AnalyticDB for PostgreSQL uses a primary/secondary architecture to provide multi-replica redundancy. AnalyticDB for PostgreSQL also uses triplicate storage technology of disks to ensure data reliability. The system switches over to a secondary node when the primary node fails and recovers a node when the node fails unexpectedly. This ensures high availability. You can also restore data by using the backup and restoration feature.
Primary/secondary multi-replica redundancy and triplicate storage technology of disks
AnalyticDB for PostgreSQL is a distributed database service that uses the massively parallel processing (MPP) architecture. Distribution allows performance to be enhanced. However, a distributed database system may encounter high fault probability on nodes. Typically, a distributed database system provides multiple replicas to ensure service availability when some nodes fail. AnalyticDB for PostgreSQL uses a dual-replica architecture. If a single replica set provides available compute nodes, the corresponding instance can provide services normally based on the primary/secondary dual-replica architecture. To ensure data reliability, AnalyticDB for PostgreSQL uses triplicate storage technology of disks for public cloud scenarios. A data file is discretely stored in triple-modular redundancy to prevent data loss. AnalyticDB for PostgreSQL uses disks on physical machines to create a RAID array for data redundancy.
AnalyticDB for PostgreSQL allows you to create multiple primary/secondary replicas in multiple data centers across zones and regions. This way, you can implement disaster recovery across data centers.
Automatic failover and recovery
AnalyticDB for PostgreSQL checks the health status of nodes on an instance in real time. If the primary node is abnormal, a hot standby node takes over from the abnormal node to ensure instance availability. If a node is abnormal, AnalyticDB for PostgreSQL recovers the node to ensure multiple-replica redundancy for an instance. The recovery methods vary based on scenarios. For example, if nodes are abnormal due to exceptions on a physical machine, the system migrates the nodes to another machine. If nodes are abnormal due to a suspended process, the system restarts the process.
Backup and restoration
In some cases, data is lost due to data center-level failures or accidental deletion of the data. AnalyticDB for PostgreSQL can restore data for instances by using remote data backups. AnalyticDB for PostgreSQL supports full logical backup and restoration and is about to support incremental backup and restoration. Compared with full logical backup, incremental backup does not lock data and has little impact on your business. AnalyticDB for PostgreSQL allows you to restore data at the instance, database, and schema levels.