Grunt watch error - Waiting...Fatal error: watch ENOSPC

After doing some research found the solution. Run the below command.

echo fs.inotify.max_user_watches=524288 | sudo tee -a /etc/sysctl.conf && sudo sysctl -p

For Arch Linux add this line to /etc/sysctl.d/99-sysctl.conf:

fs.inotify.max_user_watches=524288

Any time you need to run sudo something ... to fix something, you should be pausing to think about what's going on. While the accepted answer here is perfectly valid, it's treating the symptom rather than the problem. Sorta the equivalent of buying bigger saddlebags to solve the problem of: error, cannot load more garbage onto pony. Pony has so much garbage already loaded, that pony is fainting with exhaustion.

An alternative (perhaps comparable to taking excess garbage off of pony and placing in the dump), is to run:

npm dedupe

Then go congratulate yourself for making pony happy.

To find out who's making inotify instances, try this command (source):

for foo in /proc/*/fd/*; do readlink -f $foo; done | grep inotify | sort | uniq -c | sort -nr

Mine looked like this:

 25 /proc/2857/fd/anon_inode:inotify
  9 /proc/2880/fd/anon_inode:inotify
  4 /proc/1375/fd/anon_inode:inotify
  3 /proc/1851/fd/anon_inode:inotify
  2 /proc/2611/fd/anon_inode:inotify
  2 /proc/2414/fd/anon_inode:inotify
  1 /proc/2992/fd/anon_inode:inotify

Using ps -p 2857, I was able to identify process 2857 as sublime_text. Only after closing all sublime windows was I able to run my node script.

After trying grenade's answer you may use a temporary fix:

sudo bash -c 'echo 524288 > /proc/sys/fs/inotify/max_user_watches'

This does the same thing as kds's answer, but without persisting the changes. This is useful if the error just occurs after some uptime of your system.