Aurora

Table of Contents

RDS Read Replica
#

cross AZ: OK
cross region: OK
promote / point-in-time recovery / restore from snapshot will create a new db instance which is not in my expectation. When DR, user need to update dns / db endpoint in application

For example:

The replication between the multi region db clusters is:

When write record to region1 master, the record will be replicated to

When writing a record to region2 master, the record will only be synchronised to region2 read-replica.

And both region1 and region2 masters are writables!!!

AWS “assume” people will use the region2 cluster as a readonly database.

This is not mentioned in aws doc. I found the answer from aws:repost instead.

when create a cross-region read-replica, it will create another db cluster in the new region
you cannot see the cross-region db in the same aws console
endpoint cannot cross region, only within the region
need to manually set binlog_format in the parameter group to before creating the read replica
need to manually set read_only true in the parameter group at second region db cluster to prevent its writer instance from writable by other
endpoint writer / reader instance depends on role so the writer endpoint in second region is incorrect and should be not used
route53 + latency based routing policy will need to be set up to route both reader endpoints to utilise those read-replicas. but the reader endpoint does not include the master. if you want to utilise the second region master, You need to create a custom endpoint for this

So AWS created Aurora Global Database to solve these drawbacks. But it does not support on small db instance type!