LXD with machine containers gives you the cloud experience locally. You can use this local cloud to build the same models with Juju that you can build on other public and private clouds. Using LXD as a method to test, verify, and replicate complex software deployments is a powerful tool that every Juju user needs. These instructions will get you up and running and deliver the best-possible experience with Juju. At the moment, that means using the latest release of Ubuntu with 'Long Term Support' (LTS): 16.04 LTS (Xenial).
Your system will need the following:
- LXD: a hypervisor for LXC, providing fast, secure containers.
- ZFS: a highly efficient and feature-rich filesystem and logical volume manager.
Begin by installing the required software. You'll need Juju and LXD on the machine.
For more options in installing Juju see the Juju install docs.
For LXD you can use:
sudo apt install lxd zfsutils-linux
In order to use LXD, your user must be a member of the
lxd group. This
should already be the case but you can confirm this by running the command:
Your groups may vary, but if
lxd is absent you can add yourself to the
Secondly, LXD includes an interactive initialisation which includes setting up a ZFS pool and appropriate networking for your LXD containers. To start this process, enter:
sudo lxd init
You will be asked several questions to configure LXD for use. Pressing Enter will accept the default answer (provided in square brackets).
Name of the storage backend to use (dir or zfs) [default=zfs]: Create a new ZFS pool (yes/no) [default=yes]? Name of the new ZFS pool [default=lxd]: Would you like to use an existing block device (yes/no) [default=no]? Size in GB of the new loop device (1GB minimum) [default=10GB]: 20 Would you like LXD to be available over the network (yes/no) [default=no]? Do you want to configure the LXD bridge (yes/no) [default=yes]?
The bridge network will then be configured via a second round of questions. Except in the case where the randomly chosen subnet may conflict with an existing one in your local environment, it is fine to accept all the default answers. IPv6 networking (the last question) is not required for Juju.
Walkthrough of network configuration
In order for networking to be established between containers and Juju, you need to set up a bridge device.
The default name for the bridge device is
lxdbr0. This name must be used
for Juju to be able to connect to the containers.
Juju will expect an IPv4 network space for the containers, so you should enable this.
The default address is chosen randomly in the 10.x.x.x space. You do not need to change this unless it conflicts with another subnet you know is on your network.
You need to enter a CIDR mask value. The default of 24 gives you a possible 254 addresses for the subnet.
You can now specify the start of the DHCP address range...
And the end address of the range...
You can also specify the total number of DHCP clients to accept.
Finally for IPv4, enable Network Address Translation (NAT) to allow the containers to communicate with the outside world.
You can continue to set up a similar IPv6 bridge device, but this is not necessary for Juju.
LXD is now configured to work with Juju.
Note: LXD adds iptables (firewall) rules to allow traffic to the
subnet/bridge it created. If you subsequently add/change firewall settings
ufw), ensure that such changes have not interfered with Juju's
ability to communicate with LXD.
Before you can start deploying applications, Juju needs to create a controller. The controller manages both the state and events for your models that host the applications.
juju bootstrap command is used to create the controller. The command
expects a name (for referencing this controller) and a cloud to use. The LXD
'cloud' is known as 'localhost' to Juju.
For our localhost cloud, we will create a controller called 'lxd-test':
juju bootstrap localhost lxd-test
This may take a few minutes as LXD must download an image for Xenial. A cache will be used for subsequent containers.
Once the process has completed you can check that the controller has been created:
This will return a list of the controllers known to Juju. You can see our 'lxd-test' controller listed.
Use --refresh flag with this command to see the latest information. Controller Model User Access Cloud/Region Models Machines HA Version lxd-test* default admin superuser localhost/localhost 2 1 none 2.0.0
A newly-created controller has two models: The 'controller' model, which should be used only for internal Juju management, and a 'default' model, which is ready for actual use. The controller model can only be destroyed by destroying the controller itself.
The following command shows the currently active controller, model, and user:
In our example, the output should look like this:
Controller: lxd-test Model: default User: admin
Juju is now ready to deploy applications from among the hundreds included in the Juju charm store. It is a good idea to test your new model. How about an MediaWiki site?
juju deploy cs:bundle/mediawiki-single
This will fetch a 'bundle' from the Juju store. A bundle is a pre-packaged set of applications, in this case 'MediaWiki', and a database to run it with. Juju will install both applications and leveraging the relation in the model between them, configures the wiki to use the newly created database instance. This is part of the magic of Juju: it isn't just about deploying software, Juju also knows how applications coordinate together.
Installing shouldn't take long. You can check on how far Juju has got by running the command:
When the applications have been installed, the output of the above command will look something like this:
There is a lot of useful information there! The important parts for now are the 'App' section, which shows that MediaWiki and MySQL are installed, and the 'Unit' section, which shows the IP addresses allocated to each. These addresses correspond to the subnet we created for LXD earlier on.
Juju models which applications have firewalls that allow external access on
clouds that support firewall controls. You can update the model by using the
juju expose <application> command. In our case we would want to
juju expose mediawiki.
From the status output, we can see that the IP address for the MediaWiki site we have created is 10.154.173.2 Open a browser and enter that address to see the site.
Congratulations, you have just deployed an application with Juju!
Note: To remove all the applications in the model you just created, it is
often quickest to destroy the model with the command
juju destroy-model default and then create a new model.
Now that you have configured Juju to work with a local LXD cloud, you can develop, test, and experiment with applications with speed, ease, and without additional public cloud costs.
We suggest you continue your journey by discovering: