Skip to content

DR Database

Regional

Overview

dr-database-regional

DB running in EC2

  • Running DB in EC2 instances will not provide HA or resiliency
    • EC2 and EBS - single AZ product by themselves

RDS

  • Runs in subnet group which spans over mutliple AZs
  • Primary instance with standby instance in different AZ
    • Synchcronous replication between primary and standby instances
    • Failover happens when primary instance goes down and standy becomes primary
  • Both primary and standby has to be down from multi-AZ failure to cause complete failure in the region

Aurora

  • Can handle multi AZ failures
  • Volume(virtual database storage) runs in cluster mode which spans multiple AZ
  • Both primary (also write) and replicas read from same cluster volume
    • upto 15 read-only replicas
    • failover process is automated if primary instance goes down
  • Only region failover would significantly impact service

DynamodDB

  • Replcations between multiple nodes in mutliple AZs
  • Fails when the region goes down

Global

Overview

dr-database-global

DynamoDB Global Tables

  • Default dynamoDB table located in single region
  • Global Table provides multi master replications between regional replicas
    • last write wins replication method
  • Same region write - strongly consistent read | Alternate region - eventual consistency read

Aurora Global Databases

  • Read/Write cluster in one region and Read cluster in different regions
    • read cluster can be promoted to read/write cluster
  • Storage level replication - 1s lag
  • NON multi master

RDS cross-region replication

  • Cross Region Read Replica
  • Asynchronous Replication - lag could be longer compared to aurora