We're in the process of getting Passenger Enterprise purchased and setup, but in the meantime we'd like to get some pseudo-"rolling restarts" setup, by issuing commands on each server to sleep X seconds and then restart passenger when we deploy. It's working great, except for one problem: Our load balancer is still sending requests to the server on which Passenger is restarting, resulting in 50% of requests being queued into the restarting Passenger (which takes a while to boot up).
The load balancer is unfortunately out of our control and basically just checks to make sure that a server can accept a request without errors (50x) before sending it off, and I guess that it thinks a restarting Passenger is fine.
My question is: Is there anything I can do, either in Passenger or nginx, to tell the load balancer that Passenger cannot accept requests while it's restarting, therefore sending all of the requests over to the other server?
I hope this makes sense! Thanks.