CloudNative

especially with a few small teams working on testing software, kubernetes can be hugely beneficial on even a small cluster

Don't need to be near Google scale to benefit from k8s. Underlying capabilities and patterns are useful at any scale, but get exponentially more so with scale.

it isn't just about scale. the reality is that everyone can benefit from this type of operating model, but it is literally the only way to run internet scale workloads. so internet companies have had to run this way.

our mission is to make the model of operating available and accessible to everyone.

the health model for pods is super helpful, who doesn't want auto-healing?

I think once you have users who expect your service to be up all the time you're at that scale. A lot of what we do is make it possible to manage and deploy software reliably at any scale.

Hmm... We developed our next generation element management software @HDSCorp as a #cloudnative app as it allows us to manage ever device in every data center for our enterprise customers

@brendandburns building apps this way allows them to scale based on adoption and not having to reengineer

personally on single small applications kubernetes proves itself from the declarative nature of orchestration alone. The best part is then when scaling that app you are already on the platform that can scale with your application needs.

#kubercon was our event at @oscon. We have done regular contributor events, meetups, etc. This was special for 1.0. #cloudnative foundation will take it forward

Also, stay tuned for upcoming #GCPNext event announcement; seems likely you'll find a few #kubernetes folks there ;)

showing your age with Barney reference

We're not working on it right now, but we'd love to extend the kubernetes (and the on-node kubelet specifically) to be able to use hyper-visor and microkernel based runtimes. PRs welcome!

for now we have been focused on stateful and stateless web app hosting; lots of our customers use the platform for scale out testing, and build their devops pipelines on the platform too.

Next up for #kubernetes will be running scale out batch workloads on the same basic infrastructure as the app hosting.

@cmcluck the microservices architectures haven't focused much on stateful data (to date). do you see that expanding with k8r going forward, or should that be solved at the container layer?

via @jeffsussna - Go directly to your nearest @kelseyhightower preso :)

I would look at the examples in the GKE docs, they have a networking and storage story, there may be similar examples on the kubernetes site.

Yes; this is critical. inside Google we run all our stateful services on the same clusters as everything else. Even today we have pretty solid volumes support for local and network attached storage. I expect more intelligence around state to come.

https://cloud.google... has a pretty gentle "how to" for Node.js developers

It the Google container optimized VM (https://cloud.google...) based on Debian with latest Docker and Kubelet installed

damn, forgot to switch accounts again.

This is kind of my point. There is a difference in relevance and demand.

busted

@cloud_opinion Um... Not buying it. ;)

there's lots of different aspects to the Google Cloud, if all you want a virtual machine running Windows or Linux you can use Google Compute Engine. If all you want is reliable storage, Google Cloud Storage, or Cloud DNS, etc.

I am cloud opinion

@furrier No, I am @cloud_opinion

We are all @cloud_opinion

@stu I'll take that for a dollar.

we all *have* cloud opinions

weird; I just circumnavigated in 8 days for #GCPNext, talked to 10k+ folks, and the demand is real. Use cases range from complex to simple, but value prop stands up across oldschool and #cloudnative

You say science like it's a bad thing :) But in all seriousness, we want to do two things - meet customers where they are today and help them with the transition to cloud native, at their own pace.

We are working hard to change this. A lot of Google technology has gone mainstream (bigtable -> nosql; mapreduce -> hadoop; cgroups -> Docker). We decided to build this one in the open with the community to make it immediately relevant to business.

> I don't know. Know a few people who spend more time in fly over country instead of flying over
@ImAmyO @Schmarzo

@cloud_opinion is your mum

@njoneja science is a bad thing to a lot of people, don't shoot the messenger :) I am an admirer of Google, met some of their team tonight, but it doesn't mean I'm blind or lacking critique

ReplicationControllers can scale any set of pods, but the policy of how you scale (e.g. how quickly) may be something custom that you want to build.

@brendandburns Are we able to run custom ReplicationControllers on GKE and if so are there any examples?

There are not any examples, but it's definitely possible. Just create your application that talks to the Kubernetes API and creates/deletes Pods. Moving toward the future, we would like to enable the API be extensible for custom user-defined types