EC2 on Rails

Deploy a Ruby on Rails app on EC2 in five minutes

Main Page: https://ec2onrails.rubyforge.org

Code on Github: https://github.com/pauldowman/ec2onrails

EC2 on Rails is an Ubuntu Linux server image for Amazon’s EC2 hosting service that’s ready to run a standard Ruby on Rails application with little or no customization. It’s a Ruby on Rails virtual appliance.

If you have an EC2 account and can start EC2 instances you’re five minutes away from deploying your Rails app.

EC2 on Rails is opinionated software: the opinion is that for many rails apps the server setup can be generalized and shared the same way as the web application framework itself. For many people (Twitter, this isn’t for you) the server image can be treated the same way as other shared libraries. And if the day comes when your needs areunique enough that EC2 on Rails can’t be configured to work for you then you can bundle your own image from it or fork the build source and customize it.

But until then, why spend your time configuring servers?

Features of the EC2 image:

Ready to deploy a Rails app with little or no configuration of the server required
Optional Amazon Elastic Block Store (EBS) setup and support for MySQL database persistence.
Automatic backup of MySQL database to S3 (EBS enabled: snapshots taken every 2 hrs; Non-EBS enabled: full backup nightly + incremental backup using binary logs every 5 minutes if not using )
Capistrano tasks to customize the server image, archive and restore the database to/from S3, and more (available as a rubygem)
Mongrel_cluster behind Apache 2.2, configured according to
Coda Hale’s excellent guide
Ruby on Rails 2.1.0, 2.0.2 and 1.2.6
Ruby 1.8.6
MySQL 5
memcached
monit configured to monitor apache, mongrel, mysql, memcached, drive space and system load
Ubuntu 8.04 LTS “Hardy” base image built using Eric Hammond’s EC2 Ubuntu script
SSL support
Amazon AMI tools installed
Apache or Nginx web proxies
MySQL, Apache/Nginx, and syslog configured to use /mnt for data and logging so you don’t fill up EC2’s small root filesystem
Automatically archives Rails and Apache logs to S3 nightly.
32-bit and 64-bit images available (supports all instance types, small to extra large).
Created using a build file, full source is available (the EC2 on Rails script is run from Eric Hammond’s EC2 Ubuntu script)
Can be used as a clustered Rails app running on multiple instances
Automatically runs hourly, daily, weekly and monthly scripts if they exist in Rails application’s script directory
Local Postfix SMTP mail server (only available from within the instance, not listening on external network interfaces)

Using the image

This documentation will be improved soon, for now hopefully this covers the basics.

The current AMI id’s are:

ami-c9bc58a0 (32-bit)
ami-cbbc58a2 (64-bit)

I will keep these images available for as long as possible, they will not be deleted for at least a few years.

1. Install the gem

sudo gem install ec2onrails

2. Add the config files to your Rails app

You will need to place Capfile in the root of your rails folder, and put deploy.rb and s3.yml in the config folder.

Within your rails directory, run
wget -q -O Capfile http://github.com/pauldowman/ec2onrails/tree/master/examples/Capfile?raw=true wget -q -O config/deploy.rb http://github.com/pauldowman/ec2onrails/tree/master/examples/deploy.rb?raw=true wget -q -O config/s3.yml http://github.com/pauldowman/ec2onrails/tree/master/examples/s3.yml?raw=true

Be sure to customize those files and read the comments.

Also, use the hostname “db_primary” in your database.yml file. After running “cap ec2onrails:server:set_roles” it will resolve to the instance defined in your Capistrano “db” role.

3. Start up one or more instances of the image.

There is nothing EC2 on Rails-specific here yet (though soon there will be a Capistrano task to do this for you), if you’ve started EC2 instances before you can skip this section. Otherwise, I’m not going to lie, this part is complicated and will take a lot more than 5 minutes the first time.

Read the running an instance section in Amazon’s getting started guide.

For the AMI id’s of the current images do cap ec2onrails:ami_ids from within the app that you configured in the previous step (they’re also listed earlier on this page).

NOTE: Only use the images that match the current version of the gem.

Please see the change log for release notes, and see the list of open issues.

As is standard for public AMI’s, password-based logins are disabled. You log in with your own public/private keypair.

Most basic things can be configured automatically by the Capistrano tasks, but if you want to you can login by ssh as a user named “admin” (has sudo ability) or as “app” (the user that the app runs as, does not have sudo ability). The Capistrano tasks automatically use the app user to deploy the app, and the admin user for server admin tasks that require sudo.

IMPORTANT: Double-check your firewall settings. Be sure that you haven’t allowed public access to any ports other than TCP 22 and TCP 80 (and possibly TCP 443 if you’re going to enable HTTPS). If you’re using multiple instances, be sure to allow them network access to each other.

4. Copy your public key from the server to keep Capistrano happy

This is a workaround for a quirk in Capistrano. Technically all you should need to connect to the server is the private key file, the public key is on the server. But for some reason Capistrano requires that you have both the public key and the private key files together on the client.

There is a Capistrano task that tries to fix this for you. From within the root of your rails app do:

cap ec2onrails:get_public_key_from_server

Note, this will only work if you have an external ssh command in the path, it won’t work for most Windows users.

5. Deploy the app with Capistrano

Now that the gem is installed, your deploy.rb is configured and you can start and stop EC2 instances, this is the only thing you’ll need to do from now on.

cap ec2onrails:setup
cap deploy:cold

Yes, it’s that easy! The setup task will set the server’s timezone, install any gems and Ubuntu packages that you specified in the config file, and create your database.

That’s it, your app is now running on EC2!!

Capistrano tasks

Capistrano is the most commonly used Rails deployment tool. It comes with many standard “tasks”, and the EC2 on Rails gem includes Capistrano tasks specifically for configuring the server instance.

Capistrano is run from the command-line using the “cap” command, with a task name given as an argument.

Commonly-used tasks

You’ll mostly need just the following Capistrano tasks:

cap ec2onrails:ami_ids Shows the AMI id’s of the images that match the current version of the gem.

cap ec2onrails:setup This task configures a newly-launched instance. This is the first thing you should do after starting a new instance. It can be run more than once without ill effect. After running “cap ec2onrails:setup” the next thing to do is run “cap deploy:cold”

cap ec2onrails:db:enable_ebs This task will move the primary mysql database onto an Amazon Elastic Storage Block (EBS) volume. You can call this task with the optional SIZE parameter defined (defaults to 10 gigs) like
```
cap ec2onrails:db:enable_ebs SIZE=10
```

You should then specify your own volume (or the one created by this task) in your capistrano deploy.rb file like so:

role :db, “ec2-xx-xxx-xx-xxx.compute-1.amazonaws.com”, :primary => true, :ebs_vol_id => ‘vol-12345abc’

NOTE MySQL EBS is not enabled by default. You may call this task at anytime to move your MySQL over to EBS, but just make sure you keep track of the volume-id that is printed out by this task and use it to modify your deploy.rb file

cap ec2onrails:server:set_roles Customizes each instance for it’s role(s) (as defined in your Capistrano deploy.rb file). Run this after starting or stopping instances. For now this just makes sure that only the appropriate services (Apache, Mongrel, and/or MySQL) are running. Eventually this will customize settings for the running services also. Note that an instance can have more than one role. If there’s only one instance it will have all roles.

Note that due to the way that Capistrano works all tasks are run against all hosts that are currently defined in the deploy.rb file. So if you start a new instance then add it to your deploy.rb you will need to run “cap ec2onrails:setup” again which will be run on all existing instances.

Database management tasks

cap ec2onrails:db:archive Archive the MySQL database to the bucket specified in your deploy.rb. This is for archiving a snapshot of your database into any S3 bucket. For example, you might want to do this before deploying.

cap ec2onrails:db:restore Restore the MySQL database from the bucket specified in your deploy.rb For example, I use this to restore the current production data (from my actual production backup bucket) onto a staging server that has the current production version of my app. I then deploy the new version which tests migrations exactly as they’ll run on the production server.

To get a full list of the Capistrano tasks at any time type cap -T from with your rails app root.

Building the image

Building the image is not required, most people will simply use the prebuilt public image, but there is also a build script that builds the image. It’s meant to be called by Eric Hammond’s EC2 Ubuntu script.

Here is how you go about setting it up:

Notes

amazon-keypair is your amazon keypair, which in this case has has had the public and private keys copied into the ~/.ssh folder
we need to copy the ec2onrails source folder onto the image instance
HOST_PUBLIC_INSTANCE is the hostname of the instance that you will soon be creating. run ec2-describe-instances to find this value. It will look something like ec2-xxx-xxx-xxx-xxx.compute-x.amazonaws.com

Getting started…

launch one of Eric’s default images (see Eric Hammond’s page for the most recent ami-images): ec2-run-instances ami-179e7a7e -k amazon-keypair (for a 64-bit image, launch: ec2-run-instances ami-f89d7991 -k amazon-keypair --instance-type m1.large)
lets get your amazon keypair up onto the instance: scp -i ~/.ssh/id_rsa-amazon-keypair ~/.ec2/{cert,pk}-*.pem root@HOST_PUBLIC_ADDRESS:/mnt/
ditto for getting the ec2onrails server code up onto your new instance: scp -rp -i ~/.ssh/id_rsa-amazon-keypair ~/your/copy/of/ec2onrails/. root@HOST_PUBLIC_ADDRESS:/mnt/ec2onrails
time to log into your instance: ssh -i ~/.ssh/id_rsa-amazon-keypair root@HOST_PUBLIC_ADDRESS
get the most recent ec2ubuntu script: curl -Lo /mnt/ec2ubuntu-build-ami http://ec2ubuntu-build-ami.notlong.com

time to get the instance ready for ec2onrails. The script will end with a command you need to run to finish creating an amazon ami-image.

bash /mnt/ec2ubuntu-build-ami \ 

—codename hardy \ 

—bucket YOURNAME_ec2onrails \ 

—prefix ec2onrails \ 

—user YOUR_AMAZON_USER_ID \ 

—access-key AMAZON_ACCESS_KEY \ 

—secret-key AMAZON_SECRET_KEY \ 

—private-key /mnt/pk-.pem \ 

-cert /mnt/cert.pem \ 

—script /mnt/ec2onrails/server/build-ec2onrails.sh

Mailing lists

There are two Google groups, one for announcements (usually just new release announcements) and one for discussion.

Comments

Comments are welcome. Send an email to Paul Dowman or to the Google group. If you find bugs please file them here or send me an email.

Change log

See the change log.

How to submit patches

Pleae read the 8 steps for fixing other people’s code. The source code can be checked out anonymously using:



svn checkout http://ec2onrails.rubyforge.org/svn/trunk ec2onrails