分类: 云计算
2017-08-11 16:36:59
一、介绍
二、安装
三、运行
四、开发
1.介绍
Tengine:轻量级、高性能、高并发、配置化、模块化、可扩展、可移植的Web和反向代理 服务器,Tengine是nginx超集,但做了很多优化,包含了很多比较有用的模块,比如直接包含了lua、proc等很有用的模块。
Lua:一个很轻量级的 脚本,也号称性能最高的 脚本。代码总共不到600k,32个C文件,23个头文件:
root@j9 ~/lua-5.1.5/src# du -sh ./ 572K ./ root@j9 ~/lua-5.1.5/src# ls *.c | wc -l 32 root@j9 ~/lua-5.1.5/src# ls *.h | wc -l 23 root@j9 ~/lua-5.1.5/src#
可以非常容易的嵌入C和C++工程中,也比较容易与C和C++互动,这也是目前Lua主要的用法。
ngx_lua:一个nginx很重要的第三方模块,作者:章亦春(agentzh、春哥),结合了nginx和Lua各自优点,把Lua嵌入nginx中,使其支持Lua来快速开发基于nginx下的业务逻辑。
[]()
2. 安装
2.1、LuaJIT
wget -c tar xzvf LuaJIT-2.0.4.tar.gz cd LuaJIT-2.0.4 make install PREFIX=/usr/local/luajit #注意环境变量! export LUAJIT_LIB=/usr/local/luajit/lib export LUAJIT_INC=/usr/local/luajit/include/luajit-2.0
2.2、Tengine
tengine最新代码中已经包含lua模块了,直接git clone下来就可以
git clone cd tengine ./configure --prefix=/opt/tengine --with-http_lua_module make make install
如果是原生nginx的话,得自行下载lua模块代码:
wget tar xvf nginx-1.7.8.tar.gz cd nginx-1.7.8 mkdir modules cd modules git clone .git cd .. ./configure --prefix=/opt/nginx --add-module=./modules/lua-nginx-module/ make make install
3. 运行
修改/opt/tengine/conf/nginx.conf:
worker_processes 1; error_log logs/error.log; pid logs/nginx.pid; events { worker_connections 1024; } http { include mime.types; default_type application/octet-stream; log_format main '$remote_addr - $remote_user [$time_local] "$request" ' '$status $body_bytes_sent "$http_referer" ' '"$http_user_agent" "$http_x_forwarded_for"'; access_log logs/access.log main; sendfile on; keepalive_timeout 65; server { listen 80; server_name localhost; location / { root html; index index.html index.htm; } location /hello_lua { content_by_lua ' ngx.say("Lua: hello world!") '; } } }
运行tengine:
root@j9 ~/tengine# /opt/tengine/sbin/nginx
curl访问一下hello_lua:
root@j9 ~/tengine# curl Lua: hello world!
运行ok。
4、开发
> 语法
> 入门
> 深入
4.1、语法
参考:
[Lua简明教程]()
[Lua在线lua学习教程]()
4.2、入门
4.2.1、API
- ngx.print
输出响应内容体;
例如:ngx.print("a", "b", "c")
- ngx.say
跟ngx.print的区别只是最后会多输出一个换行符;
例如:ngx.say("a", "b", "c")
- ngx.status
设置响应HTTP状态码;
注意,设置状态码仅在响应头发送前有效。当调用ngx.say或者ngx.print时自动发送响应状态码(默认为200);可以通ngx.headers_sent来判断是否发送了响应状态码。
例如:ngx.status = 200
- ngx.exit
设置响应HTTP状态码并退出;
注意,设置状态码仅在响应头发送前有效,并且该函数调用之后该函数后面的lua将被忽略掉,因为已经exit了。
例如:ngx.exit(200)
- ngx.header
输出响应头;
注意,头部字段中含有横杠(-)的要转换成下划线(_),ngx_lua模块自动将_转换成-。
例如:ngx.header["X-Cache"] = "HIT" 或者 ngx.header.X_Cache = "HIT"或者ngx.header.X_Cache = {"AA", "BB"}
- ngx.redirect
301或者302重定向
例如:ngx.redirect("[]()", 301)
- ngx.log
打印nginx错误日志,日志级别有:ngx.STDERR、ngx.EMERG、ngx.ALERT、ngx.CRIT、ngx.ERR、ngx.WARN、ngx.NOTICE、ngx.INFO、ngx.DEBUG
例如:ngx.log(ngx.ERR, "test: ", "ok")
例子:
server { listen 9898; location / { default_type "text/html"; content_by_lua ' local headers_sent_1 = ngx.headers_sent ngx.header["X-Cache"] = "HIT" ngx.header.Y_Cache = "MISS" ngx.header.Z_Cache = {"AA", "BB"} ngx.status = 602 local headers_sent_2 = ngx.headers_sent ngx.print("a", "b") local headers_sent_3 = ngx.headers_sent ngx.say("c", "d") ngx.say("e", "f") ngx.say("headers_sent_1: ", tostring(headers_sent_1)) ngx.say("headers_sent_2: ", tostring(headers_sent_2)) ngx.say("headers_sent_3: ", tostring(headers_sent_3)) ngx.log(ngx.ERR, "ngx.log test ok") ngx.exit(601) ngx.say("g", "h") '; } location ^~ /redirect { content_by_lua ' ngx.redirect("", 301) '; } }
测试结果:
root@j9 ~# curl "" -i HTTP/1.1 602 Server: Tengine/2.2.0 Date: Mon, 19 Oct 2015 16:10:42 GMT Content-Type: text/html Transfer-Encoding: chunked Connection: keep-alive X-Cache: HIT Y-Cache: MISS Z-Cache: AA Z-Cache: BB abcd ef headers_sent_1: false headers_sent_2: false headers_sent_3: true root@j9 ~# curl "redirect" -i HTTP/1.1 301 Moved Permanently Server: Tengine/2.2.0 Date: Mon, 19 Oct 2015 16:18:16 GMT Content-Type: text/html Content-Length: 284 Connection: keep-alive Location:301 Moved Permanently 301 Moved Permanently
The requested resource has been assigned a new permanent URI.
Powered by Tengine/2.2.0 root@j9 ~#
- ngx.var
读取nginx变量,如nginx变量为$a,则在Lua中通过ngx.var.a获取,也可以给nginx变量赋值如ngx.var.a = "aa",前提是该变量在nginx中必须存在,不能在Lua中创建nginx变量。另外,对于nginx location中使用正则捕获的捕获组可以使用ngx.var[捕获组数字]获取。
例子
server { listen 9898; location ~ /var/([^/]*)/([^/]*) { default_type "text/html"; set $a "aaa"; set $b $host; content_by_lua ' ngx.say("$a: ", ngx.var.a) ngx.say("$b: ", ngx.var.b) ngx.say("$host: ", ngx.var.host) ngx.say("$arg_id: ", ngx.var.arg_id) ngx.say("$1: ", ngx.var[1]) ngx.say("$2: ", ngx.var[2]) '; } }
测试结果:
root@j9 ~# curl "var/aaaa/bbbb?id=22" -H "Host: " $a: aaa $b: $host: $arg_id: 22 $1: aaaa $2: bbbb root@j9 ~#
- ngx.req.raw_header
未解析的请求头字符串;
例如:ngx.req.raw_header()
- ngx.req.get_headers
获取请求头,默认只获取前100个头部,如果想要获取所有头部可以调用ngx.req.get_headers(0);获取带中划线的请求头时要把中划线转换成下划线使用如headers.user_agent这种方式;如果一个请求头有多个值,则返回的是table;
例如:ngx.req.get_headers()
- ngx.req.get_uri_args
获取url请求参数,其用法与ngx.req.get_headers类似;
- ngx.req.get_post_args
获取post请求body参数,其用法与ngx.req.get_uri_args类似,但必须提前调用ngx.req.read_body();
- ngx.req.read_body
如果要获取请求的body,则需要调用ngx.req.read_body(),否则获取不到body数据,(ps:也可以在nginx配置文件中加入指令lua_need_request_body on;来开启读取body,但官方不推荐)
- ngx.req.discard_body
忽略请求的body
注意,如果处理一个包含body的请求且需要ngx.exit时,需要调用此函数来忽略body,否则nginx可能将body当成header来解析,从而导致400错误;
- ngx.req.get_body_data
获取请求body数据
例子
location ^~ /req { content_by_lua ' ngx.say("===========ngx.req.raw_header=") ngx.say(ngx.req.raw_header()) local headers = ngx.req.get_headers() ngx.say("===========headers============") ngx.say("Host: ", headers["Host"]) ngx.say("user-agent: ", headers.user_agent) ngx.say("===========all headers========") for k,v in pairs(headers) do if type(v) == "table" then ngx.say(k, ": ", table.concat(v, ",")) else ngx.say(k, ": ", v) end end ngx.say("===========args===============") local args = ngx.req.get_uri_args() for k,v in pairs(args) do ngx.say(k, ": ", v) end ngx.say("===========body===============") ngx.say("body data: ", ngx.req.get_body_data()) ngx.req.read_body() local post_args = ngx.req.get_post_args() for k,v in pairs(post_args) do ngx.say(k, ": ", v) end ngx.say("body data: ", ngx.req.get_body_data()) '; }
测试结果:
root@j9 ~# curl "req?a=11&b=22&c=33" --data "d=11&e=22&f=33" ===========ngx.req.raw_header= POST /req?a=11&b=22&c=33 HTTP/1.1 User-Agent: curl/7.22.0 (x86_64-pc-linux-gnu) libcurl/7.22.0 OpenSSL/1.0.1 zlib/1.2.3.4 libidn/1.23 librtmp/2.3 Host: 127.0.0.1:9898 Accept: */* Content-Length: 14 Content-Type: application/x-www-form-urlencoded ===========headers============ Host: 127.0.0.1:9898 user-agent: curl/7.22.0 (x86_64-pc-linux-gnu) libcurl/7.22.0 OpenSSL/1.0.1 zlib/1.2.3.4 libidn/1.23 librtmp/2.3 ===========all headers======== host: 127.0.0.1:9898 content-type: application/x-www-form-urlencoded accept: */* content-length: 14 user-agent: curl/7.22.0 (x86_64-pc-linux-gnu) libcurl/7.22.0 OpenSSL/1.0.1 zlib/1.2.3.4 libidn/1.23 librtmp/2.3 ===========args=============== b: 22 a: 11 c: 33 ===========body=============== body data: nil d: 11 f: 33 e: 22 body data: d=11&e=22&f=33 root@j9 ~#
- ngx.escape_uri/ngx.unescape_uri
uri编码解码
- ngx.encode_args/ngx.decode_args
参数编码解码
- ngx.encode_base64/ngx.decode_base64
BASE64编码解码
- ngx.md5
md5加密
例子
location ^~ /code { content_by_lua ' local request_uri = ngx.var.request_uri local args = {a=11, b=22} ngx.say("request uri: ", request_uri) ngx.say("unescape request uri: ", ngx.unescape_uri(request_uri)) ngx.say("encode args: ", ngx.encode_args(args)) ngx.say("encode base64 request uri: ", ngx.encode_base64(request_uri)) ngx.say("md5(123456): ", ngx.md5("123456")) '; }
测试结果:
root@j9 ~# curl "code?name=%E9%87%91%E4%B9%9D" request uri: /code?name=%E9%87%91%E4%B9%9D unescape request uri: /code?name=金九 encode args: a=11&b=22 encode base64 request uri: L2NvZGU/bmFtZT0lRTklODclOTElRTQlQjklOUQ= md5(123456): e10adc3949ba59abbe56e057f20f883e root@j9 ~#
- ngx.shared.DICT
共享内存接口,其中DICT为共享内存zone名称,在nginx.conf中通过指令lua_shared_dict配置,而且lua_shared_dict指令配置的共享内存大小最小值为8k。
例子
lua_shared_dict cc_shared_data 16k; server { listen 9999; default_type "text/html"; location ^~ /shared_data { content_by_lua ' local shared_data = ngx.shared.cc_shared_data local i = shared_data:get("i") if not i then shared_data:set("i", 1) end i = shared_data:incr("i", 1) ngx.say("i: ", i) '; } }
测试结果
root@j9 ~# curl "" i: 2 root@j9 ~# curl "" i: 3 root@j9 ~#
ngx.shared.DICT详细说明:[]()
**4.2.2、指令**
指令 | 阶段 | 范围 | 说明 |
init_by_lua/init_by_lua_file | loading-config | http | nginx master进程加载配置时执行;通常用于初始化全局配置/预加载Lua模块 |
init_worker_by_lua/init_worker_by_lua_file | starting-worker | http | 每个nginx worker进程启动时调用的计时器,如果master进程不允许则只会在init_by_lua之后调用;通常用于定时拉取配置/数据,或者后端服务的健康检查 |
set_by_lua/set_by_lua_file | rewrite | server,server if,location,location if | 设置nginx变量,可以实现复杂的赋值逻辑;此处是阻塞的,Lua代码要做到非常快 |
rewrite_by_lua/rewrite_by_lua_file | rewrite tail | http,server,location,location ifrewrite | 阶段处理,可以实现复杂的转发/重定向逻辑 |
access_by_lua/access_by_lua_file | access tail | http,server,location,location if | 请求访问阶段处理,用于访问控制 |
content_by_lua/content_by_lua_file | content | location,location if | 内容处理器,接收请求处理并输出响应 |
header_filter_by_lua/header_filter_by_lua_file | output-header-filter | http,server,location,location if | 设置header和cookie |
body_filter_by_lua/body_filter_by_lua_file | output-body-filter | http,server,location,location if | 对响应数据进行过滤,比如截断、替换 |
log_by_lua/log_by_lua_file | log | http,server,location,location iflog | 阶段处理,比如记录访问量/统计平均响应时间 |
更详细的解释请参考官网:
init_by_lua
每次nginx重新加载配置时执行,可以用它来完成一些耗时模块的加载,或者初始化一些全局配置;
例子:
init_by_lua ' cjson = require("cjson") ngx.log(ngx.ERR, "init_by_lua ok") '; server { listen 9292; default_type "text/html"; location / { content_by_lua ' local arg_json = cjson.decode(ngx.var.arg_json) ngx.say("aa: ", arg_json.aa) '; } }
测试结果:
root@j9 ~# curl '\{"aa":111,"bbb":222\}' aa: 111 root@j9 ~#
**init_worker_by_lua**
每个worker启动之后初始化时执行,通常用于每个worker都要做的工作,比如启动定时任务
例子:
worker_processes 2; http { #这里省略了其他配置 init_worker_by_lua ' ngx.log(ngx.ERR, "test init_worker_by_lua") -- TODO: 启动定时任务 '; }
grep一下error.log,会发现两条包含"test init_worker_by_lua"关键字的log,说明每个worker都会执行这个Lua代码。
set_by_lua
语法:set_by_lua resluascriptstr
arg1 $arg2...; 在Lua代码中可以实现所有复杂的逻辑,但是要执行速度很快,不要阻塞;
需要注意的是,这个指令需要加入模块ngx_devel_kit,否则不支持这个指令。
这个指令的Lua代码中不支持以下API:
1、输出(ngx.say、ngx.send_headers……)
2、控制(ngx.exit……)
3、子请求(ngx.location.capture、ngx.location.capture_multi……)
4、cosocket(ngx.socket.tcp、ngx.req.socket……)
5、ngx.sleep
例子:
server { listen 9393; default_type "text/html"; location /add { set $diff ''; set $double_c ''; set_by_lua $sum ' local a = ngx.var.arg_a local b = ngx.var.arg_b ngx.var.diff = a - b ngx.var.double_c = 2 * tonumber(ngx.arg[1]) return a + b; ' $arg_c; return 200 "a + b = $sum, a - b = $diff, 2 * c = $double_c"; } }
测试结果:
root@j9 ~# curl "" a + b = 33, a - b = -11, 2 * c = 176
rewrite_by_lua
执行内部URL重写或者外部重定向(301或者302),典型的如伪静态化的URL重写。其默认执行在rewrite处理阶段的最后。
需要注意的是:
1、在长连接中如果调用了ngx.exit(200)一个请求,则需要调用ngx.req.discard_body(),否则nginx可能会把当前请求的body当成header解析,从而导致400错误返回码并且长连接被关闭。
2、如果该阶段调用了ngx.exit(ngx.OK),content_by_lua阶段仍然能得到执行。
例子:
server { listen 9494; default_type "text/html"; location /rewrite_by_lua { set $a 11; rewrite_by_lua ' ngx.var.a = "aa" if ngx.var.arg_exit == "ok" then ngx.exit(ngx.OK) else ngx.exit(200) end '; content_by_lua ' ngx.say("a: ", ngx.var.a) '; } }
测试结果
root@j9 ~# curl "" a: aa root@j9 ~# curl "" root@j9 ~# access_by_lua
用于访问控制,比如IP黑白名单限制、鉴权。
例子:
server { listen 9595; default_type "text/html"; location / { access_by_lua ' local auth = ngx.var.arg_auth; local key = "alicdnj9"; if ngx.md5(key) ~= auth then return ngx.exit(403) end '; content_by_lua ' ngx.say("access ok") '; } }
测试结果:
root@j9 ~# curl ""403 Forbidden 403 Forbidden
You don't have permission to access the URL on this server. Sorry for the inconvenience.
Please report this message and include the following information to us.
Thank you very much!
URL: | |
Server: | j9 |
Date: | 2015/10/27 16:47:20 |
注意,如果在access_by_lua中调用ngx.exit(ngx.OK),content阶段仍然能得到执行。
content_by_lua
content阶段,注意在同一个Location中不要和其他content阶段指令一起使用,比如proxy_pass。
例子:略
header_filter_by_lua和body_filter_by_lua
分别为header_filter阶段和body_filter阶段,其中body_filter可能会被执行多次。
不支持以下API:
1. 输出 (ngx.say、ngx.send_headers)
2. 控制 (ngx.exit、ngx.exec)
3. 子请求 (ngx.location.capture、ngx.location.capture_multi)
4. Cosocket (ngx.socket.tcp、ngx.req.socket).
比如对后端chunked长度做限制:
server { listen 9696; default_type "text/html"; set $content_len 0; location / { header_filter_by_lua ' -- 先去掉Content-Length头部,转成Chunked传输 ngx.header.content_length = nil '; body_filter_by_lua ' local content_length = #ngx.arg[1] content_length = ngx.var.content_len + content_length ngx.var.content_len = content_length -- 最多只能传输10字节的body,否则直接关掉连接 if content_length > 10 then return ngx.ERROR end '; content_by_lua ' for i=1, ngx.var.arg_len do ngx.print("a") end '; } }
测试结果
root@j9 ~# curl "" -i HTTP/1.1 200 OK Server: Tengine/2.2.0 Date: Mon, 26 Oct 2015 01:48:23 GMT Content-Type: text/html Transfer-Encoding: chunked Connection: keep-alive aaaaaaaaaa root@j9 ~# curl "" -i curl: (52) Empty reply from server root@j9 ~#
可以看出当参数len为11时,服务器就直接不返回数据了。
**4.3、深入**
1、content_by_lua中的代码一定要注意单引号或者双引号,一般用法是外单内双,或者外双内单。
2、在nginx_lua中值为nil的变量不能与字符串或者数字相加,否则nginx会报500错误。
3、lua调试: ngx.log(ngx.ERR,xx)。(tail -f logs/error.log)
4、*_by_lua_file指令指定的文件支持绝对路径和相对路径,其中相对路径是相对nginx工作目录。
5、lua文件的require函数指定的lua模块路径查找顺序,可以从出错信息中看出来:
no file '/opt/libs/lua/a.lua'
no file './a.lua'
no file '/usr/local/luajit/share/luajit-2.0.4/a.lua'
no file '/usr/local/share/lua/5.1/a.lua'
no file '/usr/local/share/lua/5.1/a/init.lua'
no file '/usr/local/luajit/share/lua/5.1/a.lua'
no file '/usr/local/luajit/share/lua/5.1/a/init.lua'
no file './a.so'
no file '/usr/local/lib/lua/5.1/a.so'
no file '/usr/local/luajit/lib/lua/5.1/a.so'
no file '/usr/local/lib/lua/5.1/loadall.so'
其中,第一个/opt/libs/lua/a.lua为lua_package_path指定的路径:lua_package_path '/opt/libs/lua/?.lua;;';
第二个./a.lua为相对路径,相对于nginx.conf配置文件,而非包含它的lua文件。
so模块查找顺序类似,但是先查找.lua再查找.so,查找.so时先在lua_package_cpah指定的路径查找:lua_package_cpath '/opt/libs/lua_shared/?.so;;';
可以从出错信息中看出来:
no field package.preload['a']
no file '/opt/libs/lua/a.lua'
no file './a.lua'
no file '/usr/local/luajit/share/luajit-2.0.4/a.lua'
no file '/usr/local/share/lua/5.1/a.lua'
no file '/usr/local/share/lua/5.1/a/init.lua'
no file '/usr/local/luajit/share/lua/5.1/a.lua'
no file '/usr/local/luajit/share/lua/5.1/a/init.lua'
no file '/opt/libs/lua_shared/a.so'
no file './a.so'
no file '/usr/local/lib/lua/5.1/a.so'
no file '/usr/local/luajit/lib/lua/5.1/a.so'
no file '/usr/local/lib/lua/5.1/loadall.so'
6、lua代码一定要健壮,否则不管lua产生什么错,nginx都会返回500错误,这时可以从error.log中查看错误信息来定位。
7、编写lua代码时最好用local局部变量,不要用全局变量。
8、实现worker级别的全局变量:
server { listen 9797; default_type "text/html"; location / { content_by_lua ' local a = 1 local b = {b = 1} local status = require("status") ngx.say("a: ", a, ", b: ", b.b, " counter: ", status.counter) a = a + 1 b.b = b.b + 1 status.counter = (status.counter or 0) + 1 '; } }
其中status.lua为:
local m = {} m.counter = 1 return m
测试结果:
root@j9 ~# curl "" a: 1, b: 1 counter: 1 root@j9 ~# curl "" a: 1, b: 1 counter: 2 root@j9 ~# curl "" a: 1, b: 1 counter: 3 root@j9 ~#
可以看出status.counter的值一直是累加的,这是因为require一个模块只load第一次,后续require该模块都会先看全局表中是否已经load过,load过则就不需要再load了,所以status.counter累加其实是累加m.counter。
**9、定时任务**
API: ok, err = ngx.timer.at(delay, callback, user_arg1, user_arg2, ...)
例子:
local delay = 5 local handler handler = function (premature) -- do some routine job in Lua just like a cron job if premature then return end local ok, err = ngx.timer.at(delay, handler) if not ok then ngx.log(ngx.ERR, "failed to create the timer: ", err) return end end local ok, err = ngx.timer.at(delay, handler) if not ok then ngx.log(ngx.ERR, "failed to create the timer: ", err) return end
注意:在timer处理函数的上下文中不能调用ngx.var.*、ngx.req.*、子请求API、输出API,因为这些API只能在请求上下文中生效。
**10、子请求**
API:res = ngx.location.capture(uri, options?)
上下文:rewrite_by_lua*, access_by_lua*, content_by_lua*
例子:
正向代理,当源站返回301或者302时代理客户端跳转
server { listen 8181; default_type "text/html"; location /test { content_by_lua ' local res = ngx.location.capture("/get" .. ngx.var.request_uri, { method = ngx.HTTP_HEAD }) if res.status == 200 then ngx.exec("/get" .. ngx.var.request_uri) elseif res.status == 301 or res.status == 302 then location = res.header["Location"] local m, err = ngx.re.match(location, "http://([^/]+)(/.*)") if not m then ngx.exit(500) end host = m[1] uri = m[2] ngx.exec("/redirect/" .. host .. "/" .. ngx.var.request_uri) else ngx.exit(res.status) end '; } location ~ /redirect/([^/]*)/([^/]*) { proxy_pass } location /get { if ($arg_tag = "1") { return 302 ""; } return 200 "ok"; } } server { listen 8282; default_type "text/html"; location / { return 200 "redirect ok, args: $args"; } }
测试结果:
root@j9 ~# curl "" -i HTTP/1.1 200 OK Server: Tengine/2.2.0 Date: Mon, 26 Oct 2015 15:17:10 GMT Content-Type: text/html Content-Length: 2 Connection: keep-alive ok root@j9 ~# curl "" -i HTTP/1.1 200 OK Server: Tengine/2.2.0 Date: Mon, 26 Oct 2015 15:17:14 GMT Content-Type: text/html Content-Length: 19 Connection: keep-alive redirect ok, args: tag=1 root@j9 ~#
可见,当传tag为1时,返回的值就是想要的值,不需要再302重定向了。
注意,子请求只能请求本server的非@location。
另外一个需要注意的是,发起子请求之前修改的变量在子请求的location中是获取不到的,这是因为变量的上下文是在请求结构体r中,而子请求是挂在主请求下面,是两个不同的请求。
实验:
server { listen 8383; default_type "text/html"; set $cc "cc"; location /test { content_by_lua ' ngx.var.cc = "11" local res = ngx.location.capture("/get" .. ngx.var.request_uri) if res.status == 200 then ngx.say(res.body) ngx.say("test cc: ", ngx.var.cc) else ngx.exit(res.status) end '; } location /get { content_by_lua ' ngx.say("get cc: ", ngx.var.cc) ngx.var.cc = "22" '; } }
结果:
root@j9 ~# curl "" get cc: cc test cc: 11 root@j9 ~#
11、location @xx
server { listen 8484; default_type "text/html"; set $cc "2"; location / { content_by_lua ' ngx.var.cc = "5"; if ngx.var.arg_location == "at" then ngx.exec("@cc") else ngx.exec("/cc") end '; } location @cc { return 200 "this is @cc location, cc: $cc"; } location /cc { return 200 "this is /cc location, cc: $cc"; } }
测试结果:
root@j9 ~# curl "" this is @cc location, cc: 5 root@j9 ~# curl "" this is /cc location, cc: 2 root@j9 ~#
在ngx.exec跳转之前已经把变量cc的值改成5了,但可以看出这两种跳转方式变量cc的值不一样,这是因为ngx.exec跳转到@cc这个location时,从location rewrite阶段开始执行,而跳转到/cc这个location时是从server rewrite阶段开始执行,而set指令是在server块,就是在这个阶段得到执行的,所以$cc又被赋值成2了。