sysctl kern.ipc.soacceptqueue=1024
# somaxconn is the OS buffer, backlog queue depth for accepting new incoming TCP # connections. An application will have its own, separate max queue length # (maxqlen) which can be checked with "netstat -Lan". The default is 128 # connections per application thread. Lets say your Nginx web server normally # receives 100 connections/sec and is single threaded application. If clients # are bursting in at a total of 250 connections/sec you may want to set the # somaxconn at 512 to be a 512 deep connection buffer so the extra 122 clients # (250-128=122) do not get denied service since you would have 412 # (512-100=412) extra queue slots. Also, a large listen queue will do a better # job of avoiding Denial of Service (DoS) attacks if, and only if, your # application can handle the TCP load at the cost of more RAM and CPU time. # Nginx sets is backlog queue to the same as the OS somaxconn by default. # Note: "kern.ipc.somaxconn" is not shown in "sysctl -a" output, but searching # for "kern.ipc.soacceptqueue" gives the same value and both directives stand # for the same buffer value. #kern.ipc.soacceptqueue=1024 # (default 128 ; same as kern.ipc.somaxconn)