Note that the Git repository does not include the software required for the jobs. For now, grab the software tarball from http://www.isi.edu/~rynge/soybean/software.tar.gz . Untar it in the cloned workflow directory.
The workflow is controlled by a configuration file named ~/.irri-workflow.conf . Create the file with this content:
# local refers to the submit host. Specify paths to a directory
# which can be used by the workflow as work space, and locations
# for local software installs.
[local]
work_dir = /local-scratch/%(username)s/irri
irods_bin = /ccg/software/irods/3.2/bin
# tacc refers to configuration for the TACC Stampede
# supercomputer. To use this machine, you need an allocation
# (start with TG-) and you also need to know your username
# and storage group name for the system. The easiest way to
# obtain those is to log into the system, and run:
# cds; pwd
# This should return a path like: /scratch/00384/rynge. The
# storage group is the second level, and your username is
# last level.
[tacc]
allocation = TG-ABC1234
username = rynge
storage_group = 00384
Basic files are pulled from the submit host with scp. This is to keep the requirements on the submit host light, and make it easy to run the submit host as a cloud instance in for example Atmosphere. A workflow needs an ssh key:
$ mkdir -p ~/.ssh
$ ssh-keygen -t rsa -b 2048 -f ~/.ssh/workflow
(just hit enter when asked for a passphrase)
$ cat ~/workflow.pub >>~/.ssh/authorized_keys
To access data from the iPlant iRods repository, you need a file in your home directory named ~/irods.iplant.env, with 0600 permission and content like:
irodsHost data.iplantcollaborative.org
irodsPort 1247
irodsUserName YOUR_IRODS_USERNAME
irodsZone iplant
# Pegasus requirement
irodspassword 'YOUR_IRODS_PASSWORD'
To run on TACC, you need a X509 proxy. Create one with:
$ myproxy-logon -s myproxy.teragrid.org -t 144:00 -l YOUR_XSEDE_USERNAME
To generate the workflow:
$ ./workflow-generator --exec-env tacc-stampede
The workflow can also run in a distributed mode, using HTCondor to schedule the jobs. Currently, the configuration is for the ISI submit node, and the workflow can be generated with:
$ ./workflow-generator --exec-env distributed