Installing the Cooperative Computing Tools
The CCTools software runs on Linux and Mac computers, whether laptops, desktops, or large high performance clusters. There are several ways to install:
- Install From Conda is best for most users on laptops or clusters.
- Install From Github is recommended for developers or those trying out the latest features.
- Install From Binary Tarball is recommended for use on specific supported platforms.
Install From Conda
Installing via the Conda package manager will give you a personal installation of the software in your home directory. You may already have Conda installed. To check:
$ conda -V
This displays the version of conda currently installed. If it fails, then you should install Miniforge. Miniforge is a light version of Anaconda that uses the conda-forge channel by default, and we recommend it as it is much faster to install.
Once Conda is installed, then install CCTools with:
# run this command once
$ conda create -n cctools-env -y -c conda-forge --strict-channel-priority python ndcctools
# run this command every time you want to use cctools
$ conda activate cctools-env
# run this command every time you want to update your version of cctools (after `activate` as above).
$ conda update -y -c conda-forge ndcctools
Note
You could simply run the command conda install -y -c conda-forge ndcctools
without creating
a new environment. However, sometimes conda takes a long time to resolve
all dependencies when adding new packages to an already existing
environment.
And that's it! You can test your setup following the instructions here.
Note
Parrot is not included as part of the Conda installation. You should instead install from Github or source tarball instead.
Install From Github
If you wish to install the latest version of CCTools to
try a new feature or develop new code then you can install from github.
This requires that you first install the necessary
software dependencies, and the easiest way to do this is to create
a new cctools-dev
environment via Conda:
git clone git://github.com/cooperative-computing-lab/cctools.git cctools-src
cd cctools-src
unset PYTHONPATH
conda env create -y -f environment.yml
conda activate cctools-dev
This can run slowly, so, for a potential speedup, try running:
unset PYTHONPATH
conda env create -y -f environment.yml --experimental-solver=libmamba
Now that you are inside the cctools-dev
environment, you can check out
the software repository and build it:
./configure --with-base-dir $CONDA_PREFIX --prefix $CONDA_PREFIX
make
make install
The next time you log in, be sure to activate the environment again:
unset PYTHONPATH
conda activate cctools-dev
Note
To use the installed software you will need to set the environment variables PATH
, PYTHONPATH
, and PERL5LIB
as explained here.
Install From Source Tarball
In a similar way, you can download a source package from the download page. Then unpack the tarball and compile:
$ tar zxf cctools-*-source.tar.gz
$ cd cctools-src
$ ./configure --prefix $HOME/cctools
$ make
$ make install
Note
To use the installed software you will need to set the environment variables PATH
, PYTHONPATH
, and PERL5LIB
as explained here.
Install From Binary Tarball
The CCTools software is pre-built for a small number of platforms. Download a source package from the download page. And follow this recipe while logged in as any ordinary user:
$ tar zxf cctools-*-source.tar.gz
$ cd cctools-*-source
$ ./configure --prefix $HOME/cctools
$ make
$ make install
Note
To use the installed software you will need to set the environment variables PATH
, PYTHONPATH
, and PERL5LIB
as explained here.
Setting Your Environment
If you installed CCTools from github, source tarball, or from a binary tarball, you will need to set some environment variables.
First determine the python
and perl
versions you are using:
# Note: your output may vary according to your python version.
$ python -c 'from sys import version_info; print("{}.{}".format(version_info.major, version_info.minor))'
3.7
# Note: your output may vary according to your perl version.
$ perl -e 'print("$^V\n")'
5.16.3
Now update your environment variables with those versions:
$ export PATH=$HOME/cctools/bin:$PATH
# Change 3.7 to the python version you found above.
$ export PYTHONPATH=$HOME/cctools/lib/python3.7/site-packages:${PYTHONPATH}
# Change 5.16.3 to the perl version you found above.
$ export PERL5LIB=$HOME/cctools/lib/perl5/site_perl/5.16.3:${PERL5LIB}
Testing Your Installation
You can test that the key programs are available with:
$ work_queue_worker --version
work_queue_worker version 7.4.3 FINAL (released 2022-02-03 16:26:52 +0000)
$ makeflow --version
makeflow version 7.4.3 FINAL (released 2022-02-03 16:26:52 +0000)
...
You can test that the python and perl modules are available with:
$ python -c 'import work_queue; print(work_queue.WORK_QUEUE_DEFAULT_PORT)'
9123
$ perl -MWork_Queue -e 'print("$Work_Queue::WORK_QUEUE_DEFAULT_PORT\n")'
9123
If the above commands fail, please make sure that you follow one (and only one!) of the methods above. For example, if you are using a conda installation, make sure that your PYTHONPATH is unset.
Warning
Remember that for installations from source you need swig at compile time, and to set
the environment variables PATH
, PYTHONPATH
and PERL5LIB
appropriately, as explained here.
For conda installation you should not need to manually set any of these variables, and in fact setting them may produce errors.
Special Cases
The software will happily build and run without installing any external packages. Optionally, the CCTools will interoperate with a variety of external packages for security, data access, and novel research methods. To use these, you must download and install them separately:
Once the desired packages are correctly installed, unpack the CCTools and then issue a configure command that points to all of the other installations. Then, make and install as before. For example:
$ ./configure --with-globus-path /usr/local/globus
$ make && make install
SWIG
CCTools needs SWIG during compilation to provide python
and perl support. SWIG is available through conda, or as a package of many
linux distributions. Once SWIG is installed, the configure
script should
automatically find it if the executable swig
is somewhere in your PATH
.
Note
If ./configure
cannot find your SWIG installation, you can use a command line option as follows: ./configure --with-swig-path /path/to/swig
iRODS
Building Parrot with support for the iRODS service requires some custom build instructions, since Parrot requires access to some internals of iRODS that are not usually available. To do this, first make a source build of iRODS in your home directory:
cd $HOME
git clone https://github.com/irods/irods-source
cd irods-source
git checkout 4.0.3
$ packaging/build.sh --run-in-place icommands
cd ..
Then, configure and build CCTools relative to that installation:
$ git clone https://github.com/cooperative-computing-lab/cctools cctools-source
$ cd cctools-source
$ ./configure --with-irods-path ../irods-src
$ make && make install
MPI
Building with MPI requires a valid MPI installation to be in your path.
Generally CCTools compiles with both intel-ompi and MPICH. If you do not have
mpi installed already, we suggest downloading the latest MPICH from the MPICH
website. The latest known supporting version of MPICH
with CCTools is MPICH-3.2.1. Simply build MPICH as is best for your
site/system, and then place the binaries in your path. We also suggest
configuring MPICH to use gcc
as the compiling software. For example:
$ tar -xvf MPICH-3.2.1.tar.gz
$ cd MPICH-3.2.1 CC=gcc CXX=g++
$ ./configure
$ make && make install
Once MPI is in your path, configure CCTools to use MPI and then install. For example:
$ cd ~/cctools
$ ./configure --with-mpi-path=`which mpicc`
$ make && install `
Now, our tools will be MPI enabled, allowing you to run Makeflow as an MPI job, as well as using both WorkQueue as MPI jobs, and submitting Makeflow and WorkQueue together as a single MPI job.
Build on Mac OSX
In order to build CCTools on Mac OS X you must first have the Xcode Command Line Tools installed. For OS X 10.9 and later this can be done using the following command:
xcode-select --install
Then, click "Install" in the window that appears on the screen. If the command line tools are already installed, you will get an error and can proceed with the instructions in the "Installing From Source" section above. For OS X versions before 10.9, you will need to first install Xcode. Xcode can be found in the App Store or on the installation disk.
License
The Cooperative Computing Tools are Copyright (C) 2003-2004 Douglas Thain and Copyright (C) 2022 The University of Notre Dame. All rights reserved. This software is distributed under the GNU General Public License.