A Vagrant plugin to automate bindfs mounts in the VM. This allow you to change owner, group and permissions on files and, for example, work around NFS share permissions issues.
The default vagrant provider is virtualbox. It's free and works well but has some performance problems.
People often recommend switching to vagrant's VMWare-fusion provider. This reportedly has better performance, but shares with symlinks won't work. You also have to buy both the plugin and VMware Fusion.
The final recommendation, at least on Linux/OSX hosts, is to use NFS.
However, an NFS mount has the same numeric permissions in the guest as in the host.
If you're on OSX, for instance, a folder owned by the default OSX user will appear to be owned by 501:20
in the guest.
This is where vagrant-bindfs
comes in!
Simply:
- mount your share over NFS into a temporary location in the guest
- re-mount the share to the actual destination with
vagrant-bindfs
, setting the correct permissions
Note that map_uid
and map_gid
NFS options can be used to set the identity used to read/write files on the host side.
vagrant-bindfs is distributed as a Ruby gem. You can install it as any other Vagrant plugin with:
vagrant plugin install vagrant-bindfs
In your Vagrantfile
, use config.bindfs.bind_folder
to configure folders that will be bound on VM startup.
The format is:
config.bindfs.bind_folder "/path/to/source", "/path/to/destination", options
By default, all folders are bound after folders syncing between host and guest machines.
You can pass a special :after
option to the bind_folder method to choose when a folder should be bound.
Supported values are :
:synced_folders
(default):provision
, to bind a folder after provisioning occured.
Vagrant.configure("2") do |config|
[...] # Your VM configuration
## Basic usage
config.bindfs.bind_folder "source/dir", "mount/point"
## Advanced options
config.bindfs.bind_folder "source/dir", "mount/point",
perms: "u=rw:g=r:o=r", # http://bindfs.org/docs/bindfs.1.html#sect12
create_as_user: true
## Complete example for a NFS shared folder
# Static IP is required to use NFS shared folder,
# this is only required for Virtualbox provider
config.vm.network "private_network", ip: "192.168.50.4"
# Declare shared folder with Vagrant syntax
config.vm.synced_folder "host/source/dir", "/vagrant-nfs", type: :nfs
# Use vagrant-bindfs to re-mount folder
config.bindfs.bind_folder "/vagrant-nfs", "guest/mount/point"
# Bind a folder after provisioning
config.bindfs.bind_folder "/vagrant-after-provision", "another/guest/mount/point", after: :provision
end
Remember that Vagrant use /vagrant
on guest side to automatically share your project directory (the one with the Vagrantfile), in particular for provisioning and configuration purposes. Binding a folder to /vagrant
or one of its subfolders may expose unexpected behaviors.
The bind_folder
config accept any option you can pass to bindfs.
vagrant-bindfs is compatible with bindfs from version 1.9 to 1.14.7.
Check lib/vagrant-bindfs/bindfs/option_definitions.json for a complete list of supported options and default values and read the bindfs man page for full documentation.
Both long arguments and shorthand are supported. If you set both, shorthand will prevail. Long arguments can be written indifferently with underscore ('force_user') or dash ('force-user') and as strings ('force-user') or symbols (:force_user or :'force-user').
Option arguments values can be anything that can be casted to a string via to_s
.
The plugin will try to detect flag arguments values as true or false from common notations.
vagrant-bindfs detects installed version of bindfs, translates option names when needed and ignores an option if it is not supported. As we may have missed something, it will warn you when a binding command fail.
On Debian (and Ubuntu), SUSE, Fedora, CentOS, Gentoo and OS X guest systems, vagrant-bindfs will try to install bindfs automatically if it is not installed. On other system, you'll get warned.
OS X guests may need some specific options. See bindfs README for details.
This plugin supports the following configuration options:
Setting config.bindfs.debug
to true will increase the verbosity level of this plugin in Vagrant output.
You can overwrite default bindfs options via config.bindfs.default_options=
.
Vagrant.configure("2") do |config|
# These values are the default options
config.bindfs.default_options = {
force_user: 'vagrant',
force_group: 'vagrant',
perms: 'u=rwX:g=rD:o=rD'
}
end
By default, vagrant-bindfs
will check if the user and the group set for a bound folder exists on the virtual machine.
If either one, the other or both of them are missing, it will warn you and not execute any bindfs command for this folder.
To skip these validations, you can add :user
and/or :group
to the config.bindfs.skip_validations
array.
Vagrant.configure("2") do |config|
config.bindfs.skip_validations << :user
end
By default, if bindfs
needs to be installed on the virtual machine, vagrant-bindfs
will install the most recent release available in repositories (via homebrew
for OS X guests).
If no version of bindfs
is found, it will try to install the most recent supported version of bindfs
from source. (See lib/vagrant-bindfs/bindfs.rb for the exact version number).
You can force the plugin to install the version of bindfs
of your choice with the bindfs_version
configuration option.
It will then look for the specified version in repositories and install it if available. If not, it will install it from sources.
You can also force installation from sources with the install_bindfs_from_source
configuration option.
This will skip any repositories look up.
Vagrant.configure("2") do |config|
# This will force the plugin to install bindfs-1.12.2 from sources
config.bindfs.bindfs_version = "1.12.2"
config.bindfs.install_bindfs_from_source = true
end
This feature only works with exact version match and does not try to resolve dependencies. In particular, Fuse will always be installed from the latest version available in repositories (via Homebrew Cask for OS X guests).
By default, vagrant-bindfs
won't try to empty an existing mount point before mount. Hence, if you try to bind a folder to a non-empty directory without explicitly allowing Fuse to do so (with the nonempty
Fuse option), mount will fail.
You can ask vagrant-bindfs
to make sure the mount points are empty with the config.bindfs.force_empty_mountpoints
option.
Mount poitns will then be destroyed (with rm -rf
) prior to mount, unless you allow Fuse to not mind (with the nonempty
Fuse option).
Vagrant.configure("2") do |config|
config.bindfs.force_empty_mountpoints = true
# This exemple assume two directories exist in your VM:
# - `a/non/empty/mount/point/and/its/content`
# - `another/non/empty/mount/point/and/its/content`
# `a/non/empty/mount/point` will be destroyed before mount and
# all its content will be lost.
config.bindfs.bind_folder "/vagrant", "a/non/empty/mount/point"
# `a/non/empty/mount/point` will not be destroyed before mount
# but its content will not be accessible while in use by bindfs.
config.bindfs.bind_folder "/vagrant", "a/non/empty/mount/point", o: :nonempty
end
If you find this plugin useful, we could use a few enhancements! In particular, capabilities files for installing vagrant-bindfs on other systems would be useful. We could also use some more specs…
If you've made changes to this plugin, you can easily test it locally in Vagrant.
Edit Vagrantfile
and uncomment one or more of the selected test boxes.
Then, from the root of the repo, do:
bundle install
bundle exec vagrant up
This will spin up one or more VM and try to bindfs-mount some shares in it.
Feel free to modify the included Vagrantfile
or test helpers (in test/test_helpers.rb
) to add additional boxes and test cases.
If you add a new test machine, please ensure that it will stay available and regularly updated for future tests.
We recommend to use officialy supported boxes.