【nginx】常见的陷阱和错误

很多人都可以碰到一个陷阱。下面我们列出，我们经常看到的问题，以及解释如何解决这些问题。在Freenode上的#nginx IRC频道这些讨论很频繁。

1、权限

从来不要使用777权限，查看目录的权限

namei -om /path/to/check

2、root设置

BAD：

server {
    server_name www.example.com;
    location / {
        root /var/www/nginx-default/;
        # [...]
      }
    location /foo {
        root /var/www/nginx-default/;
        # [...]
    }
    location /bar {
        root /var/www/nginx-default/;
        # [...]
    }
}

GOOD：

server {
    server_name www.example.com;
    root /var/www/nginx-default/;
    location / {
        # [...]
    }
    location /foo {
        # [...]
    }
    location /bar {
        # [...]
    }
}

3、索引设置

BAD:

http {
    index index.php index.htm index.html;
    server {
        server_name www.example.com;
        location / {
            index index.php index.htm index.html;
            # [...]
        }
    }
    server {
        server_name example.com;
        location / {
            index index.php index.htm index.html;
            # [...]
        }
        location /foo {
            index index.php;
            # [...]
        }
    }
}

GOOD:

http {
    index index.php index.htm index.html;
    server {
        server_name www.example.com;
        location / {
            # [...]
        }
    }
    server {
        server_name example.com;
        location / {
            # [...]
        }
        location /foo {
            # [...]
        }
    }
}

4、Using If

if 是邪恶的 参见 If Is Evil

5、Server Name (If)

BAD:

server {
    server_name example.com *.example.com;
        if ($host ~* ^www\.(.+)) {
            set $raw_domain $1;
            rewrite ^/(.*)$ $raw_domain/$1 permanent;
        }
        # [...]
    }
}

每次都要检测主机头，这是低效的，你应该避免，推荐使用下面的

GOOD:

server {
    server_name www.example.com;
    return 301 $scheme://example.com$request_uri;
}
server {
    server_name example.com;
    # [...]
}

这样方式便于阅读，降低了nginx的处理要求，而且也避免了硬编码（http or https）

6、Check (If) File Exists

使用if来判断是可怕的，你应该使用 try_files

BAD:

server {
    root /var/www/example.com;
    location / {
        if (!-f $request_filename) {
            break;
        }
    }
}

GOOD:

server {
    root /var/www/example.com;
    location / {
        try_files $uri $uri/ /index.html;
    }
}

try_files 意味着你测试一个队列 $uri => $uri/ => index.html，这种方法简单，而且可以避免if

7、Web Apps中的控制器

Drupal, Joomla, etc. to work, just use this:

try_files $uri $uri/ /index.php?q=$uri&$args;

Note - the parameter names are different based on the package you’re using. For example:

• “q” is the parameter used by Drupal, Joomla, WordPress
• “page” is used by CMS Made Simple

一些软件不需要 query string, 可以读取 REQUEST_URI (例如，WordPress):

try_files $uri $uri/ /index.php;

如果你不关心目录是否存在，你可以移除 $uri/

8、Passing Uncontrolled Requests to PHP

很多PHP网站中，配置nginx的例子中建议使用 .php （to the PHP interpretet）作为uri的结尾，这例有一个严重的安全问题对于大多数PHP程序，因为它可能允许执行任何第三方代码

The problem section usually looks like this:

location ~* \.php$ {
    fastcgi_pass backend;
    # [...]
}

Here, every request ending in .php will be passed to the FastCGI backend. The issue with this is that the default PHP configuration tries to guess which file you want to execute if the full path does not lead to an actual file on the filesystem.

For instance, if a request is made for /forum/avatar/1232.jpg/file.php which does not exist but if/forum/avatar/1232.jpg does, the PHP interpreter will process /forum/avatar/1232.jpg instead. If this contains embedded PHP code, this code will be executed accordingly.

Options for avoiding this are:

• Set cgi.fix_pathinfo=0 in php.ini. This causes the PHP interpreter to only try the literal path given and to stop processing if the file is not found.
• Ensure that NGINX only passes specific PHP files for execution:

location ~* (file_a|file_b|file_c)\.php$ {
    fastcgi_pass backend;
    # [...]
}

• 在上传目录禁止执行任何PHP代码

location /uploaddir {
    location ~ \.php$ {return 403;}
    # [...]
}

• 使用try_files指令过滤

location ~* \.php$ {
    try_files $uri =404;
    fastcgi_pass backend;
    # [...]
}

• 使用嵌套位置过滤

location ~* \.php$ {
    location ~ \..*/.*\.php$ {return 404;}
    fastcgi_pass backend;
    # [...]
}

9、FastCGI Path in Script Filename

尽量使用 include fastcgi_params 中的变量，不管什么语言都是一样

GOOD:

fastcgi_param  SCRIPT_FILENAME    $document_root$fastcgi_script_name;

BAD:

fastcgi_param  SCRIPT_FILENAME    /var/www/yoursite.com/$fastcgi_script_name;

10、Taxing Rewrites

我们应该努力让他们保持整洁。很简单，不添加冗余代码。

BAD:

rewrite ^/(.*)$ http://example.com/$1 permanent;

GOOD:

rewrite ^ http://example.com$request_uri? permanent;

BETTER:

return 301 http://example.com$request_uri;

通过使用内置的变量$ REQUEST_URI，我们可以有效地避免做任何捕获或匹配的。

11、Rewrite Missing http://

很简单，除非你告诉NGINX他们不是重写是相对的。一个重写绝对很简单。添加一个scheme

BAD:

rewrite ^ example.com permanent;

GOOD:

rewrite ^ http://example.com permanent;

添加 http:// 到重写规则内，简单，高效

12、Proxy Everything

BAD:

server {
    server_name _;
    root /var/www/site;
    location / {
        include fastcgi_params;
        fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
        fastcgi_pass unix:/tmp/phpcgi.socket;
    }
}

Yucky. In this instance, you pass EVERYTHING to PHP. Why? Apache might do this, you don’t need to. Let me put it this way... The try_files directive exists for an amazing reason. It tries files in a specific order. This means that NGINX can first try to server the static content. If it can’t, then it moves on. This means PHP doesn’t get involved at all. MUCH faster. Especially if you’re serving a 1MB image over PHP a few thousand times versus serving it directly. Let’s take a look at how to do that.

GOOD:

server {
    server_name _;
    root /var/www/site;
    location / {
        try_files $uri $uri/ @proxy;
    }
    location @proxy {
        include fastcgi_params;
        fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
        fastcgi_pass unix:/tmp/phpcgi.socket;
    }
}

Also GOOD:

server {
    server_name _;
    root /var/www/site;
    location / {
        try_files $uri $uri/ /index.php;
    }
    location ~ \.php$ {
        include fastcgi_params;
        fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
        fastcgi_pass unix:/tmp/phpcgi.socket;
    }
}

It’s easy, right? You see if the requested URI exists and can be served by NGINX. If not, is it a directory that can be served. If not, then you pass it to your proxy. Only when NGINX can’t serve that requested URI directly does your proxy overhead get involved.

Now.. consider how much of your requests are static content, such as images, css, javascript, etc. That’s probably a lot of overhead you just saved.

12、Config Changes Not Reflected

Browser cache. Your configuration may be perfect but you’ll sit there and beat your head against a cement wall for a month. What’s wrong is your browser cache. When you download something, your browser stores it. It also stores how that file was served. If you are playing with a types{} block you’ll encounter this.

The fix:

• In Firefox press Ctrl+Shift+Delete, check Cache, click Clear Now. In any other browser just ask your favorite search engine. Do this after every change (unless you know it’s not needed) and you’ll save yourself a lot of headaches.
• Use curl.

13、VirtualBox

If this does not work, and you’re running NGINX on a virtual machine in VirtualBox, it may be sendfile() that is causing the trouble. Simply comment out the sendfile directive or set it to “off”. The directive is most likely found in your nginx.conf file.:

sendfile off;

13、Missing (disappearing) HTTP Headers

If you do not explicitly set underscores_in_headers on, NGINX will silently drop HTTP headers with underscores (which are perfectly valid according to the HTTP standard). This is done in order to prevent ambiguities when mapping headers to CGI variables as both dashes and underscores are mapped to underscores during that process.

14、Not Using Standard Document Root Locations

Some directories in any file system should never be used for hosting data from. Some of these include / androot. You should never use these as your document root.

Doing this leaves you open to a request outside of your expected area returning private data.

NEVER DO THIS!!! (yes, we have seen this)

server {
    root /;

    location / {
        try_files /web/$uri $uri @php;
    }

    location @php {
        [...]
    }
}

When a request is made for /foo, the request is passed to php because the file isn’t found. This can appear fine, until a request in made for /etc/passwd. Yup, you just gave us a list of all users on that server. In some cases, the NGINX server is even set up run workers as root. Yup, we now have your user list as well as password hashes and how they’ve been hashed. We now own your box.

The Filesystem Hierarchy Standard defines where data should exist. You should definitely read it. The short version is that you want your web content to exist in either /var/www/, /srv, /usr/share/www.

15、Using the Default Document Root

NGINX packages that exist in Ubuntu, Debian, or other operating systems, as an easy-to-install package will often provide a ‘default’ configuration file as an example of configuration methods, and will often include a document root to hold a basic HTML file.

Most of these packaging systems do not check to see if files are modified or exist within the default document root, which can result in code loss when the packages are upgraded. Experienced system administrators know that there is no expectation of the data inside the default document root to remain untouched during upgrades.

You should not use the default document root for any site-critical files. There is no expectation that the default document root will be left untouched by the system and there is an extremely high possibility that your site-critical data may be lost upon updates and upgrades to the NGINX packages for your operating system.

16、Using a Hostname to Resolve Addresses

BAD:

upstream {
    server http://someserver;
}

server {
    listen myhostname:80;
    # [...]
}

You should never use a hostname in a listen directive. While this may work, it will come with a large number of issues. One such issue being that the hostname may not resolve at boot time or during a service restart. This can cause NGINX to be unable to bind to the desired TCP socket which will prevent NGINX from starting at all.

A safer practice is to know the IP address that needs to be bound to and use that address instead of the hostname. This prevents NGINX from needing to look up the address and removes dependencies on external and internal resolvers.

This same issue applies to upstream locations. While it may not always be possible to avoid using a hostname in an upstream block, it is bad practice and will require careful considerations to prevent issues.

GOOD:

upstream {
    server http://10.48.41.12;
}

server {
    listen 127.0.0.16:80;
    # [...]
}

17、Using SSLv3 with HTTPS

由于SSLv3的POODLE 漏洞，建议使用在SSL网站禁用，仅仅使用TLS协议代替

ssl_protocols TLSv1 TLSv1.1 TLSv1.2;　

原文：https://www.nginx.com/resources/wiki/start/topics/tutorials/config_pitfalls/