Advanced application deployment
This page is dedicated to more advanced topics related to deploying applications with Juju. The main page is Deploying applications.
This page covers the following topics:
- Charmed operators and series
- Bundles and series
- Deploying to specific machines
- Deploying to specific availability zones
- Deploying to network spaces
- Trusting an application with a credential
Charmed operators and series
Charmed operators can be created that support more than one release of a given operating system distro, such as the multiple Ubuntu releases shown below. It is not possible to create a charmed operator to support multiple distros, such as one charmed operator for both Ubuntu and CentOS. Supported series are added to the charmed operator metadata like this:
name: mycharm summary: "Great software" description: It works maintainer: Some One <email@example.com> categories: - databases series: - focal - xenial provides: db: interface: pgsql requires: syslog: interface: syslog optional: true
The default series for the charmed operator is the first one listed. So, in this example, to deploy
trusty, all you need is:
juju deploy mycharm
You can specify a different series using the
juju deploy mycharm --series xenial
You can force the charmed operator to deploy using an unsupported series using the
juju deploy mycharm --series bionic --force
Here is a more complete example showing a new machine being added that uses a different series than is supported by our
mycharm example and then forcing the charmed operator to install:
juju add-machine --series bionic juju deploy mycharm --to 1 --series bionic --force
Multi-series charmed operators may encounter upgrade difficulties if support for the installed series is dropped. See Forced upgrades for details.
The series resolution order for charms:
- the ‘–series’ option provided with the
- the series stated in the charmed operator URL, e.g.
- the ‘default-series’ model key
- the top-most series specified in the charm’s metadata file
Bundles and series
The series resolution order for bundles:
- the ‘–series’ option provided with the
- the series stated in each charmed operator URL in the bundle file
- the series given at the top level in the bundle file
- the ‘default-series’ model key
Deploying to specific machines
To deploy to specific machines the
--to option is used. It is supported by commands
add-unit. This option, along with an argument, is called a placement directive.
--to option is used, unless the machine was created via
add-machine, a charmed operator has already been deployed to the machine. When multiple charmed operators are deployed to the same machine there exists the chance of conflicting configuration files (on the machine’s filesystem).
There is one type of placement directive that can also be used as a constraint: availability zones. If used together, the placement directive takes precedence. See Using constraints for details.
To apply this option towards an existing Juju machine, the machine ID is used. This is an integer that is shown in the output to
juju status (or
juju machines). For example, this partial output shows a machine with an ID of ‘2’:
Machine State DNS Inst id Series AZ Message 2 started 10.132.70.65 juju-79b3aa-0 xenial Running
The above works well with
add-unit as will be shown below. As for
--to option is limited to either pointing to a MAAS node or, starting in
v.2.5.0, to a LXD cluster node.
Assuming a MAAS cloud named ‘maas-prod’ exists and has a node called ‘node2.maas’:
juju bootstrap maas-prod --to node2.maas
Assuming a LXD cluster cloud named ‘lxd-cluster’ exists and has a node called ‘node3’:
juju bootstrap lxd-cluster --to node3
To deploy the ‘haproxy’ application to machine ‘2’ we would do this:
juju deploy --to 2 haproxy
--constraints option is used during controller creation to ensure that each workload machine will have enough memory to run multiple applications. MySQL is deployed as the first unit (in the ‘default’ model) and so ends up on machine ‘0’. Then Rabbitmq gets deployed to the same machine:
juju bootstrap --constraints mem=4G localhost lxd juju deploy mysql juju deploy --to 0 rabbitmq-server
Juju treats a container like any other machine so it is possible to target specific containers as well. Here we deploy to containers in three different ways:
juju deploy mariadb --to lxd juju deploy mongodb --to lxd:25 juju deploy nginx --to 24/lxd/3
In the first case, mariadb is deployed to a container on a new machine. In the second case, MongoDB is deployed to a new container on existing machine ‘25’. In the third case, nginx is deployed to existing container ‘3’ on existing machine ‘24’.
Some clouds support special arguments to the
--to option, where instead of a machine you can specify a zone. In the case of MAAS or a LXD cluster a node can be specified:
juju deploy mysql --to zone=us-east-1a juju deploy mediawiki --to node1.maas juju deploy mariadb --to node1.lxd
For a Kubernetes-backed cloud, a Kubernetes node can be targeted based on matching labels. The label can be either built-in or one that is user-defined and added to the node. For example:
juju deploy mariadb-k8s --to kubernetes.io/hostname=somehost
To add a unit of ‘rabbitmq-server’ to machine ‘1’:
juju add-unit --to 1 rabbitmq-server
A comma separated list of directives can be provided to cater for the case where more than one unit is being added:
juju add-unit rabbitmq-server -n 3 --to host1.maas,host2.maas,host3.maas
If the number of values is less than the number of requested units the remaining units, as per normal behaviour, will be deployed to new machines:
juju add-unit rabbitmq-server -n 4 --to zone=us-west-1a,zone=us-east-1b
Any surplus values are ignored:
juju add-unit rabbitmq-server -n 2 --to node1.lxd,node2.lxd,node3.lxd
add-unit command is often associated with scaling out. See the Scaling applications page for information on that topic.
Deploying to specific availability zones
To deploy to specific availability zones the
--constraints option is used. It is supported by commands
The constraint type that is used to do this is ‘zones’. This is not to be confused with the ‘zone’ placement directive, which happens to take precedence over the constraint.
For instance, here we create two Trusty machines in a certain zone:
juju add-machine -n 2 --series trusty --constraints zones=us-east-1a
We then deploy an application on two new machines in a different zone:
juju deploy redis -n 2 --constraints zones=us-east-1c
Finally, in order to deploy units to the two empty machines in the initial zone we first change the application constraint default (set implicitly with the
juju set-constraints redis zones=us-east-1a juju add-unit redis -n 2
When multiple (comma separated) values are used, the constraint is interpreted as being a range of zones where a machine must end up in.
Deploying to network spaces
Using spaces, administrators are able to create a more restricted network topology for applications at deployment time (see Network spaces for details on spaces). This is achieved with the use of the
The following will deploy the ‘mysql’ application to the ‘db-space’ space:
juju deploy mysql --bind db-space
For finer control, individual endpoints can be connected to specific spaces:
juju deploy --bind "db=db-space db-admin=admin-space" mysql
If a space is mentioned that is not associated with an interface then it will act as the default space (i.e. will be used for any unspecified interface):
juju deploy --bind "default-space db=db-space db-admin=admin-space" mysql
See Concepts and terms for the definition of an endpoint, an interface, and other closely related terms.
For information on applying bindings to bundles, see Binding endpoints within a bundle.
deploy command also allows for the specification of a constraint. Here is an example of doing this with spaces:
juju deploy mysql -n 2 --constraints spaces=database
See Adding a machine with constraints for an example of doing this with spaces.
You can also declare an endpoint for spaces that is not used with relations, see Extra-bindings.
This example will have MAAS as the backing cloud and use the following criteria:
- DMZ space (with 2 subnets, one in each zone), hosting 2 units of the haproxy application, which is exposed and provides access to the CMS application behind it.
- CMS space (also with 2 subnets, one per zone), hosting 2 units of mediawiki, accessible only via haproxy (not exposed).
- Database (again, 2 subnets, one per zone), hosting 2 units of mysql, providing the database backend for mediawiki.
First, ensure MAAS has the necessary subnets and spaces. Each subnet has the “automatic public IP address” attribute enabled on each:
- 172.31.50.0/24, for space “database”
- 172.31.51.0/24, for space “database”
- 172.31.100.0/24, for space “cms”
- 172.31.110.0/24, for space “cms”
- 172.31.0.0/20, for the “dmz” space
- 172.31.16.0/20, for the “dmz” space
Recall that MAAS has native knowledge of spaces. They are created within MAAS and Juju will become aware of them when the Juju controller is built (
Second, add the MAAS cloud to Juju. See Using a MAAS cloud for guidance.
Third, create the Juju controller, assuming a cloud name of ‘maas-cloud’:
juju bootstrap maas-cloud
Finally, deploy the applications into their respective spaces (here we use the constraints method), relate them, and expose haproxy:
juju deploy haproxy -n 2 --constraints spaces=dmz juju deploy mediawiki -n 2 --constraints spaces=cms juju deploy mysql -n 2 --constraints spaces=database juju add-relation haproxy mediawiki juju add-relation mediawiki mysql juju expose haproxy
Once all the units are up, you will be able to get the public IP address of one of the haproxy units (from
juju status), and open it in a browser, seeing the mediawiki page.
Trusting an application with a credential
Some applications may require access to the backing cloud in order to fulfil their purpose (e.g. storage-related tasks). In such cases, the remote credential associated with the current model would need to be shared with the application. When the Juju administrator allows this to occur the application is said to be trusted. An application can be trusted during deployment or after deployment.
To trust the AWS integrator application during deployment:
juju deploy --trust cs:~containers/aws-integrator
To trust the application after deployment:
juju deploy cs:~containers/aws-integrator juju trust aws-integrator
Last updated 19 days ago.