Missing PID file when using a process manager to start delayed job

Question

I am using BluePill to manage processes such as delayed job for Rails. In the BluePill log, I am getting this message:

W, [2010-09-27T01:23:20.447053 #19441]  WARN -- : [fsg_distro:delayed_job] pid_file /srv/fsg_distro/shared/pids/delayed_job.pid does not exist or cannot be read
W, [2010-09-27T01:23:20.447368 #19441]  WARN -- : [fsg_distro:delayed_job] Executing start command: ruby script/delayed_job -e production start
I, [2010-09-27T01:23:20.469165 #19441]  INFO -- : [fsg_distro:delayed_job] Going from down => starting

So it's claiming that a pid file is missing, which it is, but shouldn't that pid be created when BluePill starts the Delayed Job process?

Update To be a bit more clear about this error, I can successfully run the command manually but Bluepill fails to run the start command. When I run it manually, it looks like this:

rails@george:/srv/fsg_distro/current$ /usr/bin/env RAILS_ENV=production /usr/bin/ruby /srv/fsg_distro/current/script/delayed_job start
delayed_job: process with pid 17564 started. 

When I run it with Bluepill it looks like this:

W, [2010-10-03T21:24:13.943136 #17326]  WARN -- : [fsg_distro:delayed_job] pid_file /srv/fsg_distro/shared/pids/delayed_job.pid does not exist or cannot be read
W, [2010-10-03T21:24:13.943391 #17326]  WARN -- : [fsg_distro:delayed_job] pid_file /srv/fsg_distro/shared/pids/delayed_job.pid does not exist or cannot be read
I, [2010-10-03T21:24:13.943811 #17326]  INFO -- : [fsg_distro:delayed_job] Going from starting => down
W, [2010-10-03T21:24:14.945274 #17326]  WARN -- : [fsg_distro:delayed_job] pid_file /srv/fsg_distro/shared/pids/delayed_job.pid does not exist or cannot be read
W, [2010-10-03T21:24:14.945495 #17326]  WARN -- : [fsg_distro:delayed_job] pid_file /srv/fsg_distro/shared/pids/delayed_job.pid does not exist or cannot be read
W, [2010-10-03T21:24:14.945826 #17326]  WARN -- : [fsg_distro:delayed_job] Executing start command: /usr/bin/env RAILS_ENV=production /usr/bin/ruby /srv/fsg_distro/current/script/delayed_job start
W, [2010-10-03T21:24:15.049261 #17326]  WARN -- : [fsg_distro:delayed_job] Start command execution returned non-zero exit code:
W, [2010-10-03T21:24:15.049491 #17326]  WARN -- : [fsg_distro:delayed_job] {:stderr=>"", :exit_code=>1, :stdout=>""} 
I, [2010-10-03T21:24:15.049947 #17326]  INFO -- : [fsg_distro:delayed_job] Going from down => starting 

My pill looks like this:

APP_ROOT='/srv/fsg_distro'
RAILS_ROOT='/srv/fsg_distro/current'
RAILS_ENV='production'
RUBY_EXEC='/usr/bin/ruby'

Bluepill.application("fsg_distro", :log_file => "/srv/fsg_distro/shared/log/bluepill.log") do |app|
  app.process("delayed_job") do |process|
    process.working_dir = RAILS_ROOT

    process.start_grace_time    = 30.seconds
    process.stop_grace_time     = 30.seconds
    process.restart_grace_time  = 30.seconds

    process.start_command = "/usr/bin/env RAILS_ENV=#{RAILS_ENV} #{RUBY_EXEC} #{RAILS_ROOT}/script/delayed_job start"
    process.stop_command  = "/usr/bin/env RAILS_ENV=#{RAILS_ENV} #{RUBY_EXEC} #{RAILS_ROOT}/script/delayed_job stop"

    process.pid_file = "#{APP_ROOT}/shared/pids/delayed_job.pid"
    process.uid = "deploy"
    process.gid = "deploy"
  end
end 

And my delayed job script looks like this:

#!/usr/bin/env ruby
ENV['RAILS_ENV'] ||= 'production'

require File.dirname(__FILE__) + '/../config/environment'
require 'delayed/command'
Delayed::Command.new(ARGV).daemonize 

Bouton

2 questions seeing as how this is still unanswered. What permission would you need to write to the folder /srv/fsg_distro/shared/pids/ and does the user rails (from the machine george ?) have those permissions ? – Methylal

Bluepill and me logged in are the same user.... As long as the user is the owner or in the group, he can write to the pids directory – Bouton

@Bouton Same issue here, did you ever manage to solve this? – Inga

Answer 1

If you daemonize delayed_job (which you are) then you must generate a pid file for it yourself, or get delayed_job to do it for you, then bluepill expects to find the pid file once grace_start_time has passed.

If you run delayed_job in the foreground, and you tell bluepill to daemonize the process for you, then bluepill will generate the pid file.

You can't both monitor a self-daemonizing process and have bluepill create a pid file. It's either one or the other.

Lentil

Answer 2

I have no answer, but exactly the same question. I have tried many, many solutions--including creating a startup script for init.d and calling that from bluepill as described in another monit-based posting (https://mcmap.net/q/272630/-how-to-monitor-delayed_job-with-monit), but nothing has worked. Always there is a permissions issue with the PID file. How could this be the case when running as su?

I own the code, and share it with group www. Do I need to create a custom user just to run bluepill? If so, with what settings?

Stealer