Controllers

A Juju controller is the management node of a Juju cloud environment. In particular, it houses the database and keeps track of all the models in that environment. Although it is a special node, it is a machine that gets created by Juju (during the "bootstrap" stage) and, in that sense, is similar to other Juju machines.

During controller creation two models are also created, the 'controller' model and the 'default' model. The primary purpose of the 'controller' model is to run and manage the Juju API server and the underlying database. Additional models may be created by the user - see Models.

Since a controller can host multiple models, the destruction of a controller must be done with ample consideration since all its models will be destroyed along with it.

In some circumstances you may wish to share a controller or one of its associated models. Juju provides multi-user functionality for this purpose.

Minimum requirements

The minimum resources required for a local controller are 2GiB of memory and 1 vCPU.

For LXD-based controllers, those resources will need to be increased according to the projected workload. Disk space is an important factor to consider for a localhost (LXD) cloud, due to boot images and containers.

Controller management

Common tasks are summarized below.

Create a controller

Use the juju bootstrap command to create a controller.

   juju bootstrap [options] [filter pattern ...]

For examples see Creating a controller.

For complete explanation, syntax and examples see the command reference page or the juju help bootstrap command.

List controllers

Use the juju controllers command to list all controllers knowable by the current system user.

   juju controllers [options]

The currently active controller is indicated in the list with an asterisk('*').

For complete explanation, syntax and examples see the command reference page or the juju help controllers command.

Show controller details

Use the juju show-controller command to show details for a controller. Information includes UUID, API endpoints, certificates, and bootstrap configuarion.

   juju show-controller [options]

For complete explanation, syntax and examples see the command reference page or the juju help show-controller command.

Configure a controller

There are several settings which can be adjusted from their defaults, such as whether the controller will record auditing information. Configuration must happen as the controller is bootstrapped.

This topic is covered in more depth in the configuring controllers section.

Remove a controller

Use the juju destroy-controller command to remove a controller.

   juju destroy-controller [options]

For complete explanation, syntax and examples see the command reference page or the juju help destroy-controller command.

Use the juju kill-controller command as a last resort if the controller is not accessible for some reason.

The controller will be removed by communicating directly with the cloud provider. Any other Juju machines residing within any of the controller's models will not be destroyed and will need to be removed manually using provider tools/console. This command will first attempt to mimic the behaviour of the juju destroy-controller command and failover to the more drastic behaviour if that attempt fails.

   juju kill-controller [options]

For complete explanation, syntax and examples see the command reference page or the juju help kill-controller command.

Use the Juju GUI

Each Juju controller creates a web-driven GUI as an alternative method of management. The GUI is capable of deploying, scaling and monitoring applications, as well as more advanced operations.

More details on the GUI can be found in the Juju GUI section.

Back up a controller

Juju allows one to create, restore and manage backup files containing the controller configuration/metadata. If the controller or its host machine fails, it is possible to recreate the controller from the backup.

This is a complex subject. See Juju backups.

Note: coverage of client backups are included in the above resource.

Implement HA (high availability)

Each Juju controller can be made 'Highly Available' to add resilience to the operations of the controller itself. This topic is covered in more detail in the Juju HA documentation. HA for the applications deployed is a matter for the charms, and is covered in a separate topic,charm HA.

Restricting command usage

A controller administrator is responsible for granting permissions to users registered with a controller. This can be done by restricting what sorts of changes a user can make across the controller's models.

This topic is treated in Restricting changes to the running Juju environment.

© 2018 Canonical Ltd. Ubuntu and Canonical are registered trademarks of Canonical Ltd.