AOSC - Automatic Online Schema Change
AOSC moves an OpenSearch index to a pre-created target index while the source remains live. It backfills existing documents, replays source operation history, then briefly blocks source writes to finish catch-up and swap an alias.
Backfill and replay run while the source index continues serving application traffic.
Source writes are blocked only for final replay, validation, and alias swap.
Status APIs expose coordinator phase, shard phase, counters, and errors.
When AOSC fits
AOSC is for same-cluster index migrations where you need a new target index for mappings, settings, shard layout, or document shape, but the source index must keep receiving traffic during most of the move.
Cutover is not zero-interruption
Applications must retry writes rejected during the brief source write block. Successful cutovers have been observed in the 2s-30s range, but your window depends on index size, shard count, write load, and cluster-manager responsiveness.
Why AOSC?
Many OpenSearch index properties cannot be changed in place. Changing field types, analyzers, routing assumptions, or shard counts usually means creating a new index and copying data.
AOSC automates the parts that are risky to do by hand:
- Backfill source documents into the target.
- Replay source-shard operation history so writes during backfill are applied to the target.
- Track per-shard progress and expose status through REST APIs.
- Perform alias cutover after a source write block and final catch-up.
AOSC still requires planning. The target index must be created first, applications must use an alias for cutover, and applications must retry writes rejected during the cutover write block.