Optimizing Shadowsocks

If you see a lot of error: too many open files in your log, you should optimize your system. This tutorial applies to all shadowsocks servers (Python, libev, etc).

On Debian 7:

Create /etc/sysctl.d/local.conf with the following content:

fs.file-max = 51200

net.core.rmem_max = 67108864
net.core.wmem_max = 67108864
net.core.rmem_default=65536
net.core.wmem_default=65536
net.core.netdev_max_backlog = 4096
net.core.somaxconn = 4096

net.ipv4.tcp_syncookies = 1
net.ipv4.tcp_tw_reuse = 1
net.ipv4.tcp_tw_recycle = 0
net.ipv4.tcp_fin_timeout = 30
net.ipv4.tcp_keepalive_time = 1200
net.ipv4.ip_local_port_range = 10000 65000
net.ipv4.tcp_max_syn_backlog = 4096
net.ipv4.tcp_max_tw_buckets = 5000
net.ipv4.tcp_fastopen = 3
net.ipv4.tcp_rmem = 4096 87380 67108864
net.ipv4.tcp_wmem = 4096 65536 67108864
net.ipv4.tcp_mtu_probing = 1
net.ipv4.tcp_congestion_control = hybla

Then:

sysctl --system

Warning: DO NOT ENABLE net.ipv4.tcp_tw_recycle!!! See this article.

If you use Supervisor, Make sure you have the following line in /etc/default/supervisor. Once you added that line, restart Supervisor (service stop supervisor && service start supervisor).

ulimit -n 51200

If you use other ways to run shadowsocks in the background, make sure to add ulimit -n 51200 in your init script.

After optimizing, a busy Shadowsocks server that handles thousands of connections, takes about 30MB memory and 10% CPU. Notice that at the same time, Linux kernel usually uses >100MB RAM to hold buffer and cache for those connections. If you want to use less RAM, reduce the size of rmem and wmem.

Before & after: