Nginx 长连接keep_alive的具体使用

如何开启并支持长连接

当使用nginx作为反向代理时，为了支持长连接，需要做到两点：

从client到nginx的连接是长连接从nginx到server的连接是长连接

从HTTP协议的角度看，nginx在这个过程中，对于客户端它扮演着HTTP服务器端的角色。而对于真正的服务器端（在nginx的术语中称为upstream）nginx又扮演着HTTP客户端的角色。

客户端（Client）和 Nginx 保持长连接，两个要求：

client发送的HTTP请求要求keep alivenginx设置上支持keep alive

一、客户端 与 Nginx 长连接配置

默认情况下，nginx已经自动开启了对client连接的keep alive支持。一般场景可以直接使用，但是对于一些比较特殊的场景，还是有必要调整个别参数。

keepalive_timeout 指令

keepalive_timeout指令的语法：

Syntax: keepalive_timeout timeout [header_timeout]; Default: keepalive_timeout 75s; Context: http, server, location

第一个参数设置keep-alive客户端连接在服务器端保持开启的超时值。值为0会禁用keep-alive客户端连接。可选的第二个参数在响应的header域中设置一个值“Keep-Alive: timeout=time”。这两个参数可以不一样。

注：默认75s一般情况下也够用，对于一些请求比较大的内部服务器通讯的场景，适当加大为120s或者300s。第二个参数通常可以不用设置。

keepalive_requests 指令

keepalive_requests指令用于设置一个keep-alive连接上可以服务的请求的最大数量。当最大请求数量达到时，连接被关闭。默认是100。

这个参数的真实含义，是指一个keep alive建立之后，nginx就会为这个连接设置一个计数器，记录这个keep alive的长连接上已经接收并处理的客户端请求的数量。如果达到这个参数设置的最大值时，则nginx会强行关闭这个长连接，逼迫客户端不得不重新建立新的长连接。

这个参数往往被大多数人忽略，因为大多数情况下当QPS(每秒请求数)不是很高时，默认值100凑合够用。但是，对于一些QPS比较高（比如超过10000QPS，甚至达到30000,50000甚至更高) 的场景，默认的100就显得太低。

简单计算一下，QPS=10000时，客户端每秒发送10000个请求(通常建立有多个长连接)，每个连接只能最多跑100次请求，意味着平均每秒钟就会有100个长连接因此被nginx关闭。同样意味着为了保持QPS，客户端不得不每秒中重新新建100个连接。因此，如果用netstat命令看客户端机器，就会发现有大量的TIME_WAIT的socket连接(即使此时keep alive已经在client和nginx之间生效)。

因此对于QPS较高的场景，非常有必要加大这个参数，以避免出现大量连接被生成再抛弃的情况，减少TIME_WAIT。

keepalive_requests 指令

默认值1000，单个连接中处理的最大请求数，超过这个数，连接销毁。

keepalive_disable 指令

不对某些浏览器建立长连接，默认：msie6

send_timeout 指令

两次向客户端写操作之间的间隔 如果大于这个时间则关闭连接 默认60s

此处有坑，注意耗时的同步操作有可能会丢弃用户连接

该设置表示Nginx服务器与客户端连接后，某次会话中服务器等待客户端响应超过10s，就会自动关闭连接。

http { include mime.types; default_type application/octet-stream; sendfile on; keepalive_timeout 65 65; #超过这个时间 没有活动，会让keepalive失效 keepalive_time 1h; # 一个tcp连接总时长，超过之后 强制失效 send_timeout 60;# 默认60s 此处有坑！！ 系统中 若有耗时操作，超过 send_timeout 强制断开连接。 注意：准备过程中，不是传输过程 keepalive_requests 1000; #一个tcp复用中 可以并发接收的请求个数

二、Nginx 与 服务端长连接配置

2.1 在 upstream 中配置

keepalive 100;

向上游服务器的保留连接数（线程池的概念）

keepalive_timeout 65

连接保留时间（单位秒）

keepalive_requests 10000

一个tcp复用中 可以并发接收的请求个数

2.2 在 server 中配置

proxy_http_version 1.1; 配置http版本号 默认使用http1.0协议，需要在request中增加”Connection： keep-alive“ header才能够支持，而HTTP1.1默认支持。 proxy_set_header Connection “”; 清楚close信息

三、压力测试

3.1 【客户端】 直连 【Nginx】

Server Software: nginx/1.21.6 Server Hostname: 192.168.44.102 Server Port: 80 Document Path: / Document Length: 16 bytes Concurrency Level: 30 Time taken for tests: 13.035 seconds Complete requests: 100000 Failed requests: 0 Write errors: 0 Total transferred: 25700000 bytes HTML transferred: 1600000 bytes Requests per second: 7671.48 [#/sec] (mean) Time per request: 3.911 [ms] (mean) Time per request: 0.130 [ms] (mean, across all concurrent requests) Transfer rate: 1925.36 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 0 0 0.4 0 12 Processing: 1 3 1.0 3 14 Waiting: 0 3 0.9 3 14 Total: 2 4 0.9 4 14 Percentage of the requests served within a certain time (ms) 50% 4 66% 4 75% 4 80% 4 90% 5 95% 5 98% 6 99% 7 100% 14 (longest request)

3.2 【客户端】 连接 【Nginx】 反向代理 【Nginx】

Server Software: nginx/1.21.6 Server Hostname: 192.168.44.101 Server Port: 80 Document Path: / Document Length: 16 bytes Concurrency Level: 30 Time taken for tests: 25.968 seconds Complete requests: 100000 Failed requests: 0 Write errors: 0 Total transferred: 25700000 bytes HTML transferred: 1600000 bytes Requests per second: 3850.85 [#/sec] (mean) Time per request: 7.790 [ms] (mean) Time per request: 0.260 [ms] (mean, across all concurrent requests) Transfer rate: 966.47 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 0 0 0.2 0 13 Processing: 3 8 1.4 7 22 Waiting: 1 7 1.4 7 22 Total: 3 8 1.4 7 22 Percentage of the requests served within a certain time (ms) 50% 7 66% 8 75% 8 80% 8 90% 9 95% 10 98% 12 99% 13 100% 22 (longest request)

3.3【客户端】 直连 【Tomcat】

Server Software: Server Hostname: 192.168.44.105 Server Port: 8080 Document Path: / Document Length: 7834 bytes Concurrency Level: 30 Time taken for tests: 31.033 seconds Complete requests: 100000 Failed requests: 0 Write errors: 0 Total transferred: 804300000 bytes HTML transferred: 783400000 bytes Requests per second: 3222.38 [#/sec] (mean) Time per request: 9.310 [ms] (mean) Time per request: 0.310 [ms] (mean, across all concurrent requests) Transfer rate: 25310.16 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 0 0 0.3 0 15 Processing: 0 9 7.8 7 209 Waiting: 0 9 7.2 7 209 Total: 0 9 7.8 7 209 Percentage of the requests served within a certain time (ms) 50% 7 66% 9 75% 11 80% 13 90% 18 95% 22 98% 27 99% 36 100% 209 (longest request)

3.4 【客户端】 连接 【Nginx】 反向代理 【Tomcat】并开启keepalive

Server Software: nginx/1.21.6 Server Hostname: 192.168.44.101 Server Port: 80 Document Path: / Document Length: 7834 bytes Concurrency Level: 30 Time taken for tests: 23.379 seconds Complete requests: 100000 Failed requests: 0 Write errors: 0 Total transferred: 806500000 bytes HTML transferred: 783400000 bytes Requests per second: 4277.41 [#/sec] (mean) Time per request: 7.014 [ms] (mean) Time per request: 0.234 [ms] (mean, across all concurrent requests) Transfer rate: 33688.77 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 0 0 0.2 0 9 Processing: 1 7 4.2 6 143 Waiting: 1 7 4.2 6 143 Total: 1 7 4.2 6 143 Percentage of the requests served within a certain time (ms) 50% 6 66% 7 75% 7 80% 7 90% 8 95% 10 98% 13 99% 16 100% 143 (longest request)

3.5 【客户端】 连接 【Nginx】 反向代理 【Tomcat】不开启keepalive

Server Software: nginx/1.21.6 Server Hostname: 192.168.44.101 Server Port: 80 Document Path: / Document Length: 7834 bytes Concurrency Level: 30 Time taken for tests: 33.814 seconds Complete requests: 100000 Failed requests: 0 Write errors: 0 Total transferred: 806500000 bytes HTML transferred: 783400000 bytes Requests per second: 2957.32 [#/sec] (mean) Time per request: 10.144 [ms] (mean) Time per request: 0.338 [ms] (mean, across all concurrent requests) Transfer rate: 23291.74 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect: 0 0 0.2 0 9 Processing: 1 10 5.5 9 229 Waiting: 1 10 5.5 9 229 Total: 1 10 5.5 9 229 Percentage of the requests served within a certain time (ms) 50% 9 66% 10 75% 11 80% 11 90% 13 95% 14 98% 17 99% 19 100% 229 (longest request)