- Newest
- Most votes
- Most comments
Hi Pia
Generally, replacing an instance to change its' size and/or instance class requires we goes through the EC2 provisioning control plane and often takes many minutes. It would be best to create a new "reader" instance of the appropriate new size and/or instance class in advance and then failover the writer role to it. This generally takes mid-10's of seconds for the failover, worst case is about 60 seconds. If you are using topology-aware native drivers for JDBC, ODBC, or Python right now (see https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/Aurora.Connecting.html), that failover time can be sub-10 seconds or so... After that, you can delete the original instance to save costs.
Hope that helps...
Hi Pia.
The Single-AZ Aurora cluster, it is generally better to directly modify the instance size of the writer rather than creating a reader and promoting it. Since you are not leveraging Multi-AZ benefits, promoting a reader won’t add much value and will introduce unnecessary complexity. If high availability and seamless scaling are important, consider moving to a Multi-AZ Aurora setup. This allows for easier scaling and failover between instances without significant downtime.
https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/Aurora.Connecting.html
Thank you for the help!
Relevant content
- AWS OFFICIALUpdated 3 years ago
Thank you for the help!
I wanted to add that my bias tends to favor solutions with lower downtime. @Parthasaradi's answer below is also accurate and favors simplicity. If you use Infrastructure-as-Code (IaC) in your development and deployment, that might be an important consideration.