Skip to content

DC2 to RA3 migration, resize failing silently after multiple attempts

0

AZ is us-east-1e

I'm trying to migrate my Redshift DC2 cluster to RA3 before the EOL deadline early next year, but the resize operation keeps failing immediately with no error messages.

I've been trying classic resizes from my 2-node dc2.large to a 2-node ra3.large. The resize gets acknowledged, cluster restarts, but within a minute or two its status changes to "cancelling-resize" and then rolls back to dc2.large with the message "the requested resize operation was cancelled in the past. Rollback completed." and that's it.

I've tried 2 different ways:

  1. Scheduled resize during maintenance window (confirmed queued but it never executed)
  2. Force immediate resize via CLI (tried this a couple of times) Cloudwatch events show the cancellation but no error explaining why for both approaches.

Has anyone experienced this? Is there a known issue with DC2 to RA3 migrations in certain AZs? Any hidden requirements I'm missing?

The only other option I haven't tried is creating a new cluster off of a snapshot and then terminating the DC2 cluster, but I'm worried this wouldn't qualify for the RA3 upgrade credits that AWS is offering for direct DC2 to RA3 migrations due to he EOL migration.

Any help is appreciated!

1 Answer
0

When migrating from DC2 to RA3 nodes in Amazon Redshift, there are several potential reasons why your resize operations might be failing silently:

  1. Availability Zone Compatibility: While you mentioned us-east-1e, there's no known specific AZ issue with DC2 to RA3 migrations. However, RA3 nodes might have different availability requirements than DC2 nodes.

  2. Cluster Configuration: For a 2-node DC2 cluster, the recommended migration path to RA3 is typically 1:1 for DC2.large to RA3.large (1 node of RA3.large for every 1 node of DC2.large). Your approach of going from 2-node DC2.large to 2-node RA3.large aligns with the recommended sizing guidance.

  3. Single-Node vs. Multi-Node Considerations: If you were trying to go from a multi-node to a single-node configuration, this would require a classic resize rather than elastic resize, but you're maintaining the same node count, so this shouldn't be the issue.

  4. Alternative Migration Options: Since both classic resize and scheduled resize are failing, you might need to consider the snapshot restore method. This involves:

  • Creating a snapshot of your existing cluster
  • Restoring to a new cluster with RA3 nodes
  • Updating your endpoint references

Note that with the snapshot restore method, your cluster endpoint will change, and any data written to the source cluster after the snapshot is taken would need to be manually copied over.

  1. RA3 Upgrade Credits: Regarding your concern about qualifying for RA3 upgrade credits, I would recommend contacting AWS Support directly to confirm if the snapshot restore method would still qualify for the upgrade credits being offered for DC2 to RA3 migrations.

  2. Reserved Nodes Consideration: If you have reserved nodes for your DC2 cluster, note that reserved nodes are only available for RA3 types now, so you would need to complete the migration to RA3 before purchasing reserved nodes.

Since you've already tried multiple approaches without success, opening a support case with AWS might be the most efficient path forward to identify the specific cause of the silent failures and determine the best migration strategy for your situation.
Sources
Amazon Redshift DC2 to RA3 Migration best practices | AWS re:Post
Classic Resize Redshift Cluster status on modifying | AWS re:Post
CancelResize - Amazon Redshift
Changing cluster from dc2.large 4 nodes to ra3.xlplus 1 node | AWS re:Post
Reserved nodes in Redshift having On-demand cluster | AWS re:Post

answered a year ago

AWS
EXPERT

reviewed a year ago

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.