Control plane high availability requirements

Red Hat OpenStack Services on OpenShift (RHOSO) high availability (HA) uses Red Hat OpenShift Container Platform (RHOCP) operations to orchestrate failover and recovery deployment.

Install the following Operators on the RHOCP cluster to assist with the implementation of HA for the RHOSO control plane. These Operators work together to provide increased resiliency to help reduce the failover times of workloads and protect the operation of the control plane from node failures.

Installing and configuring the Node Health Check Operator in conjunction with the Self Node Remediation Operator, the Fence Agents Remediation Operator, or both allows these Operators to reduce the time a pod remains bound to a failed worker node.

  • The Node Health Check Operator. For information, see Remediating Nodes with Node Health Checks in the Workload Availability for Red Hat OpenShift Remediation, fencing, and maintenance guide.
  • The Self Node Remediation Operator. For information, see Using Self Node remediation in the Workload Availability for Red Hat OpenShift Remediation, fencing, and maintenance guide.
  • The Fence Agents Remediation Operator. For information, see Using Fence Agents Remediation in the Workload Availability for Red Hat OpenShift Remediation, fencing, and maintenance guide.

The Galera, RabbitMQ, and memcached control plane services are required to implement HA. These services are run as pods that are managed by their own Operators and are monitored by RHOCP. For more information about monitoring these HA services in a deployed RHOSO cloud, see Monitor cloud high availability.

To improve the isolation and scalability of RabbitMQ communication, RHOSO has revised the RabbitMQ interface for OpenStack services from a single setting that specified the RabbitMQ cluster to an expandable structure. In addition to specifying the RabbitMQ cluster, you can specify a dedicated RabbitMQ user for both the communication methods facilitated by RabbitMQ. For more information, see RabbitMQ interface for OpenStack services.

Note
You can choose to implement an improved failover behavior for a deployed RabbitMQ cluster that bypasses the default availability checks of OpenShift, which can wait up to five minutes to declare a service dead. The configuration of this improved RabbitMQ failover behavior is explained in step 5 of Create the OpenStackControlPlane CR.

To ensure compliance with your organization’s security standards, RHOSO provides the RHOSO administrator with easy to perform procedures that can rotate the following user credentials without causing service disruption:

  • To rotate the user credentials of the MariaDB database account of an OpenStack service.
  • To rotate the password of the root database account of a Galera database.
  • To rotate the user credentials of a dedicated RabbitMQ user.

For more information, see the Rotate user credentials.