In an automated environment where new instances are added automatically and manged by puppet it is a great problem when the puppet master has some issues. It can act as a SPOF.
I happened as a accidental problem that puppet master had a 100% disk usage. As a result the requests from puppet clients of new instances were failing with 503 error. On checking the puppet master I could see the following error in puppet master error log.
Exception PhusionPassenger::UnknownError in PhusionPassenger::Rack::ApplicationSpawner (header too long (OpenSSL::X509::CRLError)) (process 598, thread #