FIX hard coded vnc port
- better use graphics_port = -1 so libvirt auto assign it
and its possible to use more that one vm
Signed-off-by: Thomas Heil <heil@terminal-consulting.de>
Commonly found in other Vagrant providers, a MAC address format without
colon delimiters is now accepted for better cross-compatibility of
Vagrantfiles.
This provides the same behaviour seen in other providers (e.g.
virtualbox).
Files are placed into a "_include" subdirectory and the user provided
Vagrantfile (if any) will be loaded from the main Vagranfile in the box.
Also fixed a typo in the `assemble_box` method.
If I suspend using managedsave suspend_mode, then reboot the hypervisor,
and try to resume the box, then it fails because the management network
is not started (of course unless it's configured to autostart, which is
not the default behaviour).
This patch makes vagrant to start (and even recreate) the management
network if it's down.
Fixesvagrant-libvirt/vagrant-libvirt#761
Switch from using the default human readable format to request the json
format from qemu-img as this will provide disk sizes in bytes.
This avoids assuming that the returned size is in GigaBytes.
Fixes#743
Allows `vagrant up` command to handle whether or not to run the
provisioners should the machine already be up. This allows re-running
certain provisioners in parallel using:
vagrant up --provision --provision-with puppet_server
Which in turn allows for certain scripts that are bootstrap related
only to be skipped.
In general however it is better to let the middleware from vagrant
decided whether provisioners should run or not, as enhancements to that
functionality will then be automatically picked up.
With this change if a user runs `vagrant up` on already running
machines, without the `--provision` option they will be greeted with a
message from vagrant stating "Machine already provisioned. Run ...",
basically letting them know what to do next.
Without this change you simply see a message about "Bringing machine
'<name>' up with 'libvirt' provider..." and then nothing further.
Change-Id: I73ca4262abaa6431a838e4cf6468c06223643dea
Pxe nodes are not provisioned, mgmt network is not required for them.
Provide new domain option 'mgmt_attach' which one can use to remove mgmt
interface from VM. Default behaviour is not changed.
Also take into account bridge interfaces when set boot order, to support
scenario when we would like to use vagrant VM to pxe boot physical nodes
in existing network.
