京公网安备11010502063487号“502 Bad Gateway” 页面上什么原因都没有,真正的线索只在 error.log 里。这篇在本机起一个自己的 nginx 和几个行为不同的上游,把上游没起、崩溃、复位、回非 HTTP、响应头太大这几种情形在 error.log 里的样子逐个对照,并测了多台上游全挂之后的恢复过程。
客户端:HTTP 502,页面标题 "502 Bad Gateway",Server: nginx/1.31.7
error.log: [error] *2 kevent() reported that connect() failed (61: Connection refused) while connecting to upstream, request: "GET /refused HTTP/1.1", upstream: "http://127.0.0.1:9791/refused", host: "127.0.0.1:9780"
同一个请求看 access.log:127.0.0.1 - - [29/Oct/2026:15:12:46 +0800] "GET /refused HTTP/1.1" 502 157 "-" "Python-urllib/3.14"页面上没有原因,error.log 有;access.log 只记 502。
上游收到请求就关闭连接:客户端 HTTP 502
error.log: [error] *4 upstream prematurely closed connection while reading response header from upstream, request: "GET /close HTTP/1.1", upstream: "http://127.0.0.1:9792/close", host: "127.0.0.1:9780"
上游发 RST 复位:客户端 HTTP 502
error.log: [error] *6 kevent() reported about an closed connection (54: Connection reset by peer) while reading response header from upstream, request: "GET /rst HTTP/1.1", upstream: "http://127.0.0.1:9793/rst", host: "127.0.0.1:9780"
上游回的不是 HTTP:客户端 HTTP None,收到的内容 'BadStatusLine: THIS IS NOT HTTP\r\n'
error.log: [error] *8 upstream sent no valid HTTP/1.0 header while reading response header from upstream, request: "GET /garbage HTTP/1.1", upstream: "http://127.0.0.1:9794/garbage", host: "127.0.0.1:9780"
上游响应头 20000 字节:客户端 HTTP 502
error.log: [error] *10 upstream sent too big header while reading response header from upstream, request: "GET /bighdr HTTP/1.1", upstream: "http://127.0.0.1:9795/bighdr", host: "127.0.0.1:9780"没起、收到请求就关、复位、响应头太大,日志措辞各不相同;回非 HTTP 内容时客户端甚至不是 502。
上游自己回 502:客户端 HTTP 502,页面内容 "upstream says 502",Server 头 nginx/1.31.7,error.log 里 0 条错误 nginx 生成的 502:HTTP 502,页面标题 "502 Bad Gateway",Server 头 nginx/1.31.7,error.log 里 1 条错误 对照,正常上游:HTTP 200 ok
上游自己回的 502 被原样透传、error.log 没记录;Server 头都是 nginx。
nginx -t:nginx: configuration file <tmp>/nginx.conf test is successful proxy_buffer_size 32k 之后,上游响应头 20000 字节:HTTP 200 ok error_page 502 之后,上游没起:HTTP 502,页面 "服务暂时不可用,请稍后再试"
proxy_buffer_size 调大;error_page 换页面。
nginx -t:nginx: configuration file <tmp>/nginx.conf test is successful
两台都没起,第 1 次请求:HTTP 502(1 ms)
error.log: [error] *2 kevent() reported that connect() failed (61: Connection refused) while connecting to upstream, request: "GET / HTTP/1.1", upstream: "http://127.0.0.1:9798/", h
error.log: [warn] *2 upstream server temporarily disabled while connecting to upstream, request: "GET / HTTP/1.1", upstream: "http://127.0.0.1:9798/", host: "127.0.0.1:9780"
error.log: [error] *2 kevent() reported that connect() failed (61: Connection refused) while connecting to upstream, request: "GET / HTTP/1.1", upstream: "http://127.0.0.1:9799/", h
error.log: [warn] *2 upstream server temporarily disabled while connecting to upstream, request: "GET / HTTP/1.1", upstream: "http://127.0.0.1:9799/", host: "127.0.0.1:9780"
紧接着第 2 次请求:HTTP 502
error.log: [error] *5 no live upstreams while connecting to upstream, request: "GET / HTTP/1.1", upstream: "http://pool/", host: "127.0.0.1:9780"
9798 已恢复。之后每 0.5 s 请求一次(fail_timeout = 2 s),看第几次开始成功:
恢复后 0.0 s:HTTP 502 error.log: [error] no live upstreams while connecting to upstream, client: 127.0.0.1, server: , reque
恢复后 0.5 s:HTTP 502 error.log: [error] no live upstreams while connecting to upstream, client: 127.0.0.1, server: , reque
恢复后 1.0 s:HTTP 502 error.log: [error] no live upstreams while connecting to upstream, client: 127.0.0.1, server: , reque
恢复后 1.5 s:HTTP 502 error.log: [error] no live upstreams while connecting to upstream, client: 127.0.0.1, server: , reque
恢复后 2.0 s:HTTP 502 error.log: [error] no live upstreams while connecting to upstream, client: 127.0.0.1, server: , reque
恢复后 2.5 s:HTTP 200 error.log: [warn] upstream server temporarily disabled while connecting to upstream, client: 127.0.0.标为暂时不可用,fail_timeout 内直接 502,恢复后也要等。