Create Zombie process on ruby 2.1.0 after start

[boris]

I'm using rbenv ruby 2.1.0, bluepill 0.0.68 with this pill file:

Bluepill.application("resque") do |a|
  a.working_dir = "/path/to/current"
  a.process("resque") do |p|
    p.start_command = "bundle exec rake workers:start RAILS_ENV=production WORKERS_ENV=true"
    p.pid_file = "/path/to/shared/pids/workers.pid"
    p.uid = "deploy"
    p.gid = "deploy"
  end
end

Everytime I run a bluepill restart resque, the process restart properly but it creates a zombie child process: http://cl.ly/image/0V3o0U1V3015

[jschneiderhan]

I'm seeing this too on 0.0.69, although with a different configuration (unicorn - deamonize=true), and I think I know what's going on, at least in my instance.

The Bluepill::System#execute_blocking method forks https://github.com/bluepill-rb/bluepill/blob/v0.0.69/lib/bluepill/system.rb#L154 and Exec's the process command (in my case unicorn).

The parent calls waitpid() on the child https://github.com/bluepill-rb/bluepill/blob/v0.0.69/lib/bluepill/system.rb#L194. If this waitpid() call is not made, the child stays in the zombie state and is never released.

The Bluepill::Process class calls #execute_blocking in a few places, such as https://github.com/bluepill-rb/bluepill/blob/v0.0.69/lib/bluepill/process.rb#L296, but when it does so, it does within a with_timeout block.

If the timeout value is reached before the appicatioin process (unicorn) completes, waitpid https://github.com/bluepill-rb/bluepill/blob/v0.0.69/lib/bluepill/system.rb#L194 is never called, so the child is a zombie.

I ended up solving this by increasing my start_grace_time value.