Moving a self-hosted node to Cloud
1 Jun 2026
Take a snapshot on the self-managed node. Restore it into an empty Cloud collection of the same vector size and distance.
Point a single worker at the new URL and compare hit ids for a recorded query log. When the ids match within your recall budget, move the rest of the clients.
The open engine and Cloud use the same write path, so the application does not grow a second schema for the move.