Red Hat Training

A Red Hat training course is available for Red Hat OpenStack Platform

Chapter 2. About Overcloud Components

For the purposes of backing up the OpenStack overcloud, consider the following components:

  • Pacemaker: Pacemaker is a high-availability cluster resource manager. Pacemaker achieves maximum availability for cluster services by detecting and recovering from node- and resource-level failures. Pacemaker uses messaging and membership capabilities provided by Corosync.
  • MariaDB: MariaDB is the SQL database service for the OpenStack overcloud.
  • Galera: Galera is a synchronous, multi-master clustering solution for MariaDB. In the context of OpenStack, Galera provides high availability for the database service in HA deployments of OpenStack controller nodes.

On a typical HA cluster with 3 controller nodes, OpenStack services such as OpenStack Compute (nova) or OpenStack Networking (neutron) access the MariaDB database via HAProxy over a virtual IP address. HAProxy balances all incoming traffic to a single active controller node among the available controller nodes in the cluster. In a 3-node cluster, the other 2 controller nodes act as hot standby nodes, which synchronize with the first node using Galera’s synchronous replication. If HAProxy detects that the first target node is "unavailable," it will select one of the two remaining hot standby nodes and balance incoming traffic to the selected node. When clients attempt to use an existing database connection from a failed node, HAProxy will migrate the database connections to the new node. As a side effect of this lazy migration approach, each client will recreate a SQL connection.

Consequently, backing up a running cluster requires identifying one of the idle controller nodes as the backup target to avoid any impact on the controller node that is currently servicing clients.