Jonathan Boulle

etcd is an open source distributed consistent key-value store written in Go. It has become a mature cornerstone of a variety of distributed systems for networking, service discovery, configuration management and load balancing. This talk will describe etcd, its history, its new V3 API, and some of its production use cases. It will also dive into some of the ups and downs of using Go to develop software that powers some of the most reliable distributed systems.

The ability to run resilient, hyperscale infrastructure today means running it in a distributed way. But building distributed systems is complex and can be a challenge because one of the essential elements includes removing any single point of failure. The answer? Powering these systems with a reliable, distributed database, designed to ensure your infrastructure stays up and running even when individual servers fail.

etcd is an open source distributed consistent key-value store written in Go that was introduced by the team at CoreOS. Since its release in 2013 it has become a mature cornerstone of a variety of distributed systems for doing networking, service discovery, configuration management and load balancing. This talk will describe etcd, its history, its new V3 API, and some of its production use cases. It will also dive into some of the ups and downs of using Go to develop software that powers some of the most reliable distributed systems.

Linux containers as a virtualization solution are becoming more prevalent. An example of a container solution is rkt, whose development started end of 2014. Under the hood, container software on Linux uses the same cgroups and namespaces API. Both container software and kernel code for containers continue to be developed.

Alban will first describe the Linux API that make containers possible: namespaces and cgroups. Then he will explain how rkt containers use this Linux API.

At CoreOS, we chose to write everything in Go at a time when Python was the language of choice for Linux distributions. It had been almost 10 years since a new Linux distro was created. With the entrance of the cloud era, moving deeply into distributed computing based on Linux containers and multi-core machines, we needed a language that would allow our developers to iterate rapidly and create self-contained components that were not tangled in a web of software dependencies on the operating system. Go was the clear choice because it would enable us to build applications quickly and produce standalone binaries.

With thousands of hours of Go experience, this session will dive into the trenches on Go at CoreOS. It will explore four major projects written in Go: etcd, fleet, rocket and flannel. Attendees will learn tips and tricks for using Go for building and shipping products. Learn the ways Go enables CoreOS to build simple tools to solve a large variety of infrastructure automation tasks, and create high performance components for distributed systems.

This talk will cover some of the major projects written in Go at CoreOS, including etcd, fleet, rocket and flannel. etcd is a highly-available key value store for shared configuration and service discovery, that can be used as a lock service (e.g. for leader election) or for distributed configuration. etcd uses the Go standard library extensively, and the session will detail how we used Go to make our raft implementation lightweight and easy to navigate. Fleet is a distributed init system that extends systemd to the cluster level. The session will explore how fleet uses Go to provide both event-driven and reconcilation-based orchestration of units in a cluster. Rocket is a new container runtime and the first implementation of the appc specification. Developing in Go allowed Rocket to go from concept to functional product in a rapid timeframe. Finally, flannel is an etcd-backed overlay network fabric for Linux containers. Flannel leverages Go’s ability to seamlessly integrate with C code, to provide maximal performance in critical paths but retain the safety of higher level abstractions and ability to leverage existing Go libraries.