And if maxcoon in global doesn't help investigate if kernel is not a problem. It may happen that kernel won't accept more connections
________________________________ Od: PiBa-NL <piba.nl....@gmail.com> Wysłane: 27 października 2017 19:46:19 Do: kushal bhattacharya; haproxy@formilux.org Temat: Re: confusion regarding usage of haproxy for large number of connections Hi, Op 27-10-2017 om 14:58 schreef kushal bhattacharya: Hi, I am confused regarding the readme text ' This is a development version, so it is expected to break from time to time, to add and remove features without prior notification and it should not be used in production' .Here I am testing for 8000 connections being distributed to three virtual mqtt brokers having same ip address but three different ports.I am getting a maximum threshold of 2000 connections being handled in this setup.Haproxy is listeneing to a port for incoming client connections and distributing it to the 3 mqtt brokers with the configuration file given below defaults mode tcp maxconn 8000 timeout connect 5000s timeout client 5000s timeout server 5000s frontend localnodes bind *:9875 log global log 127.0.0.1:514<http://127.0.0.1:514> local0 info option tcplog default_backend nodes backend nodes mode tcp balance roundrobin server web01 192.168.0.5:9878<http://192.168.0.5:9878> maxconn 3000 server web02 192.168.0.5:9877<http://192.168.0.5:9877> maxconn 3000 server web03 192.168.0.5:9876<http://192.168.0.5:9876> maxconn 2000 With this configuration can i undergo my setup with 8000 connection load distribution or do i have to undergo some changes here Thanks, Kushal Add a 'maxconn 8000' in 'global' section? Regards, PiBa-NL