When we want to take an upstream server down for maintenance, we need to do it gradually.
Ideally, we would mark it down, and it would stop accepting new connections but would continue service active users (using ip_hash). We could then see when the upstream server is no longer serving requests and make changes to it before bringing it back up.
If I mark an upstream server as down, reload the nginx upstream configuration, will all users with an ip_hash for that server immediately be routed to another server?
Rob