This topic describes three-zone instances and the benefits of three-zone instances.
What is a three-node instance?
PolarDB-X supports the following deployment modes for instances: single-zone deployment and three-zone deployment. Compared with single-zone instances, three-zone instances provide the cross-zone disaster recovery capability.
Enterprises can use three-zone instances to ensure high availability of their business applications. Three-zone instances are deployed in a three-node architecture that uses one leader node and two follower nodes. This architecture provides multiple replicas to ensure strong data consistency and high reliability that are required in financial scenarios.
| Disaster recovery level | Single-zone instance | Three-zone instance |
| Node-level (physical machine-level) disaster recovery | Supported | Supported |
| Cross-zone disaster recovery | Not supported | Supported |
Architecture
- A three-zone instance contains standby compute nodes. This improves the availability of compute nodes. In normal conditions, the standby compute nodes do not process workloads. If the zone where the primary compute node is deployed fails to provide services or the primary compute node fails, a standby compute node is elected as the new primary compute node and workloads are switched to the node.
- The data nodes and Global Meta Service (GMS) nodes of a three-node instance are deployed in three zones. The leader data node and leader GMS node are deployed in the primary zone, the follower data nodes and follower GMS nodes are deployed in the secondary zones, and the Logger node is deployed in the third zone that you specify when you create the instance.
Disaster recovery
Disaster recovery capability of compute nodesDisaster recovery after single zone failures
- If the primary zone fails to provide services, all compute nodes that are deployed in the primary zone become unavailable. In this case, you need to manually elect a secondary zone as the new primary zone. In most cases, the switchover operation can be completed within 30 seconds.
- If a secondary zone fails, the availability of the instance is not affected.
- The system checks the status of each node. If a node fails, the system automatically builds a new node in the same zone to maintain the three-node mode.
Disaster recovery after single zone failures
- A three-zone instance contains three data nodes and three GMS nodes that are deployed in three zones.
- If the leader data node or GMS node fails, a follower data node or GMS node is elected as the leader node. The instance temporarily becomes unavailable during the switchover process. This process does not cause data loss, and the RPO is 0.
- If a follower data node or GMS node, or the logger node fails, the availability of the instance is not affected, and no data is lost.
- The system checks the status of each node. If a node fails, the system automatically builds a new node in the same zone to maintain the three-node mode.
Benefits
High reliability
- Three-node instances provide high reliability for enterprise with an RPO of 0.
- Nodes in a three-node instance are deployed in three different data centers in the same region to support the cross-zone disaster recovery feature.
Create a three-zone instance
When you purchase a PolarDB-X instance, you can set the Topology parameter to Single-zone Deployment or Three-zone Deployment. To create a three-zone instance, set the Topology parameter to Three-zone Deployment. For more information, see Create PolarDB-X instances.
View information about a three-zone instance
After a three-zone instance is purchased, you can view basic information about the instance on the Instance List page and the details of the instance on the instance details page.