Running Jobs on M1 from M2

MASSIVE is made up of two parts, M2 (Monash) and M1 (Australian Synchrotron), which are connected by 6 x 10G links. The clusters can be considered them same and users can submit jobs to either, however there are a number of things to consider when doing so:

1. The compute nodes of each cluster half cannot see the filesystem of the other half, so files must be copied between clusters before processing can be done (although a automated mechanisim is being developed)
2. Each cluster has it's own home directory so files in your home dir on M2 will not be visible on M1
3. Only M2 has project directories (which are backed up) so in most cases you will need to use the scratch space on M1 (which is not backed up) and then copy the results you wish to keep back to M2
4. M1 and M2 have the same software installed, with M2 being the "master", meaning software is installed and tested on M2 and then copied to M1. Please to resolve any issues with software.


Considering the above, here is the general process for running jobs on M1 from M2:

Take Monash001 project for example.

1. Log into

ssh <userID>

2. Copy the files you need to run the job to m1-login2:

scp -r ~/Monash001/job_dir m1-login2:~/Monash001_scratch/

3. Log into m1-login2

ssh <userID>

4. Run your jobs:

cd ~/Monash001_scratch/job_dir
sbatch job.script

4. Save your results file (e.g. result.txt) back to your project directory on M2

scp -r ~/Monash001_scratch/job_dir/results.txt m2-login2:~/Monash001/job_dir/


Copyright © 2016 MASSIVE. All Rights Reserved.