RDS Blue/Green Deployments now have a downtime of less than 5 seconds

This is Oniumi (@fat47) from the Service Reliability Group (SRG) of the Media Headquarters.
#SRG(Service Reliability Group) is a group that mainly provides cross-sectional support for the infrastructure of our media services, improving existing services, launching new ones, and contributing to OSS.
This article summarizes how the RDS Blue/Green Deployments feature now takes less than 5 seconds to switch over.
I hope this helps in some way.
 

What are RDS Blue/Green Deployments?


RDS Blue/Green Deployments is a feature announced in 2022 that allows you to easily create a Green cluster, automatically establish replication with the Blue environment, and complete the switchover with the push of a button.
Please refer to our previous blog post for more details.
 
When switching between B/GMaximum 1 minute downtimeHowever, it was a very useful feature that could also be used for major version upgrades of the cluster.

Feature Updates for RDS Blue/Green Deployments


The following updates were made on January 20, 2026.
Amazon RDS Blue/Green Deployments Reduce Downtime to Less Than 5 Seconds
 
With this update,Downtime during B/G switching reduced to less than 5 secondsIt has been decided that this will be done.
 

I measured B/G switching downtime using a shell script


While running the following script, we performed an RDS B/G switch and measured the downtime.
Measurement script ( measure_aurora_downtime.sh )
The target environment is Aurora MySQL 3.04.
Set the necessary environment variables and run it.
 
Performs B/G switching during execution.
 
 
Measurement results
 
In this script, I connect to MySQL for each command, so in the verification environmentDowntime is 1.6 secondsIt was.
If the switchover is this quick, it seems that the impact on services will be minimal.

Conclusion


RDS Blue/Green Deployments was already a useful feature, but this update makes it even more useful.
I was currently upgrading Aurora MySQL from 3.04 LTS to 3.10 LTS, so the upgrade should go smoothly.
 
One thing to note:Depending on the application, downtime may take up to 30 seconds.So be careful.
The pattern I encountered was in an environment where the JVM's DNS TTL and connection timeout values ​​were set to the default of 30 seconds, and this caused the problem to occur.
Be sure to test B/G switching in the actual service environment and measure downtime.
 
If you are interested in SRG, please contact us here.