to submit the multijob. This argument is ignored if there is only one single job host clause (i.e. no '-:' arguments).
-factory-type: Specifies the factory resource ID for the target factory service. Supported values are "Fork", "Multi", "Condor", "LSF", and "PBS". For single jobs the the default is "Fork". For multijobs the default is "Multi".
-z|-authorization: Specifies the type of authorization to do against the service. Valid values are "host", "self", or a certificate subject (i.e. identity authorization). The default is "host" if not specified.
-delegation-epr-file: Specifies a file with an EPR to a delegated credential that is to be used for the job, staging, and transfer credential. If you need more flexibility over these different credentials separately, please use globusrun-ws directly.
-gridftp: Specifies the base URL for the GridFTP server used local to the client for staging files to and from the server.
-jobid: Specifies the ID of the job. This overrides the automatically generated UUID.
-np|-count: Specifies the number of processes to spawn for this job.
-host-count: Specifies the number of processing elements to employ for running the job process(es).
-max-time: Specifies the maximum amount of time the scheduler should allow for running the job process(es).
-max-memory: Specifies the maximum amount of memory the scheduler should allow for running the job process(es).
-min-memory: Specifies the minimum amount of memory the job needs to run.
-project: Specifies the project ID associated with the job.
-queue: Specifies the target queue that the job should be placed in when the job is submitted to the scheduler.
-dir: Specifies the working directory for the job process(es).
-job-type: Specifies the type of job this is. Valid values are "single", "multiple", "condor", and "mpi". The default value is "multiple".
-stdin: Specifies the stdin input file path local to the job process(es).
-stdout: Specifies the stdout output file path local to the job process(es).
-stderr: Specifies the stderr output file path local to the job process(es).
-library-path: Specifies the a path to be added to the library path of the job. This argument can be repeated.
-env: Specifies name/value pairs separated by '=' to be used as environment variable assignments for the job process(es). The name/value pair values can be repeated until a dashed argument is specified or there are only two arguments left on the command line (i.e. ending host clause).
-args: Specifies default command arguments for the subjobs. The values can be repeated until a host clause is encountered. Arguments specified with the executable in the host clause are appended to these values.

Options

These options can be specified before each subjob host clause when using the multijob syntax.

-factory-type: Specifies the factory resource ID for the target factory service. Supported values are "Fork", "Condor", "LSF", and "PBS". The default is "Fork". A value of "Multi" is not allowed in the subjob options because the service semantics do not allow this.
-np|-count: Specifies the number of processes to spawn for this job.
-host-count: Specifies the number of processing elements to employ for running the job process(es).
-max-time: Specifies the maximum amount of time the scheduler should allow for running the job process(es).
-max-memory: Specifies the maximum amount of memory the scheduler should allow for running the job process(es).
-min-memory: Specifies the minimum amount of memory the job needs to run.
-project: Specifies the project ID associated with the job.
-queue: Specifies the target queue that the job should be placed in when the job is submitted to the scheduler.
-dir: Specifies the working directory for the job process(es).
-job-type: Specifies the type of job this is. Valid values are "single", "multiple", "condor", and "mpi". The default value is "multiple".
-stdin: Specifies the stdin input file path local to the job process(es).
-stdout: Specifies the stdout output file path local to the job process(es).
-stderr: Specifies the stderr output file path local to the job process(es).
-library-path: Specifies the a path to be added to the library path of the job. This argument can be repeated.
-env: Specifies name/value pairs separated by '=' to be used as environment variable assignments for the job process(es). The name/value pair values can be repeated until a dashed argument is specified or there are only two arguments left on the command line (i.e. ending host clause).

Host Clause

The host clause specifies the service contact string and executable.

contact string

Specifies the target service. Valid forms of the contact string are as follows:

                               hostname
                               hostname:port
                               hostname:port/service
                               hostname/service
                               protocol://hostname
                               protocol://hostname:port
                               protocol://hostname:port/service
                               protocol://hostname/service

-l|-s

Specifies whether the executable is already local to the execution machine (-l) or should be staged (-s). The default is -l.

executable

The local path to the program or script to be executed.

Notes

Job stdin defaults to /dev/null. Job stdout/stderr defaults to "stdout/err" in the scratch directory. The file modifiers -l and -s specify different filespaces: -l[ocal] file is relative to working directory of job (DEFAULT) -s[tage] file relative to job request is staged to job host The working directory of the submitted job defaults to $HOME. More than one host clause can be specified, in which case the job runs on all of the hosts simultaneously. The host clause begins with '-:' for a multi-request, and the contact string of the remote resource for a single request. It allows any of the specified options in any order. The first unrecognized (non-flagged) term is treated as the executable, with arguments following immediately after. In the multi-request case, arguments may be specified specifically for each host-clause by appending them after the name of the executable, or job-wide with the -args flag. In either case, arguments continue until the end of the command-line or until the next '-:' flag signifying a new host-clause. Two '-:' flags in a row are interpreted as an escaped '-:' and the host- clause continues until the next unpaired '-:' flag or end of input. Except for -library-path, if more than one copy of an argument is specified (such as two -stdin definitions), the setting defined last on the command line will be used.