Kubernetes Seattle: Managing and Debugging Etcd & App Deployment with Pulumi

Amazon - 1918 8th Ave. Seattle - View Map Seattle
Wed, Oct 17, 2018, 6:30 PM (PDT)

About this event

Kubernauts!

In October we are featuring two talks:

Etcd - Usage, Status, and Roadmap.

Kubectl is Not Enough: Reliable App Deployments on Kubernetes

The meet-up will be on Wednesday, October 17, 2018 and starts at 6:30 with light food/beverages, speakers start at 7:00.

If you would like to present or host feel free to reach out to me via email (baldwin_at_stackpointcloud.com) or Twitter @baldwinmathew.

Agenda:

Meeting Room: Puget Sound (3rd Floor)
Event Date: Oct 17th
Event Time: 6:30 – 8:30pm

6:30 – 6:50 pm – Arrive and Mingle

6:50 pm – Meeting Start: Intro, Code of Conduct, Guest Speakers Intro, Host Intro

7:00 - 7:30 - Gyuho Lee from the AWS EKS OSS Team will be talking about etcd.

Etcd is core to Kubernetes control plane, providing various distributed system primitives for high availabilities. Etcd is known to be hard to manage and debug. In this talk we will review some challenges, provide some operational tips, and discuss the roadmap to making etcd more accessible and reliable.

7:30 - 8:00 - Alex Clemmer (@hausdorff_space) will be presenting Pulumi.

Brendan Burns has said that Kubernetes competes mainly with bash — that it is aimed at users who historically would (e.g.) scp application files to servers and then start them with ssh. Perhaps more recently they might run a Docker image instead.

But while Kubernetes has modernized this workflow, many things are still relegated to bash. This includes both simple things, like manually rolling out changes to a ConfigMap's data (since changing the data in a ConfigMap does not trigger a rollout of Pods that reference it), as well as complex things like A/B traffic splitting, staged rollouts, etc.

In this talk we present Pulumi, an OSS cloud configuration SDK, which is aimed at solving these problems. We will demonstrate the CLI interface, which proactively displays information about Kubernetes resources that fail to initialize correctly (e.g., when a Deployment rollout fails). We will then show how we can use the real languages (e.g., TypeScript, Python Go) to model these deployments in a way that is reliable and predictable, without sacrificing the ability to reason about the blast radius of a change.

8:00 – 8:30 pm – Wrap up and head out.

Bios:

Gyuho is working on AWS EKS open source engineering. He is the lead maintainer of etcd. He loves to talk about distributed systems, and is passionate about making complex systems easier to understand.

Alex works at Pulumi. He was an early employee at Heptio. Before that, he worked at Microsoft on Mesos, the Bing index core, and a large-scale stream processing system.

Location:

We'll be hosting at Amazon this month!

Amazon Blackfoot Building
1918 8th Ave
2nd Floor - Room 201
Seattle, WA

When

Wednesday, Oct 17
6:30 PM - 9:00 PM (PDT)

Where

Amazon
1918 8th Ave. Seattle