Getting Started with Load Balancing on an LKE Cluster
Traducciones al EspañolEstamos traduciendo nuestros guías y tutoriales al Español. Es posible que usted esté viendo una traducción generada automáticamente. Estamos trabajando con traductores profesionales para verificar las traducciones de nuestro sitio web. Este proyecto es un trabajo en curso.
The Linode Kubernetes Engine (LKE) is Linode’s managed Kubernetes service. When you deploy an LKE cluster, you receive a Kubernetes Master which runs your cluster’s control plane components, at no additional cost. The control plane includes Linode’s Cloud Controller Manager (CCM), which provides a way for your cluster to access additional Linode services. Linode’s CCM provides access to Linode’s load balancing service, Linode NodeBalancers.
NodeBalancers provide your Kubernetes cluster with a reliable way of exposing resources to the public internet. The LKE control plane handles the creation and deletion of the NodeBalancer, and correctly identifies the resources, and their networking, that the NodeBalancer will route traffic to. Whenever a Kubernetes Service of the LoadBalancer
type is created, your Kubernetes cluster will create a Linode NodeBalancer service with the help of the Linode CCM.
NoteAdding external Linode NodeBalancers to your LKE cluster will incur additional costs. See Linode’s Pricing page for details.
NoteAll existing LKE clusters receive CCM updates automatically every two weeks when a new LKE release is deployed. See the LKE Changelog for information on the latest LKE release.
NoteThe Linode Terraform K8s module also deploys a Kubernetes cluster with the Linode CCM installed by default. Any Kubernetes cluster with a Linode CCM installation can make use of Linode NodeBalancers in the ways described in this guide.
In this Guide
This guide will show you:
- manifest file configurations needed to add Linode NodeBalancers to your LKE cluster.
- annotations available to further configure your Linode NodeBalancers behavior and how to incorporate them into a manifest file.
- prerequisites and annotations needed to configure TLS termination on your cluster’s NodeBalancers.
- how to configure session affinity for the Pods in a cluster.
Before You Begin
This guide assumes you have a working Kubernetes cluster that was deployed using the Linode Kubernetes Engine (LKE). You can deploy a Kubernetes cluster using LKE in the following ways:
The Linode Cloud Manager.
Terraform, the popular infrastructure as code (IaC) tool.
Note
An LKE cluster will already have Linode’s Cloud Controller Manager installed in the cluster’s control plane. If you did not deploy your Kubernetes cluster using LKE and would like to make use of the Linode Cloud Controller Manager, see Installing the Linode CCM on an Unmanaged Kubernetes Cluster - A Tutorial.
Adding Linode NodeBalancers to your Kubernetes Cluster
To add an external load balancer to your Kubernetes cluster you can add the example lines to a new configuration file, or more commonly, to a Service file. When the configuration is applied to your cluster, Linode NodeBalancers will be created, and added to your Kubernetes cluster. Your cluster will be accessible via a public IP address and the NodeBalancers will route external traffic to a Service running on healthy nodes in your cluster.
NoteBilling for Linode NodeBalancers begin as soon as the example configuration is successfully applied to your Kubernetes cluster.
In any NodeBalancer configuration, users should keep in mind that NodeBalancers have a maximum connection limit of 10,000 concurrent connections.
1 2 3 4 5 6 7
spec: type: LoadBalancer ports: - name: http port: 80 protocol: TCP targetPort: 80
- The
spec.type
ofLoadBalancer
is responsible for telling Kubernetes to create a Linode NodeBalancer. - The remaining lines provide port definitions for your Service’s Pods and maps an incoming port to a container’s targetPort.
Viewing Linode NodeBalancer Details
To view details about running NodeBalancers on your cluster:
Get the services running on your cluster:
kubectl get services
You will see a similar output:
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE kubernetes ClusterIP 10.128.0.1 none 443/TCP 3h5m example-service LoadBalancer 10.128.171.88 45.79.246.55 80:30028/TCP 36m
- Viewing the entry for the
example-service
, you can find your NodeBalancer’s public IP under theEXTERNAL-IP
column. - The
PORT(S)
column displays theexample-service
incoming port and NodePort.
- Viewing the entry for the
View details about the
example-service
to retrieve information about the deployed NodeBalancers:kubectl describe service example-service
Name: nginx-service Namespace: default Labels: app=nginx Annotations: service.beta.kubernetes.io/linode-loadbalancer-throttle: 4 Selector: app=nginx Type: LoadBalancer IP: 10.128.171.88 LoadBalancer Ingress: 192.0.2.0 Port: http 80/TCP TargetPort: 80/TCP NodePort: http 30028/TCP Endpoints: 10.2.1.2:80,10.2.1.3:80,10.2.2.2:80 Session Affinity: None External Traffic Policy: Cluster Events: <none>
Configuring your Linode NodeBalancers with Annotations
The Linode CCM accepts annotations that configure the behavior and settings of your cluster’s underlying NodeBalancers.
- The table below provides a list of all available annotation suffixes.
- Each annotation must be prefixed with
service.beta.kubernetes.io/linode-loadbalancer-
. For example, the complete value for thethrottle
annotation isservice.beta.kubernetes.io/linode-loadbalancer-throttle
. - Annotation values such as
http
are case-sensitive.
Annotations Reference
Annotation (suffix) | Values | Default Value | Description |
---|---|---|---|
throttle | • integer • 0 -20 • 0 disables the throttle | 20 | The client connection throttle limits the number of new connections-per-second from the same client IP. |
default-protocol | • string • tcp , http , https | tcp | Specifies the protocol for the NodeBalancer to use. |
default-proxy-protocol | • string • none , v1 , v2 | none | Enables Proxy Protocol on the underlying NodeBalancer and specifies the version of Proxy Protocol to use. The Proxy Protocol allows TCP client connection information, like IP address and port number, to be transferred to cluster nodes. See the Using Proxy Protocol with NodeBalancers guide for details on each Proxy Protocol version. |
port-* | A JSON object of port configurations For example: { "tls-secret-name": "prod-app-tls", "protocol": "https"}) | None | • Specifies a NodeBalancer port to configure, i.e. port-443 .• Ports 1-65534 are available for balancing.• The available port configurations are: "tls-secret-name" use this key to provide a Kubernetes secret name when setting up TLS termination for a service to be accessed over HTTPS. The secret type should be kubernetes.io/tls ."protocol" specifies the protocol to use for this port, i.e. tcp , http , https . The default protocol is tcp , unless you provided a different configuration for the default-protocol annotation. |
check-type | • string • none , connection , http , http_body | None | • The type of health check to perform on Nodes to ensure that they are serving requests. The behavior for each check is the following:none no check is performedconnection checks for a valid TCP handshakehttp checks for a 2xx or 3xx response codehttp_body checks for a specific string within the response body of the healthcheck URL. Use the check-body annotation to provide the string to use for the check. |
check-path | string | None | The URL path that the NodeBalancer will use to check on the health of the back-end Nodes. |
check-body | string | None | The string that must be present in the response body of the URL path used for health checks. You must have a check-type annotation configured for a http_body check. |
check-interval | integer | None | The duration, in seconds, between health checks. |
check-timeout | • integer • value between 1 -30 | None | Duration, in seconds, to wait for a health check to succeed before it is considered a failure. |
check-attempts | • integer • value between 1 -30 | None | Number of health checks to perform before removing a back-end Node from service. |
check-passive | Boolean | false | When true , 5xx status codes will cause the health check to fail. |
preserve | Boolean | false | When true , deleting a LoadBalancer service does not delete the underlying NodeBalancer |
nodebalancer-id | string | None | The ID of the NodeBalancer to front the service. When not specified, a new NodeBalancer will be created. This can be configured on service creation or patching. |
hostname-only-ingress | Boolean | false | When true , the LoadBalancerStatus for the service will only contain the Hostname. This is useful for bypassing kube-proxy’s rerouting of in-cluster requests originally intended for the external LoadBalancer to the service’s constituent Pod IPs. |
NoteTo view a list of deprecated annotations, visit the Linode CCM GitHub repository.
Configuring Linode NodeBalancers for TLS Encryption
This section describes how to set up TLS termination on your Linode NodeBalancers so a Kubernetes Service can be accessed over HTTPS.
Generating a TLS type Secret
Kubernetes allows you to store sensitive information in a Secret object for use within your cluster. This is useful for storing things like passwords and API tokens. In this section, you will create a Kubernetes secret to store Transport Layer Security (TLS) certificates and keys that you will then use to configure TLS termination on your Linode NodeBalancers.
In the context of the Linode CCM, Secrets are useful for storing Transport Layer Security (TLS) certificates and keys. The linode-loadbalancer-tls
annotation requires TLS certificates and keys to be stored as Kubernetes Secrets with the type tls
. Follow the steps in this section to create a Kubernetes TLS Secret.
NoteThe steps in this section will create a self-signed TLS certificate. To learn how to create a TLS certificate from the Let’s Encrypt certificate authority (CA) and apply it to an application running on Kubernetes, see the Configuring Load Balancing with TLS Encryption on a Kubernetes Cluster.
Generate a TLS key and certificate using a TLS toolkit like OpenSSL. Be sure to change the
CN
andO
values to those of your own website domain.openssl req -newkey rsa:4096 \ -x509 \ -sha256 \ -days 3650 \ -nodes \ -out tls.crt \ -keyout tls.key \ -subj "/CN=mywebsite.com/O=mywebsite.com"
Create the secret using the
create secret tls
command. Ensure you substitute$SECRET_NAME
for the name you’d like to give to your secret. This will be how you reference the secret in your Service manifest.kubectl create secret tls $SECRET_NAME --cert tls.crt --key tls.key
You can check to make sure your Secret has been successfully stored by using
describe
:kubectl describe secret $SECRET_NAME
You should see output like the following:
kubectl describe secret docteamdemosite Name: my-secret Namespace: default Labels: <none> Annotations: <none> Type: kubernetes.io/tls Data ==== tls.crt: 1164 bytes tls.key: 1704 bytes
If your key is not formatted correctly you’ll receive an error stating that there is no PEM formatted data within the key file.
Configuring TLS within a Service
By default, Kubernetes does not expose Services with TLS termination over HTTPS. In order to use https
you’ll need to instruct the Service to use the correct port using the required annotations. You can add the following code snippet to a Service file to enable TLS termination on your NodeBalancers:
- File: example-service.yaml
1 2 3 4 5 6
... metadata: annotations: service.beta.kubernetes.io/linode-loadbalancer-default-protocol: http service.beta.kubernetes.io/linode-loadbalancer-port-443: '{ "tls-secret-name": "example-secret", "protocol": "https" }' ...
The
service.beta.kubernetes.io/linode-loadbalancer-default-protocol
annotation configures the NodeBalancer’s default protocol.service.beta.kubernetes.io/linode-loadbalancer-port-443
specifies port443
as the port to be configured. The value of this annotation is a JSON object designating the TLS secret name to use (example-secret
) and the protocol to use for the port being configured (https
).
If you have multiple Secrets and ports for different environments (testing, staging, etc.), you can define more than one secret and port pair:
- File: example-service.yaml
1 2 3 4 5 6 7
... metadata: annotations: service.beta.kubernetes.io/linode-loadbalancer-default-protocol: http service.beta.kubernetes.io/linode-loadbalancer-port-443: '{ "tls-secret-name": "example-secret", "protocol": "https" }' service.beta.kubernetes.io/linode-loadbalancer-port-8443: '{ "tls-secret-name": "example-secret-staging", "protocol": "https" }' ...
Configuring Session Affinity for Cluster Pods
kube-proxy
will always attempt to proxy traffic to a random backend Pod. To direct traffic to the same Pod, you can use the sessionAffinity
mechanism. When set to clientIP
, sessionAffinity
will ensure that all traffic from the same IP will be directed to the same Pod. You can add the example lines to a Service configuration file to
1 2 3 4 5 6 7 8
spec: type: LoadBalancer selector: app: example-app sessionAffinity: ClientIP sessionAffinityConfig: clientIP: timeoutSeconds: 100
Removing Linode NodeBalancers from your Kubernetes Cluster
To delete a NodeBalancer and the Service that it represents, you can use the Service manifest file you used to create the NodeBalancer. Simply use the delete
command and supply your file name with the f
flag:
kubectl delete -f example-service.yaml
Similarly, you can delete the Service by name:
kubectl delete service example-service
After deleting your service, its corresponding NodeBalancer will be removed from your Linode account.
NoteIf your Service file used thepreserve
annotation, the underlying NodeBalancer will not be removed from your Linode account. See the annotations reference for details.
This page was originally published on