Setting up the development environment: needlework circle (Part 1)

Hello dear reader!
Today I want to share my experience in setting up a personal environment for working with various PHP-based projects. This article describes the experience of manually setting the environment.

This article was written for my beloved to show how to simplify or complicate your life. I do not want to call for action, but only share what I encountered and what I did.

I will automate the process of deploying the environment in one of the following articles.
')
The article will be very long with a bias towards the technical side. Please under the "cat".

Prehistory

Before I begin, I want to share a little story about how it all began.

For a long time, the development process was conducted mainly on one project, with an emphasis on long-term prospects (there was a lot of work: 1 year in advance). And so it was from company to company. So it wasn’t too difficult for me to add a couple of “configs” and reload services on the local machine to add a new project. When changing the company, I just set up everything from scratch adapting to the project.

At the end of 2016, I had to switch from one project to another due to a series of very unpleasant events. Turning to a new project, it was faced with the fact that the project architecture was built in such a way that you need to simultaneously support at least 5 web services that work in conjunction. I also still have to support previous projects from time to time. And do not forget about R & D projects, Open Source projects, “hack-work”, family and children.

This set of circumstances influenced the subsequent history of my development environment and everything connected with it.

Purpose: flexible development environment for PHP-based projects

I need to get a working environment that allows me to develop, debug and research in multiple PHP-based projects with reference to the company's infrastructure.
Note: The following description refers to two projects that are now in my area of ​​responsibility: Warface Hub is what is left of it, and CryEngine Hub is the result of lightning-fast implementation, “successful” architecture and “proper” planning.

So i got the following

• Windows 10 - Host with VMWare or VitrualBox
• Ubuntu - Guest c:
  
  • Samba - for working with project files
  • Bind9 - local DNS server
  • Nginx, PHP FPM, MySQL and everything else
  
  
• Other virtual machines

Resources

In large companies, Microsoft solutions are very often used to provide internal communication between employees (Outlook, Lync, AD) and even heaps of self-written applications. Since I am not too lazy to search and configure clients for Linux or use Wine, I successfully started using the “hybrid” of two OSs and virtualization:

• Windows 10 (Host OS)
• Ubuntu 16.04 (Guest OS - VM)
• SSD - as fast as possible. With this implementation, it is very important to have a fast disk.
• 16Gb ram - The more the better. I have to allocate up to 12Gb for all guest OS, since I have to solve cross-browser problems that are not reproduced on emulators.

Installing services, applications and components.

At this stage, we install everything (Guest OS) that we may need in order to raise a simple stack from Nginx, PHP and MySQL, as well as everything else that can be useful in projects (NPM, Yarn, Composer). Do not forget about our development environment and add Samba and Bind9 there.

Samba server

The first thing I reached out for was the Samba server , which was configured so that my IDE could work directly with the guest OS file system.

All my projects will be located on the path / var / www / {project-name}.
Also in the process of setting up the guest OS, I forcibly configured a DHCP server so that it has a static IP and hides everything behind NAT: 172.16.126.130.


In order for the IDE to work with this resource, I mount this Netbios folder as a remote local disk.


Why did I start using Samba and not Shared Folders?
At the time when I asked this question, it turned out that Shared Folders was a directory to be mounted, and NGINX did not want to use them as the root directory of the project. The second reason was that from time to time my Shared Folders fell off for no apparent reason and without the ability to quickly restore everything. And also there is no possibility to work with symlinks, which are necessary for some npm packages and not only.

Are there any problems with IDE?
Yes, and different, but the SSD saves:

• sometimes project indexing takes a long time
• when quickly changing a large number of files (merge, rebase, changing brunch), changes are not immediately picked up. I look towards rsync.
• If to disconnect a disk earlier, than to close IDE, then, when closing the project, IDE will hang up
• If the network falls off, the project will most likely need to be opened again.

DNS server

Dear reader can imagine how difficult it is in a large company to request small changes in network settings. And even if you successfully complete the first stage of an explanation of why you need it, the second stage (implementation) may take a very long time, which is easier to do yourself.

Since I am very lazy, after a while I’m tired of adding a record file to the hosts:
172.16.126.130 project1.lo
172.16.126.130 project2.lo


It was decided that we need a local DNS server, which can be configured faster than I can drink a bottle of German strong beer. After which a request was sent to the league of laziness. The result was not long in coming: BIND9

Setup is very simple: we describe the configuration of the forward and reverse pool of domains, and also do not forget to connect them:


The server is configured. Now, for complete joy, it remains to solve a few more problems:

• force the Host system to use our DNS server for the .lo domain. Therefore, we change the parameters of the virtual network on the Host machine so that our DNS server is the first, and we clean the DNS cache.
• force Guest to use our DNS server too. It's easier here: edit the network interface and register our DNS server there

The result is not long in coming:


That's just the "nslookup example.lo" does not work. Why, I don’t know, but any browser on the Host machine will normally go to the corresponding server.

PHP-FPM and Pools

It seems everything works. Now you need to start setting up PHP for each of the projects.
Working on the latest projects, it turned out that one project can consist of various services or collect many other projects in itself. That's why I got the following structure:

• warface.com
• cryengine.com
• shop.cryengine.com
• forum.cryengine.com

In the course of work, you have to use API calls between projects and it would be very good to be able to debug this all with the help of XDebug and your favorite IDE. When trying to use one PHP Pool, I encountered the following problems:

• when debugging on the same domain with a common session, PHP-FPM blocks the execution of other requests from the same sessions (as I could and described :)). Thus, staying somewhere on cryengine.com I will not be able to work with forum.cryengine.com until the previous request is completed.
• It is inconvenient to debug two projects simultaneously with different XDebug keys. Yes, yes, I try to avoid changing the environment settings from PHP as much as possible.

To solve this, I created my PHP Pool using a template for each project for each version of PHP. And connect them to 'include = / etc / php / 7.0 / fpm / pool.d / *. D / *. Conf'


It looks like this:


Now each project has its own PHP Pool and xdebug.idekey, which allows you to perform continuous debugging between several projects (do not forget to configure php.ini)

Nginx

Setting up the server is pretty simple. For each project, a separate configuration file is created in 'sites-available' , which describes the vhost configuration. In the configuration, we include the appropriate configuration files for each project.

Based on the above architecture, I implemented the following structure:


Thus, in /etc/nginx/{project-nameki.d/ all configurations are stored that relate to the project, as well as related projects. Each project has its own PHP Pool while each version of PHP, which allows you to quickly switch between versions and only for one project.

Results

Thank you for reading to the end. I sincerely hope that manual labor has not yet died out, because the developer must be able to set up the environment himself and without automation.
As a result, I received a hybrid between the two operating systems, which allows you to fully use all the internal services of the company, as well as conduct development and debugging in an environment as close as possible to LIVE.

I understand that this is not an ideal solution, but this is one of the ways how to prepare a development environment in a short period of time.

Cons :

• Long indexing in IDE
• When reinstalling / migrating a virtual machine, you need to change the configuration for Bind9 and the disk connection script
• This is all not automated.

Pros :

• the working environment can be transferred from machine to machine
• the developer remains in the Windows environment and it is possible to use MS services that are installed in the company.
• The developer works with a Linux environment that can be easily configured as a LIVE environment.

Update:
Setting up the development environment: coffee grounds (Part 2)