In this Episode, Shannon, co-founder of Rancher Labs, and Rahul discuss
Kubernetes, a container orchestration tool,
and
Rancher, a multi-cluster container deployment tool, in detail.

They discuss the challenges faced in setting up Kubernetes
clusters for production and how rancher(2.0) helps easing the Kubernetes deployments.

Further, they discuss some of the challenges like Deploying stateful apps
on container platforms and how we can solve them.
They also talk about serverless and how container ecosystem
is an easy platform to develop serverless applications.
Shannon thinks that equipping yourself with container tools is a need
of an hour as most of the enterprise companies have started migrating to container deployments.

Key Points From This Episode

• Introduction to Rancher 2.0.


• Key Features added in Rancher 2.0 for kubernetes.


• Storage and Networking with Rancher and Kubernetes.


• Running stateful applications on kubernetes.


• Serverless on top of container ecosystem.

Links Mentioned in This Episode

• BigBinary on Twitter


• Shannon on Twitter


• Rahul on Twitter


• Kubernetes


• Rancher


• Serverless


• CRI-O


• Rkt


• AWS Lambda

Transcript

[0:00:00 ] RAHUL: Hello and welcome to BigBinary’s podcast; All Things Devops and today we have Shannon Williams, the co-founder and VP of sales from Rancher Technologies and I would like to welcome you on behalf of BigBinary.

Hi Shannon, thanks for taking out time from your busy schedule and we really love to hear from your Rancher kubernetes and all the ecosystems. So, Shannon, do you mind introducing yourself?
[0:00:30 ] SHANNON: No, not at all Rahul. Thanks a lot for having me. It’s great to chat, I’m a fan of your podcast. So we started Rancher in 2014 so about three years ago now, and, at its core Rancher is an open-source software company. We create open source projects and develop, really, to help people manage and put containers into production. Four of us started the company after working for about 6 years prior to this on building infrastructure as a service in cloud computing.

We really set out with the goal of answering some questions about how people were going to really run applications in a world where infrastructure was plentiful and available but still quite different, whether you’re using all premise resources or cloud providers, you have this very plentiful resources, but there wasn’t a lot of standardization and so we were really excited about the potential of Docker, because it really opened up the idea that maybe we could get to a point where the design and development and even operations of containers or applications could be standard, even if you’re running on wildly different infrastructure. And so, that’s really driven us for a long time. Our primary project that’s gotten so popular is called Rancher, and Rancher is an opensource container-management platform that really makes it incredibly easy to build and deploy container clusters across any cloud, any infrastructure anywhere.

We’re a relatively mature team, so anyone who’s looking to use Kubernetes, use Docker they tend to find Rancher to be the layer that ties it all together. You know, Dockers has the runtime and the daemon at the base layer, I think Kubernetes is just the kernel of orchestration, so Rancher’s kind of like an operating system, you know, it sort of, it puts everything together, it simplifies how you sort of, compose the application and manage the storage and deal with the networking and control user access and just, everything, to make it work. So that’s a little bit about us and what we do.
[0:02 31] RAHUL: Sounds really interesting. Rancher and your job name and brand in general and the company and how you’re product Rancher which is really evolving as an best orchestration tool for different orchestrators like Kubernetes, Cattle, Docker-Swarm, or Mesos. So we have been using Rancher and it has really helped us solving some real challenges.

But I’ve got an email with a couple of weeks back you released Rancher 2.0, even though I haven’t got really, a chance to try it out, but as we are Kubernetes shop and we try to build everything on Kubernetes, I would like to understand like, what is a breaking feature in Rancher 2.0 because, as a tag-line says, run Rancher on your Kubernetes so would you just tell us, in short, about that?

[0:03:20 ] SHANNON: Yeah, sure. So 2.0 is a big release for us. We released 1.0 sort of, towards the beginning of 2016, so about 18 months ago, and as you said, the defining feature of 1.0 was a real powerful, sort of multi-cluster management layer that would you know, kind of orchestrate your ops on multiple different orchestrators. So we deployed an operated upgrade of Kubernetes or Docker Swarm, or Mesos.

We had our own orchestrator called Cattle that was based on Docker Compose and so we had these different orchestrators, we could deploy them, we could operate them and run them and then manage sort of centralised user administration and logging and policy and things like that, across them. Over time though, what we found is two orchestrators were dramatically more popular than the rest, so despite supporting Swarm and Mesos, most of our users weren’t using those, they were using Kubernetes and they were using Cattle, that we had developed ourselves. And when we looked at it, we were kind of like ‘Ok, so what’s the reason that these are so much more popular?’ and when I say ‘so much’ I mean, you know, 95, 98% of people were using one of those. What it turned out is that Kubernetes had grown and grown in popularity since we introduced it in 1.0 based on the scalability, the maturity and the reliability of Kubernetes, really, is what attracted users to it. And we had always been pretty big fans of Kubernetes, so it made a lot of sense, it was sort of, reflected what we saw as well, that Kubernetes was quite production-grade, production quality.

But the other orchestrator that was really popular was Cattle. And Cattle was interesting. Cattle was sort of our initial orchestrator that we built with Rancher at the very beginning. And we built Cattle, actually only because we could never really get Swarm to stabilise and be sort of, a standard orchestrator for people who liked Docker Compose and liked the Docker framework, so we had a lot of trouble making Swarm really work well with network and storage on any kind of scale and so we decided to take and sort of build out some simple orchestration around Docker Compose, and as we built it, Cattle sort of got better and better and it ended up being, in a lot of ways, pretty similar to Kubernetes. Not as feature-full and not as robust but from an architecture perspective, actually quite similar. It had concepts like health checks, like you know, daemon sides, it had very similar concepts. And so that actually was even more popular than Kubernetes among Rancher users.

Probably about 75% of all the deployments of Rancher were using our Cattle orchestration, and as we looked at it, what people liked about Cattle was the fact that it used Docker Compose and was very much the API, CLI were all Docker based, so you could use all your native Docker commands very easily, and they like the user experience, they liked the, you know, simplicity of being able to really easily understand all the relationship. If you do understand Docker, you can understand Cattle really quickly. So as we look at 2.0, we have to decide what we wanted to do. Do we want to support lots of orchestrators or not? And we decided that from our perspective, Kubernetes had really gotten to the point where both its momentum and its quality and reliability were so high, there was just no good case to use anything else in production. We felt that it was just the best orchestration for production, and what we really wanted to do is we wanted to use Kubernetes as the standard implementation of orchestration across whatever we did, and so everything we were doing on the orchestration side in Cattle was already being done as well in Kubernetes, so we really felt like it was a lot of waste of resources to keep adding the same features or similar features to what was going into Kubernetes. So what we decided to do was standardise on Kubernetes but exceptionally the Cattle user experience. And so, what that meant was you know, taking everything we knew, and everything people liked about Cattle and implementing it around Kubernetes.

And so what Rancher 2.0 is, you know, it’s really 3 things. The first is something we’ve always done, which is a really reliable implementation of a Kubernetes distribution. You turn on Rancher and it automatically builds and deploys a very good upstream deployment of Kubernetes with good networking, and storage drivers ready to go and full implementation of Kubernetes. That’s at the core of Rancher, is that distro component to run Kubernetes really well anywhere, on VMware, on bare metal, on AWS, on any cloud.

But then the other piece is what we consider Kubernetes operation and Kubernetes container operation, which is really towards sort of, the IT department. And this is where you sort of think of Rancher’s sort of, multi-cluster management. That really gets to kind of, ‘Ok, we as an organisation are going to run