Bringing machine 'default' up with 'libvirt' provider...
/home/teto/vagrant-libvirt/vendor/bundle/ruby/1.9.1/gems/fog-1.20.0/lib/fog/libvirt/models/compute/volume.rb:84:in `default_pool_name': No storage pools are defined (Fog::Errors::Error)
from /home/teto/vagrant-libvirt/vendor/bundle/ruby/1.9.1/gems/fog-1.20.0/lib/fog/libvirt/models/compute/volume.rb:33:in `initialize'
from /home/teto/vagrant-libvirt/vendor/bundle/ruby/1.9.1/gems/fog-1.20.0/lib/fog/core/collection.rb:114:in `new'
- The `vagrant ssh_config` command was including the `Host` option, which for
OpenSSH_6.5p1 at least is throwing an error:
```
Host directive not supported as a command-line option
```
- Also changing the local bind_address in the ssh command to 'localhost'
instead of '0.0.0.0' in order to not expose the VM to other hosts on the host
machine's network. If this is desired, then the user can manually override
it in the Vagrantfile config for port forwarding.
If ssh inherits stdout from vagrant, then when you pipe vagrants output
to another program ssh will also use that pipe. The receiving program
will hang waiting on ssh to exit. To make reasoning about things easier,
I have ruby start ssh directly instead of via a shell. Then I have
spawn set stdin/stdout to /dev/null.
Network device creation logic did not properly handle forwarded_ports;
instead it created connections to the management network for each one.
We depended on this in order to have at least one connection to the
management network. This commit explicitly handles the management
network and ignores forwarded ports.
The use of the term default makes discussing this network confusing.
This network must exist and be configured a certain way in order for vagrant
to manage a VMs configuration, so lets call it the management network instead.
This commit
* renames the option default_network to default_network_name
* introduces the option default_network_address, since we need an
address in order to create a network
* handles creation of the default network similarly to other private
networks if needed
This patch adds support for a provider specific default_prefix option.
When set, this value is used instead of the project's dir name as the
prefix for the machines being created.
