Open-sourcing LogDevice, a distributed data store for sequential data · LogDevice


We are excited to announce that today we are open-sourcing LogDevice! The code is now available on GitHub under the permissive 3-clause BSD license.

LogDevice is a distributed data store for sequential data originally developed within Facebook. This 2017 blog post in the Facebook Engineering blog explains the basic components of LogDevice. You can also refer to the recording of a presentation from the @Scale conference for more detail.

LogDevice is widely deployed for use at Facebook. Existing use cases include:

  • Stream processing pipelines
  • Distribution of index updates in large distributed databases
  • Machine learning pipelines
  • Replication pipelines
  • Durable reliable task queues

Over the past few years we have built a large number of internal tools to manage LogDevice clusters. We are working on slowly shifting some of that automation to be shipped in the open-source version as well. For this initial release, we are including a bare-bones toolset that forms the core of LogDevice operational infrastructure. In a large-scale production environment it will need to be supplemented by other tools for tasks like monitoring and cluster topology management.

LogDevice comes with a command-line administration tool called LDShell (ldshell). LDShell is the primary management tool that we use here at Facebook as well. Stay tuned as we add more commands and automation to it. This is one of the areas where we expect support from the open-source community.

We are excited to see your contributions to LogDevice, especially those related to integration with cloud computing environments, monitoring systems, and configuration management. Getting LogDevice to run on Kubernetes is a good starting point if you are interested in helping.

To get started with LogDevice, dive into the overview of the system to learn how it works, or check out the code on GitHub

By Ahmed Soliman and Andrejs Krasilnikovs, LogDevice team.