February 2023 Summaries

This post provides guidance on how to make a simple Django application production-ready and deploy it to Fly.io. The author walks through the process of setting up the local Django app for production, including using PostgreSQL as the database, managing environment variables with django-environ, handling static files with WhiteNoise, and ensuring security measures are in place. They then cover deploying the production-ready Django app to Fly.io using flyctl, a command-line utility provided by Fly.io. The author also discusses additional steps for deployment, such as setting up database migrations and updating settings.py with the dedicated subdomain chosen during deployment. Finally, they provide resources for further reading on deploying Django apps and encourage readers to share their experiences in the Fly.io Community.

The text discusses a series of distributed systems challenges called Gossip Glomers, developed by Fly.io in collaboration with Kyle Kingsbury, author of Jepsen. These challenges combine real-world coding experience with academic rigor to teach people about distributed systems theory. Maelstrom, a tool built on top of Jepsen, is used to run toy distributed systems on a simulated network. The challenges start off easy and increase in difficulty, covering topics such as unique ID generation, message broadcasting, g-counters, replicated logs, and transactional databases. These challenges were initially designed for evaluating staff engineers at Fly.io but have been released for anyone to try.

Fly.io is a platform that runs apps globally with just a few commands, generating a lot of log output. Centralizing these logs is important for various use cases such as debugging, tracking, collating, correlating, coalescing, and condensing the happenings of your code into human-parsable information. Fly.io's logging system works by grabbing stdout from processes run in apps and sending it to a centralized location via NATS (a "fancy, clustered pubsub service"). Flaps, a proxy that sits in front of NATS, ensures users only see their own logs. To ship logs, an instance of the Fly Log Shipper can be run, which configures a Vector sink and runs Vector to send logs to various destinations like Loki, Datadog, or Cloudwatch. The process is similar regardless of the log aggregator used.

The first alpha version of Django 4.2 has been released, with an expected final release in early April 2023. This long-term support (LTS) release includes several new features and improvements such as psycopg Version 3 Support, database comments on columns and tables, and lookups on field instances. The new version also offers enhanced accessibility in the admin site and forms, and improved asynchronous support. Django 4.2 is expected to be a solid release for one of the most popular web frameworks.

Fly.io is a global sandwich-rating company with a hosting problem. They built a platform using Docker images and an orchestrator called flyd, which operates like a market where requests to schedule jobs are bids for resources. The system has three main components: flyd, flaps, and flyctl. Flyd is the source of truth for all VMs running on a particular worker, while flaps acts as a broker between users and flyd instances. Flyctl, their CLI, takes over complex policy-heavy functionality in the platform. The new design, called "AppsV2", exposes basic primitives directly and allows them to write code to configure it the way they want.