Function data_migration_run 

pub async fn data_migration_run(
    configuration: &Configuration,
    workspace: &str,
    data_migration_request: DataMigrationRequest,
) -> Result<QueuedDataMigrationRun, Error<DataMigrationRunError>>

Pushes the migration onto the job queue and returns the job id immediately (#1212). It does not wait for the copy: the run’s progress, logs, result and cancel control are reached through the ordinary job endpoints under that id – /jobs_u/get/{id}, /jobs/get_progress/{id} (a progress_perc percentage, updated after each committed batch), /jobs_u/queue/cancel/{id} and the job SSE stream. Cancelling is not a rollback. The engine has no cross-batch transaction and no checkpoint (#1216), so cancelling an in-flight run stops it at the next batch boundary and leaves every batch that already committed in the target. The job’s error message names how many rows were written. A client must not present this operation as atomic. The dry-run operation stays synchronous – it is bounded by a sample row limit and is what an interactive mapping UI calls.