GC3Pie is a python package for executing computational workflows consisting of tasks with complex inter-dependencies. GC3Pie accomplishes this by defining a run-time task list in which a task can only be executed once all upstream task dependencies have been successfullly completed. In contrast to other workflow managers, GC3Pie accplications are written in python and not a markup language. The advantage is that this makes it trivial to write highly complex workflows. GC3Pies roots are in shared-nothing achitectures (server-less for example) and can be configured to use backends such as batch clusters or clouds.
GC3Pie is a suite of Python classes (and command-line tools built upon them) to aid in submitting and controlling batch jobs to clusters and grid resources seamlessly. GC3Pie aims at providing the building blocks by which Python scripts that combine several applications in a dynamic workflow can be quickly developed.
The GC3Pie suite is comprised of three main components:
- GC3Libs: A python package for controlling the life-cycle of a Grid or batch computational job
- GC3Utils: Command-line tools exposing the main functionality provided by GC3Libs
- GC3Apps: Driver scripts to run large job campaigns
GC3Libs provides services for submitting computational jobs to Grids and batch systems and controlling their execution, persisting job information, and retrieving the final output.
GC3Libs takes an application-oriented approach to batch computing. A
generic Application class provides the basic operations for
controlling remote computations, but different Application
subclasses can expose adapted interfaces, focusing on the most
relevant aspects of the application being represented. Specific
interfaces are already provided for the GAMESS and Rosetta suites;
new ones can be easily created by subclassing the generic
Application class.
Most of the time users have lots of different accounts on several diverse resources. The idea underlying GC3Utils is that a user can submit and control a computational job from one single place with a few simple commands.
Commands are provided to submit a job (gsub), check its running
status (gstat), get a snapshot of the output files (gget,
gtail), or cancel it (gkill).
There is a need in some scientific communities, to run large job campaigns to analyze a vast number of data files with the same application. The single-job level of control implemented by GC3Utils in this case is not enough: you would have to implement "glue scripts" to control hundreds or thousand scripts at once. GC3Pie has provisons for this, in the form of re-usable Python classes that implement a single point of control for job families.
The GC3Apps scripts are driver scripts that run job campaigns using the supported applications on a large set of input data. They can be used in production as-is, or adapted to suit your data processing needs.
There are several examples and a tutorial in the examples directory.
import sys
import time
import gc3libs
class GdemoSimpleApp(gc3libs.Application):
"""
This simple application will run `/bin/hostname`:file: on the remote host,
and retrieve the output in a file named `stdout.txt`:file: into a
directory `GdemoSimpleApp_output`:file: inside the current directory.
"""
def __init__(self):
gc3libs.Application.__init__(
self,
# the following arguments are mandatory:
arguments = ["/bin/hostname"],
inputs = [],
outputs = [],
output_dir = "./GdemoSimpleApp_output",
# the rest is optional and has reasonable defaults:
stdout = "stdout.txt",)
# Create an instance of GdemoSimpleApp
app = GdemoSimpleApp()
# Create an instance of `Engine` using the configuration file present
# in your home directory.
engine = gc3libs.create_engine()
# Add your application to the engine. This will NOT submit your
# application yet, but will make the engine awere *aware* of the
# application.
engine.add(app)
# in case you want to select a specific resource, call
# `Engine.select_resource(<resource_name>)`
if len(sys.argv)>1:
engine.select_resource(sys.argv[1])
# Periodically check the status of your application.
while app.execution.state != gc3libs.Run.State.TERMINATED:
print "Job in status %s " % app.execution.state
# `Engine.progress()` will do the GC3Pie magic:
# submit new jobs, update status of submitted jobs, get
# results of terminating jobs etc...
engine.progress()
# Wait a few seconds...
time.sleep(1)
print "Job is now terminated."
print "The output of the application is in `%s`." % app.output_dirThis is what it looks like when the code is run:
$ python gdemo_simple.py localhost
gdemo_simple.py: [2019-01-21 14:37:53] INFO : Computational resource 'localhost' initialized successfully.
Job in status NEW
gdemo_simple.py: [2019-01-21 14:37:55] INFO : Successfully submitted GdemoSimpleApp@7f07aa094a90 to: localhost
Job in status SUBMITTED
Job is now terminated.
The output of the application is in `./GdemoSimpleApp_output`.
The output file looks as follows:
$ cat GdemoSimpleApp_output/stdout.txt
93607d089233
For up-to-date information, please read the GC3Pie documentation at: http://gc3pie.readthedocs.io/
Installation instructions are in the INSTALL.rst file (in this same directory), or can be read online at: http://gc3pie.readthedocs.io/en/latest/users/install.html
The GC3Pie library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.
A copy of the GNU Lesser General Public License is in file LICENSE.md.