Enable Coordinator Mirroring for Apache Cloudberry
Apache Cloudberry provides a series of high-availability features to make sure your database system can tolerate unexpected incidents such as a hardware platform failure and can be quickly recovered from such incidents.
This topic describes how to configure coordinator mirroring to ensure a smooth coordinator node failover.
Overview of coordinator mirroring
In addition to the primary working coordinator node, you can include a standby coordinator in your CBDB cluster, which can take the system over when the primary host is down.
The primary and standby coordinators should be deployed on different hosts so that the cluster can tolerate a single-host failure. Clients connect to the primary coordinator and queries can be run only on the primary coordinator. The standby coordinator is kept up to date with the primary coordinator using Write-Ahead Logging (WAL)-based streaming replication.
If the coordinator fails, the administrator needs to run the gpactivatestandby
utility to have the standby coordinator take over as the new primary coordinator. You can configure a virtual IP address for the coordinator and standby so that client programs do not have to switch to a different network address when the current coordinator changes. If the coordinator host fails, the virtual IP address can be swapped to the actual acting coordinator.
Configure coordinator mirroring
Take the following steps to enable and activate coordinator mirroring for your CBDB cluster: