conradirwin/cap-runit

Capistrano 3 runit support

Capistrano 3 Runit Support

cap-runit provides a helper for building runit services. It lets you easily deploy services using capistrano 3 and run them with runit.

Installation

Using bundler

1. Add gem 'cap-runit' to your Gemfile
2. Run bundle install
3. Add require 'bundler'; Bundler.require to your Capfile

Without bundler

1. Run gem install cap-runit
2. Add require 'capistrano/runit' to your Capfile`

Usage

cap-runit provides a helper method runit_service to Capistrano. As the philosophy of runit is that everything is just a shell script, cap-runit requires you to write a shell script.

For example, to run sidekiq, you might do something like this:

runit_service 'sidekiq' do

  # Sidekiq should only run on servers with role: %w{sidekiq}
  roles :sidekiq

  run <<-EOF
#!/bin/bash
cd /var/www/app/current
exec bundle exec sidekiq -e production -C ./config/sidekiq.yml 2>&1
  EOF

end

This creates three capistrano tasks:

cap sidekiq:restart    # Restart bugsnag-event-server
cap sidekiq:start      # Start bugsnag-event-server
cap sidekiq:stop       # Stop bugsnag-event-server

If you'd like to restart sidekiq when your app restarts, you also need to specify:

namespace :deploy do
  after :publishing, 'sidekiq:restart'
end

Logging

Runit sends the STDOUT of your run script to the log script. To get these logs somewhere useful, you can use any program that reads from STDIN. But svlogd is recommended as it works well with runit and handles log rotation automatically.

runit_service 'sidekiq' do

  roles :sidekiq

  run <<-EOF
#!/bin/bash -l
cd /var/www/app/current
exec bundle exec sidekiq -e production -C ./config/sidekiq.yml 2>&1
  EOF

  log <<-EOF
#!/bin/sh
mkdir -p /var/log/sidekiq
exec svlogd -tt /var/log/sidekiq
  EOF
end

N.B. For this example to work you will need to ensure that /var/log/sidekiq can be written to by the capistrano user.

Crash notifications

If you need to be notified of when a crash happens in your app, you can also configure a finish script. This is run whenever the run script exits.

runit_service 'sidekiq' do

  roles :sidekiq

  run <<-EOF
#!/bin/bash -l
cd /var/www/app/current
exec bundle exec sidekiq -e production -C ./config/sidekiq.yml 2>&1
  EOF

  log <<-EOF
#!/bin/sh
mkdir -p /var/log/sidekiq
exec svlogd -tt /var/log/sidekiq
  EOF

  finish <<-EOF
#!/bin/sh
export BUGSNAG_API_KEY=bf04f47b034e8df9d22ee75f30fdae6a
[ "$1" = 0 ] || exec bugsnag-runit "$0"
  EOF
end

Configuration

The only configuration variable you need to pass is runit_service_directory

set :runit_service_directory, "/apps/service"

This should point to a directory that is being run by runit, and which the capistrano user has permission to modify. I suggest setting up something like /apps/service run by the deploy user. This can be done by creating two files:

#/etc/service/cap-runit/run

#!/bin/bash
exec chpst -udeploy -- runsvdir /apps/service 2>&1

And

#/etc/service/cap-runit/log/run

#!/bin/sh
mkdir -p /var/log/cap-runit
exec svlogd -tt /var/log/cap-runit

These files should both be owned by root and chmod +x. This way the deploy user does not need root permissions to reboot a service on deploy.

cap-runit's Issues

runit script necessity

Just a question:

If I specify runit_service_directory, then do I still need to have the runit script inside a code block within capistrano? or does it use the scripts in the servive_directory instead?

I don't want to have to hard code huge runit scripts in capistrano....

thanks!