Zero downtime restart of unicorn/ rails - ruby-on-rails

Question

In our deploy scripts we use the following snippet to restart unicorn:
desc "Zero downtime restart of Unicorn"
task :restart do
run "kill -s USR2 unicorn_pid"
end
The master process forks, starts the new workers and then kills the old one. But now it seems the new master kills the old master and takes over any new connections before the new children are fully up. As we disabled preloading of the app using preload_app false the new workers take around 30 - 60 seconds to start. During this time the new connections/ the website hangs. How to avoid this, so making the new master only take over once the new children are fully up and ready to server requests? :)
Update:
My unicorn.rb looks like this:
# name of application
application = "myapp"
# environment specific stuff
case ENV["RAILS_ENV"]
when "integration", "staging"
worker_processes 1
when "production"
worker_processes 4
else
raise "Invalid runtime environment '#{ENV["RAILS_ENV"]}'"
end
# set directories
base_dir = "/srv/#{application}/current"
shared_path = "/srv/#{application}/shared"
working_directory base_dir
# maximum runtime of requests
timeout 300
# multiple listen directives are allowed
listen "#{shared_path}/unicorn.sock", :backlog => 64
# location of pid file
pid "#{shared_path}/pids/unicorn.pid"
# location of log files
stdout_path "#{shared_path}/log/unicorn.log"
stderr_path "#{shared_path}/log/unicorn.err"
# combine REE with "preload_app true" for memory savings
# http://rubyenterpriseedition.com/faq.html#adapt_apps_for_cow
preload_app false
if GC.respond_to?(:copy_on_write_friendly=)
GC.copy_on_write_friendly = true
end
before_exec do |server|
# see http://unicorn.bogomips.org/Sandbox.html
ENV["BUNDLE_GEMFILE"] = "#{base_dir}/Gemfile"
end
before_fork do |server, worker|
# the following is highly recomended for "preload_app true"
if defined?(ActiveRecord::Base)
ActiveRecord::Base.connection.disconnect!
end
if defined?(Sequel::Model)
Sequel::DATABASES.each{ |db| db.disconnect }
end
# This allows a new master process to incrementally
# phase out the old master process with SIGTTOU to avoid a
# thundering herd (especially in the "preload_app false" case)
# when doing a transparent upgrade. The last worker spawned
# will then kill off the old master process with a SIGQUIT.
old_pid = "#{server.config[:pid]}.oldbin"
if old_pid != server.pid
begin
sig = (worker.nr + 1) >= server.worker_processes ? :QUIT : :TTOU
Process.kill(sig, File.read(old_pid).to_i)
rescue Errno::ENOENT, Errno::ESRCH
# someone else did our job for us
end
end
end
after_fork do |server, worker|
if defined?(ActiveRecord::Base)
ActiveRecord::Base.establish_connection
end
end
I think the main problem is that there's no child_ready hook. The before_fork and after_hook are called "too early". I think I could add some logic in the after_fork hook to somehow detect when a child is ready...but I hope there's an easier solution? :)