Juju is a model-driven Operator Lifecycle Manager (OLM). Juju greatly improves the experience of running Kubernetes operators, especially in projects that integrate many operators from different publishers.
A Kubernetes operator is a container that drives the config and operation of a workload. By encapsulating ops code as a reusable container, the operator pattern moves beyond traditional config management to allow much more agile operations for complex cloud workloads.
Shared, open source operators take infra as code to the next level with community-driven ops and integration code. Reuse of ops code improves quality and encourages wider community engagement and contribution. Operators also improve security through consistent automation. Juju operators are a community-driven devsecops approach to open source operations.
Juju implements the Kubernetes operator pattern, but is also a universal OLM that extends the operator pattern to traditional applications (without Kubernetes) on Linux and Windows. Such machine operators can work on bare metal, virtual machines or cloud instances, enabling multi cloud and hybrid cloud operations. Juju allows you to embrace the operator pattern on both container and legacy estate. An operator for machine-based environments can share 95% of its code with a Kubernetes operator for the same app.
Juju excels at application integration. Instead of simply focusing on lifecycle management, the Juju OLM provides a rich application graph model that tells operators how to integrate with one another. This dramatically simplifies the operations of large deployments.
A key focus for Juju is to simplify operator design, development and usage. Instead of making very complex operators for specific scenarios, Juju encourages devops to make composable operators, each of which drives a single Docker image, and which can be reused in different settings. Composable operators enable very rich scenarios to be constructed out of simpler operators that do one thing and do it well.
The OLM provides a central mechanism for operator instantiation, configuration, upgrades, integration and administration. The OLM provides a range of operator lifecycle services including leader election and persistent state. Instead of manually deploying and configuring operators, the OLM manages all the operators in a model at the direction of the administrator.
The world's largest collection of operators all use Juju as their OLM. The Charmhub community emphasizes quality, collaboration and consistency. Publish your own operator and share integration code for other operators to connect to your application.
The Open Operator Manifesto outlines the values of the community and describe the ideal behaviour of operators, to shape contributions and discussions.
The Juju OLM supports AWS, Azure, Google, Oracle, OpenStack, VMware and bare metal machines, as well as any conformant Kubernetes cluster. Integrate operators across clouds, and across machines and containers, just as easily. A single scenario can include applications on Kubernetes, as well as applications on a range of clouds and bare metal instances, all integrated automatically.
Juju operators support multiple CPU architectures. Connect applications on ARM with applications on x86 and take advantage of silicon-specific optimisations. It is good practice for operators to adapt to their environment and accelerate workloads accordingly.
The Python Operator Framework makes it easy to write an operator. The framework handles all the details of communication between integrated operators, so you can focus on your own application lifecycle management.
Code sharing between operator publishers is simplified making it much faster to collaborate on distributed systems involving components from many different publishers and upstreams. Your operator is a Python event handler. Lifecycle management, configuration and integration are all events delivered to your charm by the framework.
The Juju client, server and agent are all written in Golang. The standard Juju packaging includes an embedded database for centralised logging and persistence, but there is no need to manage that database separately.
Operators can be written in any language but we do encourage new authors to use the Python Operator Framework for ease of contribution, support and community participation.
The Juju server has built-in support for high availability when scaled out to three instances. It can monitor itself and grow additional instances in the event of failure, within predetermined limits. Juju supports backup, restore, and rolling upgrade operations appropriate for large-scale centralised enterprise grade management and operations systems.
Our community hangs out at the Charmhub discourse which serves as a combination mailing list and web forum. Keep up with the news and get a feel for operator engineering and usage there. Get the Juju CLI on Windows, macOS or Linux with the install instructions and try the tutorials. All you need is a small K8s cluster, or an Ubuntu machine or VM to run MicroK8s.
Read the documentation for a comprehensive reference of commands and usage.
Follow our code and contribution guidelines to learn how to make code changes. File bugs in Launchpad or ask questions on our Freenode IRC channel.