The utcondor library provides Python tools for distributed computing on the Condor platform at UTCS. These tools probably won't be useful to you unless you're a computer science graduate student at the University of Texas at Austin.

The goal of utcondor is a reliable implementation of a simple distributed computing model. It requires little boilerplate, and switches easily between local and remote execution.

The utcondor library supports distributed computing tasks that can be cast as a one-level parallel map: a function executed over multiple inputs on multiple machines. This model is primitive, but easy to apply and often good enough. For a trivial example, to square a range of numbers in distributed fashion:

import condor

def f(x):
    return x**2

def main():
    calls = [(f, [x]) for x in range(16)]

    for (call, result) in condor.do(calls, 4):
        print call.args, result

if __name__ == "__main__":
    main()

Any arguments passed to the remotely-executed callable must be pickleable.

First, make sure that the condor binaries are accessible. Does "condor_q" run? If not, you may want to add /lusr/opt/condor/bin to your PATH.

Then you can install the package. The latest release is pip-installable, so

$ pip install utcondor

should work.

You're running inside a virtualenv, right?

The default Condor match expression is set in condor/defaults.py:

condor_matching = "InMastodon && (Arch == \"X86_64\") && (OpSys == \"LINUX\") && (Memory > 1024)"

Note that this includes the UT-specific "InMastodon" requirement. If you are attempting to use this library outside of UT, or if you want greater control over job matching, you can globally override this setting. For example, including the statement

condor.defaults.condor_matching = (
    "regexp(\"rhavan-.*\", ParallelSchedulingGroup)"
    " && (Arch == \"X86_64\")"
    " && (OpSys == \"LINUX\")"
    " && (Memory > 1024)")

in your code will require your jobs to run only on nodes with names that begin with "rhavan-".

Be careful. Pay attention to whether Condor jobs are being successfully cleaned up. Use at your own risk.

The primary author is Bryan Silverthorn bcs@cargo-cult.org.

This software package is provided under the non-copyleft open-source "MIT" license. The complete legal notice can be found in the included LICENSE file.