juju is devops distilled.

juju-core is written in Go (http://golang.org), 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, https://launchpad.net/~juju/+archive/stable, and can be installed with:

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

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.

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 http://golang.org/pkg/go/build/ 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:

PATH="$PATH:$GOPATH/bin"

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

go get -v github.com/juju/core/...

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/github.com/juju/core. The source for any dependent packages will also be available inside $GOPATH. You can use bzr pull, or the less convenient go get -u github.com/juju/core/... 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.

go install -v github.com/juju/core/...

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.

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}

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 , juju ssh , juju terminate-machine <machine#>, etc), by parsing cached juju status output for speedup. It also does command flags completion by parsing juju help ... output.

testing testing