Chapter 6: Container Orchestration

Question 1

What questions/concerns do container orchestration tools address? Name 5 tasks

Answer 1

• How to group multiple host together to form a cluster and manage them as a single compute unit?
• How to schedule containers to run on specific hosts?
• How can containers running on one host communicate with containers running on other hosts?
• How to **provide **containers with dependent storage when it is scheduled on a specific host?
• How to access a container through a server name instead of accessing it directly through their IP addresses?

Question 2

What is container orchestration? What does it encompass?

Answer 2

• umbrella concept, encompassing container scheduling and cluster management

Question 3

How is container orchestration achieved?

Answer 3

• through policy-driven mechanism
• that automates the decision process that distributes containers across the nodes of the cluster
• this decision process is aided by cluster management aspects like state of the exisitng workload and cluster node resource availability

Question 4

What is enabled through the cluster management orchestrators?

Answer 4

• management of resources of cluster nodes
• as well as adding and deletion of node through cluster scaling mechanism

Question 5

What is Kubernetes?

Answer 5

• Apache 2.0 licensed open source project for automating deployment, operations and scaling of containerized applications
• started in 2014 by Google, IP transferred 2015 to Cloud Native Computing Foundation (CNCF), non-profit organization

Question 6

What container runtimes are supported by Kubernetes?

Answer 6

• CRI-O
• containerd
• Docker Engine
• Miratins Container Runtime
• may change with matury of Kubernetes project

Question 7

What are the main components of the kubernetes architecture?

Answer 7

• cloud-control-manager (optional)
• kube-api-server
• scheduler
• etcd (persistent store)
• Controller Manager
• kubelet
• kube-proxy
• CRI
• pod(s)

Question 8

How is a Cluster made up?

Answer 8

• Control Plane containing cloud-control-manager, kube-api-server, Controller Manager, scheduler and etcd
• nodes each containing kubelet, kube-proxy and CRI, which contains pods
• from each node kubelet and kube-proxy communicate with kube-api-server
• cloud control manager, controller manager, scheduler and etcd communicate with kube-api-server
• cloud control manager communicates with the outside Cloud provider API

Question 9

What are the key components/elements of kubernetes?

Answer 9

• Cluster
• Control-Plane-Node
• Worker-Node
• Namespace

Question 10

What is a Cluster?

Answer 10

• collection of systems (bare-metal or virtual) and other infrastructure resources used by Kubernetes to run containerized applications

Question 11

What is a Control-Plane Node? What does it do?

Answer 11

• a system that:
• takes containerized workload scheduling decisions
• manages worker nodes
• enforces access control policies
• reconciles changes in the state of the cluster
• delegates container management tasks to worker node agens

Question 12

What are the main components of a Control-Plane Node?

Answer 12

• kube-api server
• etcd
• kube controller manager
• cloud controller manager
• kube-scheduler

Question 13

How many Control Plane nodes are usually found within a container?

Answer 13

• multiple can be found as a solution for High Availability

Question 14

What are Worker Nodes?

Answer 14

• system where containers are schedules to run workload management units called pods

Question 15

How do Worker nodes function?

Answer 15

• runs a kubelet named daemon responsible for intercepting instructions related to container deployment and lifecycle management
• these instructions are comming from kube-apiserver
• the node delegates such tasks to the container runtime, found on the node
• implements container health checks, enforces resource utilization limits and reports node status back to the kube-apiserver

Question 16

What function does the kube-proxy of a Worker node have?

Answer 16

• is a network proxy
• enables applications running in the cluster to be accessible by external requests

Question 17

Where can kubelets and kube-proxies as well as container runtimes be found?

Answer 17

• on worker nodes
• on control-plane nodes

Question 18

What does a namespace enable?

Answer 18

• allows logical partition of the cluster into virtual sub-clusters by segregating the cluster’s resources
• addressing the multi-tenancy requirements of enterprises requiring an isolation method for their projects, applications, users, and teams

Question 19

What are the key API resources of the Kubernetes architecture?

Answer 19

• Pod
• ReplicaSet
• Deployment
• DaemonSet
• Service
• Label
• Selector
• Volume

Question 20

What is a pod?

Answer 20

• smalles deployment unit in Kubernetes
• logical workload management unit
• enabling co-location of a group of containers with shared dependencies such as storage volumes
• often managing a single container and its dependencies such as Secrets or ConfigMaps
• can be created independently
• but lacks than self-healing, scaling and seamless updates capabilities
• usually managed by controller programs or operators such as ReplicaSet, Deployment, DaemonSet or StatefulSet

Question 21

What is ReplicaSet API resource?

Answer 21

• mid-level controller or operator
• manages the lifecycle of pods
• rolls out desired amount of pod replicas
• uses state reconcilliation to ensure that desired number of application pods is running at all times
• enables self-heal if an application pod is lost unexpectedly

Question 22

What is the Deployment API resource?

Answer 22

• top-level controller
• allows declarative updates for pods and ReplicaSets
• can be defined to create new resources or to replace existing ones with new ones
• represents the default stateless application rollout mechanism

Question 23

What is the DaemonSet API resource?

Answer 23

• controller or operator
• that manages the lifecycle of node agent pods
• rolls out a desired amount of pod replicas while ensuring that each cluster node will run exactly one application pod replica
• uses reconciliation to ensure that the desired number of application pod replicas is running at all times
• self-heal the application if a pod replica is unexpectedly lost due

Question 24

What are typical Deployment use cases?

Answer 24

• Create a Deployment to roll out a desired amount of pods with a ReplicaSet
• Check the status of a deployment to see if the rollout was successful or not
• later update the deployment to recreate the pods (use a new image) - through the Rolling Update Mechanism
• roll back to an earliert Deployment revision if the current Deployment isnt stable
• scale, pause and resume a deployment

Question 25

What is the Service API resource?

Answer 25

- traffic routing unit - implemented by kube-proxy - provides a load-balancing access interface to a logical grouping of pods, typically managed by the same operator - enables applicatons with a DNS name registrations, name resolution to a private/cluster internal static IP - can reference a single pod or set of pods managed by ReplicaSet, Deployments, DaemonSets or StatefulSets

Question 26

What is the Label API resource?

Answer 26

- arbitrary key-value pair attached to resources - typically used to tag resources of a particular application, such as pods or deployments to logically group them for management purposes - for updates, scaling, or traffic routing

Question 27

What is the Selector API resource?

Answer 27

- allows controllers or operators to seach for resources or groups of resources described by a desired key-value pair labels

Question 28

What is the Volume API resource?

Answer 28

- abstraction layer implemented through Kubernetes plugins and third-party driver - aimed to provide simplified and flexible method of container storage management in Kubernetes - Volume containers are able to mount local host storage, network storage, distributed storage clusters, cloud storage services seamlessly

Question 29

What options do exist for Kubernetes in public clouds?

Answer 29

- clusters can be managed by yourself - hosted (managed) Kubernetes services, where management tasks are performed by cloud services providers

Question 30

What are examples for hosted solutions available for Kubernetes?

Answer 30

- Amazon Elastic Kubernetes Services - Azure Kubernetes Services (AKS) - Google Kubernetes Engine - IBM Cloud Kubernetes Services - Red Hat OpenShift

Question 31

What do Kubernetes Platform solutions offer?

Answer 31

- a managed Kubernetes environment flexible enough, to be deployed on any public cloud, multi-cloud or on-premises/private cloud - eliminating vendor lock-in

Question 32

What are examples for Kubernetes Platforms?

Answer 32

- Managed Kubernetes - Kubermatic Kubernetes Platform - Platform9 Managed Kubernetes (PMK) - Ranchers Kubernetes Engine (RKE)