Using the Download Partition
Compute nodes on SuperCloud do not have internet access.
This means that if you need to check out git repositories, download packages or download small data sets, you will need to do this on the login node before you start or launch your job. The login nodes have internet access for pip installs, git repo checkouts, etc.
To accommodate large data downloads, we have added a separate partition named "download" with data transfer nodes. This node is schedulable through this new download partition.
You can request use of this partition by requesting it with your job submission. If you are using sbatch or sbatch flags in your job script you can use the -p
or --partition
flag to specify the download partition:
--partition=download
-p download
If you are using LLsub
the flag to select a partition is -q
:
-q download
This enables you to launch a data download job before your job starts or after your job completes in order to sync whatever data you need.
Examples
Interactive Job with LLsub
LLsub -i -q download
Job Submission Script
where URL is the URL to the file you are trying to download. You would launch this job with either:
or
Monitoring your download job
You can execute the LLstat
command at the Linux command line to see if
your download job is running. We recommend checking the log file to
check for errors and to track the progress of your download.