是否可以在不停机的情况下扩展/缩减 Aurora RDS 集群? [英] Is it possible to scale up/down Aurora RDS Cluster without downtime?

问题描述

我有两个实例的 RDS Aurora PostgreSQL 集群:

I have RDS Aurora PostgreSQL cluster with two instances:

cluster
├── instance_1 [writer] [no multiAZ]
└── instance_2 [reader] [no multiAZ]

当我更改 instance_1 故障转移操作的实例类型时，故障转移操作正常，但我的停机时间约为 1~2 分钟.我通过运行检查了停机时间

When I changing instance type for instance_1 failover operation working correct but I have downtime about 1~2 minutes. I checked downtime by running

watch -n 3 "psql -h db.cluster.url -p 5432 -d postgres -U postgres -c 'select ID from TABLE limit 1'"

之后 instance_1 成为 reader.

有没有什么办法可以手动将instance_1改成reader，改一下类型再恢复到writer，不用长时间停机(不停机最好，5~10秒也可以)

Is there any way to change instance_1 to reader manually, change it type and revert to writer without long downtime (no downtime is the best, but 5~10 seconds also acceptable)

我知道我可能会使用多可用区实例，但成本会高两倍.

I know that I may use multiAZ instances but it will be cost twice expensive.

推荐答案

使用 RDS 代理 可以大大减少故障转移期间的停机时间:

Using RDS Proxy can greatly reduce downtime during failover:

借助 RDS Proxy，Aurora 和 RDS 数据库的故障转移时间最多可减少 66%

With RDS Proxy, failover times for Aurora and RDS databases are reduced by up to 66%

大量看似漫长的故障转移由

A big amount of the seemingly long failover is taken by

1. 从连接丢失中恢复的客户端库
2. 读写器开关的 DNS 传播

RDS 代理处理读取器/写入器切换，因此无需将 DNS 更改传播到客户端，它始终使用相同的端点.

RDS Proxy handles the reader/writer switch so that no DNS changes have to be propagated to the client, it uses always the same end point.

有一个好文章 RDS 代理，其中显示了使用 RDS 代理时平均故障转移恢复时间从 24 秒提高到 3 秒.

There is a good article RDS Proxy which shows the speedup from 24 to 3 seconds average failover recovery time when using RDS proxy.

这篇关于是否可以在不停机的情况下扩展/缩减 Aurora RDS 集群?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！