Nginx in UDP load balancing continue to send to an unreachable upstream server

I use Nginx to load balance traffic coming from udp syslog sources to logstash Configuration : user nginx; worker_processes auto; error_log /var/log/nginx/error.log; pid /run/nginx.pid; worker_rlimit_nofile 1000000; # Load dynamic modules. See /usr/share/doc/nginx/README.dynamic. include /usr/share/nginx/modules/*.conf; events { worker_connections 1024; } include /etc/nginx/conf.d/*.conf; #UDP syslog load balancing stream { server { listen 514 udp; proxy_pass logstash_servers; proxy_timeout 1s; proxy_responses 0; proxy_bind $remote_addr transparent; } upstream logstash_servers { server 192.168.2.90:514 fail_timeout=10s; server 192.168.2.95:514 fail_timeout=10s; } } It works fine but if one of my upstream logstash server is down, Nginx does not take it into account and I receive the half of message on the remaining upstream logstash server. How can I tell to Nginx to use only the remaining logstash server if one is down ?