Skip to content

Aurora SOA

Links: 114 AWS SOA Index

Introduction

HA and Read Scaling

Backup, Backtracking and Restores

  • Automatic Backups
    • Retention period 1-35 days (CANT be disabled)
    • PITR, restore your DB cluster within 5 minutes of the current time
    • Restore to a new DB cluster
  • Aurora Backtracking
    • Rewind the DB cluster back and forth in time (up to 72 hours)
    • DOESN'T create a new DB cluster (in-place restore)
    • Supports Aurora MySQL only
  • Aurora Database Cloning
    • Creates a new DB cluster that uses the same DB cluster volume as the original cluster.
    • Uses copy-on-write protocol (use the original/single copy of the data and allocate storage only when changes made to the data)
    • Example: create a test environment using your production data
You can migrate an RDS MySQL snapshot to Aurora MySQL Cluster.
You have an Aurora DB Cluster where the automatic backup is enabled with a retention period of 10 days. You're using this Aurora DB Cluster for testing purposes, so you want to disable automatic backups to reduce costs. What should you do?

You can't disable Aurora DB Cluster Automatic Backups.

If the question asks for a particular Aurora engine type, always go for MySQL.

CloudWatch Metrics

  • AuroraReplicaLag: amount of lag when replicating updates from the primary instance.
    • AuroraReplicaLagMaximum: max. amount of lag across all DB instances in the cluster
    • AuroraReplicaLagMinimum: min. amount of lag across all DB instances in the cluster
If replica lag is high, that means the users will have a different experience based on which replica they get the data from (eventual consistency)

Like for example if a user adds an item in a cart and then refreshes the page then he may not see the new item added to the cart.

  • DatabaseConnections: current number of connections to a DB instance
  • InsertLatency: average duration of insert operations

Miscellaneous

Consider this scenario - the primary instance of an Amazon Aurora cluster is unavailable because of an outage that has affected an entire AZ. The primary instance and all the reader instances are in the same AZ. As a SysOps Administrator, what action will you take to get the database online?

You must manually create one or more new DB instances in another AZ.

Last updated: 2023-03-20