An application is typically a long-running service that is accessible over the network. Applications are the centre of a Juju deployment. Everything within the Juju ecosystem exists to facilitate them.
It’s easiest to think of the term “application” in Juju in the same way you would think of using it day-to-day. Middleware such as database servers (PostgreSQL, MySQL, Percona Cluster, etcd, …), message queues (RabbitMQ) and other utilities (Nagios, Prometheus, …) are all applications. The term has a special meaning within the Juju community, however. It is broader than the ordinary use of the term in computing.
A Juju application is more than a software application
Juju takes care of ensuring that the compute node that they’re being deployed to satisfies the size constraints that you specify, installing them, increasing their scale, setting up their networking and storage capacity rules. This, and other functionality, is provided within software packages called charmed operators.
Alongside your application, Juju executes charm code when triggered. Triggers are typically requests from the administrator, such as:
|“The configuration needs to change”|
|description||The spark charm provides the ability to dynamically change the memory available to the driver and executors|
|“Please scale-up this application”|
|description||The postgresql charm can detect when its scale is more than 1 and automatically switches itself into a high-availability cluster|
|“Allocate a 20GB storage volume to the application unit 0”|
|description||The etcd charm can provide an SSD-backed volume on AWS to the etcd application with|
The Juju project uses an active agent architecture. Juju software agents are running alongside your applications. They periodically execute commands that are provided in software packages called charmed operators.
Differences between a stock software application and a Juju application
A. Applications are scale-independent
An application in the Juju ecosystem can span multiple operating system processes. An HTTP API would probably be considered a Juju application, but that might bundle together several other components.
- A Ruby on Rails web application might be deployed behind Apache2 and Phusion Passenger.
- All workers within a Hadoop cluster are considered a single application, although each worker has its unit.
A Juju application can also span multiple compute nodes and/or containers.
Within the Juju community, we use the term machine to cover physical hardware, virtual machines and containers.
To make this clearer, consider an analogy from the desktop. An Electron app is composed of an Internet browser, a node.js runtime and application code. Each of those components is distinct, but they exist as a single unit. That unit is an application.
A final iteration of scale-independence is that Juju will maintain a record for applications that have a scale of 0. Perhaps earlier in the application’s life cycle it was wound down, but the business required that the storage volumes were to be retained.
B. Applications are active
Applications automatically negotiate their configuration depending on their situation. Through the business logic encoded within charmed operators, two applications can create user accounts and passwords between themselves without leaking secrets.
C. Applications are responsive
Juju applications can indicate their status, run actions and provide metrics. An action is typically a script that is useful for running a management task.
Application management tasks
Applications can have their configuration options set during, or after, deployment. The Configuring applications page explains how this is done.
Credentials and application trust
Some applications may require access to the backing cloud to fulfil their purpose. In such cases, the credential associated with the current model would need to be shared with the application. See section Trusting an application with a credential for details.
The Deploying applications page covers an array of methods for getting your applications deployed.
The Deploying applications - advanced page contains more advanced use cases.
See the Deploying charms offline page for guidance when deploying in a network-restricted environment.
Applications can be deployed and configured as a collection of charmed operators. This subject is treated on the Charmed bundles page.
Display information for deployed applications
Once an application is deployed it is possible to display information on it. This is done with the show-application command. For example, you could deploy an application with:
juju deploy postgresql pgsql
After it is deployed, you can get information about it:
juju show-application pgsql
pgsql: charm: postgresql series: bionic channel: stable principal: true exposed: false remote: false endpoint-bindings: coordinator: "" data: "" db: "" db-admin: "" local-monitors: "" master: "" nrpe-external-master: "" replication: "" syslog: ""
When an application requires another application to fulfil its purpose they need to be logically linked together. In Juju, such a link is called a relation . The Managing relations page explains this important concept.
Removing an application is a simple process. See the Removing things page for guidance.
Juju horizontally scales applications up and down by adding and removing application units. See the Scaling applications page for details.
Upgrade an application
Upgrading an application in Juju means to upgrade the application’s charm. See the Upgrading applications page for in-depth coverage.
Actions are charm-specific code which can be called at will from the command line. Working with actions page provides full coverage of the subject.
Application groups allow an administrator to manage groups of the same application by providing custom application names during deployment. See the Application groups page for details.
Application high availability
Application high availability pertains to the distribution of units over availability zones. See the Application high availability page for full information.
Metrics for applications can be collected for model-level assessment of application utilisation and capacity planning. See the Application metrics page to learn more.
Constraints are used to specify minimum requirements for the machine that will host an application. See page Using constraints for details.
A Juju resource is additional content/files that a charm can make use of, or may even require to run. See the Working with resources page for details.
Certain applications can benefit from advanced storage configurations and if a charm exists for such an application Juju can declare such requirements both at deploy time and during the lifetime of the application. See the Using storage page for more detail.