best practices
NEVER run jobs on the gateway. Always either reserve an interactive node, or submit your jobs through batch scripts, as described here.
Avoid running disk I/O heavy jobs from your home directory. Additional options can be found in disk storage.
You'll soon realize you need to write the same commands over and over while using discovery cluster. Defining aliases for such commands can speed up your workflow. You need to define aliases in your ~/.bashrc file. Structure looks like this:
alias aliasname='commands'So whenever you call aliasname, commands will be executed. Some useful examples are:
alias sq="squeue -l -u ($username)"
alias si="sinfo -Nle"More examples of useful aliases can be found in this .bashrc file example.
To avoid entering your password every time you connect to Discovery, you can generate an SSH key and use it for authentication during your connection. Instructions on how to do so can be found here.
You can also create an alias for Discovery and declare your username by editing your SSH configuration file. In particular, by adding these lines
Host xfer
HostName xfer.discovery.neu.edu
User YOURUSERNAME
Host discovery
HostName login.discovery.neu.edu
User YOURUSERNAMEto your (local) .ssh/config file. After these aliases are set up, you can subsequently log in to discovery by typing:
ssh discovery
and transfer files via, e.g.:
scp LOCALFILE xfer:/scratch/REMOTEFILE
