Remove puma from eye config.

Share queues between workers This reduces the memory footprint of ontohub drastically: Each worker runs an instance of the app, consuming as much memory as the rails server. Fewer workers means less memory consumption.

Add a sidekiq queue for migrations Use it by specifically enqueuing a job to this queue, e.g.: Sidekiq::Client.enqueue_to(:'hets-migration', SomeWorker, args_for_worker)

Use only one worker (with several threads) for the hets queue.

Add puma to Eye config.

Add eye config.