Kubernetes and Helm Deployments Session One

Kubernetes and Helm Deployments Session One

Helm

• Helm is a package manager
• To learn helm, the first things you should start learning is kubernetes resources
• Helm chart supports almost all of the kubernetes resources, which is impressive

Flavors of Kubernetes

1. EKS on amazon
2. on-prem
3. Google cloud

Helm Chart: Mapping information

• Ingress Controller: defines what network is enabled and describes how to send traffic from external world to the current cluster
• Rest of the charts: what API and features are enabled, you can even enable beta features
• You can deploy all of blueapron.com in one single package

Helm common charts

• templated, abstracted common logic
• shared helm charts
• all our helm charts rely on this chart

Defaults

• Memory
• CPU
• Instance - How many replicas do we need to deploy to a cluster
• Hostname

Overrides

• Works like CSS, by convention over configuration

Where does docker fitting into all of this?

Kubernetes doesn’t require docker, but it’s flexible enough to support the docker environment

What is a deployment

1. creates replica-sets
2. Kubernetes provides stateful-sets. You can ask kubernetes not terminate the previous replica-set. As a best practice here, it’s always better to deploy the persistentance layer elsewhere such as RDS.
3. Ingress, to route the external traffic to it - between pod and deployment replica-sets there is this layer of abstraction

DNS components

• DNS in AWS, Kubernetes, and in console in legacy
• Currently uses IP Table, needs to be modernized

What do you need to understand as a non-DevOps Application Developer?

Developers need to understand

1. Docker
2. Pod - A group of containers that share resources. They will be deployed together, and allow them to communicate

How do we start a new application?

External example

Internal example

Use wms-server was a starter template

1. First build a Dockerfile - pro-tip: optimize until you reduce the container to a smaller size
2. Create a Jenkinsfile
   • Jenkins file is written in a combination of Groovy and Jenkins language. Blue Apron provided an superset of features, and you can find the documentation of this superset here
3. Helm See the helm folder in each project
   • ingress you almost never have to touch
   • replace the application name in the Chart.yml and in each process yaml. The web process yaml is the Rails application.
4. Knobs Q: How do we increase the number of replicas, memory, and CPU? A: You would reference the values inside a cluster. CPU is unlimited right now

Example of a cluster:

us-east-1/blue

Upcoming Features

1. removal of Jenkins Legacy EKS references
2. improved syntax sugar
3. IP Table modernization

Possibly upcoming features

• moving linkerd to istio, and into its own cluster

Covered in This Class

1. High-level overview of Helm as it related to Docker and Kubernetes
2. A case study of a Rails application wms-server, and how kafka, sidekiq, and rails process are deployed using helm in this application
3. Some knobs on how to tune the process memory and instance

Next Class

1. Further explain the health check and how they’re used
2. Perhaps go over two other flavors of deployment, one ember and one react
3. Develop and deploy jenkins-ci-library-test together
4. See how the staging and production process works in the Jenkins web UI