AWS Redshift: Upgrade DC2 clusters to Graviton-powered RG instances

RG instances feature a built-in data lake engine that processes queries directly on cluster compute nodes, eliminating the per-TB scanning cost associated with Amazon Redshift Spectrum on DC2 clusters. This integrated engine natively supports Apache Iceberg queries and automatically collects statistics using JIT-Analyze, without requiring Spectrum or manual collection, consolidating warehouse and data lake workloads.

Upgrading to RG instances provides access to managed storage that decouples storage from compute, allowing independent scaling of data volume and node count. RG instances run on AWS Graviton processors, offering higher throughput and lower cost per vCPU compared to RA3. They also include enhanced features like data sharing, zero-ETL integrations, streaming ingestion, concurrency scaling for writes, cross-AZ cluster relocation, multi-AZ deployments, and faster query compilation.

The post provides a table detailing RG instance configurations (rg.large, rg.xlarge, rg.4xlarge, rg.12xlarge) with their respective vCPUs and memory. It also includes a node mapping table to guide users in determining the recommended starting RG node configuration based on their current DC2 node type and count, emphasizing that extra nodes might be needed based on specific workload requirements.

Before upgrading, users must ensure snapshot availability for their DC2 cluster, verify network configuration (VPC, subnet groups, security groups), and confirm the cluster is running a supported Amazon Redshift version. Three upgrade methods are available: Elastic resize (recommended for speed and minimal downtime), Classic resize, and another method depending on operational constraints like write access during migration and elastic resize support.