'omarchy update' refuses to start: you need at least 10 GiB free (btrfs snapshots eating the root subvolume)
Re-audited on omarchy 4.0.2-1, kernel 7.1.9. The first-pass note is still true line for line: I diffed /usr/share/omarchy/bin/omarchy-update-requires-free-space, bin/omarchy-update-pkg-prune and default/snapper/root on this machine against the quattro tree and all three are byte identical, so df --output=avail --block-size=1 /, the (( available_bytes < 10 * 1024 * 1024 * 1024 )) test, the exact message, the OMARCHY_UPDATE_FORCE=1 early exit, paccache -rk2, and NUMBER_LIMIT=5 / NUMBER_LIMIT_IMPORTANT=5 / TIMELINE_CREATE=no all hold. bin/omarchy-update still calls omarchy-update-pkg-prune before omarchy-snapshot create. paccache -rk1 and paccache -ruk0 are on the Arch Wiki Pacman page verbatim. The rollback claim in danger checks out: limine-snapper-sync 1.31.0-1 is installed from the omarchy repo and omarchy-snapshot restore calls limine-snapper-restore. What the first pass did not cover is the filesystem the check does not measure. The script looks at / only, and on this install /boot is a separate vfat ESP (findmnt reports dmask=0077, 2.0 GiB, 428 MiB used), so a full ESP sails past the 10 GiB check and kills the update later with the same No space left on device the symptom attributes to a full root. That case is Omarchy specific rather than generic: /etc/limine-snapper-sync.conf as shipped sets LIMIT_USAGE_PERCENT=85 with MAX_SNAPSHOT_ENTRIES left at auto, and the upstream comments say new snapshot entries stop being added at that limit, so the same snapshots the record is about also consume ESP space. /usr/lib/snapper/plugins/10-limine-snapper-sync acts on the delete operation and re-runs limine-snapper-sync, which is why deleting snapshots is the safe way to reclaim the ESP and hand deletion of files there is not. I rewrote symptom, fix, danger and verify for that branch, added the warning that paccache -rk1 and yay -Sc spend the cache that omarchy-update-pkg-prune's own comment calls the only offline downgrade path, corrected the fix's claim that yay -Sc only clears build trees (it also clears pacman's cache of packages no longer installed), and reworded "df lies on btrfs" into what actually happens, which is that df reports against allocated chunks. The cause is correct as written and I left it alone, so cause_reconciled should stay unset. Two things I did not exercise: I ran no update and deleted no snapshot, cache or journal file, so the remedies are verified by reading sources and not by execution, and I could not list /boot/EFI/Linux because the ESP is dmask=0077 and I am not permitted to use sudo here. One thing I cannot fix from a verdict: applies_to lists arch, cachyos and endeavouros, but the symptom, the threshold and the bypass variable are all omarchy update and exist nowhere else, so only the btrfs and snapper halves carry over to those systems.
snapper delete is permanent, and those snapshots are the rollback targets you would boot to from Limine if the update breaks the desktop, so never delete the newest one and never delete the snapshot you are currently booted into. Never delete files from /boot by hand to make room: the ESP holds the kernel or UKI you are running and limine.conf, and removing the wrong file leaves the machine unbootable with no recovery entry to fall back to. Reclaim ESP space by deleting snapshots and letting limine-snapper-sync remove their copies. paccache -rk1 and yay -Sc shrink the pacman cache, which omarchy-update-pkg-prune calls the only offline downgrade path, so after them you have nothing to downgrade to if the update itself is what breaks the machine. OMARCHY_UPDATE_FORCE=1 on a genuinely full root is how you get a truncated kernel or initramfs written mid-transaction and an unbootable machine, so only use it when you have verified the free space yourself. btrfs balance is I/O heavy and must not be interrupted by a power loss, so run it on AC.
Symptom
omarchy update exits immediately, before it even asks for confirmation:
You need at least 10 GiB free to safely update Omarchy.
On a btrfs root the confusing part is that df -h / can still show a couple of gigabytes free while sudo btrfs filesystem usage / shows the device is effectively full, and an update that is forced through then dies later with No space left on device out of mkinitcpio or pacman.
A second case produces the same No space left on device and never produces this message. omarchy-update-requires-free-space measures / and nothing else, so when the vfat ESP at /boot is the filesystem that is full the check passes, the update starts, and mkinitcpio or the Limine hook dies while df -h / still shows hundreds of gigabytes free. Look at both filesystems before believing either.
Cause
omarchy-update-requires-free-space runs df --output=avail --block-size=1 / and aborts below 10 GiB (10737418240 bytes). On Omarchy's default btrfs layout that space is usually not "used by files" at all: install/config/snapper.sh installs a snapper root config with NUMBER_LIMIT=5 / NUMBER_LIMIT_IMPORTANT=5, and omarchy update creates a pre-update snapshot on every run. Those five snapshots pin every block that any deleted file used to occupy, including /var/cache/pacman/pkg, which lives on the same snapshotted subvolume. That is exactly why omarchy-update-pkg-prune runs paccache -rk2 *before* the snapshot rather than after. Deleting files inside the live root therefore frees nothing until the snapshots holding them age out.
Fix
# 0. Which filesystem is actually full? The 10 GiB check only ever looks at /.
df -h / /boot
If / is the tight one, work through steps 1 to 5. If /boot is the tight one, skip to the boot partition section at the end.
# 1. Get the truth on root. On btrfs, df reports free space from the chunks
# already allocated, so it can read low or fail with ENOSPC while raw device
# space is unused. Use btrfs's own accounting.
sudo btrfs filesystem usage /
df -h /
sudo du -xhd1 /var | sort -h | tail
# 2. Pacman and AUR caches. omarchy-update-pkg-prune runs `paccache -rk2` and
# keeps two versions on purpose, because the cache is the only offline
# downgrade path, so dropping to one costs you the spare.
sudo paccache -rk1
sudo paccache -ruk0 # drop every cached version of uninstalled packages
yay -Sc --noconfirm # yay's build trees under ~/.cache/yay, and also
# pacman's cache of packages no longer installed
# 3. Journal
journalctl --disk-usage
sudo journalctl --vacuum-size=200M
# 4. The usual real culprit: old snapshots. Keep at least the newest one.
sudo snapper -c root list
sudo snapper -c root delete --sync 12 13 14 # --sync releases the space now
sudo snapper -c root delete --sync 20-24 # a range works too
sudo snapper -c root cleanup number # apply NUMBER_LIMIT=5 retention
# 5. Re-check and update
sudo btrfs filesystem usage /
omarchy update
If btrfs still reports the device full after deleting snapshots (allocated-but-unused chunks):
sudo btrfs balance start -dusage=20 -musage=20 /
sudo btrfs filesystem usage /
**When /boot is the full one (Omarchy 4).** Omarchy installs limine-snapper-sync, which copies each snapshot's kernel and initramfs onto the ESP so that snapshot can be selected from Limine's Snapshots menu. /etc/limine-snapper-sync.conf ships LIMIT_USAGE_PERCENT=85 and leaves MAX_SNAPSHOT_ENTRIES at auto, so it drops the oldest snapshot boot entries once the ESP passes 85 percent used and warns only when there are none left to drop. The ESP is mounted dmask=0077, so listing it needs root.
df -h /boot
sudo ls -la /boot/EFI /boot/EFI/Linux
Free it by deleting snapshots, not by deleting files on the ESP. The snapper plugin at /usr/lib/snapper/plugins/10-limine-snapper-sync runs on delete and re-syncs the ESP, so the copies go with the snapshots:
sudo snapper -c root list
sudo snapper -c root delete --sync 12 13 14
df -h /boot
On plain Arch with no limine-snapper-sync, the ESP holds only the current kernels and there is nothing snapshot-related to reclaim. Check for images belonging to kernels that pacman -Q no longer lists before touching anything.
Only when you are certain the 10 GiB figure is wrong for your layout (for example /var is a separate filesystem with plenty of room), bypass the check:
OMARCHY_UPDATE_FORCE=1 omarchy update
Verify
df -h / /boot shows more than 10 GiB available on / and /boot below 85 percent used, sudo btrfs filesystem usage / shows free (estimated) well above that, and omarchy update reaches its confirmation prompt.
Sources
- https://raw.githubusercontent.com/basecamp/omarchy/quattro/bin/omarchy-update-requires-free-space
- https://raw.githubusercontent.com/basecamp/omarchy/quattro/bin/omarchy-update
- https://raw.githubusercontent.com/basecamp/omarchy/quattro/bin/omarchy-update-pkg-prune
- https://raw.githubusercontent.com/basecamp/omarchy/quattro/default/snapper/root
- https://raw.githubusercontent.com/basecamp/omarchy/quattro/install/config/snapper.sh
- https://raw.githubusercontent.com/basecamp/omarchy/quattro/docs/update-process.md
- https://wiki.archlinux.org/title/Snapper
- https://gitlab.com/Zesko/limine-snapper-sync/-/raw/master/install/arch-linux/etc/limine-snapper-sync.conf
- https://wiki.archlinux.org/title/Pacman