CDB for MySQL Read-Only Instances and RO Group Architecture and Features

As user business grows, read scalability often becomes essential. CDB for MySQL supports read-only (RO) instances and introduces the concept of RO groups to enable read-write separation and scale read capacity.

An RO group is a collection of read-only instances sharing a single access address via VIP:VPORT. Weights can be assigned to individual RO instances to achieve automatic load balancing. If an RO instance fails, traffic is redistributed transparently based on configured weights.

RO instances synchronize data with the primary instance via Binlog. While RO instance specifications may differ from the primary, it is recommended that they be equal or greater.

Key product features include:

Read-Write Address Separation : Primary and RO instances have distinct addresses; only the primary supports writes.

On-Demand Scaling : RO groups and instances can be horizontally scaled to meet increasing read demand.

Automatic Load Balancing : RO groups expose a single read endpoint; traffic is distributed by weight, with automatic rebalancing on instance failure.

Health Checks : CDB provides active probing, monitoring (down to minute-level, second-level soon), and alerting. Failed or isolated RO instances are automatically removed.

Per-RO-Group Parameter Configuration : Each RO group supports custom settings including:

Automatic Removal of High-Delay Instances : When enabled, instances exceeding the configured replication lag threshold are isolated; they are re-added once lag normalizes.

Rebalancing on Weight Changes : New weights take effect for new connections immediately. To force immediate rebalancing, existing connections are terminated and redistributed per the new weights.

Known Limitations : RO instances may lag behind the primary, so strong consistency reads must be performed on the primary.

Usage Guidance : For implementation details, refer to the official Tencent Cloud documentation:

Read-Only Instances

RO Group