Page Discussion History


Revision as of 14:19, 20 January 2012 by Dandv (Talk | contribs)



Controls core features of Nginx's HTTP processing.



syntax: aio [on|off|sendfile]

default: off

context: http, server, location

This directive is usable as of Linux kernel 2.6.22. For Linux it is required to use directio, this automatically disables sendfile support.

location /video {
    aio on; 
    directio 512; 
    output_buffers 1 128k;

In FreeBSD before 5.2.1 and Nginx 0.8.12 you must disable sendfile support.

location /video {
    aio on; 
    sendfile off;
    output_buffers 1 128k;

As of FreeBSD 5.2.1 and Nginx 0.8.12 you can use it along with sendfile.

location /video {
    aio sendfile; 
    sendfile on;
    tcp_nopush on;


syntax: alias file-path|directory-path

default: no

context: location

This directive assigns a path to be used as the basis for serving requests for the indicated location. Note that it may look similar to the root directive at first sight, but the document root doesn't change, just the file system path used for the request. The location part of the request is dropped in the request Nginx issues. Let's see this in action. Consider the following example.

location  /i/ {
  alias  /spool/w3/images/;

A request for "/i/top.gif" will instruct Nginx to serve the file "/spool/w3/images/top.gif". As you can see, only the part of the URI after the location is appended. The location itself, in this case "/i/", is dropped. With a root directive the full path is appended, i.e., in the above example it would have been, "/spool/w3/images/i/top.gif" — hence including also the location "/i/".

Aliases can also be used in a location specified by a regex.

For example:

location ~ ^/download/(.*)$ {
  alias /home/website/files/$1;

The request "/download/book.pdf" will return the file "/home/website/files/book.pdf". Note again that only part of the request URI after the location is appended to the path defined by alias.

It is possible to use variables in the replacement path.


syntax: chunked_transfer_encoding on|off

default: on

context: http, server, location

This directive (0.7.66+) sets whether chunked encoding is enabled in responses (only valid for connections using HTTP 1.1 or later).


syntax: client_body_in_file_only on|off

default: off

context: http, server, location

The directive forces nginx to always store a client request body into a temporary disk file even if the body is actually of 0 size.

Please note that the file will not be removed at request completion if the directive is enabled.

This directive can be used for debugging and for the $r->request_body_file method in the Embedded Perl module.


syntax: client_body_in_single_buffer

default: off

context: http, server, location

The directive(0.7.58+) specifies whether to keep the whole body in a single client request buffer. The directive is recommended when using the variable $request_body to reduce the operations of copying.

Note that when the request body cannot be hold in a single buffer (see client_body_buffer_size), the body will still touch the disk.


syntax: client_body_buffer_size the_size

default: 8k/16k

context: http, server, location

The directive specifies the client request body buffer size.

If the request body is more than the buffer, then the entire request body or some part is written in a temporary file.

The default size is equal to two pages size, depending on platform it is either 8K or 16K.

When the Content-Length request header specifies a smaller size value than the buffer size, then Nginx will use the smaller one. So Nginx will *not* always allocate a buffer of this buffer size for every request.


syntax: client_body_temp_path dir-path [ level1 [ level2 [ level3 ]

default: client_body_temp

context: http, server, location

The directive assigns the directory for storing the temporary files in it with the body of the request.

In the dir-path a hierarchy of subdirectories up to three levels are possible.

For example

client_body_temp_path  /spool/nginx/client_temp 1 2;

The directory structure will be like this:



syntax: client_body_timeout time

default: 60

context: http, server, location

Directive sets the read timeout for the request body from client.

The timeout is set only if a body is not get in one readstep. If after this time the client send nothing, nginx returns error "Request time out" (408).


syntax: client_header_buffer_size size

default: 1k

context: http, server

Directive sets the headerbuffer size for the request header from client.

For the overwhelming majority of requests it is completely sufficient a buffer size of 1K.

However if a big cookie is in the request-header or the request has come from a wap-client the header can not be placed in 1K, therefore, the request-header or a line of request-header is not located completely in this buffer nginx allocate a bigger buffer, the size of the bigger buffer can be set with the instruction large_client_header_buffers.


syntax: client_header_timeout time

default: 60

context: http, server

Directive assigns timeout with reading of the title of the request of client.

The timeout is set only if a header is not get in one readstep. If after this time the client send nothing, nginx returns error "Request time out" (408).


syntax: client_max_body_size size

default: client_max_body_size 1m

context: http, server, location

Directive assigns the maximum accepted body size of client request, indicated by the line Content-Length in the header of request.

If size is greater the given one, then the client gets the error "Request Entity Too Large" (413).

It is necessary to keep in mind that the browsers do not know how to correctly show this error.


syntax: connection_pool_size the_size

default: connection_pool_size the_size 256

context: http (more?)

The directive is used to allocate memory per connection. The pool is used for small allocations. If a block is bigger than pool size or bigger than page size, then it is allocated outside the pool. If there is not enough memory for small allocation inside pool, then a new block of the same pool size is allocated. This directive has only a very small effect. (source http://markmail.org/message/b2kmrluscevimpba)


syntax: default_type MIME-type

default: default_type text/plain

context: http, server, location

Assigns the default MIME-type to be used for files where the standard MIME map doesn't specify anything.

See also types


syntax: directio [size|off]

default: directio off

context: http, server, location

The directive enables use of flags O_DIRECT (FreeBSD, Linux), F_NOCACHE (Mac OS X) or directio() function (Solaris) for reading files with size greater than specified. This directive disables use of sendfile for this request. This directive may be useful for big files:

    directio  4m;


syntax: error_page code [ code... ] [ = | =answer-code ] uri | @named_location

default: no

context: http, server, location, if in location

The directive specifies the URI that will be shown for the errors indicated.


error_page   404          /404.html;
error_page   502 503 504  /50x.html;
error_page   403          http://example.com/forbidden.html;
error_page   404          = @fetch;

Furthermore, it is possible to change the status code of the answer to another, for example:

error_page 404 =200 /empty.gif;
error_page 404 =403 /forbidden.gif;

Additionally you can have your designated error handler determine the returned status code by using = without specifying a status code.

error_page   404 = /404.php;

If there is no need to change URI during redirection it is possible to redirect processing of error pages into a named location:

location / (
    error_page 404 @fallback;
location @fallback (
    proxy_pass http://backend;


syntax: if_modified_since [off|exact|before]

default: if_modified_since exact

context: http, server, location

The directive (0.7.24) defines how to compare time of file modification and time in request header "If-Modified-Since":

  • off — don't check "If-Modified-Since" request header (0.7.34);
  • exact — exact match;
  • before — file modification time should be less than time in "If-Modified-Since" request header.


syntax: internal

default: no

context: location

internal indicates that the matching location can be used only for so called "internal" requests.

For external requests it will return the error "Not found" (404).

Internal requests are the following:

  • requests redirected by the instruction error_page
  • subrequests created by the command include virtual of the "ngx_http_ssi_module" module
  • requests changed by the instruction rewrite of the "ngx_http_rewrite_module" module

An example to prevent clients fetching error pages directly:

error_page 404 /404.html;
location  /404.html {


syntax: keepalive_disable [ msie6 | safari | none ]...

default: msie6 safari

context: http, server, location

Disable keepalive for certain user agents (0.9.0+). By default keepalive is disabled for MS Internet Explorer (older than 6.0 service pack 2) after POST requests, and for Safari. This is because both browsers have issues with handling POST requests with keepalives. If you are running a site that does not use POST anywhere, you may optionally choose to enable keepalive in these browsers.


syntax: keepalive_timeout [ time ]

default: keepalive_timeout 75

context: http, server, location

The first parameter assigns the timeout for keep-alive connections with the client. The server will close connections after this time.

The optional second parameter assigns the time value in the header Keep-Alive: timeout=time of the response. This header can convince some browsers to close the connection, so that the server does not have to. Without this parameter, nginx does not send a Keep-Alive header (though this is not what makes a connection "keep-alive").

The parameters can differ from each other.

Notes on how browsers handle the Keep-Alive header:

  • MSIE and Opera ignore the "Keep-Alive: timeout=<N>" header.
  • MSIE keeps the connection alive for about 60-65 seconds, then sends a TCP RST.
  • Opera keeps the connection alive for a long time.
  • Mozilla keeps the connection alive for N plus about 1-10 seconds.
  • Konqueror keeps the connection alive for about N seconds.


syntax: keepalive_requests n

default: keepalive_requests 100

context: http, server, location

Number of requests which can be made over a keep-alive connection.


syntax: large_client_header_buffers number size

default: large_client_header_buffers 4 4k/8k

context: http, server

Directive assigns the maximum number and size of buffers for large headers to read from client request.

The request line can not be bigger than the size of one buffer, if the client send a bigger header nginx returns error "Request URI too large" (414).

The longest header line of request also must be not more than the size of one buffer, otherwise the client get the error "Bad request" (400).

Buffers are separated only as needed.

By default the size of one buffer is equal to the size of page, depending on platform this either 4K or 8K, if at the end of working request connection converts to state keep-alive, then these buffers are freed.


syntax: limit_except methods {...}

default: no

context: location

Directive limits HTTP-methods, accessible inside location.

For the limitation can be used the directives of modules ngx_http_access_module and ngx_http_auth_basic_module:

limit_except  GET {
  deny   all;


syntax: limit_rate speed

default: no

context: http, server, location, if in location

Directive assigns the speed of transmission of the answer to client. Speed is assigned in the bytes per second. Limitation works only for one connection, i.e., if client opens 2 connections, then total velocity will be 2 times higher then the limit set.

If it is necessary to limit speed for the part of the clients at the server level, based on some kind of condition - then this directive does not apply. Instead you should specify the limit by assigning the value to the $limit_rate variable, as shown below:

server {
  if ($slow) {
    set $limit_rate  4k;

You can also control the rate of individual responses returned by a proxy_pass response (HttpProxyModule) by setting the X-Accel-Limit-Rate header (XSendfile). This can be done without a X-Accel-Redirect header.


syntax: limit_rate_after time

default: limit_rate_after 1m

context: http, server, location, if in location

The directive limits speed only after the first part was sent.


syntax: lingering_close on|off|always

default: lingering_close on

context: http, server, location

Sets SO_LINGER on sockets.


syntax: lingering_time time

default: lingering_close 30s

context: http, server, location

Sets SO_LINGER on sockets.


syntax: lingering_timeout time

default: lingering_timeout 5s

context: http, server, location

Sets SO_LINGER on sockets.


syntax: listen address:port [ default(deprecated in 0.8.21) | default_server | [ backlog=num | rcvbuf=size | sndbuf=size | accept_filter=filter | deferred | bind | ipv6only=[on|off] | ssl ] ]

default: listen 80

context: server

The listen directive specifies the address and port accepted by the enclosing server {...} block. It is possible to specify only an address, only a port, or a server name as the address.

listen 8000;
listen *:8000;
listen localhost:8000;

IPv6 address(0.7.36) are set in square brackets:

listen [::]:8000; 
listen [fe80::1];

In Linux by default any IPv6 TCP socket also accepts IPv4 traffic using the IPv4 to IPv6 mapped address format, i.e., ::ffff:<IPv4 adddress in dotted decimal notation>. E.g., ::ffff: maps the IPv4 address to an IPv6 address.

When you enable the address [::]:80, binding port 80 using IPv6, in the listen directive, in Linux, by default, the IPv4 port 80 is also enabled. Meaning that nginx listens for both IPv4 and IPv6 incoming traffic. Therefore if you erroneously specify also a IPv4 address you'll get an already bind address error when reloading nginx configuration.

In Linux the separation of the IPv6 and IPv4 stacks is controlled through the runtime parameter: net.ipv6.bindv6only which has the value 0 by default.

If you want to use separate sockets for IPv4 and IPv6 you should set this parameter to 1 using sysctl.

Note that any nginx instance that was running before you made the change will continue to accept IPv4 traffic. Therefore you should edit your nginx configuration to reflect the new setup for IPv6 and IPv4 packet handling and do a restart.

If on the other hand you launched another server instance (vhost) and you expect it to also handle IPv4 traffic by using only, for example:

listen [::]:80;

the binding of the IPv4 address will fail. The correct way to to this is by using the "ipv6only=on" option in the IPv6 listen directive and also specifying a IPv4 listen directive in the respective server block.

This re-editing of the configuration must be done after you changed your kernel runtime parameter. This is the most generic situation in that case (separation of IPv6 and IPv4 sockets):

listen [::]:80 ipv6only=on; # listen for IPv6 only traffic on IPv6 sockets
listen 80; # listen also for IPv4 traffic on "regular" IPv4 sockets

In FreeBSD the default is separate IPv4 and IPv6 sockets. Therefore "listen [::]:80" only binds port 80 for listening to IPv6 traffic. It's always necessary to specify also IPv4 listen directives if you wish to also handle IPv4 traffic.

It's possible to specify only IPv6 addresses in the listen directive. Using the "default_server ipv6only=on" option. Specific IPv6 addresses can be used with a IPv6 only default directive. Other server directives can also specifiy listen directives with IPv4 addresses. The uniqueness of the IPv6 handling concerns only the same server {...} block.

listen [2a02:750:5::123]:80;
listen [::]:80 default_server ipv6only=on;

If only the address is given, the default port nginx binds to is 80.

If the directive has the default_server parameter, then the enclosing server {...} block will be the default server for the address:port pair. This is useful for name-based virtual hosting where you wish to specify the default server block for hostnames that do not match any server_name directives. If there are no directives with the default_server parameter, then the default server will be the first server block in which the address:port pair appears. The default_server parameter appeared in version 0.8.21 thus deprecating the parameter default.

The listen directive accepts several parameters, specific to the system calls listen(2) and bind(2). These parameters must follow the default parameter.

backlog=num -- is assigned parameter backlog in call listen(2). By default backlog equals -1.

rcvbuf=size -- assigned to the parameter SO_RCVBUF for the listening socket.

sndbuf=size -- assigned to the parameter SO_SNDBUF for the listening socket.

accept_filter=filter -- is assigned name accept-filter.

. It works only to FreeBSD, it is possible to use two filters -- dataready and httpready. On the signal -HUP accept-filter it is possible to change only in the quite last versions FreeBSD: 6.0, 5.4-STABLE and 4.11-STABLE.

deferred -- indicates to use that postponed accept(2) on Linux with

. the aid of option TCP_DEFER_ACCEPT.

bind -- indicates that it is necessary to make bind(2) separately

. for this pair of address:port. The fact is that if are described several directives listen with the identical port, but by different addresses and one of the directives listen listens to on all addresses for this port (*:port), then nginx will make bind(2) only to *:port. It is necessary to consider that in this case for determining the address, on which the connections arrive, is done the system call getsockname(). But if are used parameters backlog, rcvbuf, sndbuf, accept_filter or deferred, then it is always done separately for this pair of address:port bind(2).

ssl -- parameter (0.7.14) not related to listen(2) and bind(2) syscalls

. but instead specifies that connections accepted on this port should work in SSL mode. This allows to specify compact configurations for servers working with both HTTP and HTTPS. For example:
listen  80;
listen  443 default_server ssl;

Example of the use of the parameters:

listen default_server accept_filter=dataready backlog=1024;

Since version 0.8.21 nginx is able to listen on unix sockets:

listen unix:/tmp/nginx1.sock;


syntax: location [=|~|~*|^~|@] /uri/ { ... }

default: no

context: server

This directive allows different configurations depending on the URI. It can be configured using both literal strings and regular expressions. To use regular expressions, you must use a prefix:

  1. "~" for case sensitive matching
  2. "~*" for case insensitive matching
  3. there is no syntax for NOT matching a regular expression. Instead, match the target regular expression and assign an empty block, then use location / to match anything else.

The order in which location directives are checked is as follows:

  1. Directives with the "=" prefix that match the query exactly (literal string). If found, searching stops.
  2. All remaining directives with conventional strings. If this match used the "^~" prefix, searching stops.
  3. Regular expressions, in the order they are defined in the configuration file.
  4. If #3 yielded a match, that result is used. Otherwise, the match from #2 is used.

Details below.

To determine which location directive matches a particular query, the literal strings are checked first. Literal strings match the beginning portion of the query - the most specific match will be used. Afterwards, regular expressions are checked in the order defined in the configuration file. The first regular expression to match the query will stop the search. If no regular expression matches are found, the result from the literal string search is used.

For case-insensitive operating systems, like Mac OS X or Windows with Cygwin, literal string matching is done in a case insensitive way (0.7.7). However, comparison is limited to single-byte locale's only.

Regular expression may contain captures (0.7.40), which can then be used in other directives.

It is possible to disable regular expression checks after literal string matching by using "^~" prefix. If the most specific match literal location has this prefix: regular expressions aren't checked.

The "=" prefix forces an exact (literal) match between the request URI and the location parameter. When matched, the search stops immediately. A useful application is that if the request "/" occurs frequently, it's better to use "location = /", as that will speed up the processing of this request a bit, since the search will stop after the first comparison.

On exact match with literal location without "=" or "^~" prefixes search is also immediately terminated.

It is important to know that nginx does the comparison against decoded URIs. For example, if you wish to match "/images/%20/test", then you must use "/images/ /test" to determine the location.


location  = / {
  # matches the query / only.
  [ configuration A ] 
location  / {
  # matches any query, since all queries begin with /, but regular
  # expressions and any longer conventional blocks will be
  # matched first.
  [ configuration B ] 
location ^~ /images/ {
  # matches any query beginning with /images/ and halts searching,
  # so regular expressions will not be checked.
  [ configuration C ] 
location ~* \.(gif|jpg|jpeg)$ {
  # matches any request ending in gif, jpg, or jpeg. However, all
  # requests to the /images/ directory will be handled by
  # Configuration C.   
  [ configuration D ] 

Example requests:

  • / -> configuration A
  • /documents/document.html -> configuration B
  • /images/1.gif -> configuration C
  • /documents/1.jpg -> configuration D

Note that you could define these 4 configurations in any order and the results would remain the same. While nested locations are allowed by the configuration file parser, their use is discouraged and may produce unexpected results.

The prefix "@" specifies a named location. Such locations are not used during normal processing of requests, they are intended only to process internally redirected requests (see error_page, try_files).


syntax: log_not_found [on|off]

default: log_not_found on

context: http, server, location

The directive enables or disables messages in error_log about files not found on disk.


syntax: log_subrequest [on|off]

default: log_subrequest off

context: http, server, location

The directive enables or disables messages in access_log about sub-requests such as rewrite rules and/or SSI requests.


syntax: merge_slashes [on|off]

default: merge_slashes on

context: http, server

version: 0.6.15

variables: no

Enables merging adjacent slashes when parsing the request line. For example, a request for http://www.example.com/foo//bar/ will produce the following values for $uri:

  • on: /foo/bar/
  • off: /foo//bar/

Be aware that static location matching is performed as a string compare, so if merge_slashes is turned off, a request for /foo//bar/ will *not* match location /foo/bar/.


syntax: msie_padding [on|off]

default: msie_padding on

context: http, server, location

This directive enables or disables the the msie_padding feature for MSIE browsers, and Chrome (as of nginx 0.8.25+). When this is enabled, nginx will pad the size of the response body to a minimum of 512 bytes, for responses with a status code above or equal to 400.

The padding prevents the activation of "friendly" HTTP error pages in MSIE and Chrome, so as to not hide the more-informative error pages from the server.


syntax: msie_refresh [on|off]

default: msie_refresh off

context: http, server, location

This directive allows or forbids issuing a refresh instead of doing a redirect for MSIE.


syntax: open_file_cache max = N [inactive = time] | off

default: open_file_cache off

context: http, server, location

The directive sets the cache activity on. These information can be stored:

  • Open file descriptors, information with their size and modification time;
  • Information about the existence of directories;
  • Error information when searches for a file - no file, do not have rights to read, etc. See also open_file_cache_errors

Options directive:

  • max - specifies the maximum number of entries in the cache. When the cache overflows, the least recently used(LRU) items will be removed;
  • inactive - specifies the time when the cached item is removed, if it has not been downloaded during that time, the default is 60 seconds;
  • off - prohibits the cache activity.


 open_file_cache max=1000 inactive=20s; 
 open_file_cache_valid    30s; 
 open_file_cache_min_uses 2;
 open_file_cache_errors   on;


syntax: open_file_cache_errors on | off

default: open_file_cache_errors off

context: http, server, location

The directive specifies to cache errors or not when searching a file.


syntax: open_file_cache_min_uses number

default: open_file_cache_min_uses 1

context: http, server, location

The directive defines the minimum use number of a file within the time specified in the directive parameter inactive in open_file_cache. ?If use more than the number, the file descriptor will remain open in the cache.


syntax: open_file_cache_valid time

default: open_file_cache_valid 60

context: http, server, location

The directive specifies the time when need to check the validity of the information about the item in open_file_cache.


syntax: optimize_server_names [ on|off ]

default: optimize_server_names on

context: http, server

Directive activates or deactivates optimization of host name checks for name-based virtual servers.

In particular, the check influences the name of the host used in redirects. If optimization is on, and all name-based servers listening on one address:port pair have identical configuration, then names are not checked during request execution and redirects use first server name.

If redirect must use host name passed by the client, then the optimization must be turned off.

Note: this directive is deprecated in nginx 0.7.x, use server_name_in_redirect instead.


syntax: port_in_redirect [ on|off ]

default: port_in_redirect on

context: http, server, location

Directive allows or prevents port indication in redirects handled by nginx.

If port_in_redirect is off, then Nginx will not add the port in the url when the request is redirected.


syntax: post_action [ uri|off ]

default: post_action off

context: http, server, location, if-in-location

Defines a URI to sub-request upon completion of current request.

location /protected_files { 
	post_action /protected_done;
# Send the post_action request to a FastCGI backend for logging.
location /protected_done {


syntax: recursive_error_pages [on|off]

default: recursive_error_pages off

context: http, server, location

recursive_error_pages enables or disables following a chain of error_page directives.


syntax: request_pool_size the_size

default: request_pool_size the_size 256

context: http (more?)

The directive is used to allocate memory per request. The pool is used for small allocations. If a block is bigger than pool size or bigger than page size, then it is allocated outside the pool. If there is not enough memory for small allocation inside pool, then a new block of the same pool size is allocated. This directive has only a very small effect. (source http://markmail.org/message/b2kmrluscevimpba)


syntax: reset_timedout_connection [on|off]

default: reset_timedout_connection off

context: http, server, location

This directive enables or disables resetting the connection on timeout. When resetting the connection, before the socket is closed, the socket SO_LINGER option is set with a 0 timeout, which forces the RST packet to be sent to the client upon closing the socket, thus freeing all memory associated with it. This prevents the socket in the FIN_WAIT1 state, along with the buffers associated with it from lying around.

Note that sockets with keepalive connections, after the defined timeout, are closed in the usual way.


syntax: resolver address

default: no

context: http, server, location

Directive defines DNS server address, e.g.



syntax: resolver_timeout time

default: 30s

context: http, server, location

Directive defines timeout for name resolution, e.g.

resolver_timeout 5s;


syntax: root path

default: root html

context: http, server, location, if in location

root specifies the document root for the requests. For example, with this configuration

location  /i/ {
  root  /spool/w3;

A request for "/i/top.gif" will return the file "/spool/w3/i/top.gif". You can use variables in the argument.

note: Keep in mind that the root will still append the directory to the request so that a request for "/i/top.gif" will not look in "/spool/w3/top.gif" like might happen in an Apache-like alias configuration where the location match itself is dropped. Use the alias directive to achieve the Apache-like functionality.


syntax: satisfy [ all | any ]

default: satisfy all

context: http, server, location

version: ≥ 0.6.25

This determines the adopted access policy when directives from multiple access phase handlers, such as the Access and Auth Basic modules, are defined in a context:

  • all - All access phase handlers must grant access to the context
  • any - Any access phase handler may grant access to the context
location / {
  satisfy any;
  deny all;
  auth_basic "closed site";
  auth_basic_user_file conf/htpasswd;


syntax: satisfy_any [ on | off ]

default: satisfy_any off

context: http, server, location

deprecated: 0.6.25 -- Use the satisfy directive instead


syntax: send_timeout the time

default: send_timeout 60

context: http, server, location

Directive assigns response timeout to client. Timeout is established not on entire transfer of answer, but only between two operations of reading, if after this time client will take nothing, then nginx is shutting down the connection.


syntax: sendfile [ on|off ]

default: sendfile off

context: http, server, location

Directive activate or deactivate the usage of sendfile().

sendfile() copies data between one file descriptor and another. Because this copying is done within the kernel, sendfile() is more efficient than the combination of read(2) and write(2), which would require transferring data to and from user space.

Read more at: https://www.kernel.org/doc/man-pages/online/pages/man2/sendfile.2.html

How does sendfile helps : http://www.techrepublic.com/article/use-sendfile-to-optimize-data-transfer/1044112


syntax: server {...}

default: no

context: http

Directive assigns configuration for the virtual server.

There is no separation of IP and name-based (the Host header of the request) servers.

Instead, the directive listen is used to describe all addresses and ports on which incoming connections can occur, and in directive server_name indicate all names of the server.


syntax: server_name name [... ]

default: server_name ""

context: server

This directive performs two actions:

  • Compares the Host header of the incoming HTTP request against the server { ... } blocks in the Nginx configuration files and selects the first one that matches. This is how virtual servers are defined. Server names are processed in the following order:
  1. full, static names
  2. names with a wildcard at the start of the name — *.example.com
  3. names with a wildcard at the end of the name — www.example.*
  4. names with regular expressions
If there is no match, a server { ... } block in the configuration file will be used based on the following order:
  1. the server block with a matching listen directive marked as [default|default_server]
  2. the first server block with a matching listen directive (or implicit listen 80;)


server {
  server_name   example.com  www.example.com;

The first name becomes the basic name of server. By default the name of the machine (hostname) is used.

It is possible to use "*" for replacing the first or the last part of the name:

server {
  server_name   example.com  *.example.com  www.example.*;

The first two of the above names (example.com and *.example.com) can be combined into one:

server {
  server_name  .example.com;

It is also possible to use regular expressions in server names, prepending the name with a tilde "~" like so:

server {
  server_name   www.example.com   ~^www\d+\.example\.com$;

Since nginx 0.7.12, an empty server name is supported to catch the requests without "Host" header, please note that most browsers will always send a Host header, if accessed by IP the Host header will contain the IP. To specify a catch-all block please see the default_server flag of the listen directive.

Since nginx 0.8.25 named captures can be used in server_name:

server {
  server_name   ~^(www\.)?(?<domain>.+)$;
  root  /sites/$domain;

Some older versions of PCRE may have issues with this syntax. If any problems arise try this following syntax:

server {
  server_name   ~^(www\.)?(?P<domain>.+)$;
  root  /sites/$domain;

Since nginx 0.9.4, $hostname can be used as a server_name argument:

server {
  server_name $hostname;


syntax: server_name_in_redirect on|off

default: server_name_in_redirect on

context: http, server, location

If server_name_in_redirect is on, then Nginx will use the first value of the server_name directive for redirects. If server_name_in_redirect is off, then nginx will use the requested Host header.

Note: for Location headers coming from an upstream proxy (via proxy_pass for example) this may not be the only directive you need. In fact, it seems to be ignored a lot of the time. If you are seeing the upstream's server name come through and not be rewritten, you will need to use proxy_redirect to rewrite the upstream's provided hostname to what you want. Something like proxy_redirect http://some.upstream.url/ / - you will want to rewrite it to a / relative path.


syntax: server_names_hash_max_size number

default: server_names_hash_max_size 512

context: http

The maximum size of the server name hash tables. For more detail see the description of tuning the hash tables in Nginx Optimizations.


syntax: server_names_hash_bucket_size number

default: server_names_hash_bucket_size 32/64/128

context: http

Directive assigns the size of basket in the hash-tables of the names of servers. This value by default depends on the size of the line of processor cache. For more detail see the description of tuning the hash tables in Nginx Optimizations.


syntax: server_tokens on|off

default: server_tokens on

context: http, server, location

Whether to send the Nginx version number in error pages and Server header.


syntax: tcp_nodelay [on|off]

default: tcp_nodelay on

context: http, server, location

This directive allows or forbids the use of the socket option TCP_NODELAY. Only included in keep-alive connections.

You can read more about the TCP_NODELAY socket option here.


syntax: tcp_nopush [on|off]

default: tcp_nopush off

context: http, server, location

This directive permits or forbids the use of the socket options TCP_NOPUSH on FreeBSD or TCP_CORK on Linux. This option is only available when using sendfile.

Setting this option causes nginx to attempt to send it's HTTP response headers in one packet on Linux and FreeBSD 4.x

You can read more about the TCP_NOPUSH and TCP_CORK socket options here.


syntax: try_files path1 [path2] uri

default: none

context: server, location

availability: 0.7.27

Checks for the existence of files in order, and returns the first file that is found. A trailing slash indicates a directory - $uri /. In the event that no file is found, an internal redirect to the last parameter is invoked. The last parameter is the fallback URI and *must* exist, or else an internal error will be raised. Unlike rewrite, $args are not automatically preserved if the fallback is not a named location. If you need args preserved, you must do so explicitly:

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

Example use in proxying Mongrel:

try_files /system/maintenance.html $uri $uri/index.html $uri.html @mongrel;
location @mongrel {
  proxy_pass http://mongrel;

Note that you can specify an HTTP status code as the last argument to try_file since Nginx version 0.7.51. Here's an example:

location / {
  try_files $uri $uri/ /error.php?c=404 =404;

When all other attempts to serve the content corresponding to the request fail issue a 404 Not Found.

Example of use with Drupal / FastCGI:

# for Drupal 6 or 7:
try_files $uri $uri/ /index.php?q=$uri&$args;
# a better version for Drupal 7 since it doesn't need q=$uri:
try_files $uri $uri/ /index.php?$args;
location ~ \.php$ {
  fastcgi_param  SCRIPT_FILENAME $document_root$fastcgi_script_name; # if not already defined in the fastcgi_params file
  # any other specific fastcgi_params

In this example, the directive try_files

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

Is basically the same as this:

location / {
  error_page     404 = @drupal;
  log_not_found  off;
location @drupal {
  rewrite ^ /index.php?q=$uri last; # for drupal 6

Or this:

if (!-e $request_filename) {
   rewrite ^ /index.php?q=$uri last;

try_files is basically a replacement for the typical mod_rewrite style file/directory existence check. It is supposed to be more efficient than using if - see IfIsEvil

Examples of use with Wordpress and Joomla (typical "Front controller pattern" packages)

# wordpress (without WP Super Cache) - example 1
try_files $uri $uri/ /index.php?q=$uri&$args;
# wordpress (without WP Super Cache) - example 2 
# It doesn't REALLY need the "q" parameter, but without an explicit $args php 
# gets an empty QUERY_STRING, breaking generated responses that don't use a 
# permalink, such as search results.
try_files $uri $uri/ /index.php?$args;
# joomla
try_files $uri $uri/ /index.php?q=$uri&$args;
location ~ \.php$ {
  fastcgi_param  SCRIPT_FILENAME $document_root$fastcgi_script_name; # if not already defined in the fastcgi_params file
  # any other specific fastcgi_params

WP Super Cache requires a bunch of static file checks. Those are not shown here.


syntax: types {...}

context: http, server, location

Directive assigns the correspondence of expansion and MIME-types of answers. To one MIME- type can correspond several expansions. By default it is used these correspondences:

types {
  text/html    html;
  image/gif    gif;
  image/jpeg   jpg;

The sufficiently complete table of mappings is included and is located in the file conf/mime.types.

So that for that determined location's for all answers would reveal MIME- type application/octet-stream, it is possible to use the following:

location /download/ {
  types         { }
  default_type  application/octet-stream;


syntax: underscores_in_headers on|off

default: off

context: http, server

Allows or disallows underscores in headers.


syntax: variables_hash_bucket_size size

default: variables_hash_bucket_size 64

context: http

variables: no

phase: configuration

Assigns the key bucket size for the variables hash table.


syntax: variables_hash_max_size size

default: variables_hash_max_size 512

context: http

variables: no

phase: configuration

The maximum size of the variables hash table. For more detail see the description of tuning the hash tables in Nginx Optimizations.


The core module supports built-in variables, whose names correspond with the names of variables in Apache.

First of all, there are variables which represent header lines in the client request, for example, $http_user_agent, $http_cookie, and so forth. Note that because these correspond to what the client actually sends, they are not guaranteed to exist and their meaning is defined elsewhere (e.g. in relevant standards).

Furthermore, there are other variables:


This variable contains the value of the GET request variable PARAMETER if present in the query string


This variable is the GET parameters in request line, e.g. foo=123&bar=blahblah; This variable could be changed.


The address of the client in binary form;


The amount of bytes sent as part of the body of the response. Is accurate even when connections are aborted or interrupted.


This variable is equal to line Content-Length in the header of request;


This variable is equal to line Content-Type in the header of request;


The value of the cookie COOKIE;


This variable is equal to the value of directive root for the current request;


The same as $uri.


This variable is equal to line Host in the header of request or name of the server processing the request if the Host header is not available.

This variable may have a different value from $http_host in such cases: 1) when the Host input header is absent or has an empty value, $host equals to the value of server_name directive; 2)when the value of Host contains port number, $host doesn't include that port number. $host's value is always lowercase since 0.8.17.


Set to the machine's hostname as returned by gethostname


The value of the HTTP request header HEADER when converted to lowercase and with 'dashes' converted to 'underscores', e.g. $http_user_agent, $http_referer...;


The value of the HTTP response header HEADER when converted to lowercase and with 'dashes' converted to 'underscores', e.g. $sent_http_cache_control, $sent_http_content_type...;


Evaluates to "?" if $args is set, "" otherwise.


This variable allows limiting the connection rate.


The version of Nginx that the server is currently running;


The same as $args except that this variable is readonly.


The address of the client.


The port of the client;


This variable is equal to the name of user, authenticated by the Auth Basic Module;


This variable is equal to path to the file for the current request, formed from directives root or alias and URI request;


This variable(0.7.58+) contains the body of the request. The significance of this variable appears in locations with directives proxy_pass or fastcgi_pass.


Client request body temporary filename;


Set to "OK" if request was completed successfully. Empty if request was not completed or if request was not the last part of a series of range requests.


This variable is equal to the method of request, usually GET or POST.

This variable always evaluates to the method name of the main request, not the current request, when the current request is a subrequest.


This variable is equal to the *original* request URI as received from the client including the args. It cannot be modified. Look at $uri for the post-rewrite/altered URI. Does not include host name. Example: "/foo/bar.php?arg=baz"


The HTTP scheme (i.e. http, https). Evaluated only on demand, for example:

rewrite  ^  $scheme://example.com$uri  redirect;


The server address. Generally nginx makes a system call to obtain this value. To improve efficiency and avoid this system call, specify an address with the listen directive and to use the bind parameter.


The name of the server.


This variable is equal to the port of the server, to which the request arrived;


This variable is the protocol of the request. Common examples are: HTTP/1.0 or HTTP/1.1


This variable is the current request URI, without any arguments (see $args for those). This variable will reflect any modifications done so far by internal redirects or the index module. Note this may be different from $request_uri, as $request_uri is what was originally sent by the browser before any such modifications. Does not include the protocol or host name. Example: /foo/bar.html


Original Documentation