Welcome to MARCC.

Overview

This web page contains the most basic information any user should be aware of before using this system. Advanced information on several topics is available on different web pages within this web site.

Obtaining access

PIs (any faculty member) and users must first request access and be given an allocation, which determines the computational resources (measured in core-hours –wall time hours of computer time) they can use. Request depends on the user’s role:

Logging in

Once the account has been created with MARCC, you may connect to the login nodes using the password you provided in the request form. A secure shell (SSH) must be used to access the main nodes for job submission; ssh implementations come with Linux and Mac terminals, while we recommend PuTTY or Cygwin for Windows users.

In addition, users will also need to set a Two Factor Authentication code.

From a terminal, you can connect with the command

$ ssh [userid]@gateway2.marcc.jhu.edu

Please refer to the Connecting to MARCC page for further information.

Running jobs

The login nodes are only meant for job submission and setting up files (copying/moving files and basic text editing). Anything requiring more processing power, such as compiling programs, running simulations, mathematical software, etc., must be submitted to the queuing system. The queuing system requires a script that specifies the resources requested (nodes, cores, time, etc) and the commands that must be run (including the environment variables).

To queue your job, run the sbatch command with your script:

$ sbatch script-name

To view your jobs in the queue:

$ squeue -u [userid]

And to cancel a job:

$ scancel job-id

Interactive Processing

If users need to run application that need interactive processing, like applications that need a GUI, post-processing, or visualization, MARCC provides an “interact” command that will allow users to request an interactive session hosted on a compute node. Please type:

$ interact -usage

for additional information.

Storing and accessing data

Users can access their data through a CIFS/SMB mount, or copy it across the network using secure copy (scp) or rsync from within a Linux/Mac/Cygwin terminal. A simple scp command from a terminal would be

$ scp sourcedir/filename.ext [userid]@gateway2.marcc.jhu.edu:/home/[userid]/targetdir/

To transmit more than a few files at a time, we recommend using rsync, which only transfers the differences between files if partial data already exists at the target file system.

For large transfer files ( > 1 GByte) please use Globus Connect through the transfer node marcc#dtn2.

Mounting file systems via sshfs

Please use this option only if  needed. Different file systems can be mounted to your local desktop or workstation using libfuse/sshfs. Make sure you install the proper version according to your local OS (Mac, Windows, or Linux). Here is an example to mount “scratch” on a Mac desktop:

my-mac. mkdir MYMountPoint

my-mac> sshfs -p 22 userid@dtn2.marcc.jhu.edu:/scratch/users/userid ~/MYMountPoint -oauto_cache,reconnect,defer_permissions,noappledouble,negative_vncache,volname=myVolname