Nginx 关于 Rewrite 执行顺序详解

Nginx关于Rewrite执行顺序详解

第一篇：break和last的区别

Rewrite 模块概述

REFER: http://wiki.nginx.org/NginxHttpRewriteModule#rewrite

If the directives of this module are given at the server level,then they are carried out before the location of the request is determined. Ifin that selected location there are further rewrite directives, then they alsoare carried out. If the URI
changed as a result of the execution of directivesinside location, then location is again determined for the new URI.

This cycle can be repeated up to 10 times, after which Nginxreturns a 500 error.

Rewrite（URL重写）指令可以出现在server{}下，也可以出现在location{}下，它们之间是有区别的！对于出现在server{}下的rewrite指令，它的执行会在location匹配之前；对于出现在location{}下的rewrite指令，它的执行当然是在location匹配之后，但是由于rewrite导致HTTP请求的URI发生了变化，所以location{}下的rewrite后的URI又需要重新匹配location，就好比一个新的HTTP请求一样（注意由location{}内的rewrite导致的这样的循环匹配最多不超过10次，否则nginx会报500错误）。总的来说，如果server{}和location{}下都有rewrite，依然是先执行server{}，然后进行location匹配，如果被匹配的location{}之内还有rewrite指令，那么继续执行rewrite，同时因为location{}内的rewrite改变了URI，那么重写后的结果URI需要当做一个新的请求，重新匹配location（应该也包括重新执行server{}下的rewrite吧）。

Last与break flag的区别

关于last flag和break flag的区别，官方文档的描述是：“last - completes processingof rewrite directives, after which searches for corresponding URI and location”和“break - completes processingof rewrite directives”，都有“不让继续执行后面的rewrite指令”的含义，但是两者的区别并没有展开。

这里我用实验来告诉大家区别。实验准备：

1、 安装nginx；（如果对安装和location不了解的，请参考： http://eyesmore.iteye.com/blog/1141660）
2、 在nginx安装目录的html子目录下创建4个文件，分别叫：aaa.html，bbb.html，ccc.html和ddd.html，文件内容分别是各自的文件名（例aaa.html文件内容不妨写aaa html file）。

3、 Nginx配置文件初始化是：

error_log logs/error.log info; #URL重写模块的日志会写入此文件

server {

listen 9090;

server_name localhost;

root html;

rewrite_log on; #打开URL重写模块的日志开关，以便写入error_log

location /aaa.html {

rewrite"^/aaa\.html$" /bbb.html;

rewrite"^/bbb\.html$" /ddd.html;

}

location /bbb.html {

rewrite "^/bbb\.html$"/ccc.html;

}

}

上述配置注意两点：1、打开rewrite模块的日志开关，以便rewrite执行日志写入error_log（注：rewrite日志写入error_log的级别是notice，所以要注意error_log日志级别，此处用info）；2、定义了两个location，分别是/aaa.html和/bbb.html，但是在/aaa.html中，把/aaa.html重写成/bbb.html，接着又把/bbb.html重写成/ddd.html；在/bbb.html中，把/bbb.html重写成/ccc.html。

[测试1] 没有last和break标记时：请求aaa.html

[root@web108 ~]# curlhttp://localhost:9090/aaa.html

dddhtml file

[root@web108 ~]#

Error_log的日志内容：

2011/08/07 22:13:23 [notice] 9066#0: *85 "^/aaa\.html$" matches "/aaa.html",client: 127.0.0.1, server: localhost, request: "GET /aaa.htmlHTTP/1.1", host: "localhost:9090"

2011/08/07 22:13:23 [notice] 9066#0: *85 rewritten data: "/bbb.html",args: "", client: 127.0.0.1, server: localhost, request: "GET/aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/07 22:13:23 [notice] 9066#0: *85 "^/bbb\.html$" matches"/bbb.html", client: 127.0.0.1, server: localhost, request:"GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/07 22:13:23 [notice] 9066#0: *85 rewritten data: "/ddd.html",args: "", client: 127.0.0.1, server: localhost, request: "GET/aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/07 22:13:23 [info] 9066#0: *85client 127.0.0.1 closed keepalive connection

URL重写模块的日志告诉我们：对于一个HTTP请求“GET/aaa.html”，重写过程是：先/aaa.html被重写为/bbb.html；然后rewritten data: /bbb.html，继续执行后面的rewrite指令，进而被重写为/ddd.html，然后rewritterndata: /ddd.html 后面没有重写了（其实此时/ddd.html需要再次重新匹配location的，只是日志没有体现出来，接下来的测试2会体现这点），于是输出/ddd.html的内容。

[测试2]使用last标记时：请求aaa.html

将上述location /aaa.html {} 修改成：

location /aaa.html {

rewrite "^/aaa\.html$" /bbb.html last;

rewrite "^/bbb\.html$" /ddd.html;

}

测试结果：

[root@web108 ~]# curlhttp://localhost:9090/aaa.html

ccchtml file

[root@web108 ~]#

Error_log日志：

2011/08/07 22:24:31 [notice] 18569#0: *86 "^/aaa\.html$" matches"/aaa.html", client: 127.0.0.1, server: localhost, request:"GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/07 22:24:31 [notice] 18569#0: *86 rewritten data: "/bbb.html",args: "", client: 127.0.0.1, server: localhost, request: "GET/aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/07 22:24:31 [notice] 18569#0: *86 "^/bbb\.html$" matches"/bbb.html", client: 127.0.0.1, server: localhost, request:"GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/07 22:24:31 [notice] 18569#0: *86 rewritten data: "/ccc.html",args: "", client: 127.0.0.1, server: localhost, request: "GET/aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/07 22:24:31 [info] 18569#0: *86client 127.0.0.1 closed keepalive connection

不知道读者看到GET /aaa.html显示的结果“ccc html file”会不会惊讶：“为什么结果不是bbb html file”。下面解释下整个过程：首先/aaa.html匹配了location /aaa.html {}，于是执行rewrite "^/aaa\.html$" /bbb.html last，把/aaa.html重写为/bbb.html，同时由于last flag的使用，后面的rewrite指令（指的是rewrite "^/bbb\.html$" /ddd.html）不会被执行。似乎此时应该输出“bbb
html file”才对，但是我们看看nginx官方解释：“last - completes processingof rewrite directives, after which searches for corresponding URI and location”意思是说last不再匹配后面的rewrite指令，但是紧接着需要对重写后的URI重新匹配location。让我们再看看官方的“If the directives of this module are given at the server
level, thenthey are carried out before the location of the request is determined. If inthat selected location there are further rewrite directives, then they also arecarried out.
If the URI changed as aresult of the execution of directives inside location, then location is againdetermined for the new URI. This cycle can be repeated up to 10 times,after which Nginx returns a 500 error.”因此，重新匹配的时候，匹配到了新的location/bbb.html
{}，执行“rewrite "^/bbb\.html$" /ccc.html”，最后的内容是“ccchtml file”。

[测试3]使用break标记时：请求aaa.html

将上述location /aaa.html {} 修改成使用break标记：

location /aaa.html {

rewrite "^/aaa\.html$" /bbb.html break;

rewrite "^/bbb\.html$" /ddd.html;

}

测试结果：

[root@web108 ~]# curlhttp://localhost:9090/aaa.html

bbb html file

[root@web108 ~]#

日志结果：

2011/08/07 22:37:49 [notice] 21069#0: *89 "^/aaa\.html$" matches"/aaa.html", client: 127.0.0.1, server: localhost, request:"GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/07 22:37:49 [notice] 21069#0: *89 rewritten data: "/bbb.html",args: "", client: 127.0.0.1, server: localhost, request: "GET/aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/07 22:37:49 [info] 21069#0: *89client 127.0.0.1 closed keepalive connection

我想这个结果不用多做解释了，充分体现了break和last的区别：“last - completes processingof rewrite directives, after which searches for corresponding URI and location”和“break - completes processingof rewrite directives”。Break和last都能阻止继续执行后面的rewrite指令，但是last如果在location下用的话，对于重写后的URI会重新匹配location，但是break则不会重新匹配location。简单的说，break终止的力度比last更加彻底（为了记忆的方便，我们可以把重新后的URI重新匹配location理解为“URI匹配location的循环语句的下一次迭代”，高级程序设计里面break一般用做退出循环，所以break不仅终止继续执行rewrite，而且退出URI重新匹配location的循环迭代）。

Nginx关于Rewrite的迭代第二篇

例题1

配置：

error_log logs/error.log info;

server {

listen 9090;

server_name localhost;

root html;

rewrite_log on;

rewrite "^/aaa\.html$" /bbb.html;

location /ccc.html {

rewrite "^/ccc\.html$" /eee.html;

}

location /bbb.html {

rewrite "^/bbb\.html$" /ccc.html;

rewrite "^/ccc\.html$" /ddd.html;

}

}

结果：

[root@web108 ~]# curlhttp://localhost:9090/aaa.html

ddd html file

[root@web108 ~]#

日志：

2011/08/08 10:05:41 [notice] 31592#0: *90 "^/aaa\.html$" matches"/aaa.html", client: 127.0.0.1, server: localhost, request:"GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:05:41 [notice] 31592#0: *90 rewritten data: "/bbb.html",args: "", client: 127.0.0.1, server: localhost, request: "GET/aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:05:41 [notice] 31592#0: *90 "^/bbb\.html$" matches"/bbb.html", client: 127.0.0.1, server: localhost, request:"GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:05:41 [notice] 31592#0: *90 rewritten data: "/ccc.html",args: "", client: 127.0.0.1, server: localhost, request: "GET/aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:05:41 [notice] 31592#0: *90 "^/ccc\.html$" matches"/ccc.html", client: 127.0.0.1, server: localhost, request:"GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:05:41 [notice] 31592#0: *90 rewritten data: "/ddd.html",args: "", client: 127.0.0.1, server: localhost, request: "GET/aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:05:41 [notice] 31592#0: *90 "^/aaa\.html$" does not match"/ddd.html", client: 127.0.0.1, server: localhost, request:"GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:05:41 [info] 31592#0: *90client 127.0.0.1 closed keepalive connection

解释：

GET /aaa.html请求，首先执行server级的rewrite指令，被重写为/bbb.html，然后匹配到location/bbb.html {}，接着执行location级的rewrite指令，先重写为/ccc.html，再重写为/ddd.html；由于URI被location级的rewrite指令重写了，因此需要重新进行location的匹配，相当于重写后的URI被当做一个新的请求，会重新执行server级的rewrite，然后重新匹配location，日志“2011/08/08 10:05:41
[notice] 31592#0: *90 "^/aaa\.html$" doesnot match "/ddd.html", client: 127.0.0.1, server: localhost,request: "GET /aaa.html HTTP/1.1", host: "localhost:9090"”体现了重新匹配location的流程。

例题2

配置：

error_log logs/error.log info;

server {

listen 9090;

server_name localhost;

root html;

rewrite_log on;

rewrite "^/aaa\.html$" /bbb.html;

rewrite"^/ccc\.html$" /ddd.html;

location /bbb.html {

rewrite "^/bbb\.html$" /ccc.html;

}

location /ddd.html {

rewrite "^/ddd\.html$" /eee.html;

}

}

结果：

[root@web108 ~]# curlhttp://localhost:9090/aaa.html

eee html file

[root@web108 ~]#

日志：

2011/08/08 10:21:00 [notice] 2218#0: *91 "^/aaa\.html$" matches "/aaa.html",client: 127.0.0.1, server: localhost, request: "GET /aaa.htmlHTTP/1.1", host: "localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91 rewritten data: "/bbb.html",args: "", client: 127.0.0.1, server: localhost, request: "GET/aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91 "^/ccc\.html$"does not match "/bbb.html", client: 127.0.0.1, server:localhost, request: "GET /aaa.html HTTP/1.1", host:"localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91 "^/bbb\.html$" matches"/bbb.html", client: 127.0.0.1, server: localhost, request:"GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91 rewritten data: "/ccc.html",args: "", client: 127.0.0.1, server: localhost, request: "GET/aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91 "^/aaa\.html$"does not match "/ccc.html", client: 127.0.0.1, server:localhost, request: "GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91 "^/ccc\.html$"matches "/ccc.html", client: 127.0.0.1, server: localhost,request: "GET /aaa.html HTTP/1.1", host: "localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91rewritten data: "/ddd.html", args: "", client:127.0.0.1, server: localhost, request: "GET /aaa.html HTTP/1.1",host: "localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91"^/ddd\.html$" matches "/ddd.html", client:127.0.0.1, server: localhost, request: "GET /aaa.html HTTP/1.1",host: "localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91 rewrittendata: "/eee.html", args: "", client: 127.0.0.1,server: localhost, request: "GET /aaa.html HTTP/1.1", host:"localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91 "^/aaa\.html$"does not match "/eee.html", client: 127.0.0.1, server:localhost, request: "GET /aaa.html HTTP/1.1", host:"localhost:9090"

2011/08/08 10:21:00 [notice] 2218#0: *91 "^/ccc\.html$"does not match "/eee.html", client: 127.0.0.1, server:localhost, request: "GET /aaa.html HTTP/1.1", host:"localhost:9090"

2011/08/08 10:21:00 [info] 2218#0: *91client 127.0.0.1 closed keepalive connection

解释：

第一次迭代location匹配

GET /aaa.html，首先执行server级的重写，“rewrite"^/aaa\.html$" /bbb.html”把/aaa.html重写为/bbb.html，但/bbb.html没匹配上“rewrite"^/ccc\.html$" /ddd.html”，最终保留/bbb.html；接着，匹配location/bbb.html {}，执行location级的rewrite指令，把/bbb.html重写为/ccc.html，由于URI被location级rewrite重写，因此需要重新迭代location匹配。

第二次迭代location匹配

对于第一次迭代结果/ccc.html，首先依然是执行server级的rewrite指令，“rewrite "^/aaa\.html$" /bbb.html;”跟/ccc.html不匹配，但“rewrite "^/ccc\.html$" /ddd.html;”把/ccc.html重写为/ddd.html；server级rewrite执行完后，接着location匹配，/ddd.html匹配到location /ddd.html {}，执行location级的rewrite指令，把/ddd.html重写为/eee.html。同样由于URI被location级的rewrite指令重写，于是需要重新迭代location匹配。

第三次迭代location匹配

对于第二次迭代结果/eee.html，首先依然执行server级的rewrite指令，“rewrite "^/aaa\.html$" /bbb.html;”和“rewrite "^/ccc\.html$" /ddd.html;”，只不过它们都没匹配上/eee.html，接着/eee.html进行location匹配，也没有，最终结果是/eee.html，返回“eee html file”页面。

最后说明下，如果把上述配置修改成server级rewrite和location的编辑顺序调整：

server {

listen 9090;

server_name localhost;

root html;

rewrite_log on;

location /bbb.html {

rewrite "^/bbb\.html$" /ccc.html;

}

location /ddd.html {

rewrite "^/ddd\.html$" /eee.html;

}

rewrite "^/aaa\.html$" /bbb.html;

rewrite "^/ccc\.html$" /ddd.html;

}

结果是不会受影响的，也就是说location匹配迭代总是先执行server级rewrite，再进行location匹配，再执行location级的rewrite，如果URI因location级rewrite指令重写，则需要进行下一次迭代。但总的迭代次数不超过10次，否则nginx报500错误。

简单伪代码描述下rewrite执行过程：

boolean match_finish = false;

int match_count = 0;

while(!match_finish && match_count < 10) {

match_count ++;

（1）按编辑顺序执行server级的rewrite指令；

（2）按重写后的URI匹配location；

（3）

String uri_before_location = uri;

按编辑顺序执行location级的rewrite指令；

String uri_after_location = rewrite(uri);

if(uri_before_location != uri_after_location) {

match_finish = false;

} else {

match_finish = true;

}

if(location rewrite has last flag) {

continue;//表示不执行后面的rewrite，直接进入下一次迭代

}

if(location rewrite has break flag) {

break;//表示不执行后面的rewrite，并退出循环迭代

}

}

if(match_count <= 10) {

return HTTP_200;

} else {

return HTTP_500;

}