×

OKD provides methods for communicating from outside the cluster with services running in the cluster. This method uses load balancers on AWS, specifically a Network Load Balancer (NLB) or a Classic Load Balancer (CLB). Both types of load balancers can forward the client’s IP address to the node, but a CLB requires proxy protocol support, which OKD automatically enables.

There are two ways to configure an Ingress Controller to use an NLB:

  1. By force replacing the Ingress Controller that is currently using a CLB. This deletes the IngressController object and an outage will occur while the new DNS records propagate and the NLB is being provisioned.

  2. By editing an existing Ingress Controller that uses a CLB to use an NLB. This changes the load balancer without having to delete and recreate the IngressController object.

Both methods can be used to switch from an NLB to a CLB.

You can configure these load balancers on a new or existing AWS cluster.

Configuring Classic Load Balancer timeouts on AWS

OKD provides a method for setting a custom timeout period for a specific route or Ingress Controller. Additionally, an AWS Classic Load Balancer (CLB) has its own timeout period with a default time of 60 seconds.

If the timeout period of the CLB is shorter than the route timeout or Ingress Controller timeout, the load balancer can prematurely terminate the connection. You can prevent this problem by increasing both the timeout period of the route and CLB.

Configuring route timeouts

You can configure the default timeouts for an existing route when you have services in need of a low timeout, which is required for Service Level Availability (SLA) purposes, or a high timeout, for cases with a slow back end.

Prerequisites
  • You need a deployed Ingress Controller on a running cluster.

Procedure
  1. Using the oc annotate command, add the timeout to the route:

    $ oc annotate route <route_name> \
        --overwrite haproxy.router.openshift.io/timeout=<timeout><time_unit> (1)
    1 Supported time units are microseconds (us), milliseconds (ms), seconds (s), minutes (m), hours (h), or days (d).

    The following example sets a timeout of two seconds on a route named myroute:

    $ oc annotate route myroute --overwrite haproxy.router.openshift.io/timeout=2s

Configuring Classic Load Balancer timeouts

You can configure the default timeouts for a Classic Load Balancer (CLB) to extend idle connections.

Prerequisites
  • You must have a deployed Ingress Controller on a running cluster.

Procedure
  1. Set an AWS connection idle timeout of five minutes for the default ingresscontroller by running the following command:

    $ oc -n openshift-ingress-operator patch ingresscontroller/default \
        --type=merge --patch='{"spec":{"endpointPublishingStrategy": \
        {"type":"LoadBalancerService", "loadBalancer": \
        {"scope":"External", "providerParameters":{"type":"AWS", "aws": \
        {"type":"Classic", "classicLoadBalancer": \
        {"connectionIdleTimeout":"5m"}}}}}}}'
  2. Optional: Restore the default value of the timeout by running the following command:

    $ oc -n openshift-ingress-operator patch ingresscontroller/default \
        --type=merge --patch='{"spec":{"endpointPublishingStrategy": \
        {"loadBalancer":{"providerParameters":{"aws":{"classicLoadBalancer": \
        {"connectionIdleTimeout":null}}}}}}}'

You must specify the scope field when you change the connection timeout value unless the current scope is already set. When you set the scope field, you do not need to do so again if you restore the default timeout value.

Configuring ingress cluster traffic on AWS using a Network Load Balancer

OKD provides methods for communicating from outside the cluster with services that run in the cluster. One such method uses a Network Load Balancer (NLB). You can configure an NLB on a new or existing AWS cluster.

Switching the Ingress Controller from using a Classic Load Balancer to a Network Load Balancer

You can switch the Ingress Controller that is using a Classic Load Balancer (CLB) to one that uses a Network Load Balancer (NLB) on AWS.

Switching between these load balancers will not delete the IngressController object.

This procedure might cause the following issues:

  • An outage that can last several minutes due to new DNS records propagation, new load balancers provisioning, and other factors. IP addresses and canonical names of the Ingress Controller load balancer might change after applying this procedure.

  • Leaked load balancer resources due to a change in the annotation of the service.

Procedure
  1. Modify the existing Ingress Controller that you want to switch to using an NLB. This example assumes that your default Ingress Controller has an External scope and no other customizations:

    Example ingresscontroller.yaml file
    apiVersion: operator.openshift.io/v1
    kind: IngressController
    metadata:
      creationTimestamp: null
      name: default
      namespace: openshift-ingress-operator
    spec:
      endpointPublishingStrategy:
        loadBalancer:
          scope: External
          providerParameters:
            type: AWS
            aws:
              type: NLB
        type: LoadBalancerService

    If you do not specify a value for the spec.endpointPublishingStrategy.loadBalancer.providerParameters.aws.type field, the Ingress Controller uses the spec.loadBalancer.platform.aws.type value from the cluster Ingress configuration that was set during installation.

    If your Ingress Controller has other customizations that you want to update, such as changing the domain, consider force replacing the Ingress Controller definition file instead.

  2. Apply the changes to the Ingress Controller YAML file by running the command:

    $ oc apply -f ingresscontroller.yaml

    Expect several minutes of outages while the Ingress Controller updates.

Switching the Ingress Controller from using a Network Load Balancer to a Classic Load Balancer

You can switch the Ingress Controller that is using a Network Load Balancer (NLB) to one that uses a Classic Load Balancer (CLB) on AWS.

Switching between these load balancers will not delete the IngressController object.

This procedure might cause an outage that can last several minutes due to new DNS records propagation, new load balancers provisioning, and other factors. IP addresses and canonical names of the Ingress Controller load balancer might change after applying this procedure.

Procedure
  1. Modify the existing Ingress Controller that you want to switch to using a CLB. This example assumes that your default Ingress Controller has an External scope and no other customizations:

    Example ingresscontroller.yaml file
    apiVersion: operator.openshift.io/v1
    kind: IngressController
    metadata:
      creationTimestamp: null
      name: default
      namespace: openshift-ingress-operator
    spec:
      endpointPublishingStrategy:
        loadBalancer:
          scope: External
          providerParameters:
            type: AWS
            aws:
              type: Classic
        type: LoadBalancerService

    If you do not specify a value for the spec.endpointPublishingStrategy.loadBalancer.providerParameters.aws.type field, the Ingress Controller uses the spec.loadBalancer.platform.aws.type value from the cluster Ingress configuration that was set during installation.

    If your Ingress Controller has other customizations that you want to update, such as changing the domain, consider force replacing the Ingress Controller definition file instead.

  2. Apply the changes to the Ingress Controller YAML file by running the command:

    $ oc apply -f ingresscontroller.yaml

    Expect several minutes of outages while the Ingress Controller updates.

Replacing Ingress Controller Classic Load Balancer with Network Load Balancer

You can replace an Ingress Controller that is using a Classic Load Balancer (CLB) with one that uses a Network Load Balancer (NLB) on AWS.

This procedure might cause the following issues:

  • An outage that can last several minutes due to new DNS records propagation, new load balancers provisioning, and other factors. IP addresses and canonical names of the Ingress Controller load balancer might change after applying this procedure.

  • Leaked load balancer resources due to a change in the annotation of the service.

Procedure
  1. Create a file with a new default Ingress Controller. The following example assumes that your default Ingress Controller has an External scope and no other customizations:

    Example ingresscontroller.yml file
    apiVersion: operator.openshift.io/v1
    kind: IngressController
    metadata:
      creationTimestamp: null
      name: default
      namespace: openshift-ingress-operator
    spec:
      endpointPublishingStrategy:
        loadBalancer:
          scope: External
          providerParameters:
            type: AWS
            aws:
              type: NLB
        type: LoadBalancerService

    If your default Ingress Controller has other customizations, ensure that you modify the file accordingly.

    If your Ingress Controller has no other customizations and you are only updating the load balancer type, consider following the procedure detailed in "Switching the Ingress Controller from using a Classic Load Balancer to a Network Load Balancer".

  2. Force replace the Ingress Controller YAML file:

    $ oc replace --force --wait -f ingresscontroller.yml

    Wait until the Ingress Controller is replaced. Expect several of minutes of outages.

Configuring an Ingress Controller Network Load Balancer on an existing AWS cluster

You can create an Ingress Controller backed by an AWS Network Load Balancer (NLB) on an existing cluster.

Prerequisites
  • You must have an installed AWS cluster.

  • PlatformStatus of the infrastructure resource must be AWS.

    • To verify that the PlatformStatus is AWS, run:

      $ oc get infrastructure/cluster -o jsonpath='{.status.platformStatus.type}'
      AWS
Procedure

Create an Ingress Controller backed by an AWS NLB on an existing cluster.

  1. Create the Ingress Controller manifest:

     $ cat ingresscontroller-aws-nlb.yaml
    Example output
    apiVersion: operator.openshift.io/v1
    kind: IngressController
    metadata:
      name: $my_ingress_controller(1)
      namespace: openshift-ingress-operator
    spec:
      domain: $my_unique_ingress_domain(2)
      endpointPublishingStrategy:
        type: LoadBalancerService
        loadBalancer:
          scope: External(3)
          providerParameters:
            type: AWS
            aws:
              type: NLB
    1 Replace $my_ingress_controller with a unique name for the Ingress Controller.
    2 Replace $my_unique_ingress_domain with a domain name that is unique among all Ingress Controllers in the cluster.
    3 You can replace External with Internal to use an internal NLB.
  2. Create the resource in the cluster:

    $ oc create -f ingresscontroller-aws-nlb.yaml

Before you can configure an Ingress Controller NLB on a new AWS cluster, you must complete the Creating the installation configuration file procedure.

Configuring an Ingress Controller Network Load Balancer on a new AWS cluster

You can create an Ingress Controller backed by an AWS Network Load Balancer (NLB) on a new cluster.

Prerequisites
  • Create the install-config.yaml file and complete any modifications to it.

Procedure

Create an Ingress Controller backed by an AWS NLB on a new cluster.

  1. Change to the directory that contains the installation program and create the manifests:

    $ ./openshift-install create manifests --dir <installation_directory> (1)
    1 For <installation_directory>, specify the name of the directory that contains the install-config.yaml file for your cluster.
  2. Create a file that is named cluster-ingress-default-ingresscontroller.yaml in the <installation_directory>/manifests/ directory:

    $ touch <installation_directory>/manifests/cluster-ingress-default-ingresscontroller.yaml (1)
    1 For <installation_directory>, specify the directory name that contains the manifests/ directory for your cluster.

    After creating the file, several network configuration files are in the manifests/ directory, as shown:

    $ ls <installation_directory>/manifests/cluster-ingress-default-ingresscontroller.yaml
    Example output
    cluster-ingress-default-ingresscontroller.yaml
  3. Open the cluster-ingress-default-ingresscontroller.yaml file in an editor and enter a custom resource (CR) that describes the Operator configuration you want:

    apiVersion: operator.openshift.io/v1
    kind: IngressController
    metadata:
      creationTimestamp: null
      name: default
      namespace: openshift-ingress-operator
    spec:
      endpointPublishingStrategy:
        loadBalancer:
          scope: External
          providerParameters:
            type: AWS
            aws:
              type: NLB
        type: LoadBalancerService
  4. Save the cluster-ingress-default-ingresscontroller.yaml file and quit the text editor.

  5. Optional: Back up the manifests/cluster-ingress-default-ingresscontroller.yaml file. The installation program deletes the manifests/ directory when creating the cluster.

Additional resources