Features

RAM Overlay for System Directories

• Critical system directories are overlaid in RAM using OverlayFS for maximum performance:
  • /etc - System configuration
  • /var/log - System logs
• Performance: 9.3x faster IOPS (877K vs 94.6K), 7.2x lower latency (0.52ms vs 3.74ms)
• Efficiency: Uses only ~200MB RAM for typical workloads (~1-2% of 16GB)
• Automatic: Activates on login, syncs periodically during the session (every 5 min by default) and on logout
• Easily extensible via the OVERLAY_DIRS array

Tmpfs Mounts for Temporary Data

High-speed RAM storage for temporary files (automatically cleaned on logout):

• /tmp (5 GB) and /var/tmp (1 GB)
• /var/cache (2 GB) and /home/$USER/.cache (2 GB per user)

Persistent Caches (Always on Disk)

Essential caches bind-mounted to /persist to survive reboots:

• System: /var/cache/pacman
• User: nvidia, mesa_shader_cache, mesa_shader_cache_db
• Automatically migrated on first run

Excluded from Overlay (Remain on Disk)

• /home - User data (too large/important)
• /var/lib - Databases, flatpak packages (too large)
• /opt, /usr/local - Large applications
• /proc, /sys, /dev, /run - Virtual filesystems
• /mnt, /media, /boot - Mount points

Separately, within an overlaid directory, /var/log/journal is excluded from both sync and the RAM pre-population step (EXCLUDED_FROM_SYNC) — it's high-churn, journald manages its own persistence, and copying it into RAM up front would burn time and memory for data that's never written back to disk anyway.

Intelligent Cleanup & Safety

• Periodic cleanup: Removes stale files >5 minutes old every 30 seconds
• Periodic sync: Incrementally syncs RAM changes back to disk every 5 minutes while a session is active (SYNC_INTERVAL, disable with 0), in addition to the sync at logout — bounds how much a crash or power loss mid-session could lose
• File protection: Skips in-use files (via lsof, fuser, or /proc)
• Graceful shutdown: Catches SIGTERM/SIGINT/SIGHUP and unexpected exits, syncs before exit
• Comprehensive logging: All operations tracked in /var/log/ramoverlay.log

Optional Performance Boost

Preloads mimalloc for rsync, find, and inotifywait if available, reducing memory fragmentation during sync operations and long-lived event watching.

---

Requirements

• Commands: rsync (falls back to mv), and lsof or fuser (falls back to /proc scanning) - none are hard requirements but rsync is strongly recommended for safe syncing
• RAM: 16GB+ recommended (typical usage: 200MB-2GB)
• Filesystem: Supports OverlayFS (ext4, btrfs, xfs, f2fs)
• Optional: libmimalloc.so for faster syncs
• Optional: Kernel 6.3+ for the tmpfs noswap mount option — tried first and falls back cleanly on older kernels, so it's not a hard requirement

---

Installation

sudo cp ephemeral-overlay /bin/
sudo chmod 755 /bin/ephemeral-overlay

On rc.local systems, add to /etc/rc.local:

# Start ephemeral overlay daemon
if [ -x /bin/ephemeral-overlay ]; then
    ( sleep 20; setsid nohup /bin/ephemeral-overlay >> /var/log/ramoverlay.log 2>&1 ) &
fi

Registering and enabling it (compiling the service database, adding it to whatever bundle your other longruns are in) is the part that varies by Artix's s6-rc frontend version — use the same s6 set / s6 live install workflow you've already got working for other services rather than any specific commands here, and confirm with s6-rc-db list (or your usual check) that it's actually running before trusting it.

Note: Runs as a daemon, automatically managing overlay lifecycle based on user sessions.

---

Configuration

Edit /bin/ephemeral-overlay:

Directories to overlay:

OVERLAY_DIRS=(
    "/etc"
    "/var/log"
    # "/opt"          # Add more as needed
)

Persistent system caches:

BIND_MOUNTED_VAR_CACHE=(pacman)

Persistent user caches:

BIND_MOUNTED_USER_CACHE=(nvidia mesa_shader_cache mesa_shader_cache_db)

Tmpfs sizes:

TMP_SIZE="5G"
VAR_TMP_SIZE="1G"
VAR_CACHE_SIZE="2G"
USER_CACHE_SIZE="2G"

Cleanup and sync settings:

CLEAN_INTERVAL=30        # Check every 30 seconds
STALE_MINUTES=5          # Delete files older than 5 minutes
SYNC_INTERVAL=300        # Sync RAM -> disk every 5 min while active; 0 to disable

---

Monitoring

# View active overlays
findmnt -t overlay

# Check RAM usage
df -h /ram_overlay

# See what's in RAM
sudo du -sh /ram_overlay/upper/*

# Monitor activity
tail -f /var/log/ramoverlay.log

Typical RAM usage:

• Light: 200-500 MB (~1-2% of 16GB)
• Heavy: 1-2 GB (~5-10% of 16GB)
• Allocated: 90% of total RAM (only uses what's needed)

---

How It Works

1. Wait: Daemon waits for user login
2. Activate: On login, creates RAM overlay and tmpfs mounts
3. Operate: All writes to overlaid directories go to RAM (9.3x faster)
4. Cleanup: Daemon removes stale temp files every 30 seconds
5. Sync: Incrementally syncs changed files back to disk every 5 minutes while the session is active, then fully on logout
6. Sleep: Unmounts everything, frees RAM, waits for next login

---

Performance (vs SATA SSD)

Metric	SATA Disk	RAM Overlay	Improvement
Random Write IOPS	94,600	877,000	9.3x faster
Write Latency	3.74ms	0.52ms	7.2x lower
Sequential Write	530 MB/s	2,128 MB/s	4x faster
RAM Overhead	0	~200MB	Minimal

Benchmarked on Samsung 870 EVO SATA SSD. The RAM-vs-SSD comparison itself is a hardware characteristic, unaffected by anything above.

---

Safety Features

• Best-effort persistence: Changes sync to disk periodically during the session, and on logout/shutdown; sync errors are detected and logged, but there is no retry logic or post-sync verification pass
• Signal handling: Gracefully handles SIGTERM, SIGINT, and SIGHUP, and also traps EXIT directly so an unexpected termination (not just a clean signal) still triggers a sync attempt before the process actually dies. Sleeps in the main loop are broken into 1-second increments rather than one long sleep N — bash only checks for a pending trap between commands, so a signal arriving mid-sleep would otherwise sit queued for up to CLEAN_INTERVAL (30s by default) before shutdown actually started, which is long enough that a process supervisor with a shorter kill-escalation timeout could SIGKILL it first and skip the sync entirely
• File-in-use detection: Never deletes files that are currently open
• Rsync verification: Checks rsync exit codes (captured independently of tee) and logs sync failures
• Protected directories: User data in /home is never overlaid
• OverlayFS semantics: Per-file writes are atomic via OverlayFS; however, the sync process (unmount → rsync → remount) is not crash-safe - a power loss mid-sync may leave the filesystem partially written. The overlay mount deliberately does not use the volatile option — it looked like a good fit (upper is tmpfs, never durable across a reboot anyway, so overlayfs's own sync/fsync bookkeeping on it seemed redundant) but testing showed it writes a marker into workdir that makes the kernel refuse every subsequent mount using that workdir, with or without volatile, until it's wiped — which breaks the remount this daemon does on every single sync
• Fallback mechanisms: Multiple methods for file-in-use detection