I wanted to create in Amazon RDS the following topology: master -> slave -> slave. But this topology isn't possible in Amazon RDS. I'll appreciate if somebody can explain if I'll be able to create this kind of slaves with a replication from Amazon RDS (a master) to Aurora DB cluster when an Aurora master will replicate to slave of this "cluster"? Do this topology have additional limitations or cons?
0
votes
(1) Are you using Postgres or MySQL. (2) What are you hoping to gain by this?
– kdgregory
(3) If you're using plain RDS MySQL you can definitely do it with RDS-managed replication.
– kdgregory
I'm using MySQL. I understood Amazon MySQL RDS cannot create this kind of topology: master => slave => slave. How can I achieve this goal in Amazon RDS?
– Pavel Zeger
I want to create a running replication in case of master failover (to prompt the first slave to master when it has his own slave)
– Pavel Zeger
2 Answers
1
votes
With Aurora, you can have up to 15 read-replicas (slave in your terminology). For those read replicas, you define order of priority. In case that writer (in your terminology master) fails, first read-replica, by the order of priority is promted to writer (master), and other read replicas remain what they were. In the meantime, new RDS instance will be booted automatically with all the data, and will have a position of read-replica.
From your explanation, I believe that this does exactly what you want, when you set 1 writer and 2 read-replicas.
