Doing failover in Amazon Aurora Read Replica with the highest priority

In this article, I am going to show you how to create a high-available Aurora Cluster via the Amazon RDS Management Console, add 2 Aurora Replicas with different priorities, and in different availability zones, test a fail-over scenario, and then terminate the whole infra-structure.

• By default, an Amazon Aurora cluster has only one primary compute instance performing read/write operations.

• By adding one or more Aurora Replicas to the cluster, read scalability and high availability increases.

• If the primary instance in the cluster fails, Aurora automatically promotes an existing Replica with the highest priority to be the new primary instance.

• Generally, you’ll want to create Aurora replicas in a different Availability Zone (AZ) than the primary instance.

• This way, your database can quickly fail over to the Replica in the other AZ in the event of infrastructure detects problems in the primary AZ.

• However, if you did not add Aurora Replicas to the cluster, you will have to wait for Aurora to create a new replacement primary instance for you when a failure is detected, which may take longer time.

• Amazon Aurora uses MySQL compatibility, but you can follow a similar process for the PostgreSQL edition.

Please visit my GitHub Repository for Aurora articles on various topics being updated on constant basis.

Let’s get started!

Objectives:

1. Create an Amazon Aurora cluster

2. Create an Aurora Replica for High Availability

3. Set up the database cluster for failover

4. Test database failover

5. Terminate Resources

Pre-requisites:

• AWS user account with admin access, not a root account.
• AWS CLI.

Resources Used:

Creating a RDS MySQL DB instance

Steps for implementation to this project:

1. Create an Amazon Aurora cluster

• On the Select engine page, select Standard Create.

• Create an an Amazon RDS database with the following parameters -
  Standard create Amazon Aurora engine, Dev/Test, database-1, <username> admin, <password>, db.r5.large, default vpc, DB subnet group - default, public access - no, Under Multi-AZ deployment - select No, VPC security group - create new - aurora-sg, us-east-1a, under additional configuration - myrds

• Create database.

• Wait for 3-4 minutes till the status changes to available

• Note: availability zone - us-east-1a

2. Create 2 Aurora Replicas for High Availability

• You can increase compute availability by adding Read Replicas in different Availability Zones (AZs) within an AWS region.

• Up to 15 Aurora Replicas can be added across the AZs within a region.

Create a 1st reader

• Select the radio button for the cluster and create an Aurora Replica by selecting Actions > Add reader.

• reader-1, us-east-1c, Under Additional configuration - Failover priority - 0

• Click Add reader.

Create a 2nd reader

• Select the radio button for the cluster and create an Aurora Replica by selecting Actions > Add reader.

• reader-2, us-east-1d, Under Additional configuration - Failover priority - 1

• Click Add reader.

3. Set up the database cluster for failover

Use this Step only, if you do not use priority order while creating Read replicas shown in Step 2.

• Select the radio button next to your Reader DB instance - reader-1 and select Modify.

• Set the highest priority. Under Failover, select Tier-0.

  • During a failover, Amazon RDS will promote the replica with the highest priority (starting with tier 0) to become the new primary instance.

• Select Continue, then Modify DB instance.

4. Test database failover

• Select the radio button on the target writer instance.

• Then select Actions > Failover.

• During a failover, Amazon RDS will promote the replica with the highest priority (starting with tier 0), to become the new primary instance.

• This will cause the replica - reader-1 with tier-0 to be promoted as the new primary (or writer) instance and old primary (or writer) instance becomes a new read replica.

• reader-2 remains a read replica, as we have set the priority order as - tier-1

• The time it takes for the failover to complete depends on the amount of database activity at the time of failover but is typically under 60 seconds.

• You can monitor failover process under Log & events > Recent events.

• The failover is transparent to the application by using endpoints.

• While the cluster and reader endpoints are used as DNS for the database, the instance connection will remain the same and automatically use the new db instance.

5. Terminate Resources

• Delete reader-1 and reader-2.

• Delete Amazon Aurora cluster

• Do not create a final snapshot and select the acknowledgement. Enter delete me and select Delete. This step will delete the Aurora cluster, including storage and all automated db backups.

• After some time you will this that database-1 as deleting

What we have done so far

• We have created a high-available Aurora Cluster via the Amazon RDS Management Console, added 2 Aurora Replicas with different priorities, and in different availability zones, tested a fail-over scenario, and then terminated the whole infra-structure.