Tour of KubeCon NA 2018
I went to KubeCon and all I got was… a brain bursting with exciting technology news!
If you want to drink from the firehose, check out the full playlist of KubeCon 2018 Videos.
If you’d rather have a more curated tour, checkout these talks I found interesting, grouped by topic.
Table of Topics
Kubernetes and the Path to Serverless
Kelsey Hightower, of Google and Kubernetes community fame, gives a playful demo of bootstrapping a Fortran app into AWS Lambda, triggering it with Google Cloud Pub/Sub, and storing its output in Google Cloud Storage (GCS) (Video)
Deploying Serverless Apps to Kubernetes with Knative
Google hosted a workshop on Knative, which I found an excellent introduction to Knative Build and Knative Serving, two of the three main Knative components. Knative on GKE is in alpha, on top of Istio on GKE, which is in beta. There’s no published video for the workshops, but the Knative Intro Code Lab is online if you want to check it out.
Intro: CNCF Serverless WG / CloudEvents
Kubernetes, Istio, Knative: The New Open Cloud Stack
Is Istio the Most Next Gen Next Gen Firewall Ever Created?
Deep Dive: Linkerd
Machine Learning Model Serving and Pipeline Using Knative
Machine Learning as Code: Kubernetes with Kubeflow
Jason Smith, from Google, and David Aronchick, from Microsoft, talk about using Kubeflow to achieve machine learning as code. Also mentions integrations with Katib hyperparameter tuning, NVIDIA TensorRT Inference Server, Argo CD, Jupyter Notebook Spawner, Seldon model serving, Kubebench job benchmarks, Tensorflow Operator, PyTorch Operator, Pyflow, and Kubebench, as well as a demo of the new one-click installer (kfctl.sh). (Video)
Spawning Kubernetes in CI for Integration Tests
CI/CD in Light Speed with K8s and Argo CD
Billy Yuen and Parin Shah, from Intuit, talk about how they wrote Argo CD to replace their usage of Spinnaker in order to provide a Kubernetes-native continuous deployment solution optimized for DevOps use cases that abstracts away infrastructure concerns without hiding platform functionality. (Video)
Operators & Custom Resources
CRDs Aren’t Just for Add-Ons Anymore
Tim Hockin, Principal SWE on Kubernetes at Google, describes how the API Machinery and Architecture SIGs are planning to use Custom Resource Definitions (CRDs) for all new resources and eventually maybe rewrite existing APIs to use them too, leveling the playing field for extensions and operators. (Video)
The Future of Your CRDs
Stefan Schimanski, (a former colleague of mine) from Red Hat, and Mehdy Bohlool (Google) give an update from the API Machinery SIG, talking about how to version Custom Resource Definitions (CRDs) and use mutating webhooks to migrate between versions at runtime. (Video)
Rightsize Your Pods with Vertical Pod Autoscaling
Use the Cluster API to Deploy Clusters On-Prem and in Public Clouds
Loc Nguyen, from VMware, and Kris Nova, from Heptio (now VMware), talk about the new Kubernetes Cluster API and how to use it to deploy and manage upgrades and scaling of Kubernetes components. (Video)
Multi-Cluster & Multi-Tenancy
Running a Distributed System Across Kubernetes Clusters (Regions)
Intro: Kubernetes SIG Multicluster
Daneyon Hansen, from Cisco, and Quinton Hoole and Irfan Ur Rehman, from Huawei, discuss the state of the three multi-cluster projects: Federation v2, Cluster Registry API, and Multicluster Ingress (kubemci). There’s also proof of concept demo of Federated Istio spanning multiple regional K8s clusters on top of Federation v2. (Video)
Using a Kubernetes Operator to Manage Application Tenancy in a B2B SaaS App
Mike Arpaia, creator of osquery and co-founder of Kolide, gives an overview of how they wrote a “Tenant Operator” to manage what he calls “Application Tenancy” in order to provide a hosted SaaS where each customer gets their own full application stack on top of multiple shared Kubernetes clusters. (Video)
Tips & Tricks
Developing Kubernetes Services at Airbnb Scale
Audit in Kubernetes, the Future is Here
Stefan Schimanski (who had like 4 talks scheduled) and Maciej Szulik, both from Red Hat, talk about the functionality of Kubernetes audit logging from alpha in v1.7 through general availability in v1.12. (Video)
Fine-Grained Cost Allocation in Multi-Tenant Kubernetes Clusters
Yuri Shkuro, of Uber, and Pavol Loffay, of Red Hat, demo Jaeger, an open source distributed tracing framework for debugging distributed systems using call graphs, transaction monitoring, performance measurement, and service dependency analysis. (Video)