Overview

Teaching: 20 min
Exercises: 15 min

Questions

Objectives

• Understand how to log in to the Supercomputing Wales hubs

• Understand the difference between the login node and each cluster’s head node.

Logging in

Before we can log in to Supercomputing Wales, we need to set a password. The Supercomputing Wales clusters use a separate password database to the four partner institutions, so the password must be set separately from (and should be different to) your institutional login. To do this visit MySCW, log in with your institutional credentials, and use the “Reset SCW Password” link in the left sidebar.

While you are visiting the MySCW dashboard, take a note of your Supercomputing Wales username. Your username is usually your institutional ID prefixed by a. for Aberystwyth users, b. for Bangor users, c. for Cardiff users, and s. for Swansea users. (Occasionally the part of the username after the . will vary from this pattern, but the prefix should always match this.) External collaborators will have a username beginning with x..

Aberystwyth and Swansea users (and their external collaborators) should log in to the Swansea Sunbird system by typing:

$ ssh username@sunbird.swansea.ac.uk

Bangor and Cardiff Users (and their external collaborators) should log in to the Cardiff Hawk system by typing:

$ ssh username@hawklogin.cf.ac.uk

If you use Windows and haven’t installed the Git bash shell, you can instead use PuTTY and enter either sunbird.swansea.ac.uk or hawklogin.cf.ac.uk in the hostname box.

Lockout

Since the Supercomputing Wales clusters allow connections from anywhere in the world, they need to be relatively tough to avoid unscrupulous hackers gaining access to your research data. This means that if you enter a wrong password (or SSH key) too many times in short succession, then your computer will be blocked from accessing the cluster for a few hours.

To avoid this happening in a workshop, we recommend raising a hand or otherwise letting an instructor know if your first attempt to log in fails.

If you have been locked out and urgently need to get back in without waiting for the block to expire, then contact the support desk. During working hours a member of the support team will be available to assist.

What’s available?

Supercomputing Wales

These figures may still be subject to some change and might have been sourced from out of date documents.

Aberystwyth and Swansea users are expected to use the Swansea system and will need to make a case for why they would need to use the Cardiff system. Bangor and Cardiff users are expected to use Cardiff, and external users are expected to use the same system as the owner of the project of which they are a member.

(There are also a number of GPU nodes available; these will be discussed in Running on GPUs.)

Slurm

Slurm is the management software used on Supercomputing Wales. It lets you submit (and monitor or cancel) jobs to the cluster and chooses where to run them.

Other clusters might run different job management software such as LSF, Sun Grid Engine or Condor, although they all operate along similar principles.

How busy is the cluster?

The sinfo command tells us the state of the cluster. It lets us know what nodes are available, how busy they are and what state they are in.

Clusters are sometimes divided up into partitions. This might separate some nodes which are different to the others (e.g. they have more memory, GPUs or different processors).

PARTITION     AVAIL  TIMELIMIT  NODES  STATE NODELIST
compute*        up 3-00:00:00      1   fail scs0042
compute*        up 3-00:00:00      1 drain* scs0004
compute*        up 3-00:00:00      2    mix scs[0018,0065]
compute*        up 3-00:00:00     86  alloc scs[0001-0003,0005-0017,0019-0035,0043-0046,0049-0064,0066-0072,0097-0122]
compute*        up 3-00:00:00     32   idle scs[0036-0041,0047-0048,0073-0096]
development     up      30:00      1   fail scs0042
development     up      30:00      1 drain* scs0004
development     up      30:00      2    mix scs[0018,0065]
development     up      30:00     86  alloc scs[0001-0003,0005-0017,0019-0035,0043-0046,0049-0064,0066-0072,0097-0122]
development     up      30:00     32   idle scs[0036-0041,0047-0048,0073-0096]
gpu             up 2-00:00:00      4   idle scs[2001-2004]

• The * after compute means that this is the default partition.
• AVAIL tells us if the partition is available.
• TIMELIMIT tells us if there’s a time limit for jobs
• NODES is the number of nodes in this partition in this particular state.
• STATE describes what these nodes are doing:
  • drain means that the node will become unavailable once the jobs currently running on it complete.
  • down means that the node is is powered off or otherwise unavailable for use.
  • alloc means that the node is fully allocated; all CPU cores are in use running users’ software.
  • mix means that some of the CPU cores on a node are allocated to a user, and others are available for use.
  • idle means that the node is not currently allocated, and is available for use.

Exercises

Logging into Supercomputing Wales

If you haven’t already:

1. In your web browser go to My Supercomputing Wales and log in with your university username and password.
2. Click on “Reset SCW Password” and choose a new password for logging into the HPC. Your username is displayed in the “Account summary” box on the main page. Its usually a./b./c./s. and your normal university login details.
3. Log in to sunbird.swansea.ac.uk or hawklogin.cf.ac.uk using your SSH client.
4. Run the sinfo command to see how busy things are.
5. Try sinfo --long, what extra information does this give?

Key Points

• ssh sunbird.swansea.ac.uk or ssh hawklogin.cf.ac.uk to log in to the system

• sinfo shows partitions and how busy they are.

• slurmtop shows another view of how busy the system is.