“Read replicas include single-node read replicas and HA read replicas.”
[HA Read Replica slide] If the physical server hosting an HA read replica fails, the standby read replica takes over. If a single-node read replica fails, the database proxy routes requests to other read replicas or the primary instance.
Huawei source: Training PDF p. 287
Technical Deep Dive
Read replicas offload read-heavy database traffic from the primary instance, improving read scalability while retaining a primary location for writes. Huawei V3.5 distinguishes single-node read replicas from high-availability read replicas; B is correct. HA read replicas include active and standby replica nodes so a standby can take over after the physical server hosting the active read node fails; C is correct.
Option D is correct because the database proxy can redirect traffic away from an unavailable single-node read replica to other healthy replicas or, depending on the configured policy, the primary instance. This is an important service-continuity mechanism for applications that should not manage every read replica address themselves.
Option A is false. The course does not impose an unconditional requirement that read-replica instance specifications be identical to those of the primary. Database feature compatibility, engine constraints and settings such as case sensitivity matter, but equality of every vCPU and memory dimension is not a universal availability rule. An architect must also account for asynchronous replication lag. A user who writes a record and immediately reads from a replica may not see the latest committed change; strict read-after-write workloads should use suitable primary routing or consistency controls. Monitor delay, read/write splitting policies, proxy state, health checks and read capacity. Any answer key listing only B and C omits Huawei's explicit support for D.
Submit