deployment

Inherit local-only options with 'parent'

Parent is not working across files. Filed support request: http://drupal.org/node/1262230

In an aliases file that lives with your project, so for our SDL project, sdl.aliases.drushrc.php, put in all settings that are true no matter who is using Drush and from where. This is the settings for an externally hosted development collaboration site.

ben@ubuntu:~/code/oa/web$ rake prod vlad:update
(in /home/ben/code/oa)
Everything up-to-date
Already on 'master'
Your branch is ahead of 'agaric/master' by 10 commits.
From git.agaric.com:/srv/git/oa
 * branch            master     -> FETCH_HEAD
error: unable to unlink old 'web/sites/default/default.settings.php' (Permission denied)
Updating 183cb44..744b38c
rake aborted!
execution failed with status 1: ssh oa@sojourner.mayfirst.org cd /var/local/drupal/oa && git checkout master && git pull agaric master

It would be nice to package our Drush aliases with our projects.

Where does drush look for aliases?

It is possible to have your site-specific aliases in sites/default, such as sites/default/aliases.drushrc.php.

Somewhat Related

• issue to have aliases only loaded when in context: http://drupal.org/node/1132906

• Helper function to generate an alias record for the specified site: drush alias-add @aliasname

As when working on any Agaric project that may not have had maintenance done on it in a while, update the Rakefile to be in line with current processes.

First, make sure your Vlad checkout is up-to-date. Go to it cd ~/vlad (in my case, i symlinked Vlad's true location to my home folder so he can find it. On my system Vlad's true location is /usr/local/lib/site_ruby/1.8/i686-linux/vlad/) and git pull.

cd ~/code/projectname
mv Rakefile RakefileOLD
cp /home/ben/vlad/Rakefile Rakefile
diff -up

This is the once per server or computer setup required for Deploying the Agaric Way.

Requirements

To ease deployment the remote rake tasks from Vlad the Deployer are used. Before using it the following software must be installed:

Stefan figured out how to include the config/deploy.rb from somewhere else.

Usually the ruby load path includes something like /usr/local/lib/site_ruby/1.8//. You can verify that by typing "puts $LOAD_PATH" in an irb session.

Just move the config folder into it.

This should be done by cloning the central git repository of this folder:

cd /usr/local/lib/site_ruby/1.8/x86_64-linux/
git clone -o agaric git.agariclabs.org:/srv/git/agaric/vlad.git vlad

How we would probably want this to work is that "rake prod vlad:drush XXX" call "ssh prodservername drush -r /path/to/project/drupal XXX"

For instance: ssh prodservername drush -r /path/to/project/drupal enable bad_judgement

An extremely abbreviated summary of developer's summit deployment breakout.

Josh Koenig summarized the discussion so far and tried to put a direction on it: we want to separate configuration from code and have a couple approaches.

Greg Dunlap (heyrocker) of Deploy module disagreed completely: there is no separation of configuration and content in Drupal, there won't be, and there probably shouldn't be.

And lots of clients, what they really care about, is content deployment.

UUIDs in core would be huge.

Background

Where we come from

After journeying from Capistrano to Vlad we eventually use pure Rake tasks for our deployment. While Vlad was a great improvement it also had downsides. Lack of feedback from tasks made it hard to tell why they failed. Also the most interesting feature of those deployment tools is not required in our case: The ability to roll out to many servers at the same time.

General

• PDF of an old but excellent Jeff Eaton presentation
• A surprisingly informative deployment discussion on Stack Overflow, but best comment is Jeff Eaton's mostly reprising his session above.
• Fairly balanced 2008 overview of deployment options by Greg Dunlap (author of Deploy module).

Plain Everything-in-code