View a markdown version of this page

How Serverless differs - Migration Assistant for Amazon OpenSearch Service

How Serverless differs

Amazon OpenSearch Serverless NextGen removes the cluster-management surface you may rely on with an Amazon OpenSearch Service domain. Account for these differences before you plan the migration:

  • No snapshot restore. You cannot restore a source snapshot directly into a collection. Migration Assistant still uses a snapshot as the read source for RFS, but it bulk-indexes documents through the OpenSearch API rather than restoring.

  • No ISM/ILM policies, ingest pipelines, or custom plugins. These are not migrated and have no equivalent to configure on the collection. Plan replacement logic in your ingestion path (for example, OpenSearch Ingestion pipelines) before cutover.

  • No _cluster/settings and no node-level tuning. Capacity scales automatically through OpenSearch Compute Units (OCUs), so the cluster-tuning knobs you set on a domain do not apply.

  • SigV4 only. All access is IAM-based. There is no basic-auth or FGAC user on the collection.

  • Index size limit. A single index in a collection has a maximum size (1 TB for the search and vector-search collection types). Reduce oversized source shards or split large source indexes before backfill.

  • Bulk payload limit. Keep bulk requests modest. documentsSizePerBulkRequest caps the aggregate document bytes per bulk request (default 10 MiB), and documentsPerBulkRequest caps the document count. Lower the byte-size limit first if the collection rejects large requests.