JMeter Test Interface-Open a number of TCP connections time_wait status (Linux environment) solutions that cause errors

A problem found:

The server is a Linux system, with JMeter test interface, found to open a lot of TCP connections, [[email protected] bin]# ulimit-n 65535 with this command set the total number of connections, when the pressure test, the number of connections may reach more than 50000, Easily error-prone; View the number of TCP in each state:
Netstat-an | awk '/^tcp/{++s[$NF]} END {for (a-in s) print A,s[a]} ', the state of the connection status time_wait is found to be many, (count 80 port connections Netstat-nat|grep-i 80|WC -L Use this statistic as well). The interface without error, the number of open TCP connections is less, about 15000-25000 this way.

The initial suspicion is that the TCP time_wait state is too many, causing the connection time is too long, the system's port resource is occupied for a long time, the new request, and the allocation of resources.

Two-way solution:

After more than 3 days, finally found an effective solution, you can refer to the Micro-blog: http://blog.oldboyedu.com/tcp-wait/:

In the sysctl.conf configuration file # vim/etc/sysctl.conf, modify the configuration as follows:

1 net.ipv6.conf. default 1  111

Again, the number of TCP connections significantly decreased, and the interface TPC up, not easy to error, as shown in:

JMeter Test Interface-Open a number of TCP connections time_wait status (Linux environment) solutions that cause errors