29 Jan 2020
As I mentioned in my first post, my main project for the past couple of months has been building a Wagtail blog (note: it's this one, the one you're looking at right now). It is finished, and it runs, but deploying it has been a bit of a sticking point. Now that I’m in a place where I could theoretically do so, I don’t want to, mainly because I want to test it properly and learn good habits.
Still, I’ve learnt a lot through this process so far and wanted to share as I go along. Also, I’d rather have at least some content on my blog than wait until absolutely everything is perfect.
I’d been an admirer of 🦈Digital Ocean🦈 long before I even considered learning to code. A few years ago, in my capacity as a content writer, I went to a conference about diversity in tech, and they had some speakers from DO. I immediately loved their branding and picked up some swag on the way out. When I started to get curious about building my own web apps, I dived into their approachable, community-focused documentation. I never actually saw myself using DO’s services, though, because I assumed it was reserved for heavyweight products.
However, when I was looking to host my Wagtail blog, I tried out PythonAnywhere, Heroku, and Divio in succession. These are commonly used for hobby and personal projects, but for various reasons, they each failed to work for me. I’m not dunking on any of these services; I’d absolutely consider using them in the future, but they just didn’t deliver what I needed at that point. Exasperated, I saw that DO were giving free credit to new users, so I thought I’d give them a go. I’m really happy with them so far and have really enjoyed learning about stuff that wouldn’t have occurred to me before, such as…. Ubuntu.
Somewhat naively, I’d assumed I would never have to give Linux OS — which Ubuntu is distributed from — the time of day. I mean, I’m a Mac OS user so I’m basically set for development, right? Wrong. If you’re serious about Python/Django, you will most likely have a brush with Linux at some point.
Thankfully, you don’t need to obtain a whole new Linux machine, or even install an Ubuntu virtual machine, because DO does that for you (well, after you’ve done a bit of setup, of course).
Please note that this article is not an comprehensive walkthrough, but a compilation of observations and extra notes accompanying the Digital Ocean tutorials that I reference here.
Follow the instructions on Initial Server Setup with Ubuntu 18.04 to get the ball rolling.
After successfully SSH’ing into your Ubuntu server, you will most likely receive a message saying packages need to be updated. Run the following:
sudo apt upgrade sudo apt update
And if you get the Permission denied (publickey) error, run this, thus opening the config file: sudo nano /etc/ssh/sshd_config
Permission denied (publickey)
sudo nano /etc/ssh/sshd_config
Make sure PermitRootLogin and PasswordAuthentication are both set to yes, then reboot the server with sudo service ssh restart. Log in with your username again.
sudo service ssh restart
After you’ve finished that, go straight onto the next tutorial: How to Set Up Django with Postgres, Nginx, and Gunicorn on Ubuntu 18.04.
Of course, I used Wagtail instead of the regular Django framework. I’d recommend first familiarising yourself with starting a project on Wagtail — perhaps first on your own local machine — then you can substitute the Django commands with Wagtail ones.
I must say, I got quite a kick out of seeing the Wagtail welcome screen finally pop up, dancing egg and all, after months in vain of trying to get it going!
Now it’s time to meet Gunicorn, the WSGI application server. 👋🏻🦄
You’ll first need to add your app name to base.py in order to get it running. Then, go to the directory where wsgi.py is stored, otherwise you will keep getting worker errors. Basically, as some commenters on the DO tutorial have observed, you will probably need to add an extra directory to the end of the path given in the tutorial — i.e. the directory where the wsgi location is specified). For example, if your project directory is called mynewblog and your app directory is mysite, you will need to enter the path /home/username/mynewblog/mysite/mysite.
When you get to the part where you’re testing Gunicorn, make sure you’re in /home/username/mynewblog/mysite, and from there, run gunicorn --bind 0.0.0.0:8000 mysite.wsgi to start the server and see if the Wagtail site pops up in your browser.
gunicorn --bind 0.0.0.0:8000 mysite.wsgi
It was plain sailing for me up until the point where I had to run the curl command mentioned in the tutorial — curl --unix-socket /run/gunicorn.sock localhost — throwing me the 7 and 56 errors respectively.
curl --unix-socket /run/gunicorn.sock localhost
Once again, when configuring gunicorn.service, be sure to specify the right directory. For example, if your project directory is mynewblog and your app directory is mysite, you will need to make the path /home/username/mynewblog/mysite/mysite and mysite.wsgi:application (i.e. here, the second mysite is the directory holding the wsgi file). Then, exit and restart the SSH connection like so:
sudo systemctl start gunicorn.socket sudo systemctl enable gunicorn.socket curl --unix-socket /run/gunicorn.sock localhost
For some reason I was under the impression that once I came to the end of the last tutorial, I’d be looking at my new site.
Downloading FileZilla was slightly scary, not only because I think the website looks a bit “🤨” and the name evokes a Limewire-era file-sharing site, but also because I had my reservations about giving my private SSH key to a third-party app. Still, there were lots of trusted reviews all over the internet, so I went for it.
At first I couldn’t log in with my (non-root) username. You have to specify the port on FileZilla, and I had assumed it to be either 80 or 8000. Only when I tried 22 did it work; indeed, this is the SFTP (secure file transfer protocol) port.
I then tested this by running netstat -tulpn, which allows you to see the ports you’re connected to.
This is a good article to look at if you get stuck!
If you get this error when trying to run the development server:
django.db.utils.OperationalError: could not connect to server: Connection refused Is the server running on host "<000>.<000>.<000>.<000>" and accepting TCP/IP connections on port 5432?
Go to /etc/postgresql/10/main and run sudo nano postgresql.conf, then add port 5432 (it should be around line 64 of the file).
sudo nano postgresql.conf
If that doesn’t work, run service postgresql status — hopefully the port will be active and everything will look okay. You can also run netstat -na to show your list of current, active internet connections, and if that doesn’t work, try ufw allow 5432/tcp to open the port directly.
service postgresql status
ufw allow 5432/tcp
All this being said, in one case, it took me a while to realise that the problem was not to do with ports and hosts, but in fact to do with my database. I had set an environment variable for it, based on dj-database-url, but for some reason it didn’t take. (I’ll be talking more about environment variables in another post).
Another recurring error was the static files not loading. I’d had this problem before; for some reason, I’d kept the static folder within my app directory, not the project directory — this makes a lot of sense, as it holds the static files (including styling files) for the entire project. Although it was annoying, as always, I learnt something new about directory paths. Check the static path in /etc/nginx/sites-available, then open the file with the same name as your project. You can also mv "static" "/home/username/mynewblog" to move your static directory into the right place.
mv "static" "/home/username/mynewblog"
Here’s some more info about static files.
I’m currently learning about test-driven development (TDD), which is one of the “good habits” I mentioned, so the next part of this series will be about tests, security, and with a bit of hard work and luck, deployment…