Skip to content

Latest commit

 

History

History
91 lines (63 loc) · 2.63 KB

nci_computing.md

File metadata and controls

91 lines (63 loc) · 2.63 KB

Links

CLEX CMS wiki: http://climate-cms.wikis.unsw.edu.au/Home
NCI CMIP community page: https://opus.nci.org.au/display/CMIP/CMIP+Community+Home
NCI Data Training Page: https://nci-data-training.readthedocs.io/en/latest/index.html
NCI data catalogue: https://geonetwork.nci.org.au/

Help

Website: https://track.nci.org.au/servicedesk/customer/portals
Email: [email protected] (i.e. CMS team) or [email protected] (general NCI help)
Slack: https://arccss.slack.com
Request access to different projects here: https://my.nci.org.au/mancini/

CWS Virtual Desktop

VDI User Guide

Access requirements

  1. An NCI login with access to compute and storage resources
  • username: dbi599
  • compute/storage: I'm on r87 project, which means I should write to /local/r87/dbi599/tmp
    • Can also write to /g/data/r87/dbi599
  1. A client application that enables connection using the secure shell protocol and a VNC client application

Software solution

Install miniconda at /g/data/r87/dbi599 and use conda environments.

Version control

Need to use ssh (not https) when cloning.

The following seems to help with a git push ERROR: Permission to git denied to deploy key:

eval "$(ssh-agent -s)"  
ssh-add ~/.ssh/id_rsa_nci_virtuallab

Moving files to local computer

$ scp [email protected]:/g/data/r87/dbi599/figures/tauuo-zm/* .

If you want to transfer a file from the VDIs, replace gadi with the output from $ hostname (e.g. vdi-n1).

Creating symlinks

To create a new symlink (will fail if symlink exists already):

ln -s /path/to/file /path/to/symlink

To create or update a symlink:

ln -sf /path/to/file /path/to/symlink

Queuing system (on Gadi or Raijin)

To submit to the queue, run qsub job.txt where:

$ cat job.txt

#!/bin/bash
#PBS -P r87
#PBS -q normal
#PBS -l walltime=02:00:00
#PBS -l storage=gdata/r87+gdata/oi10
#PBS -l mem=60GB
#PBS -l wd

command...

Use qstat -u dbi599 -sw to check status of jobs (job state definitions here) and qdel <jobid> to kill one.

For large memory, single core jobs it's better to submit to the hugemem instead of normal queue.

Pangeo

User guide: https://opus.nci.org.au/display/Help/5.+Pangeo+on+Gadi
Notebooks: https://github.com/NCI-data-analysis-platform or https://nbviewer.jupyter.org/github/NCI-data-analysis-platform/