|
|
<!DOCTYPE html>
|
|
|
<html lang="en">
|
|
|
<head>
|
|
|
<meta charset="UTF-8">
|
|
|
<title>How to Solve The Django Deployment Puzzle | tait.tech</title>
|
|
|
<link rel="stylesheet" href="/assets/css/style.css">
|
|
|
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
|
|
<meta name="author" content="Tait Hoyem">
|
|
|
<meta name="keywords" content="">
|
|
|
<meta name="description" content="">
|
|
|
</head>
|
|
|
<body>
|
|
|
<div id="wrapper">
|
|
|
<header>
|
|
|
<h1>tait.tech</h1>
|
|
|
|
|
|
<nav>
|
|
|
|
|
|
<a href="/" class="nav-link" >Home</a>
|
|
|
|
|
|
<a href="/blog/" class="nav-link" >Blog</a>
|
|
|
|
|
|
<a href="/ideas/" class="nav-link" >Ideas</a>
|
|
|
|
|
|
<a href="/links/" class="nav-link" >Links</a>
|
|
|
|
|
|
<a href="https://github.com/TTWNO/" class="nav-link" target="_blank" rel="noopener noreferrer" >Github</a>
|
|
|
|
|
|
</nav>
|
|
|
|
|
|
</header>
|
|
|
<main>
|
|
|
<article>
|
|
|
<header>
|
|
|
<h1 class="post-title">How to Solve The Django Deployment Puzzle</h1>
|
|
|
<time datetime="20-08-18" class="post-date">Tuesday, August 18 2020</time>
|
|
|
</header>
|
|
|
<hr>
|
|
|
|
|
|
<p>A few days ago I had a Django project I wanted to put on a real server.
|
|
|
This project is still in its infancy, but I thought it would be nice to put it on my resume and show my friends.
|
|
|
Little did I know the headache coming my way.
|
|
|
Here are some tips to help you not make the same mistakes as me.</p>
|
|
|
|
|
|
<h3 id="asgi-servers">ASGI Servers</h3>
|
|
|
|
|
|
<p>Because my project used the ASGI (Asynchronous webServer Gateway Interface),
|
|
|
I needed to find a good production ASGI server to handle all the incoming requests.
|
|
|
The best thing I found was <a href="http://www.uvicorn.org/">uvicorn</a>.
|
|
|
It focuses on speed, which is a priority, especially when using the ASGI protocol.</p>
|
|
|
|
|
|
<p>To run uvicorn on the command line for testing purposes, use something like the following:</p>
|
|
|
|
|
|
<pre class="terminal">
|
|
|
$ uvicorn --reload myapp.asgi:application
|
|
|
</pre>
|
|
|
|
|
|
<p>The <code class="language-plaintext highlighter-rouge">--reload</code> option says to reload the server if any of the files get updated.
|
|
|
This is not recommended in production.
|
|
|
Sadly, I thought this meant I would need to do a hard shutdown of the server process every time I wanted to update.
|
|
|
This turned out to not be the case.</p>
|
|
|
|
|
|
<h3 id="workload-managers">Workload Managers</h3>
|
|
|
|
|
|
<p>There is another equine-named program called <a href="https://gunicorn.org/">gunicorn</a>
|
|
|
which can hold a number of processes under its control.
|
|
|
An interesting feature of <code class="language-plaintext highlighter-rouge">gunicorn</code> is that it will gracefully switch from an old to a new deployment,
|
|
|
replacing the subprocesses one-by-one and eventually having only the new deployment active on all subprocesses.
|
|
|
The greatest part? Zero down time.
|
|
|
The server keeps any old processes open if there is communication with them,
|
|
|
then shift and new connections to the new deployment.
|
|
|
This was a very cool feature I wanted to take advantage of.</p>
|
|
|
|
|
|
<p>“Now hold on!” you might protest.
|
|
|
“gunicorn is a WSGI server!” … oh you got me there!
|
|
|
Yes, that’s right, <code class="language-plaintext highlighter-rouge">gunicorn</code> is paired with <code class="language-plaintext highlighter-rouge">uvicorn</code> to serve my files.</p>
|
|
|
|
|
|
<h3 id="systemd">systemd</h3>
|
|
|
|
|
|
<p>Love it or hate it, the majority of Linux distributions use the <code class="language-plaintext highlighter-rouge">systemd</code> init system.
|
|
|
I decided it would be very convenient to have a .service file for my Django application to run automatically at boot.
|
|
|
<code class="language-plaintext highlighter-rouge">Systemd</code> allows me to do this with a file like the following one I stored in <code class="language-plaintext highlighter-rouge">/lib/systemd/system/lamegames.service</code>.</p>
|
|
|
|
|
|
<pre class="file">
|
|
|
[Unit]
|
|
|
Description=Gunicorn/Uvicorn (lamegames.io)
|
|
|
|
|
|
[Service]
|
|
|
WorkingDirectory=/home/lame/lamegames.io
|
|
|
Type=simple
|
|
|
RemainAfterExit=yes
|
|
|
ExecStart=/home/lame/lamegames.io/env/bin/gunicorn lamegames.asgi:application -w 2 -k uvicorn.workers.UvicornWorker
|
|
|
ExecStop=/bin/kill -HUP $MAINPID
|
|
|
Restart=always
|
|
|
|
|
|
[Install]
|
|
|
WantedBy=multi-user.target
|
|
|
</pre>
|
|
|
|
|
|
<h3 id="nginx">nginx</h3>
|
|
|
|
|
|
<p>NGINX (pronounced engine-X) is a performance web server designed for speed and simplicity.
|
|
|
For the front facing side of the site, I do need a production web server like nginx.
|
|
|
Gunicorn simply doesn’t need all the features that nginx provides, but I do.
|
|
|
To configure my nginx installation, I used the following few directives to:</p>
|
|
|
|
|
|
<ol>
|
|
|
<li>Redirect most traffic towards the gunicorn server.</li>
|
|
|
<li>Redirect statically served files (CSS, JS, images) to the directory specified in the STATIC_ROOT variable of my <code class="language-plaintext highlighter-rouge">settings.py</code> file.</li>
|
|
|
<li>Use TLS to enable https://</li>
|
|
|
</ol>
|
|
|
|
|
|
<p>Serving the static files from nginx as opposed to the <code class="language-plaintext highlighter-rouge">gunicorn</code> server is necessary.
|
|
|
Gunicorn and other production A/WSGI web server will not set the proper MIME type over TLS.
|
|
|
This will cause your browser to not load the Javascript/CSS.</p>
|
|
|
|
|
|
<p>This is the important part of my nginx config.</p>
|
|
|
|
|
|
<pre class="file">
|
|
|
server {
|
|
|
location / {
|
|
|
proxy_set_header Host $http_host;
|
|
|
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
|
|
|
proxy_set_header X-Forwarded-Proto $scheme;
|
|
|
# these two lines ensure that WebSocket, and HTTP2 connection are forwarded correctly
|
|
|
proxy_set_header Upgrade $http_upgrade;
|
|
|
proxy_set_header Connection "Upgrade";
|
|
|
proxy_redirect off;
|
|
|
proxy_buffering off;
|
|
|
# this forwards all traffic to the local server on port 8000
|
|
|
proxy_pass http://localhost:8000;
|
|
|
}
|
|
|
|
|
|
# This forwards all static requests to Django's STATIC_ROOT set in settings.py; it is generated using the collectstatic command.
|
|
|
location /static {
|
|
|
autoindex on;
|
|
|
alias /home/lame/lamegames.io/static_generated;
|
|
|
}
|
|
|
}
|
|
|
</pre>
|
|
|
|
|
|
<h3 id="setup">Setup</h3>
|
|
|
|
|
|
<p>After all that, I was able to do the following:</p>
|
|
|
|
|
|
<pre class="terminal">
|
|
|
# systemctl enable lamegames
|
|
|
</pre>
|
|
|
|
|
|
<p>This enabled my <code class="language-plaintext highlighter-rouge">gunicorn</code> server to run once the server started.
|
|
|
NGINX is that way be default.</p>
|
|
|
|
|
|
<p>And tada! You now have a working Django project on a production server!</p>
|
|
|
|
|
|
<h4 id="notes">Notes</h4>
|
|
|
|
|
|
<ul>
|
|
|
<li>If using ws:// websockets, change them to wss:// for secure web sockets.</li>
|
|
|
<li>Make sure to use channels.routing.get_default_application() instead of django.get_asgi_application() if your’re wanting to use channels/redis WebSockets.</li>
|
|
|
</ul>
|
|
|
|
|
|
</article>
|
|
|
|
|
|
</main>
|
|
|
<hr>
|
|
|
<footer>
|
|
|
This page will be mirrored on <a href="https://beta.tait.tech/2020/08/18/django-deployment/">beta.tait.tech</a>.
|
|
|
|
|
|
</footer>
|
|
|
</div>
|
|
|
</body>
|
|
|
</html>
|
