Welcome to the M3 user guide#
Important
**Upcoming Maintenance - 12-14 March 2024 **
Please be advised that Massive M3 will be undergoing a three-day scheduled maintenance starting Tuesday the 12th of March 2024.
Access to Massive M3 will not be available throughout this maintenance, as we will be conducting the following essential works:
upgrade the servers and clients for the Lustre /projects and /scratch file systems;
upgrade the system software on the M3 network switches; and
perform other critical updates.
For any concerns and enquiries, please contact our helpdesk at: help@massive.org.au
Important
Updates to Network Switches
Over the next few weeks (from 31 January 2024), we will be updating the network switches that underpin MASSIVE M3. Each switch takes a short time (i.e., several minutes) to update, but during this update, jobs that are sensitive to the network may be impacted. We will be placing batches of compute nodes on SLURM reservations to ensure that they will be drained of jobs before the updates.
The schedule of work and any potential impact to the service will be provided below.
[Update: Feb 09 2024]
We have scheduled the first lot of compute nodes for this update on the 13th of February 2024. Work should not take a couple of hours to complete.
[Update: Feb 16 2024]
Another 28 switches need to be updated and we will be scheduling a downtime for this. This downtime will also be used to perform necessary updates to the Lustre File system, along with other updates.
Important
Data Fluency Training - Note the date changes
We will be presenting two courses that might be of interest to new users. Please register your email at Data Fluency to be sent details on the courses.
Please see https://www.monash.edu/data-fluency/events for the link to the event and other events.
Introduction to Unix Shell (7th and 8th March 2024, 1/2 day each). To enroll please click at Data Fluency
Introduction to HPC (15th March 2024, 1 day)
Important
Hardware Refresh Plan – Update: 5 Jul 2022
Please be advised of the following M3 hardware refresh schedule. These servers are now coming into end-of-life and will be retired.
While this will result in a reduction of total CPU and GPU capacity, retiring these servers is necessary to make room for new and faster compute nodes. This round of procurement for new hardware is expected to be provisioned to include new remote desktop servers to replace those being decommissioned. In the interim, more desktops will be made available from our existing pool of GPUs. Note that only the Strudel Beta list of desktops will be updated to reflect the change in desktop options.
Compute Nodes |
Capability |
To be retired by |
---|---|---|
|
NVIDIA K1 GPU desktops |
21 Jun 2022 |
|
NVIDIA K80 GPU desktops |
18 Jul 2022 |
|
NVIDIA K80 GPU |
|
|
NVIDIA P100 GPU desktop |
|
|
NVIDIA P100 GPUs |
Late 2022 Exact date TBC |
We will be enabling the appropriate mechanisms (e.g., SLURM reservation
)
to ensure that these nodes will be idle of running jobs prior to their
retirement. Please check your job scripts to ensure they do not specify these
nodes using --nodelist
.
Important
Planned maintenance outages
We have scheduled quarterly outages for the M3 cluster. This is to ensure we communicate scheduled outages to our HPC users in advance. Where possible, we perform rolling upgrades with the cluster online. However, sometimes we have to perform upgrades that require the cluster to be taken offline. These include:
system software upgrades;
network maintenance;
bug and security patches; and
hardware maintenance
This site contains the documentation for the MASSIVE HPC systems. M3 is the newest addition to the facility.
- About M3
- Requesting an account
- Requesting help on M3
- Connecting to M3
- Connecting to M3 via ssh
- Connecting to M3 via the MASSIVE desktop
- Connecting to M3 via Strudel2
- Troubleshooting Common Issues Using Strudel
- I can login to Strudel and launch a desktop but cannot connect. I click the connect button and get an error saying “failed to connect to server”
- Strudel Web fails to connect to the desktop or Strudel2 crashes when trying to connect to desktop
- I can connect to a Desktop, but the display looks messed up, or I can see the display, but cannot interact with the Desktop!
- I’m using Strudel Desktop and get an error message “It looks like I was unable to contact the server for a list of sites to connect to…”
- File Systems on M3
- Copying files to and from M3
- Software on M3
- Running jobs on M3
- Partitions Available
- Other partitions
- Checking the status of M3
- Slurm Accounts
- Getting started with job submission scripts
- Running Simple Batch Jobs
- MPI on M3
- On CentOS 7
- On Rocky 9
- UCX_NET_DEVICES
- Running Multi-threading Jobs
- Running Interactive Jobs
- Running GPU Jobs
- Running Array Jobs
- QoS (Quality of Service)
- Features & Constraints
- Checking job status
- Project Allocation
- Lustre File System Quickstart Guide
- GPUs on M3
- MX2 Eiger
- Machine Learning
- Neuroimaging
- Cryo EM
- Bioinformatics
- The Genomics partition
- DGX
- Data Collections
- Machine learning
- ImageNet 2012 (ILSVRC2012)
- ImageNet 2015 Object Detection Data (ILSVRC2015 DET)
- International Skin Imaging Collaboration 2019 (ISIC 2019)
- NIH Chest X-ray Dataset (NIH CXR-14)
- Stanford Natural Language Inference (SNLI) Corpus
- COCO (Common Objects in Context) 2017
- AlphaFold
- AlphaFold v2 - AlphaFold-Multimer release
- Neuroimaging
- Human Connectome Project Dataset (HCP): HCP-1200
- Lifespan Human Connectome Project Development
- Lifespan Human Connectome Project Aging
- Baby Connectome Project
- Human Connectome Project for Early Psychosis
- Developing Human Connectome Project (dHCP)
- Brain Genomics Superstruct Project (GSP)
- Nathan Kline Institute Rockland Sample (NKI-RS): Neuroimaging Release
- Genomes
- Requesting a data collection
- Machine learning
- XNAT