Move AWS atlas cluster from provisioned to Non-provisioned IOPS without any downtime

I have a production cluster hosted on AWS by Atlas, consisting of two shards and using M50 (General) instance type. The current configuration uses “400 IOPS, provisioned”. However, I want to change it to “3000 IOPS, non-provisioned”, as this is the baseline for the gp3 volume, which provides better performance compared to the current 400 IOPS.


My concern is whether this change is safe to make, as I am not sure how it will affect the system’s uptime. According to my research, changing the EBS volume type from io1 to gp3 may require a rolling restart of cluster nodes, which could potentially cause some downtime.

Hey @Abdul_Rauf :wave: - Welcome to the community!

I’d recommend contacting the Atlas in-app chat support to verify if this change requires any downtime.

Regards,
Jason

For the avoidance of doubt, one of the key design principles of MongoDB Atlas is that all changes to database clusters are made in a rolling manner preserving majority quorum (save for a momentary election of at the replica set level) and hence uptime throughout: This means that changing from provisioned IOPS to general storage is a no downtime operation assuming your application is engineered to withstand elections: If you believe your application is not resilient to replica set level elections then we recommend working to change by testing: we offer our chaos testing capabilities like the Test Failover and test regional outage capabilities for these reasons.

2 Likes

This topic was automatically closed 5 days after the last reply. New replies are no longer allowed.