Shulou(Shulou.com)06/02 Report--

This article focuses on "Nginx module development and architecture analysis", interested friends may wish to have a look. The method introduced in this paper is simple, fast and practical. Next, let the editor take you to learn "Nginx module development and architecture analysis"!

I. preparatory work before studying Nginx

1.Nginx features: faster, high scalability, high reliability, low memory consumption, stand-alone support for more than 100000 concurrent connections, hot deployment, the freest BSD license agreement

two。 Exit nginx

Nginx-s stop

Nginx-s quit

3. Log rollback: nginx-s reopen

II. Configuration of Nginx

a. Relationships between running Nginx processes

In general, the number of worker processes is equal to the number of CPU on the server

Basic configuration of B.Nginx services

1. Configuration items for debugging processes and locating problems

| daemon on | whether off;: runs Nginx as a daemon. Default is on. |

| master_process on | whether off;: works in master/worker mode. Default is on. |

Settings for error_log pathfile level;:error logs

Debug_points [stop | abort]: helps users track and debug nginx. Generally, it is not used.

Debug_connection [IP | CIDR]: output debug-level logs only for specified clients. It is useful for locating problems that occur under highly concurrent requests. Add the parameter-with-debug when configure is required.

Worker_rlimit_core size;: limits the size of coredump core dump files

Working_directory path;: specifies the directory where the coredump file is generated

two。 Configuration items for normal operation

Evn VAR | VAR=VALUE: allows users to directly set environment variables on the operating system

Include pathfile;: embeds other configuration files

The path to the pid path/file;pid file

User username [groupname];: users and user groups that the Nginx worker process is running

Worker_rlimit_nofile limit;: specifies the maximum number of handle descriptors that the Nginx worker process can open

Worker_rlimit_sigpending limit;: limit signal queue

3. Optimize the configuration of performance

Worker_processes number;: defines the number of worker processes

Worder_cpu_affinity cpumask [cpumask...];: bind the worker process to the specified CPU kernel

Ssl_engine device;:SSL hardware acceleration

Timer_resolution tbot: how often the system calls gettimeofday

Worker_priority nice;:worker process priority setting

4. Event class configuration item

Accept_mutex [on | off];: whether to open the accept lock allows multiple worker processes to establish TCP connections with the new client in turn and serially.

The path to the lock_file path/file;:lock file

The delay time between the use of accept locks by accept_mutex_delay ms;: and the actual connection establishment

Multi_accept on | off;: establishes new connections in batch

Use [kqueue | rtsig | epoll | / dev/poll | select | poll | eventport];: select an event model

Maximum number of connections per worker for worker_connections number;:

c. Configure a static Web server with HTTP core module

1. Distribution of virtual hosts and requests

Listen address: Port [default | default_server | [backlog=num | rcvbuf=size | sndbuf=size | accept_filter=filter | deferred | bind | ipv6only= [on | off] | ssl]];: listening port, configured in server block

Server_name name [...];: host name, configured in the server block

Server_names_hash_bucket_size size;: sets the amount of memory consumed by each hash table, and nginx uses hash tables to store server_name

Server_names_hash_max_size size;: affects the collision rate of the hash table. The larger the hash table, the more memory is consumed, but the conflict of the hash key is reduced, and the retrieval speed is faster.

Server_name_in_redirect on | processing of off;: redirect host name

Location [= | ~ | ~ * | ^ ~ | @] / uri/ {...}: process the user's request by matching the configuration in the location {} block according to the requested URI, which is configured in the server block

two。 Definition of file path

Root path;: sets the resource path in root mode

Alias path;: alias, which maps uri to real disk files, only in location blocks

Index file.;: visit the home page

Error_page code [code...] [= | = answer-code] uri | @ named_location: return the code redirection page based on HTTP

Recursive_error_pages [on | off];: whether to allow recursive use of error_page

Try_files path2 [path3] uri;: tries to access each path sequentially

3. Allocation of memory and disk resources

Client_body_in_file_only on | clean | off;:HTTP packet is only stored in disk files

Client_body_in_single_buffer on | off;:HTTP packet body is written into a memory buffer as far as possible

Client_header_buffer_size size;: stores the memory buffer size of the HTTP header

Large_client_header_buffers number size;: defines the number of buffer and the size of each buffer that Nginx receives a very large HTTP header request.

Client_body_buffer_size size;: stores the memory buffer size of HTTP packets

Client_body_temp_path dir-path [level 1 [level 2 [level 3];: temporary storage directory for HTTP packages

Connection_pool_size size;:Nginx pre-allocates a memory pool for each successful TCP connection. This configuration specifies the initial size of a memory pool to reduce the number of times the kernel allocates small chunks of memory.

Request_pool_size size;:Nginx allocates a memory pool for each request, and the configuration specifies the initial size of this memory pool

4. Settings for network connection

Client_header_timeout time: timeout for reading HTTP headers

Client_body_timeout time: timeout for reading HTTP packets

Timeout for send_timeout time;: to send a response

Reset_timeout_connection on | after the off;: connection times out, the connection will be reset directly by sending the RST packet to the client. When this option is turned on, Nginx will send the RST reset packet directly to the user. Instead of waiting for the user to reply, the cache will be released directly.

Lingering_close off | on | always;: controls how Nginx closes user connections

Lingering_time time;: is very useful for uploading large files. When the time elapses, the connection will be closed regardless of whether it is still uploading or not.

After the lingering_close takes effect, ligering_timeout time;: detects whether any data sent by the user reaches the server before closing the connection. If there is no data to read after the expiration of the time, close the connection directly.

Keepalive_disable [msie6 | safari | none]...: disable the keepalive feature for some browsers

Keepalive_timeout time time:keepalive timeout

Keepalive_requests n: the maximum number of requests allowed to be carried on a keepalive persistent connection

Tcp_nodelay on | off;: determines whether to use the TCP_NODELAY option for keepalive connections

Tcp_nopush on | off;: determines whether to enable the TCP_CORK function on TCP_NOPUSH on FreeBSD or TCP_CORK on Linux when the sendfile option is turned on

Settings for the 5.MIME type

Mapping between type {...};: MIME type and file extension

Default_type MIME-type;: default MIME type

Types_hash_bucket_size size;: sets the amount of memory consumed by the hash table

Types_hash_max_size size;: affects the collision rate of hash tables

6. Restrictions on client requests

Limit_except method... {...}: restrict user requests by HTTP method name

Maximum value of client_max_body_size size;:HTTP request packet

Limit_rate speed;: limits the number of bytes per second to client requests

Limit_rate_after time;: said that the speed limit will not begin until the response length of nginx sending pounds to the client exceeds the limit_rate_after.

7. Optimization of file operation

Sendfile on | off;: enables sendfile system call to send files

| aio on | off;: indicates whether to enable kernel-level asynchronous Icano on FreeBSD or Linux systems, which is mutually exclusive with sendfile |

Directio size | off;: uses the O_DIRECT option to read files on FreeBSD and Linux systems, mutually exclusive with sendfile

Directio_alignment size;: works with directio to specify the alignment when reading files in directio mode

Open_file_cache max=N [inactive=time] | off;: opens file cache

| open_file_cache_errors on | whether off;: caches the error message of opening a file |

The minimum number of visits to open_file_cache_min_uses number;: that will not be eliminated. It is used in conjunction with open_file_cache 's inactive. If it is exceeded, it will not be eliminated from the cache.

How often open_file_cache_valid time;: verifies the validity of elements in the cache

8. Special handling of client requests

Ignore_invalid_headers on | off;: ignores illegal HTTP headers

Underscores_in_headers on | whether the off;:HTTP header is allowed to be underlined

If_modified_since [off | exact | before];: processing strategy for If-Modified-Since headers

Log_not_found on | whether the error log is recorded if the off;: file is not found.

Merge_slashes on | whether off;: merges adjacent /

Resolver address...;: deposed the address of the DNS name resolution server

Resolver_timeout time;:DNS parsing timeout

| server_tokens on | whether to indicate the Nginx version in Server when off;: returns an error page |

d. Configure a reverse proxy server with HTTP proxy module

1. Basic configuration of load balancing

Upstream name {...}: defines a cluster of upstream servers to facilitate the use of proxy_pass in the reverse proxy and is configured in the http block

Server name [weight=number,max_fails=number,fail_timeout=time,down,backup]: specifies the name of an upstream server, which can be a domain name, ip address port, UNIX handle, etc., configured in the upstream block

Ip_hash;: always places the request on a fixed upstream server based on the customer's IP address, and cannot be used at the same time as the weight configuration

two。 Basic configuration of reverse proxy

Proxy_pass URL;: reverse proxies the current request to the server specified by the URL parameter. URL can be configured in a location or if block using a domain name, ip address port, UNIX handle or upstream block

Proxy_set_header Host $host;: forward the Host header in the request. Proxy_pass does not forward it by default

Protocol method name when proxy_method method;: forwarding

Proxy_hide_header the_header;: can specify which HTTP header fields cannot be forwarded

Proxy_pass_header the_header;: is the opposite of proxy_hide_header

Proxy_pass_request_body on | off;: determines whether to send the HTTP packet body to the upstream server

Proxy_pass_request_headers on | off;: determines whether to forward HTTP headers

Proxy_redirect [default | off | redirect replacement];: when the upstream server returns a redirect or refresh request, you can reset the location or refresh field of the HTTP header

Proxy_next_upstream [errpo,timeout,invalid_header,http500,http_502,http503,http_504,http_404,off]: when an error occurs when an upstream server forwards a request, continue to process the request on another server

Third, develop a simple HTTP module

1. Encapsulation of integers: ngx_int_t, ngx_uint_t

two。 String: ngx_str_t

3. Linked list container: ngx_list_t

4.key/value pair: ngx_table_elt_t

5. Buffer: ngx_buf_t

6. Link structure used with ngx_buf_t: ngx_chain_t

Configuration, error log and request context

Access to third-party services

1.upstream ensures that when interacting with a third-party server (including establishing a TCP connection with a three-way handshake, sending a request, receiving a response, closing a TCP connection with a four-way handshake, etc.), it will not block the Nginx process from processing other requests

2.subrequest is a design pattern for decomposing complex requests, and it is finally implemented based on upstream.

3. When we want to return the content of the third-party service almost intact to the user, we generally use the upstream method, which can pass through HTTP; very efficiently. If we visit the third-party service only to obtain some information, and then construct the response based on the information and transmit it to the customer, we should use the subrequest method.

At this point, I believe you have a deeper understanding of "Nginx module development and architecture analysis". You might as well do it in practice. Here is the website, more related content can enter the relevant channels to inquire, follow us, continue to learn!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.