Before jumping to scripts/udev there are few things I’m taking “as is” from my vagrant setup and this Vagrantfile example will help you to understand/recreate what I’m doing.

Vagrant.configure('2') do |config|
  config.vm.hostname = '' 'private_network', ip: '' # and the IP of my local computer is
  config.vm.synced_folder '.', '/project', type: 'nfs'

  config.ssh.forward_agent = true

When OS is booting up all of the services are starting. I’m using php5-fpm and nginx and both of those services require paths to exist for proper running. NFS shares are mounted in vagrant guest os after starting the services.

I’m using very simple script to restart services:

sleep 5 # wait for a bit for NFS to make sure resources are available
systemctl restart php5-fpm > /dev/null 2>&1
systemctl restart nginx > /dev/null 2>&1

I’m saving this to /root/ file with ansible provisioning.

But to make it work we have to run this script after our directory is mounted over NFS. This is where I used udev.

First of all I wanted to know what event and subsystem is triggered when I mount directory. To get this information I opened 2 terminals and used vagrant ssh to get to guest os.

In one terminal I started udevadm monitor to get informations about events that were triggered by mount/umount of /project directory.

In second terminal I got root (with sudo su in my case), checked what is mounted with df -h (cause I’m too lazy to type this): 233G  186G   47G  80% /project

So now what I did and what was the effect:

# Terminal 1:
→ udevadm monitor
monitor will print the received events for:
UDEV - the event which udev sends out after rule processing
KERNEL - the kernel uevent

KERNEL[336.177605] remove   /devices/virtual/bdi/0:37 (bdi)
UDEV  [336.177844] remove   /devices/virtual/bdi/0:37 (bdi)

# Terminal 2:
umount /project
# Terminal 1:
mount --source --target /project/

# Terminal 2 with still running udevadm:
KERNEL[432.289098] add      /devices/virtual/bdi/0:37 (bdi)
UDEV  [438.240571] add      /devices/virtual/bdi/0:37 (bdi)

With this informations (and a wiki how to create udev rules) I could finally create rule and make it run the script to restart nginx and php5-fpm:

SUBSYSTEM=="bdi",ACTION=="add",RUN+="/bin/bash /root/"

With provisioning I’m putting this script in /root/ and udev rule in /etc/udev/rules.d/50-vagrant-mount.rules.

After vagrant up the default vagrant user has exactly the same uid and gid as my local user. No more problems with reading/writing in /project.

There are also prerequisits for this method:

  1. udev event is not started after getting machine up after vagrant suspend
  2. if you’re provisioning this within Vagrantfile you might need to do additional vagrant reload to make it work after provisioning (not sure because I’m using packer to build our own boxes with preinstalled scripts like this)