Error: ENOSPC: System limit for number of file watchers reached — how to fix

Error: ENOSPC: System limit for number of file watchers reached, watch '/home/me/app/src'
    at FSWatcher.<computed> (node:internal/fs/watchers:247:19)
    at Object.watch (node:fs:2343:34)
  errno: -28,
  syscall: 'watch',
  code: 'ENOSPC',

Quick fix: Raise the inotify limit permanently: echo fs.inotify.max_user_watches=524288 | sudo tee /etc/sysctl.d/60-inotify.conf && sudo sysctl --system. Also exclude node_modules from watchers, and close other watchers (extra IDE windows, old dev servers).

What it means

Linux tools watch files for changes using inotify. Each watched file or directory costs one watch, and there's a per-user limit, fs.inotify.max_user_watches. Older kernels default to 8192; newer ones scale it with RAM, but big JavaScript projects (node_modules can be 100,000+ files) still blow past it. When the limit is hit, the inotify_add_watch syscall returns ENOSPC — the "no space" refers to the watch table, not your disk.

Check the current values:

cat /proc/sys/fs/inotify/max_user_watches
cat /proc/sys/fs/inotify/max_user_instances

Fix 1: raise the limit (permanent)

echo fs.inotify.max_user_watches=524288 | sudo tee /etc/sysctl.d/60-inotify.conf
sudo sysctl --system

Each watch uses roughly 1 KB of kernel memory on 64-bit systems, so 524288 is about 0.5 GB worst case — only used if actually allocated. Restart your dev server afterwards.

Temporary (until reboot):

sudo sysctl fs.inotify.max_user_watches=524288

If the error mentions inotify_init or "too many instances", raise fs.inotify.max_user_instances the same way (e.g. 512).

Fix 2: stop watching node_modules

Most tools ignore it by default, but custom configs sometimes don't.

webpack:

module.exports = {
  watchOptions: { ignored: /node_modules/ },
};

Vite (vite.config.js):

export default {
  server: { watch: { ignored: ['**/node_modules/**', '**/dist/**'] } },
};

nodemon (nodemon.json):

{ "watch": ["src"], "ignore": ["node_modules", "dist"] }

VS Code (settings.json):

"files.watcherExclude": {
  "**/node_modules/**": true,
  "**/.git/objects/**": true,
  "**/dist/**": true
}

Fix 3: find who is using the watches

The limit is per user, so other programs count too: extra VS Code windows, JetBrains IDEs, Dropbox, old npm start processes. List inotify users:

for p in /proc/[0-9]*; do
  n=$(find "$p/fd" -lname 'anon_inode:inotify' 2>/dev/null | wc -l)
  [ "$n" -gt 0 ] && echo "$n $(cat $p/comm 2>/dev/null) ${p#/proc/}"
done | sort -rn | head

(That counts inotify instances per process; closing the top offenders frees their watches.)

Fix 4: use polling (last resort)

In containers or on network filesystems where you can't change sysctl, switch to polling. It uses more CPU but no inotify watches:

CHOKIDAR_USEPOLLING=true npm start      # create-react-app, chokidar-based tools
WATCHPACK_POLLING=true npm start        # webpack 5 / Next.js

Vite: server: { watch: { usePolling: true } }.

Docker and WSL

In Docker, max_user_watches is a host kernel setting — set it on the host (or Docker Desktop VM), not in the container. In WSL 2, run the sysctl commands inside the distro; to make it stick, add the line to /etc/sysctl.conf there.

Last updated 2026-11-03.