Running a Python job using HTCondor
This section assumes access to a cluster managed by the open source HTCondor job scheduler. The installation of HTCondor is not difficult (described in the administrator's manual available at https://research.cs.wisc.edu/htcondor/manual/), but it is outside the scope of this book.
HTCondor comes with a set of command-line tools that can be used to submit jobs to the cluster (condor_submit
), view the status of any submitted job (condor_q
), kill a job (condor_rm
), and view the status of all the machines in the cluster (condor_status
). There are many other tools—more than 60 in total; however, we will concentrate on the four main ones listed in this section.
Another way of interacting with an HTCondor cluster is using the Distributed Resource Management Application API (DRMAA), which comes with most (but not all) HTCondor installations and is packaged as a shared library (for example, libdrmaa.so
on Linux).
DRMAA abstracts away most of the vendor specificity...