Reputation: 35008
My Vagrant box is build from a base linux (scientific linux), during provisioning (using shell scripts), Apache is installed.
I recently changed the Vagrant file (v2) to:
config.vm.synced_folder "public", "/var/www/sites.d/example.com",
:owner => "apache", :group => "apache"
Which works well if the box is already provisioned and just rebooted.
Now, after a vagrant destroy && vagrant up
I get the error:
mount -t vboxsf -o uid=`id -u apache`,gid=`id -g apache`
/var/www/sites.d/example.com /var/www/sites.d/example.com
id: apache: User does not exist
Which is clear - as during the initial run, apache is not yet installed.
An ugly workaround would of course be to do the basic provisioning with that synced_folder
commented out, comment it in and then reboot.
Is there any clean trick to solve that? Especially in a way that vagrant up
always runs without interruptions, even if the box is new.
Upvotes: 41
Views: 9712
Reputation: 25158
In my case i don't need the synced_folder
to be mounted during provision phase. So i disable the synced_folder
if the guest is not provisioned.
Check if it's provisioned in the Vagrantfile is a hack but it works.
And for me it's fair enough doing
vagrant up # To build and provision the first time
vagrant halt # An intermediate step to mount the folder
vagrant up # To have the folder mounted
So my Vagrantfile is something like:
def provisioned?(vm_name='default', provider='virtualbox')
File.exist?(".vagrant/machines/#{vm_name}/#{provider}/action_provision")
end
Vagrant.configure(2) do |config|
[ ... ]
config.vm.synced_folder "../geoserver_data", "/var/lib/geoserver_data",
disabled: !provisioned?,
owner: "tomcat7",
group: "tomcat7"
[ ... ]
Upvotes: 0
Reputation: 4689
Ryan Sechrest has dealt extensively with this problem.
One of the presented solutions is:
Set permissions of directories to 777 and files to 666
config.vm.synced_folder "/Users/ryansechrest/Projects/Sites", "/var/www/domains", mount_options: ["dmode=777", "fmode=666"]
Upvotes: 8
Reputation: 121
This is what I did:
config.vm.synced_folder "./MyApp", "/opt/MyApp", owner: 10002, group: 1007, create: true
config.vm.provision :shell do |shell|
shell.inline = "groupadd -g 1007 myapp;
useradd -c 'MyApp User' -d /opt/MyApp -g myapp -m -u 10002 myapp;"
end
Instead of use the user name and group (as a text) use the uid and gid. Then create the group and user with those ids. This is because the error in fact is:
mount -t vboxsf -o uid=`id -u myapp`,gid=`getent group myapp | cut -d: -f3` opt_MyApp /opt/MyApp
...
id: myapp: No such user
The id command was not able to recognize the user. So, switching to uid and gid the command id won't be used by vagrant.
The only warning I got with this approach is that user home directory (/opt/MyApp) already exist, but I can live with that, or you can change the useradd command to ignore the home directory if already exists.
Before that, the workaround I used is:
vagrant up; vagrant provision; vagrant reload
But, it not nice neither clean.
Upvotes: 9
Reputation: 929
If you can fix the uid/gid values you can use these in the mount command - they don't have to relate to an existing user/group
I do this with a user that is later created by puppet using fixed (matching)uid / gid values
config.vm.synced_folder "foo", "/var/www/foo",
id: "foo", :mount_options => ["uid=510,gid=510"]
Upvotes: 27
Reputation: 876
How I solved this was that I first configure the share to Vagrantfile with no user or group information. Then in the provisioning phase I unmount the share and remount it with proper user and group information. e.g.:
exec {
'umount /share/location':
command => 'umount /share/location';
} -> exec {
'mount /share/location':
command => 'mount -t vboxsf -o uid=`id -u apache`,gid=`id -g apache` /share/name /share/location'
You can check the share name from virtualbox or by running the provisioning with debug flag and non working settings (it prints out the actual mount command). I know this is kinda workaround and might not work in every situation, but it worked for me.
Upvotes: 2