I'm trying to upgrade my Redshift DC2 cluster to RA3 before the end-of-life deadline next year, but the resize operation keeps failing without any error messages. My setup is in the us-east-1e availability zone. I attempted the classic resize from a 2-node dc2.large to a 2-node ra3.large, but as soon as the resize gets acknowledged and the cluster restarts, it changes to "cancelling-resize" and rolls back to dc2.large with a message stating that the resize operation was canceled. I've tried two approaches: a scheduled resize during the maintenance window (which confirmed queued but never executed) and forcing an immediate resize via CLI multiple times. Although Cloudwatch events note the cancellation, they don't provide any error details. Has anyone else faced this issue? Are there any known problems with migrating from DC2 to RA3 in certain availability zones or hidden requirements I might be overlooking? My only other option seems to be creating a new cluster from a snapshot and terminating the DC2 cluster, but I'm concerned that wouldn't qualify for AWS's upgrade credits since it's a direct EOL migration. Any insights would really help!
2 Answers
Have you considered using elastic resizes? They’re usually the go-to option for these migrations. We've switched many clusters and haven't run into any issues with them so far. It might help streamline your upgrade process!
Check out this blog post on best practices for the upgrade. It's got some useful tips that could shed light on your situation: [Best Practices for Upgrading from Amazon Redshift DC2 to RA3](https://aws.amazon.com/blogs/big-data/best-practices-for-upgrading-from-amazon-redshift-dc2-to-ra3-and-amazon-redshift-serverless/)
I tried an elastic resize too, but it ended up failing just like the others—keeps auto-canceling. It's frustrating!