The magic behind Juju is a collection of software components called Charms. They contain all the instructions necessary for deploying and configuring cloud-based applications. The charms publicly available in the online charm store represent the distilled DevOps knowledge of experts. Charms make it easy to reliably and repeatedly deploy applications, then scale up as required with minimal effort.
There are a variety of topics related to charms to be found in the left navigation pane under Charms. The three main topics you should get to know well are:
- Deploying charms - which covers how to use charms to deploy applications where and how you want them
- Charm relations - which covers connecting applications together
- Scaling - how to seamlessly scale up (or down) your deployed applications.
This section will take you through some of the common operations you will want to perform with Juju. By no means will we be using all the features of Juju, but you should gain an understanding of:
- How to deploy a charm
- Relating applications
- Exposing applications
- Scaling up applications
- Removing units
- Removing applications
- Destroying your model
Important: These instructions assume that you have already added credentials for your cloud. If you have not yet done this, please see Cloud credentials first. You can also learn about clouds on the central Clouds page.
A 'default' model is created automatically when you create a controller and it's the model that will be used here. The controller used here is named 'gce-test' and is based on the Google Compute Engine.
We are going to set up a simple MediaWiki site, then prepare it for high traffic, before scaling it back and finally removing it altogether.
For a MediaWiki site, we will need the 'mediawiki' charm from the charm store. We can deploy that to our model like this:
juju deploy mediawiki
Now, the Mediawiki application needs a database to store information in. There are several appropriate charms we could use, but for this walkthrough we will use MariaDB:
juju deploy mariadb
It may take a few minutes for Juju to actually fetch these charms from the store, create new machines to put them on and install them, but as soon as the command returns we are free to do other things, while Juju continues working in the background.
As we said at the beginning, we are going to scale up this application to cope with a lot of traffic. To do that we will use a common TCP/HTTP load balancing application, HAProxy, so we should also deploy a charm for that now:
juju deploy haproxy
After a while if you check what your model currently contains by running...
...you should see something like this:
Model Controller Cloud/Region Version SLA default gce-test google/us-east1 2.3.1 unsupported App Version Status Scale Charm Store Rev OS Notes haproxy unknown 1 haproxy jujucharms 41 ubuntu mariadb 10.1.30 active 1 mariadb jujucharms 7 ubuntu mediawiki 1.19.14 blocked 1 mediawiki jujucharms 19 ubuntu Unit Workload Agent Machine Public address Ports Message haproxy/1* unknown idle 5 220.127.116.11 mariadb/1* active idle 4 18.104.22.168 ready mediawiki/0* blocked idle 3 22.214.171.124 Database required Machine State DNS Inst id Series AZ Message 3 started 126.96.36.199 juju-5b2986-3 trusty us-east1-b RUNNING 4 started 188.8.131.52 juju-5b2986-4 trusty us-east1-c RUNNING 5 started 184.108.40.206 juju-5b2986-5 xenial us-east1-d RUNNING Relation provider Requirer Interface Type Message haproxy:peer haproxy:peer haproxy-peer peer mariadb:cluster mariadb:cluster mysql-ha peer
The applications may now be running, but they have no idea of the existence of each other, and aren't connected in any meaningful way. In order for MediaWiki to make use of the MariaDB database for a backend, you need to add a relation between them.
juju add-relation mediawiki:db mariadb
Normally, we only need to reference the applications by name. In this case, we also need to add the interface that we would like to connect, as there are two potential ways these charms could be connected.
Behind the scenes, the charms that control these applications will now contact each other and exchange information. In this case, the MediaWiki charm requests a new database to be created; MariaDB obliges, and gives the MediaWiki charm the credentials it needs to be able to access this database.
The other relation we need to add is between MediaWiki and the HAProxy application.
HAProxy will provide loadbalancing for traffic to MediaWiki, but it needs to know where the various MediaWiki applications are on the network. At this stage there is only one, but that will change shortly.
juju add-relation haproxy mediawiki
Now that the relations are set up, there is one more thing to do before the MediaWiki site is "live".
By default, Juju is very secure. Juju itself, and the applications it deploys can see the other applications in your cloud, but nothing and nobody else can. This isn't much use for a web application we want users to connect to, but Juju can easily make these applications public. With its understanding of the underlying cloud, Juju can make whatever firewall changes are necessary to expose these applications to the wider world:
juju expose haproxy
The below final output to
juju status represents a successful deployment of
Model Controller Cloud/Region Version SLA default gce-test google/us-east1 2.3.1 unsupported App Version Status Scale Charm Store Rev OS Notes haproxy unknown 1 haproxy jujucharms 41 ubuntu exposed mariadb 10.1.30 active 1 mariadb jujucharms 7 ubuntu mediawiki 1.19.14 active 1 mediawiki jujucharms 19 ubuntu Unit Workload Agent Machine Public address Ports Message haproxy/1* unknown idle 5 220.127.116.11 80/tcp mariadb/1* active idle 4 18.104.22.168 ready mediawiki/0* active idle 3 22.214.171.124 80/tcp Ready Machine State DNS Inst id Series AZ Message 3 started 126.96.36.199 juju-5b2986-3 trusty us-east1-b RUNNING 4 started 188.8.131.52 juju-5b2986-4 trusty us-east1-c RUNNING 5 started 184.108.40.206 juju-5b2986-5 xenial us-east1-d RUNNING Relation provider Requirer Interface Type Message haproxy:peer haproxy:peer haproxy-peer peer mariadb:cluster mariadb:cluster mysql-ha peer mariadb:db mediawiki:db mysql regular mediawiki:website haproxy:reverseproxy http regular Relation Provides Consumes Type peer haproxy haproxy peer website haproxy mediawiki regular cluster mariadb mariadb peer db mariadb mediawiki regular
Our MediaWiki site is now exposed via HAProxy. You can check this by pointing your browser to the IP address of HAProxy visible in the above output. In this example, it is 220.127.116.11. A confirmation will be in the form of MediaWiki's main page appearing, containing the message 'MediaWiki has been successfully installed'.
One of Juju's great strengths is that it makes it easy to scale your application to meet fluctuations in demand. We're going to demonstrate this with with the MediaWiki application we've just deployed. Scale is handled by adding and removing units, and you can add 5 units simply with the following command:
juju add-unit -n 5 mediawiki
When you now check the output of the
juju status mediawiki command, you'll see
that 5 newly provisioned machines have been deployed to run MediaWiki. However,
the load balancing is being performed by HAProxy, which is distributing any
incoming connections to your cluster of MediaWiki machines, and the complexity
of these added connections is being handled automatically by Juju.
Reducing the scale of a deployment is almost as simple as increasing the scale,
although you need to specify which specific units to remove. We currently have
a total of 6 units assigned to MediaWiki, for example, as can be seen in the
output of the
juju status mediawiki command:
Unit Workload Agent Machine Public address Ports Message mediawiki/0* active idle 3 18.104.22.168 80/tcp Ready mediawiki/1 active idle 6 22.214.171.124 80/tcp Ready mediawiki/2 active idle 7 126.96.36.199 80/tcp Ready mediawiki/3 active idle 8 188.8.131.52 80/tcp Ready mediawiki/4 active idle 9 184.108.40.206 80/tcp Ready mediawiki/5 active idle 10 220.127.116.11 80/tcp Ready
To scale back our deployment, use the
remove-unit command followed by the
unit ID of each unit you'd like to remove:
juju remove-unit mediawiki/3 mediawiki/4 mediawiki/5
A hidden part of the above process is that the machines the units were running on will be destroyed automatically if the machine is not a controller and not hosting other Juju managed containers.
If you no longer require MediaWiki, you can remove the entire application, along with all the units and machines used to operate the application, with a single command:
juju remove-application mediawiki
When the removal has completed, you should see no trace of 'mediawiki' in the
juju status, nor any of the units and machines that were used to
run the application.
Finally, to complete this brief walkthrough, we're going to remove the model
we've used to host our applications. If this is a fresh Juju installation, you
will have been operating within the 'default' model that's automatically
generated when you bootstrap the environment. You can check which models you
have available, along with which one is active, marked by the * symbol,
juju models command. Your output should be similar to the
Controller: gce-test Model Cloud/Region Status Machines Cores Access Last connection controller google/us-east1 available 1 4 admin just now default* google/us-east1 available 2 2 admin 4 minutes ago
To remove the 'default' model we've been using for the MediaWiki deployment,
juju destroy-model default and accept the warning that this step will
destroy all machines, applications, data and other resources associated with
that model. A few moments later you should find it no longer listed in the
output of the
juju models command.
If you want to start again with a clean model, restoring Juju to the state it
was in before we deployed the MediaWiki charm, you can re-create the 'default'
juju add-model default.