GKE 2

Question 1

What is GKE?

Answer 1

GKE is a managed Kubernetes solution offered by GCP.

Question 2

At a high level, what is the billing structure difference between GKE Standard and GKE Autopilot?

Answer 2

GKE Standard charges for all resources on nodes, regardless of Pod requests.

GKE Autopilot charges for only the resources Pods request. In most situations, you only pay for the CPU, memory, and storage that your workloads request while running on GKE Autopilot. You aren’t billed for unused capacity on your nodes, because GKE manages the nodes. Exceptions to this model apply when you run Pods on specific compute classes that let Pods use the full resource capacity of the node virtual machine (VM).

Question 3

When using GKE, are you responsible for managing the Control Plane and all of the system components that make up the Control Plane?

Answer 3

No. You are not responsible for managing the Control Plane or any of the system components that make up the Control Plane.

Question 4

Is it possible to convert from GKE Standard to GKE Autopilot after creation?

Answer 4

No

Question 5

What is “bin-packing”?

Answer 5

Bin-packing is the practice of placing as many Pods as possible onto each Kubernetes node to avoid wasting capacity. Because of the way GKE Autopilot is billed, this practice is not necessary. However, when using GKE Standard, this is a common practice.

Question 6

resources:, which specifies the amount of computing resources that are available to each Pod.

If the load on your app surpasses the configured compute resources, the horizontal autoscaling Service automatically replicates Pods to handle the increased load. Autopilot mode enables this autoscaling Service by default.

Question 7

Is GKE Autopilot regional or zonal?

Answer 7

Only regional. However, GKE Standard can be regional or zonal.

Question 8

From the following:
Zonal clusters
Multi-zonal clusters
Regional clusters

Which one has multiple Control Plane replicas?

Answer 8

Regional clusters are the only ones that support Control Plane replicas across multiple zones.

Question 9

What is a Node Pool?

Answer 9

A Node Pool is a group of nodes within a cluster that have the same configuration. Every cluster has at least 1 Node Pool. In fact, when a cluster is created a “default” Node Pool is created.

Question 10

Google recommends that you use Google Cloud’s Managed Service for Prometheus to monitor your Kubernetes applications and infrastructure

Question 11

Which OS is used for GKE Autopilot nodes? What about GKE Standard?

Answer 11

GKE Autopilot nodes use Container-optimizes OS. When using GKE Standard, you decide which OS your nodes use.

Question 12

Which Kubernetes component is responsible for running containers in a Pod?

Answer 12

Kubelet

Question 13

you are charged for node-to-node traffic across zones. For example, if a workload running in one zone needs to communicate with a workload in a different zone, the cross-zone traffic incurs cost.

Question 14

When you create an Autopilot cluster with public networking, workloads in the cluster can communicate with each other and with the internet.

Question 15

In Kubernetes, does every Pod get its own unique cluster-wide IP address?

Answer 15

Yes. This allows a Pod to communicate with any other Pod in any other node within the cluster.

Question 16

What makes it possible for containers in a Pod to communicate with each other?

Answer 16

Containers in a Pod share the same IP address. This makes it possible for those containers to communicate with each other by using localhost with different ports.

For example:
ContainerA localhost:8080
ContainerB localhost:8081

Question 17

At a high level, what are the 4 different network communication possibilities that Kubernetes provides?

Answer 17

1. Container-to-container communication within a Pod
2. Pod-to-pod communication
3. Pod-to-service communication
4. External-to-service communication

Question 18

Pods, Nodes, and Services each have their own IP address.

Question 19

What is a Service and what is it used for?

Answer 19

A Service is an Object that is used to expose an application running as one or more pods in a cluster. Clients can connect to the Service instead of trying to connect to one of the various Pods. Remember, the IP address of Pods are ephemeral. A Service abstracts access to Pods.

Question 20

If you need to automatically scale the number of Pods in your cluster, what can you do?

Answer 20

You can use Horizontal Pod Autoscaling, which can scale Pods based on the built-in CPU and memory metrics, or custom metrics from Cloud Monitoring.

Question 21

If you need to deploy some Pods to a specific node because that node has special hardware, what can you do?

Answer 21

You can give the node a label and use it to specify where the Pods should be deployed. This is known as workload separation.

Question 22

Use Pod affinity and anti-affinity to co-locate Pods on a single node or to make some Pods avoid other Pods. For example, you could tell GKE to avoid scheduling frontend Pods alongside other frontend Pods on the same nodes to improve availability in case of an outage.