延迟的工作人员不能并行运行

Question

(Re-added due to remove sensitive information)

We currently have 4 delayed job workers processing background jobs created by our main application. These are all spawned with their own instance via an upstart script. We would expect these 4 workers to work concurrently however they are now waiting on each other to complete before another job is processed. This is causing huge backlogs in our job pool as in essence only 1 job is being run at once rather than the desired 4.

Things I have tried but to no avail:

• Restart the workers
• Restart the servers (Both the delayed job & db server)
• Up the number of workers
• Run workers in separate pools for different job queues
• Spawn workers with both syntaxes ( script/delayed_job start & bundle exec rake jobs:work )

Is there something that we're missing here? How can we get the workers running in parallel again? Thanks in advance.

---

Upstart Script (this is run 4 times with a different index):

instance $index

script
# this script runs in /bin/sh by default
# respawn as bash so we can source in rvm
exec /bin/bash <<'EOT'
  # rvm
  source /home/user/.rvm/scripts/rvm

  cd /home/foo/bar/current
  exec /usr/bin/env RAILS_ENV=production bundle exec rake jobs:work
EOT
end script

• Ruby version: 2.3.1
• Rails version: 4.2.6
• Delayed Job version: 4.1.2
• Ubuntu version: 14.04.4 LTS
• Upstart version: 1.12.1
• Database: MariaDB 10.1.21

Answer 1

Open to correction on the following statements - I recently learned and implemented delayed_jobs on AWS.

In PROD I am using bin/delayed_job and daemons gem as per https://github.com/collectiveidea/delayed_job

My app is on AWS so the start/restart command is:

su -s /bin/bash -c "bundle exec bin/delayed_job -n 2 --pid-dir=$EB_APP_PIDS_DIR restart" $EB_APP_USER

rake jobs:work is usually used to start the workers in development environment from various articles I read.

To troubleshoot the problem further, try doing this in rails console:

Delayed::Worker.new

and check if the jobs get processed faster.

Another thing to try: Stop all workers via whatever script you normally use. Then do

ps -ef | grep "jobs:work"

to see if the all the worker processes actually get killed. If the worker system processes are still hanging around, you have to manually kill those system processes. Then restart the workers and check if any difference in behavior.