nginx load balancing - keep last working server - nginx

Question

I have two servers A and B.
Most of the time, both servers are not running at the same time.
So when A is running, it is most likely that B is not running.
When B is running, it is most likely that A is not running.
This switch between A and B can happen after some weeks.
So I would like that nginx redirect to the running server and keep using that server until it is down.
I have tried that solution:
upstream backend {
server SERVER_A;
server SERVER_B;
}
server {...}
This is working but I can see in the logs that it is periodically trying to connect to the "down" server.
Then I tried that:
upstream backend {
server SERVER_A;
server SERVER_B backup;
}
server {...}
This is working correctly if SERVER_A is up. But if it is SERVER_B, then it is frequently trying to access SERVER_A.
Actually, in that case, the correct configuration would be "server SERVER_A backup;" but I don't think we can do dynamic reconfiguration ;-)
Actually, it is not a very big deal that nginx is periodically trying to access the down server, but if I can avoid that using a correct configuration, it would be better.
I know about that fail-timeout argument. But I don't think it will really solve my issue, and it might even add some down time during switching.
Thanks in advance.

Answer 1

According to controlled mechanism for server switch a hook to mark an individual server down is only required:
sed -i 's/server SERVER_A;/server SERVER_A down;/' /etc/nginx/nginx.conf
nginx -s reload
A configuration load of standard procedure that handles graceful reload and it is safe: http://nginx.org/en/docs/beginners_guide.html#control
Once the master process receives the signal to reload configuration,
it checks the syntax validity of the new configuration file and tries
to apply the configuration provided in it. If this is a success, the
master process starts new worker processes and sends messages to old
worker processes, requesting them to shut down. Otherwise, the master
process rolls back the changes and continues to work with the old
configuration. Old worker processes, receiving a command to shut down,
stop accepting new connections and continue to service current
requests until all such requests are serviced. After that, the old
worker processes exit.