A model is upgraded by upgrading the version of Juju running within it, in the form of machine agents and unit agents. Agents are pieces of software that run on each machine spawned by Juju, including controllers. See the Concepts page for more information on agents.
A model upgrade is performed with the
upgrade-model command and acts on all
machine and unit agents running on all machines in the model.
Note: The client can also be upgraded. See the Juju client page for how to do that.
Several noteworthy points regarding backups:
- Upgrades must be applied to the 'controller' model before upgrading other models. An error will be emitted if this is not done.
- By default, the latest stable version will be selected.
- Juju machines request the new agent software from the controller. If the latter's cache cannot satisfy the request the controller will attempt a download from the internet.
- A controller admin can upgrade any model within that controller.
- A model owner can upgrade that model.
See Notes on upgrading models for upgrading details, including what to do when the controller lacks internet access.
This is a lit of some pre-upgrade tasks that should be considered:
- Perform a backup (see Controller backups).
- Ensure the models that are to be upgraded are in good working order by
statuscommand to them.
- See what version is currently running with
juju model-config agent-versionor
This section contains examples on upgrading a model.
To upgrade the 'controller' model for the 'aws' controller with the latest stable version:
juju upgrade-model -m aws:controller
To upgrade the current model by specifying the version:
juju upgrade-model --agent-version 2.4.1
To upgrade the current model by specifying the agent stream:
juju upgrade-model --agent-stream proposed
The verification of a successful upgrade is obtained by the
Here we check whether a 'controller' model was properly upgraded:
juju status -m aws:controller
If this command does not complete properly or if there are errors displayed in its output then proceed to the next section.
A model upgrade that does not lead to 100% success will require attention. See Troubleshooting model upgrades.