Red Hat OpenShift Container Platform on HPE SimpliVity

# Load balancer configuration

The solution supports a number of load balancer configuration options:

  • Use the playbooks to configure two load balancers, for highly available, production deployments.
  • Use the playbooks to configure a single load balancer, useful for proof of concept deployments.
  • Deploy the solution using your own load balancers.

# Quick start

A sample configuration for deploying two load balancers is shown below. This extract from the Ansible hosts inventory file shows the entries defining the nodes used for the load balancers:

[loadbalancer]
ocp-lb1 ansible_host=10.15.155.7 frontend_ipaddr=10.15.156.40/24 api_int_preferred=
ocp-lb2 ansible_host=10.15.155.8 frontend_ipaddr=10.15.156.41/24 api_preferred=

This extract from the configuration file group_vars/all/vars.yml shows the networking configuration required for the two load balancer scenario:

frontend_vm_portgroup: 'extVLAN2968'
frontend_gateway: '10.15.156.1''

loadbalancers:
  apps:
    vip: 10.15.156.42/24
  backend:
    vip: 10.15.155.9/24
    interface: ens192
    vrrp_router_id: 54
  frontend:
    vip: 10.15.156.42/24
    interface: ens224
    vrrp_router_id: 54

This configuration will be expanded on in the following sections.

# Deploying two load balancers

With this option, two load balancers are deployed in an active-active configuration to provide highly-available access. These nodes run keepalived and HAproxy. The load balancers are hosted on two VMs that are prevented from running on the same ESXi host through the use the following anti-affinity rule:

{{cluster_name}}-loadbalancer-anti-affinity-rule-001

where cluster_name is defined in the configuration file group_vars/all/vars.yml.

Two virtual machines are configured in the hosts inventory file, with each VM connected to an external (frontend) network and an internal (backend) network. To achieve this, each load balancer must have two IP addresses specified, one on the internal network using the ansible_host variable and a second address on the external network using frontend_ipaddr. The latter should be specified in CIDR notation (for example 10.10.174.165/22). The internal network is determined by the vm_portgroup variable in your configuration file, while the external network is set using the frontend_vm_portgroup variable.

You configure one of the VMs as the preferred VM for hosting the internal VIP for the OCP API, by adding api_int_preferred= to the definition. Similarly, you configure the other VM as the preferred VM for hosting the external VIP for the OCP API by adding api_preferred= to the entry. Note that the = sign is required for both settings. A sample [loadbalancer] group is shown below:

[loadbalancer]
ocp-lb1 ansible_host=10.15.155.7 frontend_ipaddr=10.15.156.40/24 api_int_preferred=
ocp-lb2 ansible_host=10.15.155.8 frontend_ipaddr=10.15.156.41/24 api_preferred=

The following variables must also be declared in your configuration file group_vars/all/vars.yml. The frontend network must be defined, similar to the example below:

frontend_vm_portgroup: 'extVLAN2968' # Name of the portgroup / external network
frontend_gateway: '10.15.156.1'      # gateway for the external network

The loadbalancers variable in your configuration file group_vars/all/vars.yml is used to define the networking configuration. A virtual IP (VIP) is specified for external access to applications, and for external (frontend) and internal (backend) access to the OCP API. A sample loadbalancers definition is shown below:

loadbalancers:
  apps:
    vip: 10.15.156.9/24
  backend:
    vip: 10.15.155.9/24
    interface: ens192
    vrrp_router_id: 51
  frontend:
    vip: 10.15.156.9/24
    interface: ens224
    vrrp_router_id: 51

The names of the interfaces are OS-dependent and depend on how the VMs are built. Using the playbooks from the OpenShift-on-SimpliVity repository with Red Hat Enterprise 7.6, the values of ens192 (backend) and ens224 (frontend) should be appropriate. The vrrp_router_id is used to differentiate between multiple deployments on the same networking infrastructure, for example, in proof of concepts. If you have multiple deployments, ensure that each deployment uses unique VRRP router IDs.

# Deploying one load balancer

If you do not require high availability, you can deploy a single load balancer to reduce complexity and resource requirements. In this instance, you only specify a single entry in the [loadbalancer] group in your hosts file. The load balancer must have two IP addresses specified, one on the internal network using the ansible_host variable and a second address on the external network using frontend_ipaddr.

[loadbalancer]
ocp-lb1 ansible_host=10.15.155.7 frontend_ipaddr=10.15.156.7/24

The frontend network must be defined in your configuration file group_vars/all/vars.yml, similar to the example below:

frontend_vm_portgroup: 'extVLAN2968' # Name of the portgroup / external network
frontend_gateway: '10.15.156.1'      # gateway for the external network

The definition for the loadbalancers variable is now simplified. No virtual IPs need to be specified. Instead, the two IP addresses of the load balancer VM are now used. From the example hosts file, 10.15.155.7 is used for internal access to the OCP API (from ansible_host=10.15.155.7) while 10.15.156.7 is used for external access (from frontend_ipaddr=10.15.156.7).

A sample loadbalancers definition is shown below:

loadbalancers:
  apps:
  backend:
    interface: ens192
  frontend:
    interface: ens224

Once again, the names of the interfaces are OS-dependent and depend on how the VMs are built. Using the playbooks from the OpenShift-on-SimpliVity repository with Red Hat Enterprise 7.6, the values of ens192 (backend) and ens224 (frontend) should be appropriate. The vrrp_router_id variable is no longer needed in this instance.

# Using your own load balancers

To use your own load balancers, you should delete any entries in the [loadbalancer] group in your hosts file. Note that you must still include the [loadbalancer] group in the file. Your load balancers should be configured to meet the requirements as specified in the OpenShift installation documentation at https://docs.openshift.com/container-platform/4.2/installing/installing_vsphere/installing-vsphere.html

The frontend network must be defined in your configuration file group_vars/all/vars.yml, similar to the example below:

frontend_vm_portgroup: 'extVLAN2968' # Name of the portgroup / external network
frontend_gateway: '10.15.156.1'      # gateway for the external network

The definition for the loadbalancers variable is again simplified but this time you just specify the virtual IPs to be used. A sample loadbalancers definition is shown below:

loadbalancers:
  apps:
    vip: 10.15.156.9/24
  backend:
    vip: 10.15.155.9/24
  frontend:
    vip: 10.15.156.9/24

# Load balancer variables

The following table contains definitions for all the variables used when configuring your load balancers.

Variable File Description
frontend_vm_portgroup group_vars/all/vars.yml Name of the portgroup connected to the access/public network
frontend_gateway group_vars/all/vars.yml Access network gateway
loadbalancers group_vars/all/vars.yml A dictionary containing entries for apps, backend and frontend
loadbalancers.apps group_vars/all/vars.yml A dictionary containing entries specific to apps access
loadbalancers.apps.vip group_vars/all/vars.yml If ommited, defaults to the internal IP address of the first load balancer (ie no VIP, no HA)
loadbalancers.backend group_vars/all/vars.yml A dictionary containing entries specific to backend access
loadbalancers.backend.vip group_vars/all/vars.yml If ommited, defaults to the internal IP address of the first load balancer (ie no VIP, no HA)
loadbalancers.backend.interface group_vars/all/vars.yml Name of the internal (backend) interface. For example, ens192 if using RHEL 7.6 with these playbooks
loadbalancers.backend.vrrp_router_id group_vars/all/vars.yml If you are deploying multiple clusters in the same environment, it is important to keep them separated using distinct VRRP IDs
loadbalancers.frontend group_vars/all/vars.yml A dictionary containing entries specific to frontend access
loadbalancers.frontend.vip group_vars/all/vars.yml If omitted, defaults to the external IP address of the first load balancer (i.e. no VIP, no HA)
loadbalancers.frontend.interface group_vars/all/vars.yml Name of the external (frontend) interface. For example, ens224 if using RHEL 7.6 with these playbooks
loadbalancers.frontend.vrrp_router_id group_vars/all/vars.yml If you are deploying multiple clusters in the same environment, it is important to keep them separated using distinct VRRP IDs

← Inventory group variables →