Adam Monsen email@example.com @meonkeys
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.
Daniel Martins firstname.lastname@example.org
Machines count: ~4 nodes, ~50 containers
Use Case: 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. 🎉
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.
Tom Benner @tombenner
Use case: Data pipeline scaling system designed for use by engineering team. Custom built deployment tools are built on top of Kubernetes by the Entelo team.
Eli Mallon email@example.com @iameli
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.