Load Balancer Package#

The Load Balancer Package enables load balancer and Domain Name System (DNS) services for workloads running on the edge. This package allows accessing applications deployed on edge nodes using a Fully Qualified Domain Name (FQDN) instead of the IP address.

External IP address resolution is achieved by associating FQDN and assigning a routable IP address to your application.

The Load Balancer Package contains the following:

MetalLB - Load balancer for bare-metal Kubernetes clusters; delivers routable IP (ExternalIP) address assignment for services (type=LoadBalancer) exposing your workloads.
MetalLB-Config - Configures MetalLB to fit Edge Platform clusters.
EdgeDNS - Converts DNS name to the IP address of a workload running inside a cluster based on the annotation of a resource.
Ingress-NGINX - Kubernetes ingress controller to support the ingress Kubernetes service.

Configuration Options#

MetalLB-Config, EdgeDNS, and Ingress-NGINX components in the Load Balancer package require network configurations for user applications to be deployed to edge clusters. When deploying the Load Balancer package, there are multiple fields in the Override Profile Values.

Override Profile Values > MetalLB-Config

dnsIPAddress: 10.1.0.80/32
ingressIPAddress: 10.1.0.81/32
staticIPs: "" // example - "10.1.0.82/32,10.1.0.83/32"
ipAddressRange: 10.1.0.30-10.1.0.79
L2Advertisement.enabled: false // false - not to set L2Advertisement for a specific interface; true - to set L2Advertisement for a specific interface
L2Advertisement.interface: "" // the network interface name for L2Advertisement; example - "enp138s0f0"

The dnsIPAddress is the field to set the EdgeDNS IP address.
The ingressIPAddress is for the ingress NGINX controller IP address.
The staticIPs should have the list of required static load balancer IP addresses for exposing Kubernetes services.

This should be written in the format IP1/32,IP2/32,… splitting with the , separator.
The ipAddressRange field has the IP range value defined in the format <startIP_in_range>-<endIP_in_range> accepted by MetalLB for exposing Kubernetes services.
The L2Advertisement.enabled is the flag to enable L2Advertisement.
The L2Advertisement.interface requires the network interface name for L2Advertisement. If L2Advertisement.enabled is set to true, it should be set such as enp138s0f0.

Override Profile Values > EdgeDNS

staticIP: 10.1.0.80

The staticIP is the field to set EdgeDNS IP address. It should be matched with the value dnsIPAddress in Override Profile Values > MetalLB-Config. Note that it does not need /32.

Override Profile Values > Ingress-NGINX

controller.service.loadBalancerIP: 10.1.0.81

The controller.service.loadBalancerIP field is for Ingress-NGINX controller IP address exposing with the MetalLB. It should be matched with the value ingressIPAddress in Override Profile Values > MetalLB-Config. Note that it does not need /32.

How to Apply Load Balancer, DNS, and Ingress#

In many cases, you need to communicate with the edge applications from the systems outside the edge cluster. This may include IoT sensors, media displays, or interactive users that need to connect to the applications deployed onto the edge node.

Though there are different ways, Intel recommends that you use the Load Balancer Package, which can provide the ability to assign externally accessible IP addresses and user-friendly FQDN for the edge application services. Or, even it supports the Ingress Kubernetes service too to access application services.

Apply Load Balancer and DNS#

Do the following to use the Load Balancer package in the application Helm chart to expose services outside of the cluster:

Ensure that the Load Balancer Package is deployed. Ensure that the Load Balancer IP pool range is configured properly.

Add a service with the LoadBalancer type in the application Helm chart. The following is an example of a simple manifest:

apiVersion: v1
kind: Service
metadata:
   name: example-service
spec:
   selector:
   app: example
   ports:
   - port: 8765
         targetPort: 9376
   type: LoadBalancer
   # optionally add below value to set a static LoadBalancer IP
   loadBalancerIP: 10.1.0.45

Typically, you can configure the type and loadBalancerIP fields using chart overriding values.

If you want to assign user-friendly FQDN instead of static Load Balancer IP, add the following annotation to the Service specification. Annotation value can be any domain name that you like. Also, include the annotation value variable in the chart.
```
annotations:
external-dns.alpha.kubernetes.io/hostname: "example.myedge.company.org"
```
Once the application is deployed successfully, you can access the service using the Load Balance IP address from your client system at the Edge without additional effort.

You can access the service with the FQDN in either of the following ways:

Add staticIP of EdgeDNS configured during the Load Balancer Package deployment to the domain name server list of the client system.
If there is an existing local name server in your edge network, configure a forwarding zone for the domain. In this example, myedge.company.org, with EdgeDNS staticIP in the local name server.

Apply Ingress#

To use Ingress, the Ingress resource should be defined in the application Helm chart.

Ensure that the Load Balancer package is deployed. Ensure that the Load Balancer IP pool range is configured properly.
Ensure that the application Helm chart has the Kubernetes service to expose user application(s) already.

Add an ingress with the Ingress type in the application Helm chart. Example:

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
   name: ingress-example
spec:
   rules:
   - host: foo.local # the hostname to access example-service-name Kubernetes service
      http:
         paths:
         - backend:
            service:
               name: example-service-name # target Kubernetes service
               port:
               name: http
         path: /bar # path to access the target Kubernetes service
         pathType: ImplementationSpecific

Then, the client machine to access the application via Ingress should have the DNS/Host rule.

There are at least two ways to follow:

We can just adjust the /etc/hosts file in the client machine to have the right hostname like <ingress Controller IP> <host in Ingress yaml>, such as 10.1.0.81 foo.local for the above example.
We can add the entry <ingress controller IP, host in Ingress yaml>` to the DNS server that the client machine is using.

Once DNS/Host rule is set, access the application with the url host/path such as foo.local/bar for the above example.

Note

The above snippet is just a simple example. When the source YAML files for Kubernetes service and Ingress in a Helm chart are following the official Kubernetes document, the Helm chart should work.