Accessing a cluster


* Compute Canada documentation:

Logging in

A secure shell client must be used to connect to the head nodes of a cluster. No interactive access to the compute nodes is provided to users. In order to access the head node, use the Compute Canada credentials (user name and password).

Linux users

ssh -l username

(You need to add "-X" option after "ssh" if you expect to output any graphics to your screen)

Windows Users

You have to install special software that allows secure shell connections. There are a number of free, third party packages available to accomplish this. The most popular SSH clients for Windows are PuTTY and MobaXterm.

A particularly comprehensive package is Cygwin. It represents a full-featured Linux-type environment inside your windows session. You may need to install this environment if you expect to output graphic objects on your screen. This tutorial explains the steps required for its installation.

Once the ssh client is installed, you may need to create a profile. If so, you will need the following information:

Host name :
User name : Enter your user name
Port number: 22

NOTICE #1: There are 5 general-access login nodes on Guillimin cluster. When you use the given above login address (, you login to a "round-robin" device which further forwards you to the login node with the lowest load at that particular time. This means that every time you log in to the cluster you may find yourself on one of 5 login nodes. There are also specific login names and groups:

  •,,,, - login to specific login nodes.
  • - round-robin for any "phase 1" login node with a Westmere processor, presently only
  • - round-robin for any "phase 2" login node with a Sandy Bridge processor, presently all other 4 login nodes.

The last two aliases are useful if you compile your software and it by default generates code for the native processor, in which case code compiled on a Sandy Bridge processor will not run on an older Westmere processor (any "phase 1" worker node). By compiling on a Westmere node your code will then run anywhere in the cluster.

HostnameAddressPhase and CPU typeGroup AddressGlobal Address
lg-1r14-n04 Phase 1: Westmere processor
lg-1r17-n01 Phase 2: Sandy Bridge processor


NOTICE #2: The login nodes are not at anytime to be used for running jobs or workloads other than only for a few minutes. The nodes are shared with other users and any excessive use of CPU beyond a few minutes, or large amounts of memory, will result in the user processes being terminated without notice. Please use the batch queue system as this is the best and only method by which jobs should be submitted. Please see this documentation for more information.

File transfer