KubeCon 2019 Recap

KubCon 2019 in the review. The biggest announcements and what's coming down the pipeline in the form of releases from Helm, Visual Studio Code, and the Service Mesh Interface.

KubCon 2019 in review. The biggest announcements and what's coming down the pipeline in the form of releases from Helm, Visual Studio Code and the Service Mesh Interface.

- Last Week in AWS - https://www.lastweekinaws.com/
- Corey Quinn - https://twitter.com/QuinnyPig
- Rancher Rio - https://github.com/rancher/rio
- Helm - https://cloudblogs.microsoft.com/opensource/2019/05/21/helm-3-release/
Spotify Deleting Production K8S Cluster - https://kccnceu19.sched.com/event/MQbb/keynote-how-spotify-accidentally-deleted-all-its-kube-clusters-with-no-user-impact-david-xia-infrastructure-engineer-spotify
Service Mesh Interface - https://cloudblogs.microsoft.com/opensource/2019/05/21/service-mesh-interface-smi-release/

Episode Transcription

Welcome back to The Byte. In this episode we're going to do a KubeCon recap. KubeCon 2019. It's been a heck of a week last week. We were traveling to Austria for some customer visits, and I just completely missed out on recording some episodes. So, I'm back. And I had plenty of time to actually review all of the news coming out of KubeCon. It was an incredible amount of news. Corey Quinn, you know, from last week and AWS was actually attending KubeCon, which is great 'cause he's a really great person to listen to and understand his viewpoint on the technologies, and he really is critical on the technologies, and he several times said, "Everyone's trying to manage Kubernetes. Everyone's trying to roll their own version of it," et cetera. So he has a very interesting viewpoint on the Kubernetes world and how that's going...

But the announcements out of KubeCon were... They're slowly becoming more standard releases. There are no shockers coming out anymore because the community is maturing. That's really the key message here, is the community is still growing rapidly, unbelievably, but we're starting to see, not even a plateau, but we're starting to get to a top of the curve where we're... The height cycle is not completely there yet but you can definitely see that the technology is mature enough that people are using it, and it's becoming more stable.

Now some key announcements out from KubeCon... Actually, before KubeCon actually launched was BitNami, the provider of all the Docker images, and packaged software was purchased by VMware shortly before the conference. And that's pretty big news because BitNami's a nice service. I use them for several different projects, because they package together, for example, Wordpress. They document the heck out of it. They tell you all the ins and outs, where you should look, how you should operate it, how you should grow it. If you want to do high availability... And they maintain it, which is really awesome. It's a great, great product. I recommend if you ever look for an image, check out BitNami Images, 'cause they're very well documented as I said, and they're battle-tested.

Another announcement out of KubeCon was Rancher launched Rio. So Rancher launched not so long ago k3s, which is the slimmed-down version of Kubernetes. Now, on top of that, they've launched Rio, which is a micro-platforms and service, based on top of k3s. So it's a micro-platform as a service, and the idea is to get closer to the edge and start bringing more services to the edge. I think it's a brilliant model, and Rancher continually surprises us on the features they keep announcing.

But the real shocker here was Microsoft. Microsoft is the one that had all the announcements. First, they announced visual studio code Kubernetes 8... a new Kubernetes extension which is now all supported. It's actually a certified extension, and this is quite big. Visual studio code is now becoming the standard, and now they're really throwing their weight behind creating the toolset to actually support it.

The next thing out of Microsoft camp was the virtual cubelet, Hit 1.0. That's basically a server-less Kubernetes distribution. It allows you to run Azure container instances and bringing server-less as a complete package offering within Kubernetes. I find this quite cool. I haven't played with it yet, so it's new to me as well. I'll have to dig into it a little bit more.

Helm 3. Helm is obviously the de facto standard for packaging and deploying Kubernetes applications, and Microsoft announced the first alpha of Helm 3. Also a big announcement because they're throwing all their resources behind making RBAC and CRDs, they're making it part of Helm 3, which is what everyone's been asking for.

Now what really shook everybody, the biggest announcement from the conference, was Service Mesh Interface, and what that is, is it's a new standard interface across all service meshes. So Istio, Envoy, it doesn't matter the service mesh, but we're going to have a standardized interface to all these. And why is that important? Because all these service meshes popped up so quickly and grew so quickly, they developed their own ecosystems, their own APIs. Now if you're a company that runs several service meshes, you also have to integrate all these different APIs. Well this new service mesh allows us to actually... The Service Mesh Interface really standardizes across all service meshes and allow us to define traffic polices, traffic telemetry, traffic management, across all these different service meshes. If you look at it, it's like standardizing the gas nozzle in cars. Every car has a gas nozzle, but if they're all different, it's a bit difficult to operate. You can obviously operate it, no problem individually, but it'd be much easier if they're all standardized. And that's what SMI is doing. It's really standardizing the Istio interface to make it easier to operate and manage.

That's really the KubeCon recap. There was some great news coming out of the ecosystem. The ecosystem's continuing to grow. One of my favorites so far, what I've seen so far is my favorite talk from KubeCon is the Spotify talk, where Spotify talked about on the keynote how they accidentally deleted all its cube clusters in production. Yeah. 

They deleted... What'd they say, about 50 nodes or something like that? 50-node cluster, and zero user impact. They were actually trying to do a migration from one of their other cloud providers to Google Compute Engine, and it just went horribly wrong. They deleted the cluster, then it walked you through how they recovered, how long it took to recover, how a lot of scripts weren't ready and things were not in place to actually do this recovery, and how they went for it and is learning. And this just shows you the Spotify culture is all about learning and embracing failure, and I think more companies can learn from this as well. It was a brilliant talk, it's really nice to hear somebody, "Hey, we're not doing everything right. We do fail once in a while, and this is how we did it and this is what we learned."
It brings us back to the Kubernetes fail stories. The same type of situation, but they're actually talking about KubeCon, and I thought that was quite nice.

That's all we have for this episode of The Byte. Look forward to some more episodes coming up. I have a whole queue in my queue to get up and running, so bear with us, have a great day, and we'll see you next episode.
 
Brian Christner