June 2021 Summaries

The text discusses how Fly, an AI platform, enables developers to deploy globally distributed applications with replicated databases for both reads and writes using standard tools and features. It explains that while it's easy to scale out a database only used for reading, handling updates across multiple instances is challenging. The article then introduces a hack that simplifies this process by allowing the application to handle read/write operations based on exceptions generated when attempting to write to a read replica. This approach involves minimal code and is easy to understand.

Fly.io is a platform that simplifies deploying and running Elixir applications by converting Docker containers into fleets of Firecracker micro-vms around the world. It also provides built-in Prometheus metrics for monitoring purposes. The integration with PromEx, an Elixir library, makes it easy to export various metrics from your Elixir applications and visualize them in Grafana dashboards. This combination allows developers to quickly set up application monitoring and proactively address issues without wasting time on extensive testing of all application components after each deployment.

Livebook, initially Elixir's version of Jupyter Notebooks, has evolved into a powerful tool for various uses. One such use is creating runbooks, which are routine procedures and operations carried out by system administrators or operators. By connecting a local running Livebook to a remote live Elixir system on Fly.io, users can perform tasks like finding an account and making it admin, generating revenue reports, and more. The notebooks can be version controlled, kept up-to-date, and shared among the team as they are plain markdown files.

Fly.io is currently hiring for their SRE team and offers an interesting platform that converts container images into fleets of Firecracker VMs running on their hardware worldwide. The company values a strong ops culture, with everyone contributing to operations work. They use Linux, HashiCorp stack, Firecracker, WireGuard, Rails-based GraphQL API, Prometheus-style metrics cluster, ElasticSearch cluster for logging, and Sensu Go monitoring system. Fly.io is a remote team with members in various locations and offers health care for US employees, flexible vacation time, hardware/phone allowances, and other standard benefits. They have an unusual hiring process that focuses on discovering new talent rather than relying solely on resumes or interviews. The role involves wearing multiple hats such as managing server fleets, configuring alert systems, shipping features with canaries or blues and greens, diagnosing networking issues, and expanding data centers globally. Candidates should be comfortable coding in Python, Ruby, or Go. Interested applicants can email [email protected] for more information.

Mark Ericksen introduces a method to use Fly networking and Observer, a WxWidgets graphical interface for observing running Erlang nodes, to monitor Elixir applications in production. By setting up a secure WireGuard tunnel between the local machine and the servers on Fly, developers can run Observer locally and connect to remote Elixir nodes. This allows them to inspect individual processes or even kill them if necessary. The process is demonstrated using a multi-region Tic-Tac-Toe game created with Elixir. By observing the application in production, developers can identify issues, test system recovery from unexpected failures, and improve user experience during crash recovery.