Job Manipulation

There are several ways to control and manipulate your jobs within Ganga.

Copying Jobs

You can copy jobs using the copy method or using the cop-constructor in the Job creation. The job status is always set to new:

j = Job(name = 'original')
j2 = j.copy()
j2.name = 'copy'
j.submit()
j3 = Job(j, name = 'copy2')
jobs
Ganga Out [3]:
Registry Slice: jobs (4 objects)
--------------
    fqid |    status |      name | subjobs |    application |        backend |                             backend.actualCE |                       comment
-------------------------------------------------------------------------------------------------------------------------------------------------------------
       0 | completed |           |         |     Executable |          Local |                       epldt017.ph.bham.ac.uk |
       1 | completed |  original |         |     Executable |          Local |                       epldt017.ph.bham.ac.uk |
       2 |       new |      copy |         |     Executable |          Local |                                              |
       3 |       new |     copy2 |         |     Executable |          Local |                                              |

Accessing Jobs in the Repository

As shown before, you can view all the jobs that Ganga is aware of using the jobs command. To access a specific job from the repo with the parentheses, use it’s id number or:

jobs(2)

You can also use the square bracket ([]) notation to specify single jobs, lists of jobs or a job by a (unique) name:

jobs[2]
jobs[2:]
jobs['copy2']

Resubmitting Jobs

Jobs can fail for any number of reasons and often it’s a transient problem that resubmitting the job will solve. To do this in Ganga, simply call the resubmit method:

jobs(0).resubmit()

Note that, as above, this can also be used on completed jobs, though it’s backend and application dependent.

Forcing to Failed

Sometimes you may encounter a problem where the job has been marked completed by the backend but you notice in the logs that there was a problem which renders the output useless. To mark this job as failed, you can do:

jobs(1).force_status('failed')

Note that there are PostProcessors in Ganga that can help with a lot of these kind of problems.

Removing Jobs

As you submit more jobs, your Ganga repository will grow and could become quite large. If you have finished with jobs it is good practise to remove them from the repository:

jobs(2).remove()

This will remove all associated files and directories from disk.

Performing Bulk Job Operations

There are several job operations you can perform in bulk from a set of jobs. To obtain a list of jobs, you can either use the array syntax described above or the select method:

# can select on ids, name, status, backend, application
jobs.select(status='new')
jobs.select(backend='Local')
jobs.select(ids=[1,3])

# can restrict on min/max id
jobs.select(1,3, application='Executable')

Given this selection, you can then perform a number of operations on all of the jobs at once:

jobs.select(status='new').submit()

Available operations are: submit, copy`, ``kill, resubmit, remove. These also take the keep_going argument which, if set to True will mean that it will keep looping through the jobs even if an error occurs performing the operation on one of them. These operations can also be performed on subjobs as well - see SplittersAndPostprocessors for more info.

Export and Import of Ganga Objects

Ganga is able to export a Job object (or a select``ion of Job objects) or any other Ganga object using the ``export method which will create a human readable text file that you can edit manually and then load in using load:

export(jobs(0), 'my_job.txt')
jlist = load('my_job.txt')
jlist[0].submit()

As in the above example, any jobs loaded will be put into the new state.