Couple of dudes at Kubecon 2019

Joonas Lehtimäki
Polar Squad
Published in
5 min readMay 24, 2019

It was a typical Tuesday afternoon, two guys from the best DevOps company* Polar Squad walked to the plane to Barcelona to attend KubeCon + CloudNativeCon 2019. I traveled with one colleague Juhani Atula, a senior KubeCon attendee (not really, went to last KubeCon but knew everything) but this was my first time at KubeCon. The atmosphere could be summarized with one word: excited!

Now to the actual topic, here are my thoughts about the days, techs and talks

Day 1

The day started with keynotes from people like Bryan Liles and Lucas Käldström. It’s now the 5th anniversary of Kubernetes, and it showed in the talks, but it was cool to see exciting projects get moved (and the process behind that) from incubation to graduated stage. Congrats Fluentd! It was clear that Kubernetes and CNCF community is running strong.

One keynote that inspired me the most was from Lucas Käldström and Nikhita Raghunath, their story about how to start contributing was absolutely gold, made me want to start coding and giving back to the community. On the other hand, I have to admit I was a little bit jealous towards these two, like they’re close twenties and already contributing and Lucas being SIG lead, COME ON! That’s so cool.

I then went to listen to talks about Istio, service mesh and GitOps. Istio has been hyped for a while now, and it also showed at KubeCon. I wasn’t so familiar with it and talk from Daniel Berg & Ram Vennam opened my eyes to that technology. GitOps panel discussion was also a good one. I highly recommend to check that out if you are not familiar with GitOps.

Day 1 ended to the best keynote in all of KubeCon. CERN guys Ricardo Rocha and Lukas Heinrich showed how they are using Kubernetes to do analytics. They were utilizing 25,000 cores and pods in three K8S clusters, 70TB of data to calculate Higgs boson, LIVE! This demo showed that technologies like Kubernetes and Ceph can really scale when done right!

Day 2

Our day started with a talk from David Xia, Spotify. He was giving a talk about how Spotify deleted their K8S clusters with no user impact. I love these battle stories because they remind us, that everyone, everywhere make mistakes and how we should learn from them. After that, the rest of the keynotes were a little bit too sales like for my taste. IBM announced its OpenSource solution for automating multicluster Kubernetes deployments, let’s see if it takes off, there is a lot of “competition” in that space already, but it’s pleasant to see these companies try to contribute to the community (even tho there might be a hidden agenda).

Multi-Cluster was the most used buzz word in the conference and rightfully so. I think there is a huge advantage to use multiple providers; for example, one provider can do GPU better that other provider. The problem there is the networking aspect and there technology like Istio comes in. It was nice to see different deployment methods one could use with Istio in these situations. It could, for example, be a single control plane with shared subnets or to use multi-control planes, the choice is left for the hackers to decide, which is always a beautiful thing.

Rook, how to bring storage easily to Kubernetes. This is to me, is one of the most interesting Kubernetes projects out there. There is a need for a project like Rook, storage is and has always been a pain in the ass. I bet many of you have stumbled to the problem with EKS and PVs. If you want a PV with good IOPS, you need to settle for a single AZ, which is a deal breaker for most of the companies. Of course, Rook itself doesn’t solve that issue. It needs actual storage tech behind it like Ceph or EdgeFS. Rook makes it super easy to orchestrate storage solutions for Kubernetes with ResourceClass and ResourceClaims.

WireGuard, the rethought way to do VPN. I’m seeing a lot of potential in this tech. Lucas Servén Marín showed in a cool demo how easy it is to set up and use. Wireguard could be a solution to problems like how to connect external databases securely to Kubernetes pods or if you want to use other resources in other cloud providers, i.e. better GPU VM in AWS to GKE.

THE AFTERPARTY

One word, wow! Organizers had booked Poble Espanyol for the after party. If there were problems at the actual KubeCon venue, they definitely compensated that with this event. Food and drinks were 10/10.

Day 3

After surviving the morning, it was time to concentrate on the last talks of KubeCon.

Jaeger, again a cool tech from CNCF project pool. Prithvi Raj gave useful insights on how to use Jaeger, definitely check that video on Youtube.

Serverless, another buzz word, maybe the future how we do software? Maybe. I’m not too convinced yet, the tech is there but it’ll require changing developers mentalities (again) and I don’t see that happening before the actual development process can be made more simple. Erwin Van Eyk made a good point on his talk that for example, when AWS Lambda gets new features, they’re not in the emulators at the same time. This un synchronization between features/mechanisms and tools means that developers have to develop “blindly” serverless, which is not good.

Afterthoughts

TBH I didn’t like the venue that much. There was a lot of walking between the talks and food was really bad if you could even get the lunch bag that you wanted. You could only get soft drinks at lunch if you were there 10 minutes prior. Everything ran out every day. I also didn’t like how the venue staff was pestering people.
Content-wise, KubeCon was excellent. It was a bit overwhelming when trying to select a talk when you want to see multiple. It was also nice to travel with Juhani. It’s nice to go with someone and then discuss thoughts after the talks. At Polar Squad, this is one of the core values in my opinion that we share what we learn and know, which makes us the best DevOps company*. In the end, I enjoyed my time and company. Hopefully, I get the chance to go again!

*according to us

--

--