Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit



Failed to load latest commit information.
Latest commit message
Commit time

juju is devops distilled.

Getting started

`juju-core` is written in Go (, a modern, compiled, statically
typed, concurrent language. This document describes how to build `juju-core`
from source.

If you are looking for binary releases of `juju-core`, they are
available from the Juju stable PPA, ``,
and can be installed with:

    sudo apt-add-repository ppa:juju/stable
    sudo apt-get update
    sudo apt-get install juju

Installing prerequisites

You can use `make install-dependencies` or, if you prefer to install
them manually, check the Makefile target.

This will add some PPAs to ensure that you can install the required
golang and mongodb-server versions for precise onwards, in addition to the
other dependencies.

Setting GOPATH

When working with the source of Go programs, you should define a path within
your home directory (or other workspace) which will be your `GOPATH`. `GOPATH`
is similar to Java's `CLASSPATH` or Python's `~/.local`. `GOPATH` is documented
online at `` and inside the `go` tool itself

    go help gopath

Various conventions exist for naming the location of your `GOPATH`, but it should
exist, and be writable by you. For example

    export GOPATH=${HOME}/work
    mkdir $GOPATH

will define and create `$HOME/work` as your local `GOPATH`. The `go` tool itself
will create three subdirectories inside your `GOPATH` when required; `src`, `pkg`
and `bin`, which hold the source of Go programs, compiled packages and compiled
binaries, respectively.

Setting `GOPATH` correctly is critical when developing Go programs. Set and
export it as part of your login script.

Add `$GOPATH/bin` to your `PATH`, so you can run the go programs you install:


Getting juju-core

The easiest way to get the source for `juju-core` is to use the `go get` command.

    go get -v

This command will checkout the source of `juju-core` and inspect it for any unmet
Go package dependencies, downloading those as well. `go get` will also build and
install `juju-core` and its dependencies. To checkout without installing, use the
`-d` flag. More details on the `go get` flags are available using

    go help get

At this point you will have the bzr working copy of the `juju-core` source at
`$GOPATH/`. The source for any dependent packages will
also be available inside `$GOPATH`. You can use `bzr pull`, or the less convenient
`go get -u` to update the source from time to time.
If you want to know more about contributing to `juju-core`, please read the
`CONTRIBUTING` companion to this file.

Building juju-core

    go install -v

Will build juju and install the binary commands into `$GOPATH/bin`. It is likely
if you have just completed the previous step to get the `juju-core` source, the
install process will produce no output, as the final executables are up-to-date.

Using juju-core

After following the steps above you will have the `juju` client installed in
`GOPATH/bin/juju`. You should ensure that this version of `juju` appears earlier
in your path than any packaged versions of `juju-core`, or older Python juju
commands. You can verify this using

    which juju

You should be able to bootstrap a local environment now with the following
(Note: the use of sudo for bootstrap here is only required for the local
provider because it uses LXC, which requires root privileges)

    juju init
    juju switch local
    sudo juju bootstrap


The `juju` client program, and the juju 'tools' are deployed in lockstep. When a
release of `juju-core` is made, the compiled tools matching that version of juju
are extracted and uploaded to a known location. This consumes a release version
number, and implies that no tools are available for the next, development, version
of juju. Therefore, when using the development version of juju you will need to
pass an additional flag, `--upload-tools` to instruct the `juju` client to build
a set of tools from source and upload them to the environment as part of the
bootstrap process.

    juju bootstrap -e your-environment --upload-tools {--debug}

Installing bash completion for juju

    make install-etc

Will install Bash completion for `juju` cli to
`/etc/bash_completion.d/juju-core` It does dynamic completion for commands
requiring service, unit or machine names (like e.g. juju status <service>,
juju ssh <instance>, juju terminate-machine <machine#>, etc),
by parsing cached `juju status` output for speedup. It also does
command flags completion by parsing `juju help ...` output.


Juju core







No releases published


No packages published