Stateful systems in the time of orchestrators

Danielle Lancashire (HashiCorp)

14:20–15:00 Thursday, 7 November 2019

Location: Hall A3

Building Cloud Native Systems

Average rating:

(4.00, 2 ratings)

Who is this presentation for?

Developers and operators

Level

Intermediate

Description

Stateful workloads have always been difficult to manage, often requiring a lot of domain knowledge to reliably upgrade applications, operating systems, and the underlying hardware. In many cases, especially before the move to the cloud, they were also often configured once, never to be seen again before a critical outage occurred.

Danielle Lancashire dives into how workload orchestrators can help you to reliably run your stateful workloads by exploring the container storage interface, detailing common issues and production stories of running stateful workloads in orchestrators, providing solutions to reliably running workloads in an orchestrator, and explaining when you should run stateful services yourself versus outsourcing to a managed service.

Prerequisite knowledge

Familiarity with Nomad, Mesos, or Kubernetes
A basic understanding of a stateful tool like PostgreSQL

What you'll learn

Discover the container storage interface and how to use it
Identify frequent problems with stateful workloads
Learn how operators and developers can leverage these features to reliably run their stateful services in cluster orchestrators, ranging from special cased nodes to using CSI and managing ephemerality

Danielle Lancashire

HashiCorp

Danielle Lancashire is a senior software engineer at HashiCorp where she works on the Nomad cluster orchestrator. Previously, she was a staff software engineer at CircleCI where she worked on build infrastructure and was a maintainer of CocoaPods and fastlane. She’s also been described as the queer who yells at clouds.