Letters from a Maladroit

Setting up a rails development environment with Vagrant and Ansible

Over the weekend, I started learning Ruby on Rails 4. I’ve used Ruby on Rails 2 in the past, but only for a community college course and a couple toy projects. One reason for giving Ruby on Rails another shot is simply because of its continued popularity. After working with Rails for a few days now, I can see why. The automation tools, testability, and number of user contributed gems make development go a lot faster. I don’t have to rebuild the wheel in regards to automation/scaffolding/migration/etc, and this is a huge win for my increasing “laziness” as a developer.

The first thing I needed to do was set up a rails development environment with Vagrant and Ansible. It took about four hours to get things right, but the reproducibility and environment separation is worth the effort.

Vagrant Setup

For the Vagrant box, I wanted to use Ubuntu 14 since it is the current LTS edition. One of the troubles with Vagrant boxes is that they are not all set up the same. I considered the following boxes:

The following is my Vagrantfile:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
# -*- mode: ruby -*-
# vi: set ft=ruby :

VAGRANTFILE_API_VERSION = "2"

Vagrant.configure(VAGRANTFILE_API_VERSION) do |config|

  config.vm.box = "chef/ubuntu-14.04"

  config.vm.network "forwarded_port", guest: 3000, host: 3000

  config.vm.provider "vmware_fusion" do |vf|
    vf.vmx["memsize"] = "1024"
  end

  config.vm.provision :ansible do |ansible|
    ansible.playbook = "playbook.yml"
  end

end

The Vagrantfile is straightforward. I forward port 3000 so I can view the development website. I also set the RAM at 1GB. Currently my laptop has 8GB RAM and I usually run Chrome, Evernote, Spotify, Scrivener, Sublime Text 2, and iTerm2 with no problems. I think a smaller amount of RAM can be used in this case, but I highly recommend getting lots of RAM if you plan on developing in VMs. I used to have 16GB RAM installed, but then one of the RAM sticks died.

That’s it for the Vagrantfile.

Ansible Setup

I recently started using Ansible. For the past 2-3 years I have been using Puppet with Vagrant. The downsides were:

  1. Managing modules. Somewhat solved with Puppet Librarian. It feels like a convoluted solution though.
  2. Needed to find Vagrant boxes with Puppet installed or build my own.
  3. The Puppet configuration language is better than BASH, but still not the most pleasant language to work with.

I haven’t used any advanced features of Ansible yet, but here’s what I like so far.

  1. Don’t need Ansible installed on the Vagrant box, although this is somewhat negated by obtuse SSH errors that can occur if the Vagrant box is set up in a certain way.

  2. I like the integrated Ansible Galaxy that allows “roles” to be downloaded and used. It’s cleaner than Puppet Librarian, but there is the issue of needing to automate the installation of roles for other development machines. Ansible probably already has a solution to this. I just need to look into it.

  3. I like the YML configuration. It makes a lot of sense for my relatively basic use cases so far.

Here is my playbook.yml file:

---
- hosts: all
  sudo: true
  roles:
    - role: znzj.rbenv
      rbenv_ruby_version: 2.2.1
  vars:
    - gem_path: "{{rbenv_root}}/shims/gem"
  tasks:
    - apt_repository: repo=ppa:chris-lea/node.js
    - apt: update_cache=yes
    - apt: name=nodejs
    - apt: name=vim

    - name: Install ruby dependencies
      apt: name={{item}}
      with_items:
        - git-core
        - curl
        - zlib1g-dev
        - build-essential
        - libssl-dev
        - libreadline-dev
        - libyaml-dev
        - libsqlite3-dev
        - sqlite3
        - libxml2-dev
        - libxslt1-dev
        - libcurl4-openssl-dev
        - python-software-properties
        - libffi-dev

    - name: Install Postgres dependencies
      apt: name={{item}}
      with_items:
        - postgresql
        - postgresql-contrib
        - libpq-dev
        - python-dev
        - python-pip
    - pip: name=psycopg2

    - name: Allow vagrant to sudo as postgres
      lineinfile:
        dest: /etc/sudoers
        line: vagrant ALL=(postgres) ALL
        validate: 'visudo -cf %s'

    - name: Create Helloworld user
      postgresql_user: name=helloworld password=helloworld role_attr_flags=CREATEDB,SUPERUSER
      sudo_user: postgres

    - name: Create Critiqual user
      postgresql_user: name=critiqual password=critiqual role_attr_flags=CREATEDB,SUPERUSER
      sudo_user: postgres

    - name: Install rails
      gem:
        name: rails
        version: 4.2.0
        state: present
        user_install: no
        executable: "{{gem_path}}"

This setup is somewhat based on the instructions found at https://gorails.com/setup/ubuntu/14.10.

Here are some of the issues I ran into during setup.

Gem path issue

I used a “role” to install rbenv. This was the suggested way to install the latest version of ruby. The other option was to use rvm. I had a bad experience with rvm three or four years ago and have no idea how to remove it from my laptop. This is a big reason why I only do development in virtual machines nowadays.

One issue that I had was that Ansible kept telling me that the gem executable did not exist. Apparently the solution is to provide the gem path via the executable parameter.

Another thing to be aware of is that by default, the gem will be installed in the root user’s home directory. One solution is to probably use sudo_user: vagrant. Alternatively you can do a global install with user_install: no, which is what I ended up doing. Note that you will need to delete the .gem folder from the root user’s home directory. For some reason Ansible cannot detect this change.

For reference, here’s what that snippet looks like:

- name: Install rails
  gem:
    name: rails
    version: 4.2.0
    state: present
    user_install: no
    executable: "{{gem_path}}"

PostgreSQL issue 1: Dependencies

There were a number of issues with installing PostgreSQL. The first was missing dependencies that prevented Ansible from running its PostgreSQL tasks. Ansible uses python, so it makes sense that I needed to install psycopg2, which is a PostgreSQL adapter for python. I installed this via pip, but ran into an error. I also needed to install libpq-dev via apt. It may have been easier to just do python-psycopg2 via apt. Then it would have handled the dependencies correctly. That’s definitely the one drawback of pip, the obscure error messages when a dependency is missing.

PostgreSQL issue 2: Can’t connect to psql

After resolving the first issue, I was still unable to execute the PostgreSQL tasks to create database users.

Specifically I got this message:

Sorry, user vagrant is not allowed to execute '/usr/bin/psql' as postgres on vagrant.vm.

Turns out the following command was not working: sudo -u postgres psql. This I believe was due to how the vagrant user was set up in this Vagrant box.

My eventual solution was to edit the sudoers file and add the correct privileges.

- name: Allow vagrant to sudo as postgres
  lineinfile:
    dest: /etc/sudoers
    line: vagrant ALL=(postgres) ALL
    validate: 'visudo -cf %s'

Also you need to make sure to add sudo_user: postgres to be able to login as the postgres admin/superuser account when running the postgresql_user task.

PostgreSQL issue 3: Rails can’t create test, development, production databases

Initially I created database users without a password, but I was unable to connect to the database via psql. It always checked for a password. There seems to be a way to configure PostgreSQL to not check for a password, but the pg_hba.conf file needs to be altered.

In the end it was simpler to just use passwords with the database users. This required hardcoding the passwords into playbook.yml, which I think this is OK for development. Ansible also has a feature called lookups that can keep passwords out the playbook.yml file, but I didn’t use it here.

PostgreSQL issue 4: Can’t seed test database with fixtures

I had trouble running the Rails unit tests because the fixtures would not install correctly when multiple related tables were involved. This was due to a foreign key issue, specifically, Rails needs to disable foreign key integrity checks when installing fixtures. In PostgreSQL, this requires the SUPERUSER privilege apparently.

Source: http://stackoverflow.com/questions/28046415/loading-rails-fixtures-in-a-specific-order-when-testing