KubeCon Europe 2022 Recap: Great to be back in person!

Giri Kuncoro
6 min readMay 30, 2022

--

I‘ve been part of Kubernetes and Cloud Native community since 2017, where my first PR to kubectl was reviewed by Brendan Burns, the Kubernetes creator himself. Since then, I realized how warm welcoming the community is, especially for new comers and underrepresented group. Being underrepresented myself (have you worked together with Indonesian engineers in open source project? probably not), it was really hard to meet fellow Indonesians in community meeting, or in KubeCon and other international conferences. Indonesia was not part of the committers table, but today, we are ranked 23rd on top country committers through various projects, including Indonesian localized docs.

I’m writing this blog to share my recent experience attending KubeCon Europe in Valencia in-person (finally!!), and how I met amazing friends I usually collaborate with online. This is my 4th KubeCon (or 7th, if virtual KubeCon is counted), I usually didn’t see other fellow Indonesians in the conference. This time, I’m very proud that there are at least 4 engineers from Indonesia out of 7,000 attendees.

At least 4 engineers from Indonesia attended KubeCon Europe 2022

It was sad that my proposal got waitlisted, but fortunately as underrepresented group, I could attend through Dan Kohn diversity scholarship. I was lucky that I had a chance to talk with Dan Kohn back in 2019, a very humble person and amazing leader (R.I.P Dan). I hope many more Indonesians can apply the scholarship and more people get exposure to this amazing community.

Meeting Dan Kohn in Open Source Summit Japan 2019

Amazing Conversations I had — The Good People

My goal was obviously to support fellow Indonesians who were speaking at conference: Iqbal & Gio, and Mas Aris.

Gio & Iqbal giving chaos talk at KubeCon EU 2022
Mas Aris giving security talk at KubeCon EU 2022

However, my primary goal of attending conference in-person was to make new friends and to make good conversations (there’s a term for this in KubeCon: Hallway Track), because otherwise, how is it different from watching recorded videos on Youtube?

With Daniel Borkmann, the Cilium creator
With Bill Mulligan, we worked together to organize Kubernetes Community Days Indonesia 2021 (the largest KCD in the world)
With Suzanne, Backstage Maintainer, I was able to resolve all my doubts related to backstage
With Tim Bannister, my techlead in Kubernetes docs project
With Kunal (100k Youtube subs within 3 months — wow!), Saiyam (my fellow CNCF ambassador), and Ramiro (okteto creator)
Liz Rice signing her newest eBPF book, surprised that she knows Gojek
With Natalia, eBPF security book author who was working remotely from Bali and used Gojek daily
With fellow Dan Kohn diversity scholarship recipients (there are around at least 30 recipients who gathered)

Interesting Sessions and Booths I attended

Below are some of the attended sessions I like the most:

  • 7 Years of Running Kubernetes in Mercedes-Benz
    Glad to see more and more big name end-users are adopting cloud native ecosystem. Mercedes-Benz has around 1,000 clusters managed with Cluster API and they shared success stories of improving resiliency of their systems. Unfortunately, I could not catchup with them in-person to learn the details and get inputs for my Kubernetes work at office.
  • Datadog Production Incident related to DNS, GRPC config, and Linux kernel edgecase
    This session was one of the most packed sessions at KubeCon, where Datadog engineers shared their learnings after debugging this production incident for few weeks. The symptom was DNS query spikes (triggered OOM), but they found edgecase in Linux kernel which triggered unexpected behavior of Cilium CNI (they raised PR to fix this in Cilium), and tuning GRPC client config.
  • The day we deleted production cluster — InfluxDB
    This was another production incident talk I attended. A classic story where someone raised wrong PR in the infra code, which deleted their production cluster (I did this myself last year, where we had major downtime). However, I took them almost 5 hours to fully get back, because they could not simply revert the changes and the extra complexity of having stateful workloads on Kubernetes. They also had to tune their ArgoCD system to handle the case.
  • Bringing Chaos Engineering to Cloud Native Developers — creator of LitmusChaos & Okteto
    This session gave new paradigm to “shift left” chaos testing towards developers. Instead of running chaos in CI through pre-deployment or post-deployment jobs, the idea was to bring chaos test right in developers’ laptop through ephemeral dev environment (think of running “make test”, but instead of unit tests, it runs chaos test in a remote dev environment).
  • Service Mesh at Scale: 22k Pods in Xbox Cloud Gaming with Linkerd
    Microsoft engineer shared their journey to migrate to Linkerd based service-mesh within 1 year. The decision between Istio vs Linkerd for them them: it was no brainer to setup mTLS for service-to-service communication in Linkerd (given I have Istio production experience, need to try this out with Linkerd). What impressed me the most is how fast they could rollout Linkerd for 22k pods, unfortunately I forgot to ask the details around this. However, they face the same issue I encountered at work where we need to balance tradeoff between rich mesh metrics and high cardinality to avoid prometheus issues.
  • Simplifying Service Mesh Operations with Flux and Flagger — by Flux Creator
    I learned from this session that starting this year, we can jump 2 minor versions when upgrading Istio (need to double check starting from which version). Istio community is “partnering” with Flux/Flagger to make Istio upgrading easy, through GitOps approach and canaries. At work, we have been religiously upgrading Istio using Istio Operator, so we don’t really need this. But, this new direction is interesting and worth to keep an eye on.
  • Guided Tour of Cilium Service Mesh — Liz Rice
    Liz is one of the open source leaders I look up to. Some of you might know here from her amazing talk: container from scratch. I learned Cilium Service Mesh is proxyless mesh, still using Envoy (1 envoy per node), and they are able to optimize kernel space calls thanks to eBPF technology, which significantly improved the latency performance. It is still early days for Cilium Service Mesh, but worth to keep an eye on and see if there’s an easy way to migrate from Istio. Cilium claims Istio control plane can still be used to control their proxyless workloads.

Impressions & The Next Step

First, I’ve been part of KubeCon program committee for at least 7 times and I am super happy to see the talks I selected for KubeCon EU were really good talks.

Second, as part of diversity scholarship recipient, there was no networking session and opportunity to get to know each other which was really sad (I hope I didn’t miss this if there was actually one).

Finally, I really hope this blog and pics I shared are motivating you to attend KubeCon next time, or even submit proposals to speak at this amazing event.

Cheers,
Giri, CNCF Ambassador from Indonesia

--

--

Giri Kuncoro

Gojek Engineering. Kubernetes and Containers. Formerly Cornell and VMware/Pivotal in Silicon Valley.