To deploy Kubernetes, RKE deploys several core components or services in Docker containers on the nodes. Based on the roles of the node, the containers deployed may be different.

All services support additional custom arguments, Docker mount binds and extra environment variables.

Component Services key name in cluster.yml
etcd etcd
kube-apiserver kube-api
kube-controller-manager kube-controller
kubelet kubelet
kube-scheduler scheduler
kube-proxy kubeproxy

etcd

Kubernetes uses etcd as a store for cluster state and data. Etcd is a reliable, consistent and distributed key-value store.

RKE supports running etcd in a single node mode or in HA cluster mode. It also supports adding and removing etcd nodes to the cluster.

You can enable etcd to take recurring snapshots. These snapshots can be used to restore etcd.

By default, RKE will deploy a new etcd service, but you can also run Kubernetes with an external etcd service.

Kubernetes API Server

Note for Rancher 2 users If you are configuring Cluster Options using a Config File when creating Rancher Launched Kubernetes, the names of services should contain underscores only: kube_api. This only applies to Rancher v2.0.5 and v2.0.6.

The Kubernetes API REST service, which handles requests and data for all Kubernetes objects and provide shared state for all the other Kubernetes components.

services:
  kube-api:
    # IP range for any services created on Kubernetes
    # This must match the service_cluster_ip_range in kube-controller
    service_cluster_ip_range: 10.43.0.0/16
    # Expose a different port range for NodePort services
    service_node_port_range: 30000-32767
    pod_security_policy: false
    # Enable AlwaysPullImages Admission controller plugin
    # Available as of v0.2.0
    always_pull_images: false

Kubernetes API Server Options

RKE supports the following options for the kube-api service :

  • Service Cluster IP Range (service_cluster_ip_range) - This is the virtual IP address that will be assigned to services created on Kubernetes. By default, the service cluster IP range is 10.43.0.0/16. If you change this value, then it must also be set with the same value on the Kubernetes Controller Manager (kube-controller).
  • Node Port Range (service_node_port_range) - The port range to be used for Kubernetes services created with the type NodePort. By default, the port range is 30000-32767.
  • Pod Security Policy (pod_security_policy) - An option to enable the Kubernetes Pod Security Policy. By default, we do not enable pod security policies as it is set to false. > Note: If you set pod_security_policy value to true, RKE will configure an open policy to allow any pods to work on the cluster. You will need to configure your own policies to fully utilize PSP.
  • Always Pull Images (always_pull_images) - Enable AlwaysPullImages Admission controller plugin. Enabling AlwaysPullImages is a security best practice. It forces Kubernetes to validate the image and pull credentials with the remote image registry. Local image layer cache will still be used, but it does add a small bit of overhead when launching containers to pull and compare image hashes. Note: Available as of v0.2.0

Kubernetes Controller Manager

Note for Rancher 2 users If you are configuring Cluster Options using a Config File when creating Rancher Launched Kubernetes, the names of services should contain underscores only: kube_controller. This only applies to Rancher v2.0.5 and v2.0.6.

The Kubernetes Controller Manager service is the component responsible for running Kubernetes main control loops. The controller manager monitors the cluster desired state through the Kubernetes API server and makes the necessary changes to the current state to reach the desired state.

services:
    kube-controller:
      # CIDR pool used to assign IP addresses to pods in the cluster
      cluster_cidr: 10.42.0.0/16
      # IP range for any services created on Kubernetes
      # This must match the service_cluster_ip_range in kube-api
      service_cluster_ip_range: 10.43.0.0/16

Kubernetes Controller Manager Options

RKE supports the following options for the kube-controller service:

  • Cluster CIDR (cluster_cidr) - The CIDR pool used to assign IP addresses to pods in the cluster. By default, each node in the cluster is assigned a /24 network from this pool for pod IP assignments. The default value for this option is 10.42.0.0/16.
  • Service Cluster IP Range (service_cluster_ip_range) - This is the virtual IP address that will be assigned to services created on Kubernetes. By default, the service cluster IP range is 10.43.0.0/16. If you change this value, then it must also be set with the same value on the Kubernetes API server (kube-api).

Kubelet

The kubelet services acts as a “node agent” for Kubernetes. It runs on all nodes deployed by RKE, and gives Kubernetes the ability to manage the container runtime on the node.

services:
    kubelet:
     # Base domain for the cluster
     cluster_domain: cluster.local
     # IP address for the DNS service endpoint
     cluster_dns_server: 10.43.0.10
     # Fail if swap is on
     fail_swap_on: false

Kubelet Options

RKE supports the following options for the kubelet service:

  • Cluster Domain (cluster_domain) - The base domain for the cluster. All services and DNS records created on the cluster. By default, the domain is set to cluster.local.
  • Cluster DNS Server (cluster_dns_server) - The IP address assigned to the DNS service endpoint within the cluster. DNS queries will be sent to this IP address which is used by KubeDNS. The default value for this option is 10.43.0.10
  • Fail if Swap is On (fail_swap_on) - In Kubernetes, the default behavior for the kubelet is to fail if swap is enabled on the node. RKE does not follow this default and allows deployments on nodes with swap enabled. By default, the value is false. If you’d like to revert to the default kubelet behavior, set this option to true.

Kubernetes Scheduler

The Kubernetes Scheduler service is responsible for scheduling cluster workloads based on various configurations, metrics, resource requirements and workload-specific requirements.

Currently, RKE doesn’t support any specific options for the scheduler service.

Kubernetes Network Proxy

The Kubernetes network proxy service runs on all nodes and manages endpoints created by Kubernetes for TCP/UDP ports.

Currently, RKE doesn’t support any specific options for the kubeproxy service.