Storage

This page describes the physical disks (hard drives) inside nfc3008, the computer this lab uses to run OptoFly experiments — which cameras connect to it, which software runs on it, and where all of the data it produces lives. If you're reading this on a different computer, none of this applies; ask a labmate which machine is nfc3008 if you're not sure.

The exact amount of free space on each disk changes constantly as experiments run — don't worry if the numbers below don't match what you see when you check yourself. What doesn't change is the layout: which disk holds what kind of data, and why.

The disks, in plain terms

A computer's storage is split across separate physical disks, and each disk is attached to the file system at a mount point — the folder name Linux uses as the "front door" into that disk. When you look inside a mount point's folder, you're looking at files on that specific physical disk, not the computer's main disk. This machine has four storage areas, described below in the order you're likely to care about them.

The main drive (/)

Mount point: / (a single forward slash — this is the "root" of the whole file system, not a specific folder name).

This drive holds the operating system (Ubuntu Linux), every program installed on the machine, and the nfc user's home folder (/home/nfc — code checkouts, this wiki, editor settings, and so on). It's a 1TB NVMe SSD (a fast type of solid-state drive that plugs directly into the motherboard rather than connecting with a cable) — a Samsung 970 EVO Plus.

The working data disk (/mnt/data)

Mount point: /mnt/data.

This is where optofly writes recordings — video files and .braidz tracking files — while an experiment is running, and where they stay immediately afterward. It's a separate 2TB NVMe SSD (a Samsung 990 PRO), chosen specifically because it's fast enough to keep up with several cameras recording video at once, which the main drive doesn't need to be optimized for.

The long-term storage disk (/mnt/storage)

Mount point: /mnt/storage.

Once a recording is finished, a copy of it ends up here permanently (see Backups for exactly how and when). Unlike the other disks, this one isn't a single physical drive — it's two matched 8TB hard disk drives (traditional spinning-platter drives, not solid-state) set up as what's called a RAID1 mirror: every file written here is automatically duplicated onto both drives at once, so if one drive physically breaks, nothing is lost and the computer keeps working normally off the other drive. Linux presents both drives to you as a single folder, /mnt/storage — you never interact with the two drives separately.

The backup disk (/mnt/system_backups)

Mount point: /mnt/system_backups.

A 1TB SATA SSD (a solid-state drive that connects via a cable, rather than plugging directly into the motherboard like the main and working drives do) dedicated entirely to holding a backup of the nfc user's home folder. See Backups for what that means and how it works — this page is only about the disk itself.

At a glance

Disk	Size	Type	Mount point	What it's for
Samsung 970 EVO Plus	1TB	NVMe SSD	/	Operating system, programs, home folder
Samsung 990 PRO	2TB	NVMe SSD	/mnt/data	Active/recent experiment recordings
2× 8TB HDD (RAID1 mirror)	7.3TB usable	HDD	/mnt/storage	Permanent storage of finished recordings
Samsung 860 EVO	1TB	SATA SSD	/mnt/system_backups	Backup of the home folder

Checking how full a disk is

Open a terminal and run:

df -h

df stands for "disk free" — it lists every disk and how much space is used. The -h flag means "human-readable," so it shows sizes like 1.8T instead of a huge number of bytes. You'll see a line for each disk in the table above, plus some you can ignore (they're internal to the system). As of 2026-08-13, the ones that matter looked like this:

Filesystem      Size  Used Avail Use% Mounted on
/dev/nvme1n1p2  915G  203G  666G  24% /
/dev/nvme0n1    1.8T  1.3T  500G  72% /mnt/data
/dev/md0        7.3T  6.1T  803G  89% /mnt/storage
/dev/sdc1       916G   46G  824G   6% /mnt/system_backups

The Use% column is the one to watch — it's how full that disk currently is.

Common failures

⚠️ Common failure: /mnt/storage (the long-term storage disk) is already at 89% full as of 2026-08-13, and it only ever grows — the daily copy job that fills it (see Backups) never deletes anything, even if the original file is deleted from /mnt/data. Nothing on this machine automatically frees up space. If a backup or copy job starts failing with a message like "No space left on device" in its log file, this disk is likely full — someone needs to manually go through old recordings and delete or move ones that are no longer needed.

⚠️ Common failure: one of the two drives in the /mnt/storage mirror fails. Check its health with:

cat /proc/mdstat

A healthy mirror shows [UU] (both drives Up). A failed drive shows something like [U_] (one drive missing) instead. This machine does try to alert about it automatically — mdadm (the tool that runs the RAID1 mirror) emails the machine's root account when a drive fails — but in practice nobody reads that mailbox, so treat this as something a person has to actively check rather than wait to be told about. If you see anything other than [UU], treat it as urgent: with one drive already down, a second failure would lose everything on /mnt/storage.

⚠️ Common failure: after a reboot, one of the /mnt/* disks seems to be "missing" — but the machine boots up normally with no error message at all. This machine's disk configuration (/etc/fstab) is deliberately set to boot normally even if a disk fails to attach, rather than getting stuck on an error screen. That means a missing disk fails silently: a folder like /mnt/data will still exist and look empty, instead of showing an error. If a mount point looks unexpectedly empty, that's the first thing to suspect — compare against the "at a glance" table above, and ask a labmate for help checking the physical drive connection.