raggleton / SFrameBatch

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

SFrame Batch

This is a script to split the SFrame XML files and send the jobs to the batch with qsub, while still beeing compatible with plain SFrame. After the jobs are back it also takes care to hadd the jobs if you want it. There are some open issues and possible extensions.

##Install -> git clone this repo anywhere :)

-> For convenience you might have to chmod a+x sframe_split.py

##How to use -> SFrame creates jobTemp_* directories in your workdir. Take care to have enough disk space.

-> Setup sframe & cmssw for a 2.7 python version

-> export PATH=$PATH:/DirToScript/SFrameBatch/

-> Take the ConfigParse & ConfigSGE part from the example file and add it to your xml file. It has to stay commented at the beginnign of the file otherwise sframe would read it and not work.

-> Go to the directory where you run sframe_main from. This is not requiered by sframe_batch.py but if you use relative paths, it will not be possible to resolve them correctly and errors will be thrown.

-> Usage: sframe_batch.py [options] File.xml

-> sframe_batch.py File.xml creates the xml files, the needed sh files and tells you how many jobs you are going to submit.

-> You can still run all the xml files with sframe_main in the workdir.

-> To submit the jobs use the -s option. Pls make sure that you don't submit too many jobs.

-> For more have a look at the help: sframe_batch.py --help

-> Split by file and by events is in.

-> All the information of stdout and stderr is in workdir/Stream_xml-File-Name

-> For ppl interested in new features pls check the brach development. Since coding is most of the time rather easy and debbuging is not. At least some ppl using this branch would be very good.

Issues

-> very few safety & sanity checks

-> Code documantation is missing

-> optparser deprecated should be changed to argparser

Extensions & improvements

-> gui e.g. see https://github.com/wardi/urwid, it is shipped with cmssw

About


Languages

Language:Python 100.0%