Published on

KubeCon 2023: My notes & TLDRs (Day 3)

Notes & TLDRs

Day: Wednesday (19/04/2023)

Keynotes

  • Main Keynotes
    • CNCF has 159 projects!! (mindblowing, considering that they were founded in 2015)
    • 50% of the attendees are new starters this years.
    • KubeCon in 2024 will be in Paris (March)
    • New certifications added - KCSA, KCNA, CGOA
    • New documentary - Inside Envoy.
    • New newsletter - Wisdom of the Cloud
    • A LOT of talk on sustainability... People are not happy how much electricity Kubernetes uses.
  • Graduated project updates
    • Linkerd users doubled in the past year.
    • Linkerd now holds Linkerd day.

Talks

  • Learn the Helm Code Base and PR Review Process - Scott Rigby, Independent; Andrew Block & Karena Angell, Red Hat

    • HIPs sounds like a good idea. The idea behind is that if you have an improvement proposal (Functional, Documentation or any other type of improvements) - then write HIP and get feedback before you touch the code.
    • They have roles in their organization (e.g. Subproject Maintainer,Triage Maintainer*, or Community Maintainer*). (Which, by the way, I think that are a must in any shared/common library, even if it's within the company you are working on)
    • They have bot, which assigns size label (i.e.small,medium, extra large).
    • Anything else than extra small (which is < 5 lines of code) requires 2 reviewers.
    • rook-ceph will support COSI in next version (v1.12), it's better than CSI apparently.
    • You can encrypt OSDs

  • 🔥 Choose Your Own Adventure: The Treacherous Trek to Development - Whey Lee, VMware & Viktor Farcic, Upbound

    • Loved this talk! The public was allowed to choose any tech, and the guy just came up with solution on-the-go! **Amazing! **
    • cdk8s looks pretty amazing! The idea is that you define k8s manifests in Go, Typescript, etc, and it generates a yaml. It can be a replacement for Helm or Kustomize.
    • DevSpace and Telepresence can be complementary to each other.

  • Rook: Intro and Deep Dive with Ceph Storage - Travis Nielsen & Blaine Gardner, IBM Storage; Alexander Trost & Deepika Upadhyay, Koor Technologies, Inc

    • P.S: For those who don't know: CEPH is a distributed software-defined storage solution.
    • CEPH is used in CERN's Large Hadron Collider 😱
    • Hmm. At work we use node-attached devices for CEPH, but folks mentioned that rook-ceph supports cloud volumes and loopback devices too. Need to think how to support all three in Redgate Clone (existing product I am working on here, in Redgate)
    • COSI is 'better' than CSI driver?
    • You can encrypt OSDs

  • 🔥 Tutorial: Hands on with WebAssembly Microservices and Kubernetes - Jiaxiao Zhou, David Justice & Kate Goldenring, Microsoft & Radu Matei, Fermyon

    • WASM has 2 runtimes - Javascript (browser) & WASI (designed to be independent of browser)
    • You can run WASM using - built-in runtime (i.e. cargo target), framework (like Spin), or runwasi with Kubernetes.
    • Folks teach you how to run app in WASI runtime, instead of a docker.
    • WASI image is 26x times smaller compared to Docker container, it's also much faster and uses way less resources.
    • Yet another amazing talk.

  • How to Blow up a Kubernetes Cluster - Felix Hoffmann, iteratec

    • When a node runs out of memory, K8s terminates pods that exceed their memory requests.
    • Kafka stores all events in memory.
    • Main takeaways: 1) Set memory requests == memory limits, 2) Dont set CPU limits.

  • Observability with Fluent Bit: Logs, Metrics & Traces - Eduardo Silva & Anurag Gupta, Calyptia