Google Kubernetes Engine

Question 1

Microservices

Answer 1

This is an architecture where an application is divided into a number of small tasks instead of lumped onto a single node, allowing for greater access and easier scalability.

Question 2

How do Microservices communicate?

Answer 2

They used APIs, which allows them to be written in different programming languages. You can also use CI/CD to deliver faster funcationalities.

Question 3

CI/CD

Answer 3

Continuous Integration and Continuous Delivery/Deployment. A methodology of streamlining software delivery.

Question 4

Continuous Integration

Answer 4

Developers frequently submit code to a common repository where it is reviewed and tested. Once validated an automatic build is triggered.

Question 5

Continuous Delivery

Answer 5

Allows for the automation of the release process so that software can be deployed to the target enviornment at any time

Question 6

Continuous Deployment

Answer 6

Changes made by a developer that pass all tests are automatically deployed to the production enviornment

Question 7

What are the advantages of containers?

Answer 7

They provide isolation between servies allowing for them to each have their own unconflicting libraries, the application’s use can be limited to each resource, they are self contained with all dependancies and don’t need an OS, and they are lightweight.

Question 8

What types of containers does GKE support?

Answer 8

GKE primarily supports Docker. This platform creates Dockerfiles that can then be hosted in Google Container Registry and accessed from GKE clusters.

Question 9

Kubernetes

Answer 9

Also known as K8s, an open source container orchestrator that was initially developed by Google that allows for the deployment, scaling, and management of containerized applications (both stateless and stateful).

Question 10

Master Nodes

Answer 10

Responsible for the management of the cluster. Takes care of the desired state using YAML files on worker nodes. Runs API server, controller manager, etcd, and scheduler.

Question 11

Worker Nodes

Answer 11

Host the workloads which are pods. Managed by the master node and runs kublet, kube proxy, and container runtime (aka Dockerfiles)

Question 12

Pods

Answer 12

The smallest unit and can contain one or more containers. Each one has a unique IP address that is shared between the contents. They are created when called upon (ephemeral).

Question 13

Kubernetes Objects

Answer 13

Records of intent defined in YAML files that are declarative and determine the state of the pods, ReplicaSets, Replication controllers, deployments, and namespaces.

Question 14

What data do Kubernetes objects need?

Answer 14

apiVersion (the version of the Kubernetes API being used), kind (the kind of object to be created), metadata (data that helps uniquely identify the object, such as name), spec (the specification of the object, which is dependant on type)

Question 15

ReplicaSets

Answer 15

Used to manage the number of pods that are running at any given time by monitoring how many pods are running and deploying new ones as needed. Defined under the replicas parameter.

Question 16

Deployments

Answer 16

Used to deploy, update, and control pods, creating ReplicaSets without the need to define them seperately. Support both Canary and Blue/Green deployment.

Question 17

Canary Deployment

Answer 17

Deploy a new app version to a subset of users and when that version is working properly, deploy to all.

Question 18

Blue/Green Deployment

Answer 18

Use two enviornments with only one active at a time. After updating inactive one to new version and testing traffic is switched to make it the active one.

Question 19

NameSpaces

Answer 19

Virtual clusters with a Kubernetes cluster, used to allow the recycling of the names of resources. Names only have to be unique within the namespace but not within the cluster. They can be isolated from each other using network policies.

Question 20

Services

Answer 20

Used to group Pods into a single endpoint. Has a stable IP address, so requests can be sent to it and then forwarded to a pod. Types include Cluster IP, NodePort, Load Balancer, ExternalName and Ingress.

Question 21

ClusterIP Service

Answer 21

Default service that uses internal ClusterIP to expose Pods, meaning services aren’t available outside of cluster. Used for internal communication between microservices within a cluster.

Question 22

NodePort Service

Answer 22

Exposes each node outside of the cluster. The pods can then be accessed using :. If there are multiple IP addresses with the same port, they will all be exposed.

Question 23

LoadBalancer Service

Answer 23

Dynamically creates a provider load balancer. In the case of GCP, a network load balancer is created for you. It does use objects outside of Kubernetes and generates additional costs.

Question 24

ExternalName Serivce

Answer 24

Service is exposed using a DNS name specified in the External Name spec.

Question 25

Ingress Service

Answer 25

Object that allows the routing of HTTP(S) traffic according to defined rules/paths. Can be associated with one or more service objects.

Question 26

How many master nodes are deployed in a zonal GKE Cluster?

Answer 26

One

Question 27

How many master nodes are deployed in a regional GKE Cluster?

Answer 27

Three. Recommended for upgrading the management plane to experince zero downtime.

Question 28

Node Pools

Answer 28

Groups worker nodes with the same configuration. Multiple node pools are used for groups with specific characteristics such as local SSDs, minimum CPU, a specific node image, or a preemptible instance.

Question 29

Node Taints

Answer 29

Schedule your workload on a particular node pool

Question 30

Container-Optimized OS

Answer 30

Default OS that your Kubernetes engine uses

Question 31

Google’s Container Optimized OS

Answer 31

a locked down version of Chromium OS which runs containerized operations. Images are maintained and updated by Google. Not good for running non containerized applications or when you need enterprise support from Linux.

Question 32

GKE Storage Suggestions

Answer 32

Persistent storage in GKE is easiest when using GCP services such as Cloud SQL, Datastore, or Cloud Storage. Expose storage directly to your pods using volumes that are backed by ephemeral or durable storage.

Question 33

Cloud Build

Answer 33

A Cloud Build was the original container builder that was intially used with the intention of turning it into a CI/CD service. It runs one or more build steps to produce artifacts, using steps defined by YAML configuration profiles usually triggered by a code repository.