Don't be Afraid of CSI: How Critical TechWorks Orchestrates Stateful Apps with Nomad

At Critical TechWorks, a BMW Group company, the team decided to build a tool to help teams quickly get going with their desired DevOps technological stack, whether that meant Jenkins or GitLab, Nexus or Artifactory, etc.

While working on it, they needed a way to deal with jobs that needed persistent storage. They started with a relatively naïve approach that would create different types of HashiCorp Nomad clients (some with external disks, some without) and manage things via Nomad host volumes, but that was cumbersome and was a pain when upgrading. However, Nomad 0.11 came out just as they were getting ready to work on their MVP, and it brought some new toys to play around with...

This talk is the story of how they went from a vague idea of how to handle orchestrating stateful applications, to a stable and functional CSI (Container Storage Interface)-based process, while being some of the world's first to push the boundaries of this innovative technology.

Key takeaways: - Nomad is a container orchestration platform that can handle stateful workloads well - It's important to give back to the community through asking relevant questions, improving documentation, etc. - Using the same technologies to deploy both third-party tools and your applications makes everyone's life easier

Highlighted tools: Nomad (main focus), with help from HashiCorp Vault, Terraform, Packer, and Consul

GitHub Repo for this talk: https://github.com/carlosrbcunha/hashitalks2021

Slides here: https://www.dropbox.com/s/h3snbaz8dzv34c4/Hashitalks2021_MaiaCunha.pptx?dl=0

Speakers: José Maia & Carlos Cunha