coreos-kubernetes icon indicating copy to clipboard operation
coreos-kubernetes copied to clipboard

kube-aws: document kube-aws users

Open philips opened this issue 9 years ago • 13 comments

We should start documenting kube-aws users. If you are a user please comment with your organization name, email (optional), how many machines, and use case.

Thank You!

philips avatar Aug 31 '16 18:08 philips

@tombenner would you mind if we added Entelo to the list?

philips avatar Aug 31 '16 18:08 philips

@thuey did you end up using kube-aws?

philips avatar Aug 31 '16 18:08 philips

\cc @pieterlange @whereisaaron @mumoshu @iameli

colhom avatar Aug 31 '16 19:08 colhom

also, @cgag. I hear you run some stuff on kube-aws as well?

colhom avatar Aug 31 '16 19:08 colhom

@philips We'd be happy to have Entelo on this list; thanks!

tombenner avatar Aug 31 '16 20:08 tombenner

Company: Stream Kitchen Email: [email protected] Machines count: Varies wildly depending on what's being processed that day Use case:

  • kube-aws provides a robust, scalable foundation for compositing live video streams in the cloud
  • utilizes spot instances to dynamically shift processing onto the cheapest hardware currently available
  • easy CloudFormation bootstrapping allows us to come up in new regions without fuss

That work?

(Ping @meonkeys to talk about C-SATS)

iameli avatar Sep 01 '16 18:09 iameli

@iameli that's perfect, thank you.

colhom avatar Sep 01 '16 18:09 colhom

Some more people that might have things to say: @rwillard, @four43, @danielfm.

cgag avatar Sep 01 '16 19:09 cgag

  • Company: C-SATS
  • Email: [email protected] or [email protected]
  • Machines count: ~4 nodes, ~100 containers
  • Use case: we originally chose Kubernetes because we needed to manage and be able to scale out Meteor apps. Galaxy did not yet exist. Kubernetes also appeared to have more (and more useful) features than Amazon ECS. We also require HIPAA-compliant storage (e.g. at-rest strong encryption), which Galaxy also does/did not offer. Today we use Kubernetes for running all our cattle and pets. We first tried Kubernetes on Ubuntu server EC2 instances, but eventually found kube-aws way easier and more stable. @iameli set up most of our cluster infrastructure, and I recall this was prompted in part by meeting with CoreOS/Tectonic folks at a Kubecon.

I hope this helps! -Adam

meonkeys avatar Sep 01 '16 21:09 meonkeys

@adamfeuer are you using kube-aws?

meonkeys avatar Sep 01 '16 21:09 meonkeys

I am at a new company, but I've put everything on Kubernetes here. We're on GCP though, so I'm using GKE. I can answer any questions about my past kube-aws setup if you'd like.

rwillard avatar Sep 01 '16 23:09 rwillard

@meonkeys I am at a new company, we're not using Kubernetes though. I can comment on my past kube-aws usage; however it was a few months ago, and kube-aws has moved on quite a bit since then.

adamfeuer avatar Sep 02 '16 04:09 adamfeuer

@cgag

We were using Elastic Beanstalk for all our applications, but it was very hard to fully utilize the available compute resources since each Beanstalk instance runs one application only. We considered using ECS as a way to run many containers per instance, but ECS still isn't available in our main AWS region (sa-east-1), so it wasn't really an option.

When first testing Kubernetes as a possible solution for that problem, I played around with kube-up but didn't quite like the way kube-up worked. After looking around a bit, I found kube-aws and although it missed a few features I wanted (i.e. cluster level logging), I really dig the fact it uses CloudFormation under the hoods. It really made pretty easy for me to change the things I wanted.

So after a few months I managed to migrate all our staging environments to Kubernetes and last week I moved the first project (a NodeJS front-end application) to production. :tada:

I also put together a AWS Lambda function (nothing fancy, just a couple hundred lines of JS code) to automate the deploy of our applications to the Kubernetes cluster via kubectl based on GitHub and CircleCI notifications.

One particularly nice component of our deployment pipeline is what we call here development environments; every time someone submits a pull request, this Lambda function creates a temporary deployment in Kubernetes so that other developers can see the change "live" and give a more accurate review on the changes being made. Then, when the pull request is merged/closed, the deployment is automatically deleted.

danielfm avatar Sep 02 '16 13:09 danielfm