AWS guide for migrating from Amazon CloudSearch to OpenSearch Serverless

Amazon OpenSearch Serverless offers modern search capabilities such as semantic search, Retrieval Augmented Generation (RAG), and agentic search, with automatic scaling and a pay-as-you-go model. The migration process from CloudSearch involves assessing existing configurations, converting data and queries, configuring security, loading data with Amazon OpenSearch Ingestion, and validating the setup.

OpenSearch Serverless scales indexing and search compute independently, potentially scaling to zero when idle, and features a layered security model with separate encryption, network, and data access policies. Prerequisites for migration include an AWS account, an active Amazon CloudSearch domain with indexed data, source data available in a durable store like S3 or DynamoDB, and appropriate IAM permissions.

Effective migration planning involves thoroughly documenting the existing CloudSearch configuration, including instance type, partition and replication counts, total document count, data size, field definitions, analyzers, and API version. The guide also helps determine if OpenSearch Serverless is the most suitable target, or if a managed OpenSearch Service cluster is better for workloads requiring very low latency or direct instance control.

When creating an index in OpenSearch Serverless, it is essential to explicitly define index mappings by converting CloudSearch field types to their OpenSearch Service equivalents. The guide recommends using the narrowest possible numeric types for efficiency and adding keyword sub-fields for text fields that require sorting or aggregation.

Since Amazon CloudSearch does not provide built-in backup or snapshot features, migrating data involves re-ingesting it from a durable source like Amazon S3 or Amazon DynamoDB. This process requires transforming CloudSearch's Source Data Format (SDF) JSON or XML documents into a format compatible with OpenSearch Service.