Dead PostgreSQL connections on worker restart

[Adam Fields <unicorn5958@street86.com>]

We have discovered that every time a unicorn worker is restarted, rails throws "PG::ConnectionBad: connection is closed” errors.

We are using preload: true, and have before and after fork rules to close and reopen the connections:

---------------------------------------------------
before_fork do |server, worker|
 defined?(ActiveRecord::Base) and
     ActiveRecord::Base.connection.disconnect!

   old_pid = "#{server.config[:pid]}.oldbin"
   if File.exist?(old_pid) && server.pid != old_pid
     begin
       sig = (worker.nr + 1) >= server.worker_processes ? :QUIT : :TTOU
       Process.kill(sig, File.read(old_pid).to_i)
     rescue Errno::ENOENT, Errno::ESRCH
       # someone else did our job for us
     end
   end
end

after_fork do |_server, _worker|
  defined?(ActiveRecord::Base) and ActiveRecord::Base.establish_connection

  child_pid = _server.config[:pid].sub('.pid', ".#{_worker.nr}.pid")
  system("echo #{Process.pid} > #{child_pid}")
end
---------------------------------------------------

Yet it seems that something is holding onto a dead connection and trying to use it. 

We have definitely correlated it to worker restarts - we have a monit process in place to restart individual workers if they exceeded a memory threshold, and when this number was too low, they were getting recycled often and we saw a very high number of these errors. When we raised the threshold, the error almost completely disappeared (but it still happens sometimes when a worker is recycled).

How can we troubleshoot this?

Thanks!