October 2020 Summaries

In this text, Dj Walker-Morgan discusses various aspects of using Fly, a platform for deploying applications. The key points include: 1. Answering questions about Fly on Twitter and in the Fly Community. 2. Using the `fly suspend` command to park an application without destroying it or losing configuration, followed by the `fly resume` command to bring it back online. 3. The `fly restart` command is used for restarting instances of an application when it gets into an indeterminate state but doesn't trigger any deployments. 4. Publish/Subscribe (Pub/Sub) is not supported on Fly's Redis implementation, along with some other unsupported commands related to server or client management and geo-based Redis commands. 5. The minimum number of instances for an app cannot be set to 0; the `fly suspend` command should be used instead. 6. The Fly CLI is being migrated from flyctl to the shorter fly command, with back-compatibility in mind. 7. The community site at community.fly.io is a great place for asking questions and getting help with Fly applications.

Fly, a platform for deploying applications, has released new features and improvements. Persistent storage is now available in preview for Fly applications with the volumes command allowing creation of persistent disks. Environment variables can be set in the fly.toml file instead of using secrets. The latest release of Fly's CLI tool includes better builtins support with settings, a Python/Procfile builtin, updated Deno and static website builtins, new commands for Volumes (persistent disk support) in preview, automatic use of default org when initiating an app, added --watch to status command for autorefresh, and fixed wildcard domain support.

The text discusses the use of eBPF (extended Berkeley Packet Filter) for deploying UDP applications on Fly.io, a content distribution network for Docker containers. It delves into the history and evolution of packet filters, from CSPF to BPF+ and eventually eBPF, highlighting their role in observability and access control. The author explains how eBPF is used in TC (traffic classifier) system and XDP (eXpress Data Path), allowing for real-time decision making, modification of packets, and routing them to the appropriate interfaces or sockets. The text also provides advice on writing BPF C code, debugging tools like dropwatch and perf, and emphasizes the potential benefits of using eBPF in various networking scenarios.

Fly has replaced its entire logging platform with a new approach that involves using Vector, an efficient log processing tool written in Rust. The previous system used seven large Graylog servers to manage logs, but the volume of logs was too high for them to process quickly enough, leading to data loss. By implementing Vector and sending logs directly to Elasticsearch, Fly has improved its logging capabilities and made searching across applications easier with a common log schema. This new architecture also allows for more flexibility in sending logs to different service endpoints in the future.

The article discusses how to turn a local ad-blocking tool called Pi-Hole into a globally available app using Fly, an edge network platform. It explains what Pi-Hole is and its benefits in blocking advertising at the DNS level. The author then describes the challenges of taking Pi-Hole on the road and how it can be made more portable by running it as a Docker image. Finally, the article demonstrates how to deploy a global version of Pi-Hole using Fly's edge network, which provides low latency and global availability. The author also shares their Dockerfile and fly.toml file configurations for deploying the app on Fly.

The text describes the process of creating an online map that tracks a marathon runner from a mobile device using Leaflet.js and deploying it with Fly. It covers drawing out the route, exporting data to KML format, displaying the map, setting up a server for location data collection, enabling the map to get updates, and finally deploying the app on Fly. The author also mentions potential improvements such as adding authentication, Redis support, or multiple runner tracking.