ADF unable to ingest partitioned Delta data from Azure Synapse Link (Dataverse/FnO)

Don’t read those folders as ordinary Parquet with recursive or wildcard traversal. A Delta table is the table root plus its _delta_log; targeting PartitionId=* bypasses the transaction log and can include obsolete files or fail metadata discovery.

In ADF, the direct option is a Mapping Data Flow with an inline Delta source pointed at the table root, not a Parquet dataset pointed at the partition folders. First confirm _delta_log exists at that root and remove the partition wildcard.

Because Synapse serverless already reads this data correctly, I’d use a serverless view for the production load: OPENROWSET(..., FORMAT='DELTA') against the root, project stable SQL types in the view, then let ADF Copy read that view into Azure SQL. This keeps Delta parsing and partition pruning in an engine that understands the log, while ADF handles orchestration and movement.

One caveat: Synapse serverless documents Delta Lake v1 support. If the producer uses newer features such as delete vectors or v2 checkpoints, use another Delta-capable engine.