nginx upstream转发连接错误情况研究

本次测试用到3台服务器：

192.168.10.115：转发服务器A

192.168.10.209：upstream下服务器1

192.168.10.210：upstream下服务器2

1台客户端：192.168.10.112

服务器A中nginx主要配置如下：

    log_format  main  '$remote_addr - $remote_user [$time_local] "$request" ''$status $body_bytes_sent "$http_referer" ''"$http_user_agent" "$http_x_forwarded_for"';keepalive_timeout  65;#gzip  on;upstream testup{server 192.168.10.209 weight=1 max_fails=1 fail_timeout=30s;server 192.168.10.210 weight=1 max_fails=1 fail_timeout=30s;}server {listen       80;server_name  localhost;#charset koi8-r;access_log  logs/host.access.log  main;location / {#root   html;#index  index.html index.htm;proxy_next_upstream http_502 http_504 error timeout invalid_header;proxy_ignore_client_abort on;proxy_send_timeout 60s;proxy_read_timeout 300s;proxy_next_upstream_tries 0;proxy_pass http://testup;proxy_set_header Host $host:$server_port;proxy_set_header X-Real-IP $remote_addr;proxy_connect_timeout 3;proxy_redirect default;proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;}}

nginx工作进程配置为2,

服务器1和服务器2都是普通的web服务配置不在此展示了

服务器1页面1：

服务器2页面2：

正常访问服务器A会在如上页面1和2之间切换

测试情况1：关闭服务器1

请求A地址，先卡顿差不多3s（应该与'proxy_connect_timeout 3'相关）然后切换成页面2内容，随后刷新无任何卡顿，30s（nginx相关配置：'max_fails=1 fail_timeout=30s'）过后再次请求仍会卡顿3s成功随后并不卡顿，nginx报错日志如下：

2024/09/21 18:23:19 [error] 6056#0: *114 upstream timed out (110: Connection timed out) while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://192.168.10.209:80/", host: "192.168.10.115"

测试情况2：不关闭服务器1，只关闭其nginx服务

请求A地址，不出现卡顿，页面一直显示的页面2内容。nginx报错日志：

2024/09/21 18:30:14 [error] 6055#0: *133 connect() failed (113: No route to host) while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://192.168.10.209:80/", host: "192.168.10.115"
2024/09/21 18:30:48 [error] 6055#0: *133 connect() failed (113: No route to host) while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://192.168.10.209:80/", host: "192.168.10.115"

此处说明'proxy_connect_timeout 3'针对的是请求能否转达服务器，与目标服务器上的nginx服务是否正常运行无关。

测试情况3：关闭服务器1和2（2选择的是屏蔽了对外端口）

请求A地址，卡顿了差不多6,7s出现如下页面：

此处应该是先后请求了两个服务器耗时3*2s，随后请求不卡顿直接返回如上页面，30s后继续请求仍会卡顿6s随后不卡顿。nginx报错日志：

此处前两条分别是两台服务器的超时日志，后续的日志“no live upstreams while connecting to upstream”是nginx在由于之前请求超时已经判断两台机器都不可用，在30s不会再去转发请求到该服务器，没有了可用的upstreams直接报该错误。

测试情况4：模拟慢网情况

恢复服务器1,2的正常访问，保证访问A地址能正常在页面1,2之前切换。

模拟网络延迟：

tc qdisc add dev ens33 root netem delay 1000ms

参考文档：1分钟学会在Linux下模拟网络延迟_linux模拟网络延迟-CSDN博客

这里将服务器A的nginx配置'proxy_connect_timeout'值改为1。如果喜欢等待的老铁也可以选择不改，但上面的延迟时间要改为3000ms了。这延迟那真的是连xshell连接服务器的输入输出都延迟了╮(╯▽╰)╭，只要是走网卡的应该都会被卡一下。

此时访问A地址，卡顿了一段时间后页面如下：

与之前的关闭服务器1,2的情况大致一样。首次卡顿了4,5s返回如上页面，随后刷新2s返回，过30s后仍会卡顿4,5s返回。错误日志输出情况也与情况3一致。

将服务器A的nginx配置'proxy_connect_timeout'值改为2。继续请求A地址，差不多4,5s返回页面1或页面2，并且页面内容正常切换。

测试情况4.2：调整proxy_next_upstream_tries

在情况4的弱网请求失败情况下，修改'proxy_next_upstream_tries'值为1，请求A地址3次。nginx错误日志：

2024/09/21 21:03:10 [notice] 6399#0: signal process started
2024/09/21 21:03:29 [error] 6401#0: *396 upstream timed out (110: Connection timed out) while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://192.168.10.209:80/", host: "192.168.10.115"
2024/09/21 21:03:40 [error] 6401#0: *396 upstream timed out (110: Connection timed out) while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://192.168.10.210:80/", host: "192.168.10.115"
2024/09/21 21:03:49 [error] 6401#0: *396 no live upstreams while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://testup/", host: "192.168.10.115

修改'proxy_next_upstream_tries'值为2，请求A地址2次，错误日志：

2024/09/21 21:23:16 [error] 6432#0: *413 upstream timed out (110: Connection timed out) while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://192.168.10.209:80/", host: "192.168.10.115"
2024/09/21 21:23:17 [error] 6432#0: *413 upstream timed out (110: Connection timed out) while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://192.168.10.210:80/", host: "192.168.10.115"
2024/09/21 21:23:22 [error] 6432#0: *413 no live upstreams while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://testup/", host: "192.168.10.115"

相较于上次修改配置，这次请求第一次就连续尝试连接209和210，都失败，归为不可用，第二次请求直接返回“no live upstreams”报错。

再次修改'proxy_next_upstream_tries'值为3，请求A地址2次，错误日志：

2024/09/21 21:28:59 [error] 6446#0: *423 upstream timed out (110: Connection timed out) while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://192.168.10.209:80/", host: "192.168.10.115"
2024/09/21 21:29:00 [error] 6446#0: *423 upstream timed out (110: Connection timed out) while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://192.168.10.210:80/", host: "192.168.10.115"
2024/09/21 21:29:12 [error] 6446#0: *423 no live upstreams while connecting to upstream, client: 192.168.10.112, server: localhost, request: "GET / HTTP/1.1", upstream: "http://testup/", host: "192.168.10.115"

跟上次测试日志输出一样，upstream下总共就2台服务器，不会因为proxy_next_upstream_tries的值大于2就循环再次请求第一个服务器。所以proxy_next_upstream_tries的值大于转发的服务器的数量时以服务器数量为准。proxy_next_upstream_tries值为0的时候从日志上看应该是按照服务器数量尝试连接的。

测试情况4.3：调整max_fails

配置nginx中max_fails值改为2，请求地址A3次，前两次请求时间明显较长，报错日志：