[Capistrano 3] - Using "extended server syntax" in deploy projects.

[Joe R]

For political reasons, I've been asked to use Capistrano to deploy Chef Solo to a list of servers and interact with an API.

I'm new to Cap and new(ish) to Ruby. I'm looking for some help re: best way to use server definitions in Cap 3.

From Cap 3 docs:

# Extended Server Syntax
    # ======================
    # This can be used to drop a more detailed server
    # definition into the server list. The second argument
    # is something that quacks like a hash and can be used
    # to set extended properties on the server.
    server 'example.com', roles: %w{web app}, my_property: :my_value

vs

# Simple Role Syntax
    # ==================
    # Supports bulk-adding hosts to roles, the primary
    # server in each group is considered to be the first
    # unless any hosts have the primary property set.
    role :app, %w{example.com}
    role :web, %w{example.com}
    role :db,  %w{example.com}

Which was the de facto standard in Cap 2.

My server list looks like this:

server '10.0.1.1', roles: [:hdp_ambari, :hdp_namenode, :hdp_hbase, :hdp_monitor_server, :hdp_zookeeper, :hdp_ganglia], internal:'ip-10-0-1-1.ec2.internal'
server '10.0.1.2', roles: [:hdp_namenode_secondary, :hdp_datanode, :hdp_zookeeper, :hdp_ganglia], internal:'ip-10-0-1-2.ec2.internal'

I want to set :options to a hash of keys and internals corresponding to a given role. What's the best way to do this?

So desired output something like:

set :options, {
  default_filesystem: "hdfs://ip-10-0-1-1.ec2.internal:8020",
  hbase_master: 'ip-10-0-1-2.ec2.internal',
  job_history_server: 'ip-10-0-1-1.ec2.internal',
  namenode: 'ip-10-0-1-1.ec2.internal',
  namenode_secondary: 'ip-10-0-1-2.ec2.internal',
  yarn_node: 'ip-10-0-1-3.ec2.internal',
  zookeepers: ['ip-10-0-1-1.ec2.internal', 'ip-10-0-1-2.ec2.internal']
}

For which I was thinking...

task :set_configuration_options do
  # Fetching custom server.properties via http://tinyurl.com/ldkm7ov
  on roles(:hdp_namenode), in: :parallel do |server|
    namenode_internal = p server.internal
    filesystem_internal = "hdfs://#{namenode_internal}:8020"
  end


options = {
  default_filesystem: "#{filesystem_internal}",
  namenode: "#{namenode_internal}"
}
end


set :options, options

But there's got to be a cleaner, better way, right? I'm having a tough time finding good examples or docs on this.

Relevant details:

Versions:

• Ruby - 2.0.0p481
• Capistrano - 3.0.1
• Rake / Rails / etc - 10.1.0, 0.9.6

Platform:

• Working on.... Mac OSX 10.9.2
• Deploying to... CentOS 6.4

Logs:

• Not relevant.  Asking for general advice, best practices re: extended server syntax.

Files:

• Capfile, deploy.rb, and stage attached.