Nginx HTTP Server - Fourth Edition

To set up a yum repository for RHEL/CentOS, create a file named /etc/yum.repos.d/nginx.repo with the following contents:

[nginx]
name=nginx repo
baseurl=http://nginx.org/packages/OS/OSRELEASE/$basearch/
gpgcheck=0
enabled=1

Replace OS with rhel or centos, depending on the distribution used, and OSRELEASE with 6 or 7, for versions 6.x or 7.x, respectively. Afterwards, Nginx can now be installed with yum:

yum install nginx

For Debian-based distributions, we need to first use their signing key to authenticate the package signatures. Download the following file first from http://nginx.org/keys/nginx_signing.key.

Then run the following command:

sudo apt-key add nginx_signing.key

With the key added, we can now add the Nginx repository to our sources.list found in /etc/apt/sources.list. For Debian, we add the following lines:

deb http://nginx.org/packages/debian/ codename nginx
deb-src http://nginx.org/packages/debian/ codename nginx

Where codename is either jessie or stretch depending on your version of Debian. For Ubuntu, we use the following dependencies:

deb http://nginx.org/packages/ubuntu/ codename nginx
deb-src http://nginx.org/packages/ubuntu/ codename nginx

Where codename is one of trusty, xenial, or zesty depending on your version of Ubuntu. Finally, we can install Nginx with the apt-get command option:

apt-get update
apt-get install nginx

Nginx is a program written in C, so you will first need to install a compiler tool such as the GNU Compiler Collection (GCC) on your system. GCC may already be present on your system, but if that is not the case you will have to install it before going any further.

First, make sure it isn't already installed on your system:

[alex@example.com ~]$ gcc

If you get the following output, it means that GCC is correctly installed on your system and you can skip to the next section:

gcc: no input files

If you receive the following message, you will have to proceed with the installation of the compiler:

~bash: gcc: command not found

GCC can be installed using the default repositories of your package manager. Depending on your distribution, the package manager will be vary-yum for a Red Hat Linux-based distribution, apt for Debian and Ubuntu, yast for SuSE Linux, and so on. Here is the typical way to proceed with the download and installation of the GCC package:

[root@example.com ~]# yum groupinstall "Development Tools"

If you use apt-get, execute the following command:

[root@example.com ~]# apt-get install build-essentials

If you use another package manager with a different syntax, you will probably find the documentation with the man utility. Either way, your package manager should be able to download and install GCC correctly, after having resolved dependencies automatically. Note that this command will not only install GCC, it also proceeds with downloading and installing all common requirements for building applications from source, such as code headers and other compilation tools.

The Perl Compatible Regular Expression (PCRE) library is required for compiling Nginx. The rewrite and HTTP core modules of Nginx use PCRE for the syntax of their regular expressions, as we will discover in later chapters. You will need to install two packages—pcre and pcre-devel. The first one provides the compiled version of the library, whereas the second one provides development headers and sources for compiling projects, which are required in our case.

Here are some example commands that you can run in order to install both the packages.

Using yum, execute the following command:

[root@example.com ~]# yum install pcre pcre-devel

Or you can install all PCRE-related packages using the following command:

[root@example.com ~]# yum install pcre*

If you use apt-get, use the following command:

[root@example.com ~]# apt-get install libpcre3 libpcre3-dev

If these packages are already installed on your system, you will receive a message saying something like nothing to do; in other words, the package manager did not install or update any component:

The zlib library provides developers with compression algorithms. It is required for the use of .gzip compression in various modules of Nginx. Again, you can use your package manager to install this component as it is part of the default repositories. Similar to PCRE, you will need both the library and its source-zlib and zlib-devel.

Using yum, execute the following command:

[root@example.com ~]# yum install zlib zlib-devel

Using apt-get, execute the following command:

[root@example.com ~]# apt-get install zlib1g zlib1g-dev

These packages install quickly and have no known dependency issues.

The OpenSSL library will be used by Nginx to serve secure web pages. We thus need to install the library and its development package. The process remains the same here – you install openssl and openssl-devel:

[root@example.com ~]# yum install openssl openssl-devel

Using apt-get, execute the following command:

[root@example.com ~]# apt-get install openssl libssl-dev

Now that you have installed all of the prerequisites, you are ready to download and compile the Nginx source code.

Although Nginx is a relatively new and growing project, there are already a good number of resources available on the World Wide Web (WWW) and an active community of administrators and developers.

The official website, which is at http://nginx.org/, currently serves as an official documentation reference, and provides links from which to download the latest version of the application source code and binaries. A wiki is also available at https://www.nginx.com/resources/wiki/ and offers a wide selection of additional resources such as installation guides for various operating systems, tutorials related to the different modules of Nginx, and more.

There are several ways to get help if you should need it. If you have a specific question, try posting on the Nginx forum—https://forum.nginx.org/. An active community of users will answer your questions in no time. Additionally, the Nginx mailing list, which is relayed on the Nginx forum, will also prove to be an excellent resource for any question you may have. And if you need direct assistance, there is always a group of regulars helping each other out on the IRC channel #Nginx on irc.freenode.net.

Another interesting source of information is the blogosphere. A simple query on your favorite search engine should return a good number of blog articles documenting Nginx, its configuration, and modules:

Personal websites and blogs documenting Nginx

It's now time to head over to the official website and get started with downloading the source code for compiling and installing Nginx. Before you do so, let us have a quick summary of the available versions and the features that come with them.

Igor Sysoev, a talented Russian developer and server administrator, initiated this open source project back in 2002. Between the first release in 2004 and the current version, the market share of Nginx has been growing steadily. It now serves nearly 15% of websites on the internet, according to a June 2015 https://www.netcraft.com/ survey. The features are numerous and render the application both powerful and flexible at the same time.

There are currently three version branches on the project:

A recurrent question regarding mainline versions is "Are they stable enough to be used on production servers?" Cliff Wells, the original founder and maintainer of the Nginx wiki https://www.nginx.com/resources/wiki/, believes so – "I generally use and recommend the latest development version. It's only bit me once!" Early adopters rarely report critical problems. It is up to you to select the version you will be using on your server, knowing that the instructions given in this book should be valid regardless of the release as the Nginx developers have decided to maintain overall backwards compatibility in new versions. You can find more information on version changes, new additions, and bug fixes in the dedicated change log page on the official website.

As of the mainline version 1.13.8, Nginx offers an impressive variety of features, which, contrary to what the title of this book indicates, are not all related to serving HTTP content. Here is a list of the main features of the web branch, quoted from the official website http://nginx.org/:

Nginx can also be used as a mail proxy server, although this aspect is not closely documented in the book:

Nginx is compatible with most computer architectures and operating systems—Windows, Linux, Mac OS, FreeBSD, and Solaris. The application runs fine on 32- and 64-bit architectures.

Once you have made your choice as to which version you will be using, head over to http://nginx.org/ and find the URL of the file you wish to download. Position yourself in your home directory, which will contain the source code to be compiled, and download the file using wget:

[alex@example.com ~]$ mkdir src && cd src
[alex@example.com src]$ wget http://nginx.org/download/nginx-1.13.8.tar.gz 

We will be using version 1.13.8, the latest stable version as of February, 2018. Once downloaded, extract the archive contents in the current folder:

[alex@example.com src]$ tar zxf nginx-1.13.8.tar.gz

You have successfully downloaded and extracted Nginx. Now, the next step will be to configure the compilation process in order to obtain a binary that perfectly fits your operating system.

If, for some reason, you do not want to bother with the configuration step, such as for testing purposes or simply because you will be recompiling the application in the future, you may simply use the configure command with no switches. Execute the following three commands to build and install a working version of Nginx:

[alex@example.com nginx-1.13.8]# ./configure

Running this command should initiate a long procedure of verifications to ensure that your system contains all of the necessary components. If the configuration process fails, please make sure you check the prerequisites section again, as it is the most common cause of errors. For information about why the command failed, you may also refer to the objs/autoconf.err file, which provides a more detailed report:

[alex@example.com nginx-1.13.8]# make

The make command will compile the application. This step should not cause any errors as long as the configuration went fine:

[root@example.com nginx-1.13.8]# make install

This last step will copy the compiled files as well as other resources to the installation directory, by default /usr/local/nginx. You may need to be logged in as root to perform this operation depending on permissions granted to the /usr/local directory.

Again, if you build the application without configuring it, you take the risk of missing out on a lot of features, such as the optional modules and others that we are about to discover.

When running the configure command, you are offered the chance to enable some switches that let you specify the directory or file paths for a variety of elements. Please note that the options offered by the configuration switches may change according to the version you downloaded. The following options listed are valid with the stable version, as of release 1.13.8. If you use another version, run the ./configure --help command to list the available switches for your setup.

Using a switch typically consists of appending some text to the command line. For instance, using the --conf-path switch:

[alex@example.com nginx-1.13.8]# ./configure --conf-path=/etc/nginx/nginx.conf

Here is an exhaustive list of the configuration switches for configuring paths:

Prerequisites come in the form of libraries and binaries. You should by now have them all installed on your system. Yet, even though they are present on your system, there may be occasions where the configuration script cannot locate them. The reasons might be diverse, for example, if they were installed in non-standard directories. In order to solve such problems, you are given the option to specify the path for prerequisites using the following switches. Miscellaneous prerequisite-related options are grouped together.

Modules, which will be detailed in Chapter 3, HTTP Configuration, and further, need to be selected before compiling the application. Some are enabled by default and some need to be enabled manually, as you will see in the following table.

Other options are available in the configuration script, for example, regarding the streaming modules, the mail server proxy feature, or event management.

Here are a few examples of configuration commands that may be used for various cases. In these examples, the path switches were omitted as they are specific to each system and leaving the default values may simply function correctly.

./configure --prefix=/usr/local/nginx-1.13.8

./configure --user=www-data --group=www-data --with-http_ssl_module --with-http_realip_module

./configure --user=www-data --group=www-data --with-http_ssl_module --with-http_realip_module --with-http_addition_module --with-http_xslt_module --with-http_image_filter_module --with-http_geoip_module --with-http_sub_module --with-http_dav_module --with-http_flv_module --with-http_mp4_module --with-http_gzip_static_module --with-http_random_index_module --with-http_secure_link_module --with-http_stub_status_module --with-http_perl_module --with-http_degradation_module --with-http_gunzip_module --with-http_auth_request_module

./configure --user=www-data --group=www-data --with-mail --with-mail_ssl_module

In some cases, the configure command may fail – after a long list of checks, you may receive a few error messages on your terminal. In most (if not all) cases, these errors are related to missing prerequisites or unspecified paths.

In such cases, proceed with the following verifications carefully to make sure you have all it takes to compile the application, and optionally consult the objs/autoconf.err file for more details about the compilation problem. This file is generated during the configure process and will tell you exactly which part of the process failed.

./configure [...] --with-openssl=/usr/lib64

The configuration process is of the utmost importance—it generates a makefile for the application depending on the selected switches and performs a long list of requirement checks on your system. Once the configure script is successfully executed, you can proceed with compiling Nginx.

Compiling the project equates to executing the make command in the project source directory:

[alex@example.com nginx-1.13.8]$ make

A successful build should result in a final message appearing: make[1]: leaving directory followed by the project source path.

Again, problems might occur at compile time. Most of these problems can originate in missing prerequisites or invalid paths specified. If this occurs, run the configure command again and triple-check the switches and all of the prerequisite options. It may also occur that you downloaded a too recent version of the prerequisites that might not be backwards-compatible. In such cases, the best option is to visit the official website of the missing component and download an older version.

If the compilation process was successful, you are ready for the next step: installing the application. The following command must be executed with root privileges:

[root@example.com nginx-1.13.8]# make install

The make install command executes the install section of the makefile. In other words, it performs a few simple operations, such as copying binaries and configuration files to the specified install folder. It also creates directories for storing log and HTML files, if these do not already exist. The make install step is not generally a source of problems, unless your system encounters an exceptional error, such as a lack of storage space or memory.

The next step is obviously to execute Nginx. However, before doing so, it's important to understand the nature of this application. There are two types of computer application—those that require immediate user input, thus running in the foreground, and those that do not, thus running in the background. Nginx is of the latter type, often referred to as daemon. Daemon names usually come with a trailing d and a couple of examples can be mentioned here—httpd the HTTP server daemon is the name given to Apache under several Linux distributions, named the nameserver daemon, or cron the task scheduler—although, as you will notice, it is not the case for Nginx. When started from the command line, a daemon immediately returns the prompt window, and in most cases, does not even bother outputting data to the terminal.

Consequently, when starting Nginx you will not see any text appear on the screen and the prompt will return immediately. While this might seem startling, it is on the contrary a good sign. It means the daemon was started correctly and the configuration did not contain any errors.

It is of the utmost importance to understand the process architecture of Nginx and particularly the user and groups its various processes run under. A very common source of troubles when setting up Nginx is invalid file access permissions—due to a user or group misconfiguration, you often end up getting 403 Forbidden HTTP errors because Nginx cannot access the requested files.

There are two levels of processes with possibly different permission sets:

The Nginx binary accepts command-line arguments for performing various operations, among which is controlling background processes. To get a full list of commands, you may invoke the help screen using the following commands:

[alex@example.com ~]$ cd /usr/local/nginx/sbin[alex@example.com sbin]$ ./nginx -h

The next few sections will describe the purpose of these switches. Some allow you to control the daemon, some let you perform various operations on the application configuration.

You can start Nginx by running the Nginx binary without any switches. If the daemon is already running, a message will show up indicating that a socket is already listening on the specified port:

[emerg]: bind() to 0.0.0.0:80 failed (98: Address already in use) [...] [emerg]: still could not bind().

Beyond this point, you may control the daemon by stopping it, restarting it, or simply reloading its configuration. Controlling is done by sending signals to the process using the nginx -s command:

When starting the daemon, stopping it, or performing any of the preceding operations, the configuration file is first parsed and verified. If the configuration is invalid, whatever command you have submitted will fail, even when trying to stop the daemon. In other words, in some cases you will not be able to even stop Nginx if the configuration file is invalid.

An alternate way to terminate the process, in desperate cases only, is to use the kill or killall commands with root privileges:

[root@example.com ~]# killall nginx

As you can imagine, this tiny bit of detail might become an important issue if you constantly tweak your configuration. The slightest mistake in any of the configuration files can result in a loss of control over the service—you are then unable to stop it via regular init control commands, and obviously, it will refuse to start again.

In consequence, the following command will be useful to you in many occasions. It allows you to check the syntax, validity, and integrity of your configuration:

[alex@example.com ~]$ /usr/local/nginx/sbin/nginx -t

The -t switch stands for test configuration. Nginx will parse the configuration anew and let you know whether it is valid or not. A valid configuration file does not necessarily mean Nginx will start though as there might be additional problems such as socket issues, invalid paths, or incorrect access permissions.

Obviously, manipulating your configuration files while your server is in production is a dangerous thing to do and should be avoided when possible. The best practice, in this case, is to place your new configuration into a separate temporary file and run the test on that file. Nginx makes it possible by offering the -c switch:

[alex@example.com sbin]$ ./nginx -t -c /home/alex/test.conf

This command will parse /home/alex/test.conf and make sure it is a valid Nginx configuration file. When you are done, after making sure that your new file is valid, proceed to replacing your current configuration file and reload the server configuration:

[alex@example.com sbin]$ cp -i /home/alex/test.conf usr/local/nginx/conf/nginx.confcp: erase 'nginx.conf' ? yes[alex@example.com sbin]$ ./nginx -s reload

Another switch that might come in handy in many situations is -V. Not only does it tell you the current Nginx build version, but more importantly it also reminds you about the arguments that you used during the configuration step – in other words, the command switches that you passed to the configure script before compilation:

[alex@example.com sbin]$ ./nginx -Vnginx version: nginx/1.13.8 (Ubuntu)built by gcc 4.8.4 (Ubuntu 4.8.4-2ubuntu1~14.04)TLS SNI support enabledconfigure arguments: --with-http_ssl_module

In this case, Nginx was configured with the --with-http_ssl_module switch only.

Why is this so important? Well, if you ever try to use a module that was not included with the configure script during the precompilation process, the directive enabling the module will result in a configuration error. Your first reaction will be to wonder where the syntax error comes from. Your second reaction will be to wonder if you even built the module in the first place! Running nginx -V will answer this question.

Additionally, the -g option lets you specify additional configuration directives, in case they were not included in the configuration file:

[alex@example.com sbin]$ ./nginx -g "timer_resolution 200ms";

Most Linux-based operating systems to date use a System-V style init daemon. In other words, their start up process is managed by a daemon called init, which functions in a way that is inherited from the old System V Unix-based operating system.

This daemon functions on the principle of runlevels, which represent the state of the computer. Here is a table representing the various runlevels and their signification:

You can manually initiate a runlevel transition: use the telinit 0 command to shut down your computer or telinit 6 to reboot it.

For each runlevel transition, a set of services are executed. This is the key concept to understand here: when your computer is stopped, its runlevel is 0. When you turn it on, there will be a transition from runlevel 0 to the default computer start up runlevel. The default start up runlevel is defined by your own system configuration (in the /etc/inittab file) and the default value depends on the distribution you are using: Debian and Ubuntu use runlevel 2, Red Hat and Fedora use runlevel 3 or 5, CentOS and Gentoo use runlevel 3, and so on, as the list is long.

So let us summarize. When you start your computer running CentOS, it operates a transition from runlevel 0 to runlevel 3. That transition consists of starting all services that are scheduled for runlevel 3. The question is—how to schedule a service to be started at a specific runlevel?

An init script, also known as the service start up script or even SysV script, is a shell script respecting a certain standard. The script will control a daemon application by responding to commands such as start, stop, and others, which are triggered at two levels. Firstly, when the computer starts, if the service is scheduled to be started for the system runlevel, the init daemon will run the script with the start argument. The other possibility for you is to manually execute the script by calling it from the shell:

[root@example.com ~]# service httpd start

Or if your system does not come with the service command:

[root@example.com ~]# /etc/init.d/httpd start

The script must accept at least the start, stop, restart, force-reload, and status commands as they will be used by the system to respectively start up, shut down, restart, forcefully reload the service, or inquire its status. However, for enlarging your field of action as a system administrator, it is often interesting to provide further options, such as a reload argument to reload the service configuration or a try-restart argument to stop and start the service again.

We will thus create a shell script for starting and stopping our Nginx daemon and also restarting and reloading it. The purpose here is not to discuss Linux shell script programming, so we will merely provide the source code of an existing init script, along with some comments to help you understand it.

Due to differences in the format of the init scripts from one distribution to another, we will here discover two separate scripts: this first one is meant for older Debian-based distributions before they were switched to Systemd.

First, create a file called nginx with the text editor of your choice, and save it in the /etc/init.d/ directory (on some systems, /etc/init.d/ is actually a symbolic link to /etc/rc.d/init.d/). In the file you just created, copy the following script carefully. Make sure that you change the paths to make them correspond to your actual setup.

You will need root permissions to save the script into the init.d directory.

Due to the system tools, shell programming functions, and specific formatting that it requires, the previously described script is only compatible with older Debian-based distributions. If your server is operated by a SystemD-based distribution such as CentOS, Fedora, newer Debian-based and many more, you will need an entirely different script.

Placing the file in the init.d directory does not complete our work. There are additional steps that will be required for enabling the service. First of all, you need to make the script executable. So far, it is only a piece of text that the system refuses to run. Granting executable permissions on the script is done with the chmod command:

[root@example.com ~]# chmod +x /etc/init.d/nginx

At this point, you should already be able to start the service using service nginx start or /etc/init.d/nginx start, as well as stopping, restarting, or reloading the service.

The last step here will be to make it so the script is automatically started at the proper runlevels. Unfortunately, doing this entirely depends on what operating system you are using. We will cover the two most popular families – Debian, Ubuntu, or other Debian-based distributions and Red Hat/Fedora/CentOS, or other Red Hat-derived systems.

[root@example.com ~]# update-rc.d -f nginx defaults

Red Hat-based distributions

For the Red Hat-based systems family, the command differs, but you get an additional tool for managing system startup. Adding the service can be done via the following command:

[root@example.com ~]# chkconfig nginx on

Once that is done, you can then verify the runlevels for the service:

[root@example.com ~]# chkconfig --list nginx
Nginx  0:off   1:off   2:on    3:off   4:on    5:on    6:off

Another tool will be useful to you for managing system services, namely, ntsysv. It lists all services scheduled to be executed on system startup and allows you to enable or disable them at will:

Note

ntsysv requires root privileges to be executed. Prior to using ntsysv, you must first run the chkconfig nginx on command, otherwise nginx will not appear in the list of services.