M3 (Monash HPC)¶

M3 is Monash University's HPC cluster. The lab project is mh42 — 500 GB shared quota across all members.

Members¶

Project: mh42 | 33 members | Last updated: 2026-05-04

Name	M3 Username	Role
Dana Kulić	`danak`	PI — project leader
Kavi Katuwandeniya	`kkatuwan`	Project leader
Lingheng Meng	`lmen0023`	Project leader
Somayeh Hussaini	`somayehh`	Visiting Post-doc
Rhys Newbury	`rnewbury`	Post-doc
Shiyu Xu	`shiyux`	Post-doc/PhD
Lirui Guo	`liruig`	PhD Student
Phan Le	`phanl`	PhD Student
Malak Sayour	`msay0013`	Research Student
Mohammad Zawari	`mzawari`	Research Student
Alana Hayes Chen	`ahayesch`	Member
Aron Thomas	`atho0057`	FYP Student
Ben McShanag	`bmcs0003`	FYP Student
Blake Matheson	`bmat0014`	FYP Student
Duleesha Gunaratne	`dgun0024`	FYP Student
Isabella Moffat	`imof0001`	FYP Student
Jack Bastasin	`jbas0020`	FYP Student
Jack Thornborrow	`jtho0062`	FYP Student
Jacinta Yao	`jyao0029`	FYP Student
Jeremia Yovinus	`jyov0001`	FYP Student
Jesse Kipchumba	`jkip0002`	FYP Student
Jordan Rozario	`jrozario`	FYP Student
Lucas Cooper	`lcoo0019`	FYP Student
Maxwell Fergie	`mfergie`	FYP Student
Michael Fitzgerald	`mfit0014`	FYP Student
Omar Salem	`osal0004`	FYP Student
Raphael Schwalb	`rschwalb`	FYP Student
Suhaas Kataria	`skat0015`	FYP Student
Taine Fryer	`tfryer`	FYP Student
Tavishi Saxena	`tsaxena`	FYP Student
Tianyi Xiao	`txiao`	FYP Student
Vedansh Malhan	`vmalhan`	FYP Student
William Ngo	`wngo`	FYP Student

Getting Access¶

Register at my.massive.org.au
Follow the Join Existing Project guide — Project ID: mh42
Notify your supervisor after submitting so they can approve promptly
Approval takes roughly 1 week

Connecting¶

SSH

ssh {your_m3_id}@m3.massive.org.au

Strudel — browser-based remote desktop

For GUI workloads (Isaac Sim, Jupyter): strudel.hpc.monash.edu

Interactive GPU session

sinteractive --partition=gpu --gres=gpu:A40:1 --time=02:00:00 --mem=32G

Warning

Isaac Sim requires an A40 GPU specifically (RT Cores required). Other GPU types on M3 (L4, T4, A100, H100) are not compatible.

More options: M3 connecting docs

Storage¶

Location	Quota	Persistent?	Use for
`/home/{your_m3_id}/`	20 GB	Yes	Config files, small scripts only
`/home/{your_m3_id}/mh42/`	500 GB shared	Yes	All large project files
`/home/{your_m3_id}/mh42_scratch`	6 TB shared	No — purge schedule undocumented; treat as volatile	Temporary large files
`/home/{your_m3_id}/mh42_scratch2`	6 TB shared	No — purge schedule undocumented; treat as volatile	Same filesystem as mh42_scratch

Danger

Your home directory has only 20 GB. Never store datasets, containers, or environments there — use the project directory.

Folder Structure¶

All files live under a two-level structure: project folder → stack folder.

/home/{your_m3_id}/mh42/
├── FYP2026S1_3487/          ← FYP project folder (created by supervisor)
│   ├── STATUS               ← project status file
│   ├── stack_alice/         ← Alice's personal stack
│   └── stack_bob/           ← Bob's personal stack
├── PhD_2023_dave/
│   ├── STATUS
│   └── stack_dave/
└── Postdoc_2026_somayeh/
    └── stack_somayeh/

Project folder naming:

Role	Format	Example
FYP student	`FYP{YYYY}S{1\\|2}_{ProjectID}`	`FYP2026S1_3487`
PhD student	`PhD_{YYYY}_{firstname}`	`PhD_2023_dave`
Postdoc	`Postdoc_{YYYY}_{firstname}`	`Postdoc_2026_somayeh`

Setting Up Your Stack Folder¶

Note

The project folder (FYP2026S1_3487/, PhD_2023_dave/, etc.) is created by your supervisor — you do not create it yourself. Your first action is to create your stack_{name}/ folder inside it.

Supervisors: see the M3 Admin Guide for the full setup procedure including default ACLs.

Create your stack folder inside your project folder:

mkdir -p /home/${USER}/mh42/{your_project_folder}/stack_{your_name}

Example for FYP student Alice in project 3487:

mkdir -p /home/${USER}/mh42/FYP2026S1_3487/stack_alice

Store all your environments, datasets, and checkpoints inside your stack folder.

Project Status Tracking¶

Each project folder has a STATUS file. Keep it up to date:

# When you start
echo "ACTIVE" > /home/${USER}/mh42/{your_project_folder}/STATUS

# When your project ends
echo "DONE-2026-06" > /home/${USER}/mh42/{your_project_folder}/STATUS

# When data is no longer needed
echo "OK-TO-DELETE" > /home/${USER}/mh42/{your_project_folder}/STATUS

Status	Meaning
`ACTIVE`	Project ongoing
`DONE-{YYYY-MM}`	Finished; data may still be needed short-term
`ARCHIVE`	Ended; kept for reference
`OK-TO-DELETE`	Safe to remove

Check all statuses at a glance:

for d in /home/${USER}/mh42/*/; do echo "$d: $(cat ${d}STATUS 2>/dev/null || echo 'NO STATUS FILE')"; done

Scratch (Temporary Storage)¶

mh42_scratch and mh42_scratch2 are both symlinks to the same filesystem (/fs04/scratch2/mh42) — use either interchangeably. The original /scratch filesystem was retired in April 2025 and migrated to /scratch2, which is why both names exist.

Follow the same folder structure as mh42/:

mkdir -p /home/${USER}/mh42_scratch/{your_project_folder}/stack_{your_name}

Quota: Default scratch2 quota is 6 TB (increased from 3 TB in May 2026). Unlike the project quota, this can be increased — submit a request at the M3 quota request form (allow up to 2 weeks; 10+ TB requests are reviewed and granted for 30 days only). The project quota (500 GB) cannot be increased.

Warning

Monash does not publish a specific purge schedule for scratch. Treat it as volatile — files could be deleted at any time without notice. Archive anything important to S Drive or move to mh42/ before it disappears.

Tip

When creating a project folder in scratch, set group write so your supervisor and teammates can create their own stacks inside it:

chmod g+w /home/${USER}/mh42_scratch2/{your_project_folder}/

Conda Environments¶

Use Miniforge3 (provided by M3) — do not install Anaconda or Miniconda yourself:

module load miniforge3

Tip

Save conda environments to your stack folder, not your home directory. See M3 Conda docs for setup.

By default your files are private. To share with all mh42 members:

chmod -R g+rwX /home/${USER}/mh42/{your_project_folder}/stack_{your_name}/

Note

There is no per-person access control — granting group access means all mh42 members can read and write those files. Do not modify other people's data even if you technically have access.

To share with a specific person only, use ACL (confirmed working on M3's Lustre filesystem):

# Grant access to existing files
setfacl -R -m u:{their_m3_id}:rwx /home/${USER}/mh42/{your_project_folder}/stack_{your_name}/

# Also inherit for new files created inside
setfacl -d -m u:{their_m3_id}:rwx /home/${USER}/mh42/{your_project_folder}/stack_{your_name}/

# Verify
getfacl /home/${USER}/mh42/{your_project_folder}/stack_{your_name}/

Archiving to S Drive¶

When M3 space is tight, archive completed runs to the lab S Drive:

# Mount S Drive (login node only — not inside a job)
gio mount 'smb://MONASH;{your_monash_id}@ad.monash.edu/Shared'
ln -s /run/user/$(id -u)/gvfs/smb-share:domain=MONASH,server=ad.monash.edu,share=shared,user={your_monash_id}/ENG-ECSE/Robotics-M3 ~/s_drive

# Archive a run, then delete from M3
rsync -av --progress /fs04/mh42/${PROJECT}/stack_${MY_NAME}/logs/<run>/ \
    ~/s_drive/${PROJECT}/stack_${MY_NAME}/logs/<run>/
rm -rf /fs04/mh42/${PROJECT}/stack_${MY_NAME}/logs/<run>

See the S Drive guide for full instructions.

Quick Reference¶

# SSH into M3
ssh {your_m3_id}@m3.massive.org.au

# Navigate to your stack
cd /home/${USER}/mh42/{your_project_folder}/stack_{your_name}

# Check quota
user_info

# Check all project statuses
for d in /home/${USER}/mh42/*/; do echo "$d: $(cat ${d}STATUS 2>/dev/null || echo 'NO STATUS FILE')"; done

# Submit a batch job
sbatch job.sh

# Check job queue
squeue -u ${USER}

# Cancel a job
scancel {job_id}