RDS Reader Auto Scaling not working as expected
Reader Average CPU in Cloudwatch
RDS Reader Autoscaling events
RDS Reader Autoscaling settings
I have set up a reader autoscaling for our RDS Cluster.
The autoscaling settings are: Target 10% CPU, Min 0 readers, Max 1 readers.
The standard cluster is 1 writer, 2 readers.
Looking at the CPU metrics it's around 30%, I'd have expected the autoscaling group to add another reader to reduce the CPU utilisation. However, instead it looks like it's trying to scale in, but failing since there are no readers in the Autoscaling group.
What am I doing wrong?
- Newest
- Most votes
- Most comments
The CPU Utilization metric shown earlier is the default reader metric provided when creating an Aurora Cluster. However, Aurora Autoscaling operates based on the CPU Utilization metric specific to the read replica created through the Autoscaling configuration, not the general Cluster Read Replica metric.
As far as my understanding, and testing, for Aurora Read Replica Autoscaling, you should configure the settings to specify a minimum of 1 instance, with the maximum number of instances based on your scaling needs.
Relevant content
- Accepted Answerasked 2 years ago
- asked a month ago
AWS OFFICIALUpdated 2 years ago- AWS OFFICIALUpdated 2 years ago
- AWS OFFICIALUpdated 8 months ago
- AWS OFFICIALUpdated a year ago
I did more testing and think I worked it out.
The minimum capacity and maximum capacity under cluster capacity settings in the autoscaling includes all readers include ones created manually / not maintained by the autoscaling policy.
This seems a bit counter intuitive to me, but when I increase it to minimum 1 and maximum 2 it works as expected.