Make rbtrace work under unicorn

The Unicorn master resets any signal handlers and prevents rbtrace
from working properly as a result. Move the `require 'rbtrace'` call
to the Unicorn `after_fork` block and keep the initializer only for Sidekiq.

config/initializers/rbtrace.rb

 # frozen_string_literal: true
 
-require 'rbtrace' if ENV['ENABLE_RBTRACE']
+# rbtrace needs to be included after the unicorn worker forks.
+# See the after_fork block in config/unicorn.rb.example.
+require 'rbtrace' if ENV['ENABLE_RBTRACE'] && Sidekiq.server?

config/unicorn.rb.example

@@ -124,6 +124,10 @@ before_fork do |server, worker|
 end
 
 after_fork do |server, worker|
+  # Unicorn clears out signals before it forks, so rbtrace won't work
+  # unless it is enabled after the fork.
+  require 'rbtrace' if ENV['ENABLE_RBTRACE']
+
   # per-process listener ports for debugging/admin/migrations
   # addr = "127.0.0.1:#{9293 + worker.nr}"
   # server.listen(addr, :tries => -1, :delay => 5, :tcp_nopush => true)

doc/administration/troubleshooting/debug.md

@@ -77,7 +77,12 @@ and more. However, this is not enabled by default. To enable it, define the
 gitlab_rails['env'] = {"ENABLE_RBTRACE" => "1"}
 ```
 
-Then reconfigure the system and restart Unicorn and Sidekiq.
+Then reconfigure the system and restart Unicorn and Sidekiq. To run this
+in Omnibus, run as root:
+
+```ruby
+/opt/gitlab/embedded/bin/ruby /opt/gitlab/embedded/bin/rbtrace
+```
 
 ## Common Problems