Using Cluster (Quick Version)

Note: This page is outdated (June 1, 2016). I will prepare a newer version.

Since most of the scripts for cluster use is located under a specific directory in the ALMF server, I recommend you to add the path to your .tcshrc as below.

set path=($path /g/almf/software/bin2)

Otherwise, please add the path above in front of script name.

Before submitting huge data with your script, you might want to test run the script interactively to see if it is indeed working. To do so, you could access cluster in interactive mode.

1. Log in to sub-master
2. then “qsub -I”
   1. wait for some seconds, you will be assigned to a specific node where you could test around.
   2. To enable multi-threaded computation, do “qsub -I -l select=ncpus=2:mem=3gb” (in this case, uses 2 CPUs and 3Gb of memory).

Script name: doIJ

Syntax

sh doIJ <fullpath to macro file> <fullpath to image file>

For example:

sh /g/almf/software/bin2/doIJ.sh /g/almf/software/ij/getArgumentMacro.ijm /g/almf/miura/blobs.tif

Use this script only for small tasks. Heavy calculation consumes the server resource where others are also using.

Script name: qsubIJ

To process image file /a/b/c.tif with an ImageJ macro file /d/e/f.ijm, like above, but with cluster

sh qsubIJ /d/e/f.ijm /a/b/c.tif

Job will be submitted to the cluster, and what should be the output from ImageJ will be saved as a text file in your ALMF home directory:
/g/almf/<Your Unix Username>/CLlogyyyymmdd_hhmm.txt.
“yyyymmdd_hhmm” should be the time of execution.

To customize arguments and so on, shell commands should be written in another shell script as follows:

#!/bin/sh
#PBS -N ImageJProcess
#PBS -l walltime=1:00:00
#
# *** Script for single image file procesing by ImageJ ***
#
#   two arguments for full paths to (2) an image file and (1) a macro file is required
#   assuming that macro file takes the fisrt argument as the 
#   Kota Miura (cmci.embl.de)
#   20101103
#   ****************************************************
#
sh /g/almf/software/ij/doIJ.sh /g/almf/software/ij/getArgumentMacro.ijm /g/almf/miura/blobs.tif

and then save the script as something like “cl_doIJ.sh”. Execute the job by following command in terminal

/usr/pbs/bin/qsub -q clusterng /g/almf/software/ij/cl_doIJ.sh

Script name: qsubIJbatch

Syntax is

sh qsubIJbatch </full/path/to/dir> <File Name of Macro File>

For example, if you want to process all images under /g/almf/miura/testsmalls using an ImageJ macro file headlesstest3.ijm, then your command would be:

sh qsubIJbatch /g/almf/miura/testsmalls headlesstest3.ijm

converting lif to tif stacks:
sh /g/almf/software/ij/lifcon.sh /g/haering/Pombescreen/test.lif
sh /g/almf/software/bin2/lifconvert /g/haering/Pombescreen/pb2/100927ExpNo2.lif

actual processing:
sh /g/almf/software/bin2/qsubIJbatch /g/haering/Pombescreen/test.lif_tifStack headlesstest3.ijm