Redshift V2

Redshift is the dominant data warehouse in AWS-native data stacks — deep integration with S3, Glue, EMR, SageMaker, and Lake Formation. Redshift Spectrum extends SQL queries to S3 data lake without data movement. Redshift Serverless eliminates cluster management: query on-demand, pay per compute-second. The Data API is the cleanest integration path for agents: execute SQL over HTTPS without managing JDBC connections, poll for results, retrieve paginated output — no connection pool management needed. Redshift ML integrates with SageMaker for in-warehouse model training and inference. Confidence is docs-derived.

Data API (primary agent interface): POST https://redshift-data.{region}.amazonaws.com ExecuteStatement (Action: ExecuteStatement, ClusterIdentifier or WorkgroupName, Database, Sql). POST DescribeStatement (polls status: PICKED/STARTED/FINISHED/FAILED). POST GetStatementResult (retrieves paginated results). POST BatchExecuteStatement (multiple SQL statements). JDBC/ODBC: traditional connection for BI tools (requires VPC or public endpoint). Python: redshift_connector or psycopg2. AWS SDK (boto3): client.execute_statement(Sql="SELECT ...", ...); client.describe_statement(Id=...); client.get_statement_result(Id=...). Redshift COPY command: bulk load from S3 (fastest ingestion path).

IAM-based auth: SigV4 request signing via AWS SDK / boto3. Redshift Data API: IAM role with redshift-data:ExecuteStatement, redshift-data:DescribeStatement permissions. Temporary credentials: aws redshift get-cluster-credentials. Username+password: cluster master credentials for JDBC. Secrets Manager integration: store Redshift credentials; Data API supports SecretArn parameter. VPC: clusters typically in private VPC — Data API enables access without VPC peering. Column-level access control: GRANT/REVOKE per-column. Row-level security: row-level security policies (RLS) in Redshift.

Data API: async execution — poll DescribeStatement until status=FINISHED. Failed statements: status=FAILED with Error field containing SQL error details. Cluster availability: provisioned clusters have startup time after resume from pause; Serverless has minimal cold start. Connection limits: JDBC connections bounded by cluster size (use connection pooling or Data API). Redshift uptime at aws.amazon.com/redshift/sla (99.9% SLA). Maintenance windows: planned maintenance may briefly pause cluster. Serverless: automatic scaling, no maintenance windows. Cost control: monitor via CloudWatch and Redshift Console query monitoring rules.

docs.aws.amazon.com/redshift covers SQL reference, Data API guide, Redshift Serverless docs, COPY command reference, Spectrum guide, and Python connector docs. Getting started: Redshift Serverless free trial ($300 credits), first Data API query via AWS Console Query Editor or boto3 in under 15 minutes. boto3 (pip install boto3) is the standard Python interface. AWS CLI (aws redshift-data execute-statement) for command-line usage. Community via AWS re:Post (formerly re:Invent), Stack Overflow (amazon-redshift tag), and AWS Discord.