Resilience

DMS for RocketMQ provides a three-level reliability architecture and uses cross-AZ DR, intra-AZ instance DR, and instance data replication to ensure service durability and reliability.

**Table 1** Reliability architecture of DMS for RocketMQ
Reliability Solution	Description
Cross-AZ DR	DMS for RocketMQ provides cross-AZ instances that support cross-AZ DR. When an AZ is abnormal, the instances can still provide services.
Intra-AZ instance DR	Each master has two slaves. To ensure strong data consistency, the slaves synchronize data from the master. When a node fault occurs, the master/slave roles are automatically switched by using the Raft protocol.
Data DR	Data DR is implemented through data replication.

Cross-AZ DR Deployment Architecture

The following section applies only in v4.8.0. v5.x is only provided with TPS.

In v4.8.0, each broker has one master node and two standby nodes. For example, to purchase one broker, two brokers will be created; To purchase two brokers (as shown in Figure 1), six brokers will be created. NameServer is deployed on two brokers.

Figure 1 Specifying the broker quantity

Single-AZ
All components of a single-AZ instance are deployed in an AZ. By default, anti-affinity deployment is enabled for single-AZ instances. For high availability purpose, anti-affinity deployment creates master and standby nodes of a broker on different physical machines.
Multi-AZ
Prerequisite: Three or more AZs are available in the region, as shown in Figure 2.

Figure 2 Selecting an AZ

Components of a multi-AZ instance are deployed in three different AZs. In this way, cross-AZ DR is achieved.
- The two brokers of NameServer are deployed in two AZs.
- The master and standby nodes of each broker are randomly and evenly deployed in three AZs.
Figure 3 Multi-AZ deployment