Amazon Redshift DC2 to RA3 Migration best practices

Why to Migrate RA3 or Serverless

RA3 nodes separates compute from storage, allowing you to pay only for the compute resources you need while scaling storage independently. This flexibility leads to better cost optimization compared to DC2's fixed compute-to-storage ratio. Redshift Serverless & RA3 node type offer these additional features such as Data Sharing, Zero-ETL, multi-AZ deployments, cluster relocation, Sagemaker Lakehouse and more over DC2

Additionally future feature development for DC2 node type will stop at Redshift release P191 and all DC2 clusters will be migrated to a new Long Term Maintenance (LTM) track. On this track, existing clusters will continue to receive critical updates until their retirement on April 23rd, 2026. Customers on the LTM track will not be able to change tracks and will receive console alert messages recommending their transition to Serverless or RA3

Choosing Between Amazon Redshift Serverless and Provisioned

Both Redshift Serverless and RA3 provisioned clusters are viable options, with the choice depending on your workload and usage patterns. Redshift Serverless offers simplified management through automatic scaling, online weekly maintenance, auto scale and the auto pause during idle periods. In contrast, RA3 provisioned clusters provide greater control over workload management and resource allocation. Choose Serverless for operational simplicity and variable workloads, or RA3 provisioned nodes when you need fine-grained control over your cluster's workload.

Dc2 to Redshift Serverless conversion Sizing Guidance

For Redshift Serverless conversion, use the following size recommendations as a starting point. However, the optimal number of RPUs will depend on your specific workload patterns. Monitor your query performance after migration and adjust the node count accordingly to meet your performance requirements.

Dc2 to Ra3 Provisioned nodes conversion Sizing Guidance

Options to upgrade to Ra3 nodes There are 3 different ways to Migrate from Dc2 to Ra3 nodes.

1. Elastic Resize ( recommended way)

   Downtime during resize is minimal( 10-15 minutes)

   Cluster remains read-only during a resize operation

   Cluster Endpoint remains the same

2. Classic Resize

   Choose Classic Resize when target configuration is not available through Elastic Resize

   For single node clusters, only a classic resize can be performed to convert the cluster into a multi-node cluster

3. Snapshot Restore

   Use this option to keep your cluster available for writes during a resize operation.

   Create a snapshot of existing cluster and restore to a new cluster.

   If data is written to the source cluster after a snapshot is taken, the data must be manually copied over after the migration is complete.

   Cluster endpoint will change

Note -> for additional details on classic and elastic resizing please refer -Amazon Redshift Resizing

Options to upgrade to Redshift Serverless

To upgrade your Redshift Dc2 nodes to serverless, you can use the Snapshot Restore to migrate your data and schema from a provisioned cluster to a Redshift Serverless endpoint.

Testing Options

While the sizing recommendations serve as a good starting point and works well in most cases, the Amazon Redshift Test Drive utility allows you to validate your specific workload performance on RA3 nodes before migration. This open-source tool features four components that help assess migration impact by replicating your production workload, automated testing, and analyzing performance metrics.

1. Workload Replicator

   It captures and replays your production workload in a test environment, mimicking real-world query patterns and usage.

2. Configuration Comparison

   Automates testing of your production workload across multiple RA3 node sizes and Serverless configurations, eliminating manual testing effort.

3. Replay Analysis

   Replay Analysis delivers a comprehensive overview of test outcomes and performance metrics, empowering you to make data-driven decisions.