The ICAT Job Portal
The ICAT Job Portal supports the submission of jobs to run on one or more batch services, applied to datasets and/or datafiles selected via the ICAT Data Service.
The web-based interface allows users to:
- select a job type to run
- search for and select datasets and/or datafiles to be supplied to the job(s)
- set other job options
- submit one or more instances of the job to the batch server(s)
- monitor the running status of each job
- view standard/error output of jobs
- cancel jobs that are in progress
- delete completed jobs
An IJP installation requires a central IJP server and one or more batch servers (one per batch system to which the central server will submit jobs). The central server's configuration includes:
- URLs of an ICAT and an IDS instance
- URLs of the batch servers to which it will submit jobs
- Definitions of the job types that it supports
- Default search filters for datasets and datafiles
A job type defines:
- a name for the job
- an executable to run on the batch system(s)
- whether job instances accept datasets, datafiles or both
- whether a single job can be run for multiple datasets / datafiles
- further (command-line) options to be passed to the job instance(s)
When the user submits a job, the IJP requests an estimate from each attached batch server, and selects one from those that return the best estimate.
Datasets and datafiles are passed to the executable by ID; the executable must use the ICAT and IDS instances to retrieve the metadata or contents. (The connection URLs and the session ID are also passed by the IJP.)
The unixbatch batch server uses the unix batch system. It can be used to test an IJP server installation (for example, it could be installed on the same machine).
There is a separate distribution, ijp.demojobs, of sample job types and scripts. The job types should be installed in the IJP server, and the scripts on the batch server(s). The scripts (mainly written in Python) include basic connectivity test scripts (e.g. report what was sent from the IJP server) as well as scripts to perform simple IDS tasks such as to create a new datafile, or to copy a datafile to another dataset. The scripts demonstrate how to check and extract the arguments supplied by the IJP, and how to use ICAT and IDS to resolve the supplied dataset and datafile IDs.