Skip to content

apache/brooklyn-docs

Brooklyn Website and Docs Source

Contributor Workflow

The contributor workflow is identical to that used by the main project, with pull requests and contributor licenses required. Therefore you should familiarise yourself with the standard workflow for Apache Brooklyn:

The documents are written in kramdown a superset of Markdown which is processed into HTML using Jekyll. In addition to the standard set of options and notation available with these platforms, a number of custom plug-ins have been implemented specifically for the Brooklyn docs. These are detailed in the contributing to docs doc.

Workstation Setup

First, if you have not already done so, clone the brooklyn repository and subprojects and set up the remotes as described in Guide for committers.

The Brooklyn documentation uses Jekyll to process the site content into HTML. This in turn requires Ruby and gems as described in the Gemfile: install RVM to manage Ruby installations and sets of Ruby gems.

curl -sSL https://get.rvm.io | bash -s stable --auto-dotfiles

Close your shell session and start a new one, to get the new environment that RVM has configured. Change directory to the location where this project is (where this file is located).

RVM should detect its configuration inside Gemfile and try to configure itself. Most likely it will report that the required version of Ruby is not installed, and it will show the command that you need to run to install the correct version. Follow the instructions it shows, typically something like rvm get master && rvm install 3.0.1.

Once the correct version of Ruby is installed, change to your home directory and then change back (cd ~ ; cd -). This will cause RVM to re-load configuration from Gemfile with the correct version of Ruby.

Finally, run this command to install all the required Gems at the correct versions:

bundle install

Any time you need to reset your Ruby environment for jekyll to run correctly, go to this directory (or the _build subdir) and re-run the above command.

On some platforms there may be some fiddling required before jekyll runs without errors, but the ecosystem is fairly mature and most problems can be resolved with a bit of googling. Some issues we've encountered are:

  • on Mac, install xcode and its command-line tools
  • if ruby gets confused about versions, clean out your gems
  • if libxml2 fails, set bundle config build.nokogiri --use-system-libraries before the install (more details here)
  • on Ubuntu, sudo apt-get install libxslt-dev libxml2-dev libcurl4-openssl-dev python-minimal
  • if openssl library headers are not found, set the dir explicitly, eg for eventmachine try gem install eventmachine -- --with-openssl-dir=/usr/local/opt/openssl@1.1
  • if libv8 complains: gem install libv8 -- --with-system-v8 or eg gem install libv8 -v '3.16.14.19' -- --with-system-v8
  • for rubyracer to find v8: gem install therubyracer -- --with-v8-dir=/usr/local/opt/v8@3.15
    • needs v8 installed by brew as: brew install v8-315

If you are building the PDF documentation, this requires wkhtmltopdf. You can download it from here or use the usual apt-get / yum / brew.

Seeing the Website and Docs

To build and see most of the documentation, run this command in this folder:

jekyll serve

This will start up a local web server. The URL is printed by Jekyll when the server starts, e.g. http://localhost:4000/ . The server will continue to run until you press Ctrl+C. Modified files will be detected and regenerated (but that might take up to 1m). Add --no-watch argument to turn off regeneration, or use jekyll build instead to generate a site in _site without a server.

This does not generate API docs and certain other material; see the notes on _build/build.sh below for that.

Project Structure

Note that there are two interlinked micro-sites in this project:

  • /website: this contains the main Brooklyn website, including committer instructions, download instructions, and "learn more" pages; this content has only one instance on the live website, and as changes are published they replace old content

  • /guide: this contains the user guide and information pertaining to a specific Brooklyn version, including code structure and API documentation; the live website contains a copy of the guide for each Brooklyn version, with the code coming from the corresponding branch in git

In addition note the following folders:

  • /style: contains JS, CSS, and image resources; on the live website, this folder is installed at the root and into archived versions of the guide.

  • /_build: contains build scripts and configuration files, and tests for some of the plugins

  • /_plugins: contains Jekyll plugins which supply tags and generation logic for the sites, including links and tables of contents

  • /_layouts: contains HTML templates used by pages

  • /_includes: contains miscellaneous content used by templates and pages

Jekyll automatically excludes any file or folder beginning with _ from direct processing, so these do not show up in the _site folder (except where they are embedded in other files).

A word on branches: The /website folder can be built against any branch; typically changes are made and published from master, to ensure that all versions are listed correctly. In contrast the /guide folder should be updated and built against the branch for which instructions are being made, e.g. master for latest snapshot updates, or 0.7.0-M2 for that milestone release. It is permitted to make changes to docs (and docs only!) after a release has been made. In most cases, these changes should also be made to master.

Website Structure

The two micro-sites above are installed on the live website as follows:

  • /: contains the website
  • /v/<version>: contains specific versions of the guide, with the special folder /v/latest containing the recent preferred stable/milestone version

The site itself is hosted at brooklyn.apache.org with a CNAME record from brooklyn.io.

Content is published to the site by updating an Apache subversion repository, brooklyn-site-public at https://svn.apache.org/repos/asf/brooklyn/site. See below for more information.

Building the Website and Guide

For most users, the jekyll serve command described above is sufficient to test changes locally. The main reason to use the build scripts (and to read this section) is to push changes to the server (requires Apache Brooklyn commit rights), or to test generated content such as API docs.

The build is controlled by config files in _build/ and accessed through _build/build.sh. There are a number of different builds possible; to list these, run:

_build/build.sh help

The normal build outputs to _site/. The three builds which are most relevant to updating the live site are:

  • website-root: to build the website only, in the root
  • guide-latest: to build the guide only, in /v/latest/
  • guide-version: to build the guide only, in the versioned namespace e.g. /v/<version>/

There are some others, including test-both, which apply slightly different configurations useful for testing. Supported options beyond that include --serve, to start a web browser serving the content of _site/, and --skip-javadoc, to speed up the build significantly by skipping javadoc generation. A handy command for testing the live files, analogous to jekyll serve but with the correct file structure, and then checking links, is:

_build/build.sh test-both --skip-javadoc --serve

And to run link-checks quickly (without validating external links), use:

htmlproof --href_ignore "https?://127.*" --alt_ignore ".*" --disable_external _site

Preparing for a Release

When doing a release and changing versions:

  • Before branching:
    • Change the brooklyn-stable-version variable in _config.yml
    • Update website/meta/versions.md with a bit of info on this release
  • In the branch, with change-version.sh run (e.g. from N.SNAPSHOT to N)
  • Ensure the guide/start/release-notes.md file is current
  • Build and publish website-root, guide-latest, and guide-version
  • In master, with change-version.sh run (e.g. to N+1-SNAPSHOT)
    • Clear old stuff in the guide/start/release-notes.md file
    • Optionally build and public guide-version

Publishing the Website and Guide

The Apache website publication process is based around the Subversion repository; the generated HTML files must be checked in to Subversion, whereupon an automated process will publish the files to the live website. So, to push changes the live site, you will need to have the website directory checked out from the Apache subversion repository. We recommend setting this up as a sibling to your brooklyn git project directory:

# verify we're in the right location and the site does not already exist
ls _build/build.sh || { echo "ERROR: you should be in the docs/ directory to run this command" ; exit 1 ; }
ls ../../brooklyn-site-public > /dev/null && { echo "ERROR: brooklyn-site-public dir already exists" ; exit 1 ; }
pushd `pwd -P`/../..

svn --non-interactive --trust-server-cert co https://svn.apache.org/repos/asf/brooklyn/site brooklyn-site-public

# verify it
cd brooklyn-site-public
ls style/img/apache-brooklyn-logo-244px-wide.png || { echo "ERROR: checkout is wrong" ; exit 1 ; }
export BROOKLYN_SITE_DIR=`pwd`
popd
echo "SUCCESS: checked out site in $BROOKLYN_SITE_DIR"

With this checked out, the build.sh script can automatically copy generated files into the right subversion sub-directories with the --install option. (This assumes the relative structure described above; if you have a different structure, set BROOKLYN_SITE_DIR to point to the directory as above. Alternatively you can copy files manually, using the instructions in build.sh as a guide.)

A typical update consists of the following commands (or a subset), copied to ${BROOKLYN_SITE_DIR-../../brooklyn-site-public}:

# ensure svn repo is up-to-date (very painful otherwise)
cd ${BROOKLYN_SITE_DIR-../../brooklyn-site-public}
svn up
cd -

# versioned guide, safe for snapshots, relative to /v/<version>/
_build/build.sh guide-version --install

# main website, if desired, relative to / 
_build/build.sh website-root --install

# this version as the latest guide, if desired, relative to /v/latest/
_build/build.sh guide-latest --install

(If HTML-Proofer find failures, then fix the links etc. Unfortunately, the javadoc build gives a lot of warnings. Fixing those is not part of this activity).

You can then preview the public site of localhost:4000 with:

_build/serve-public-site.sh

Next it is recommended to go to the SVN dir and review the changes using the usual svn commands -- status, diff, add, rm, etc. Note in particular that deleted files need special attention (there is no analogue of git add -A!). Look at deletions carefully, to try to avoid breaking links, but once you've done that these commands might be useful:

cd ${BROOKLYN_SITE_DIR-../../brooklyn-site-public}
svn add * --force
export DELETIONS=$( svn status | sed -e '/^!/!d' -e 's/^!//' )
if [ ! -z "${DELETIONS}" ] ; then svn rm ${DELETIONS} ; fi

Then check in the changes (probably picking a better message than shown here):

svn ci -m 'Update Brooklyn website'

The changes should become live within a few minutes.

SVN commits can be slow, particularly if you've regenerated javadoc. (The date is included in all javadoc files so the commands above will cause all javadoc to be updated.) Use _build/build.sh guide-version --install --skip-javadoc to update master while re-using the previously installed javadoc. That command will fail if javadoc has not been generated for that version.

More Notes on the Code

Plugins

We use some custom Jekyll plugins, in the _plugins dir:

  • include markdown files inside other files (see, for example, the *.include.md files which contain text which is used in multiple other files)
    • read_jekyll will read a file and jekyll-process it (eg for nested {% ... readj calls; it's just like include_relative but with two differences: when readj is used nested read_jekyll calls are relative to the included file where it is contained, rather than relative to the root md file which made the first read_jekyll call (so in general it's more intuitive than include_relative), and include_relative processes front matter whereas read_jekyll mangles it
    • read_literal is like read_jekyll but without the jekyll/tag processing
    • read will autodetect based on filename -- jekyll for *.md and literal for all others
  • generate the site structure / menu objects
  • parse JSON which we can loop over in our markdown docs (to build up models; previously used for the TOC in the guide, but now replaced with site_structure)
  • trim whitespace of ends of variables

Versions

Archived versions are kept under /v/ in the website. New versions should be added with the appropriate directory (_build/build.sh guide-version above will do this).
These versions take their own copy of the style files so that changes there will not affect future versions.

A list of available versions is in website/meta/versions.md.