Core Concepts

Question 1

What makes a declaritive approach different from an imperitive? Can you relate to a comparison between a yaml-definitionfile and running a command?

Answer 1

Declaritive is the desired outcome, white imperitive is focused on steps resulting in micromanagement. The definition file is seen as a declaritive approach while running commands is more imperative.

Question 2

Why should we use Kubernetes?

Answer 2

Kubernetes gives us a framework with standardised plugins for orchestrating containers. Without it we’d result to custom applications and scripts, widely different between organisations.

Question 3

In ETCD, how are the –advertise-client-urls and —-internal-cluster used?

Answer 3

–advertise-client-urls is the IP that ETCD is listening on while –internal-cluster is used for grouping cluster members in a HA setup.

Question 4

Is there something ETCD does not store?

Answer 4

No, all information about any objects in the Kubernetes clsuter is stored in the ETCD database.

Question 5

Besides being the center of attention, how does the kube-apiserver differ compared to other components?

Answer 5

The kube-apiserver also manages authentication and is in contact with ETCD.

Question 6

kube-controller-manager monitors and reacts to changes to comply with the monitoring. Describe the manger component for the nodes.

Answer 6

Nodes are monitored with a 5 sec heartbeat and have a grace period of 40 sec. Should there have been 5 min then the node is removed and any containers will be recreated on other nodes.

Question 7

How does the kube-scheduler decide which containers get placed where by the kubelet?

Answer 7

First phace filters with for eg. hardware and second looks at a 0-10 score after the container is placed on the node. The higher score decided which node gets the container which means there were more resources available after.

Question 8

Whats the difference when comparing the scheduler and kubelet components?

Answer 8

The scheduler instructs the kubelet where to load/unload containers.

Question 9

Besides aiding in communication, what is the role of kube-proxy?

Answer 9

To search for services and on each node add rules with for eg. iptables on how to reach them.

Question 10

How would you describe a POD?

Answer 10

A pod is the smallest object in Kubernetes and typically form a 1to1 relationship between pods and containers. When scaling the application, add more pods.

Question 11

Can you describe the most basic top/root-level properties below?

apiVersion:
kind:
metadata:
spec:

Answer 11

apiVersion: depends on the kind kused

kind: type of object
metadata: data about object
spec: object type configuration

Question 12

How do some kubernetes master/woker components differ when distributed with kubeadm and service?

Answer 12

kubeadm creates pods for the components with the configuration located under /etc/kubernetes/manifests/ where the service points to /etc/systemd/system/

Question 13

Both ReplicationController and ReplicaSet are used to conform and scale pods, but what does RS bring to the table?

Answer 13

RS adds the key selector: under spec: which filters pods outside of the RS. This means the RS can include pods not created by the RS itself.

Question 14

When creating a ReplicaSet the version needs to be what?

Answer 14

apps/v1

Question 15

When using a ReplicaSet with selector: key, whats critical in the creation?

Answer 15

Matching must be equal to the labels under template: key.

Question 16

Once a ReplicaSet has been updated, what must be done to any PODs within it?

Answer 16

Delete so that its recreated with the updated RS settings.

Question 17

By default you’re provided with the namespaces default, kube-system and kube-public? What’s the deal with namespaces in general?

Answer 17

They provide isolation of environments, for eg. production and development.

Question 18

Lets say you want to reach a service within your own local namespce, call it db-service. You’d simply use db-service. Now if its within another namespace like dev you’d resort to db-service.dev.svc.cluster.local. Can you elaborate on the different sections after the db-service?

Answer 18

db-service.dev.svc.cluster.local. dev is the environment, svc is for services and cluster.local is the cluster domain.

Question 19

Lets say you have three pods in the namespace dev and you accidently delete the namespace. Wha would be the outcome?

Answer 19

The namespace and all its objects would be deleted.

Question 20

Since a service object is spread across all nodes and forwards traffic to a group of pods. How does it know which pod or group of pods?

Answer 20

Just like the RS, services uses the selector: field to match labels.

Question 21

In a service NodePort; how are the NodePort, Port and TargetPorts used?

Answer 21

NodePort is what’s going to be listened to on the node itself. Port which is mandatory is what the service is actually listening to and TargetPort is what the pod is listening to.

Question 22

When would you want to use a service NodePort?

Answer 22

When you want to expose a service object within the cluster but outside of the cluster.

Question 23

Whats the deal with the service defintion using a ClusterIP?

Answer 23

ClusterIP is used to expose service objects within the cluster, representing a group of pods.