RDS Multi AZ & Disaster Recovery

When we enable "Multi-AZ deployment" on a Database configuration, in the background Amazon creates two DB instances:

1. Primary DB
2. Standby DB

in two different availability zones, like below:

Applications does NOT directly track Primary or Standby DB instances, it connects to RDS endpoint.

RDS endpoint remains same, even if there is a primay DB instance failure. This will protect RDS instance in case of a Zone failure, even AZ-1 fails now our Apps is safe , it is still can access the DB that is the one in AZ-2.

What if the region itself fails?

There are two options to recover the DB:

1. RDS Backups
   • Snapshots can be created and used in case of any RDS failure in the region.
   • Side effect: If you take a snapshot in t1 and failure occurs in t1 + t2 then you will loose t2 time data.
2. Read Replicas

• In case the RDS fails in Region-1:
  • Application needs to use the RDS endpoints that connect to Read Replica
  • We need to promote Read Replica to Master DB to accept instert/delete all requests.
    • When you promote Read Replica, it breaks the relationship between the primary DB and the secondary.
  • Because of the asyncronization replication, we will not loose any data.

Disaster Recovery Strategies

Resource: https://docs.aws.amazon.com/whitepapers/latest/disaster-recovery-workloads-on-aws/disaster-recovery-options-in-the-cloud.html

Best Practices

1- When creating a read replica, try to keep the same instance class and storage class as your primary database so that it becomes easier for you to manage. (5*)

i.e. if the primary db is db.r5.2xlarge then use this in the read replica as well.

2- On CloudWatch, it is important to monitor the following metrics: (5*)

• ReplicaLag
• NetworkTransmitThroughput
• NetworkReceiveThroughput
• WriteThroughput
• WriteLatency
• WriteIOPS

3- It is recommended you use a different account per Region to provide the highest level of resource and security isolation (in the case compromised credentials are part of your disaster recovery plans as well). (7*)

Code

1. To create a RDS using terraform -> rds.tf (6*)

2. To implement multi-region data replication for disaster recovery with AWS RDS and DMS with Terraform -> rds-dr.tf (9*)

References

1* https://docs.aws.amazon.com/prescriptive-guidance/latest/dr-standard-edition-amazon-rds/sites-strategies.html

2* RDS Setting Up Database Disaster Recovery | RDS Multi AZ | RDS Read Replica | RDS Copy Snapshots - Java Home Cloud

3* https://d1.awsstatic.com/rdsImages/IG1_RDS1_AvailabilityDurability_Final.pdf

4* “Multi-AZ” in Amazon RDS and how it may differ from High Availability - Franck Pachot

5* Designing a DR strategy on Amazon RDS for SQL Server

6* https://github.com/azeezsalu/terraform-tutorial-reference-files/blob/main/rds-refference.tf

7* https://docs.aws.amazon.com/whitepapers/latest/disaster-recovery-workloads-on-aws/disaster-recovery-options-in-the-cloud.html#warm-standby

8* AWS Disaster Recovery Strategies – PoC with Terraform - Martin Perez Rodriguez

9* Implementing multi-region data replication for disaster recovery with AWS RDS and DMS with Terraform