Troubleshooting Techniques

[ 210 ]

Network limits

If you find yourself in a situation in which no network buffers are available, you

will most likely only be able to log in at the console, if at all. This can happen when
NGINX receives so many client connections that all available network buffers are

used up. Increasing the number of network buffers is also specific to a particular
operating system and may be done as follows:

• FreeBSD
  vi /boot/loader.conf

kern.ipc.nmbclusters=262144

• Solaris

  ndd -set /dev/tcp tcp_max_buf 16777216

  
  

When NGINX is acting as either a mail or an HTTP proxy, it will need to open many
connections to its upstream servers. To enable as many connections as possible, the

ephemeral TCP port range should be adjusted to its maximum.

• Linux
  vi /etc/sysctl.conf

net.ipv4.ip_local_port_range = 1024 65535

# sysctl -p /etc/sysctl.conf

• FreeBSD
  vi /etc/sysctl.conf

net.inet.ip.portrange.first=1024

net.inet.ip.portrange.last=65535

# /etc/rc.d/sysctl reload

• Solaris

  ndd -set /dev/tcp tcp_smallest_anon_port 1024

  
  

  ndd -set /dev/tcp tcp_largest_anon_port 65535

  
  

Having adjusted these basic values, we will now take a look at more specific

performance-related parameters in the next section.