I’ve been working on an updated version of the Soil Quality website for a little while now, and recently needed to deploy the site to a staging server for testing. I ordered up a 1/4Gb Accelerator from Joyent, configured it, and deployed the app, just like I’ve done with 20+ other sites, and BOOM, straight into a brick wall.

I opened up the log files and saw this error:

** Daemonized, any open files are closed.  Look at /tmp/soilquality-mongrel.8200.pid and log/mongrel.8200.log for info.
** Starting Mongrel listening at 127.0.0.1:8200
** Define INLINEDIR or HOME in your environment and try again

Wonderfully descriptive I know, but something I’d never run into before. So, fire the the google and it turns out it’s a common error, with a common fix: just put

ENV['INLINEDIR'] = RAILS_ROOT + "/tmp"

into your config/environment.rb file, make sure the directory exists and that it’s writeable by the mongrel, restart, and away you go.

But of course, that didn’t fix it, did it.

Much hair pulling ensued. I finally enlisted the help of Darcy Laycock and together we managed to track the problem down to the attachment_fu plugin, which was causing the problem as the mongrel process booted. OK, so now we knew where the problem lay, but what was causing it.

It turned out to be the ImageScience image processor, or more specifically the way the attachment_fu plugin and ImageScience work together ON A 2ND GENERATION JOYENT ACCELERATOR – eg the ones that use pkgsrc. Didn’t seem to cause the problem on an ubuntu machine, nor on one of the older BlastWave based Accelerators. I’m not sure why as of yet, as I was more worried about getting some sleep last night when we managed to fix the problem.

And the fix? Basically, strip ImageScience out of attachment_fu. Remove

vendor/plugins/attachment_fu/lib/technoweenie/attachment_fu/processors/image_science_processor.rb

and remove “ImageScience” from this line

@@default_processors = %w(ImageScience Rmagick MiniMagick Gd2 CoreImage)

in this file

vendor/plugins/attachment_fu/lib/technoweenie/attachment_fu/attachment_fu.rb

Like I said, I have no real idea why this is happening at the moment, but I’ll try and work it out and update this post if I get anywhere.

Once again, thanks to Darcy for A) being a sounding board, B) helping fix the problem and C) being up and available when I needed him

rake rails:freeze:edge TAG=rel_2-0-1

and the appropriate version would be copied into your vendor/rails directory.

Now if you do that, rake downloads a zip of the edge release. Which is fine and all, but sometimes you don’t want to be on edge … like in any production site.

So, I found a screencast that goes through the process, but I thought I’d actually put the text into a post, mainly for my own reference more than anything else.

$ rails path_to_app

$ cd path_to_app
$ git init
$ git submodule add git://github.com/rails/rails.git vendor/rails

At this point, git will effectively clone the repository, so that you can then choose one of the branches to “freeze” to. Type “git tag” to get a list of all the available tagged branches. Choose the one you want and type

$ git checkout v2.1.0

And that’s it. Slightly more involved than the old way, but still none too shabby.

Now, I should probably also say that I already had my site up and running using subversion, capistrano and my accelerator, so this article isn’t necessarily going to help with getting everything setup the first time. For that, you probably need to read this wiki entry.

Things You Will Need

• A github.org account
• A dedicated Joyent Accelerator (I have no idea how to do all of this on a shared accelerator. Sorry.)
• Also, I’m really only talking about RubyOnRails apps here … not too sure how applicable a lot of this is to other frameworks (it will probably help at least.)

Getting started

Assuming that you have your project in a git repository, and have a github account (and obviously an Accelerator) we can start.

Compiling git on your accelerator

Unfortunately, the first step in the process was, for me at least, a total nightmare. I’m not the biggest unix-head by any stretch, but I can do some basic tasks with a degree of proficiency. Unfortunately, I went into a dark place trying to get git compiled. One thing to note is that I’m talking about setting up the git client here, not a git server. Because capistrano executes scripts on your remote server, you need to have a copy of the client software setup for capistrano to call.

So what did I do? There are a couple of helpful threads on the Joyent Forums:

• Got git?
• Git and Cap 2.0
• Installing git

Hopefully those threads will put you onto the path of successfully compiling and installing git onto your Accelerator.

Setting up SSH keys with github and your accelerator

When you setup your account on github, you need to setup an SSH key for authentication. github has a really good tutorial on how to do this. I have a user defined on my accelerator that my website “runs” under, so what I did was to create a key for that user which gets stored into the ~/.ssh directory. I then added the contents of the id_rsa.pub key to my github account, which allows that user to access the repository.

Another tip: don’t forget your passphrase. It’s needed in the next step.

Configuring capistrano

Assuming that you have your capistrano deploy.rb file setup as outlined here there are a few changes that you will need to make to get things working with git.

I’m using Capistrano 2.2 at the moment. I don’t think it will work with earlier versions because of the relatively new git support.

Here’s my deploy.rb file:

  require 'erb'
  require 'config/accelerator/accelerator_tasks'

  set :application, "website"
  set :repository, "git@github.com:your_username/website.git" 

  default_run_options[:pty] = true
  set :domain, 'XX.XX.XX.XX' #Your Accelerators public IP address
  set :deploy_to, "/var/www/apps/#{application}"
  set :user, 'website_account_username'
  set :scm, :git
  set :scm_username, "github_username"
  set :scm_passphrase, "your passphrase here" 

  role :app, domain
  role :web, domain
  role :db,  domain, :primary => true

  set :server_name, "url.for.website"
  set :server_alias, "*.url.for.website" 

  # Example dependancies
  depend :remote, :command, :gem
  depend :remote, :gem, :money, '>=1.7.1'
  depend :remote, :gem, :mongrel, '>=1.0.1'
  depend :remote, :gem, :image_science, '>=1.1.3'
  depend :remote, :gem, :rake, '>=0.7'
  depend :remote, :gem, :BlueCloth, '>=1.0.0'
  depend :remote, :gem, :RubyInline, '>=3.6.3'

  ################################
  # Some tasks for the old server
  ################################

  task :after_deploy do
    # tasks to run after deploy
  end

  ################################
  # End tasks for the old server
  ################################

  deploy.task :restart do
    accelerator.smf_restart
    accelerator.restart_apache
  end

  deploy.task :start do
    accelerator.smf_start
    accelerator.restart_apache
  end

  deploy.task :stop do
    accelerator.smf_stop
    accelerator.restart_apache
  end

  after :deploy, 'deploy:cleanup'

It appears that the important line here is ’ default_run_options[:pty] = true ’. This means that capistrano can respond automatically for the request for the SSH Key passphrase that github replies with when you try to clone the repository.

If everything is working, you can type ‘cap deploy’ and it should all deploy nicely. If you get this error:


[err] Permission denied (publickey).


then there’s a problem with your SSH key and your settings on github. Make sure the key you copied into your github account is the public key for the SSH in your .ssh directory.

Hopefully, you’ll be up and running. If you have any tips, recommendations or corrections, leave a comment.

Things to do

Do this once


gem install mongrel_cluster


then in the application directory


capify .


then remove “require ‘deprec/recipes’ from the deploy.rb file

then put the following in to the deploy.rb file


namespace :deploy do
task :start, :roles => :app do start_mongrel_cluster end
task :stop, :roles => :app do stop_mongrel_cluster end
task :restart, :roles => :app do restart_mongrel_cluster end
end


A couple of years ago, Google Maps launched in the US and opened up a massive range of possibilities for programmers to develop cool “mashups”, which took information from one site, mashed it together with maps from google, and created a whole new website. They were all really cool, but unfortunately (at least for the non-US based of us) it was nothing more than something we could sit back and watch – which was kind of weird because the team of developers that built Google Maps was based in Sydney.

But finally the Australian version of google maps was released, and then recently integrated with Local Search, which means you can search on Pizza shotps near a particular address. Which is cool.

What is even cooler (if you’re a ruby developer), though, is the release of this: GeoKit. It integrates with all of the major geocoding/mapping services and provides a huge range of options and services. So now you can do stuff like this:

add_1=GeoKit::Geocoders::GoogleGeocoder.geocode("1 St Georges Terrace, Perth, Western Australia")
add_2=GeoKit::Geocoders::GoogleGeocoder.geocode("1 York St, Albany, Western Australia")

distance = add_1.distance_from(add_2, :units => :kms)

Which returns


389.248018478531


which is, of course, how far it is between the main streets of Perth and Albany in Western Australia.

Sawwweeeeeet.

Update

Well, that didn’t take long at all. A quick loop over the existing schools in the SchoolSeek database got the lat/long information for all the addresses. Then adding this line

acts_as_mappable default_units => :kms

to my Address model allows you to do this

@addresses = Address.find(:all,  rigin => "18 Bland St, Ashfield, New South Wales", :conditions => "distance < 10")

which gives me all of the addresses within 10Kms of 18 Bland St, Ashfield. Nice.

So, basically GeoKit let me add distance searching to the site within about 45 minutes (allowing 25 minutes for me to read through the examples and such!)

When looking at my own Rails code and that of the community as a whole, I often see places where certain Rails techniques could have been used, but weren’t. As much for my own memory as yours, I thought I’d list down some Rails tricks and tips that can make your application or code more efficient: