Merge pull request #112098 from kromerm/adfdocsmark

Update data-flow-alter-row.md

Author: PRMerger9

articles/data-factory/data-flow-alter-row.md

@@ -7,7 +7,7 @@ ms.reviewer: daperlov
 ms.service: data-factory
 ms.topic: conceptual
 ms.custom: seo-lt-2019
-ms.date: 01/08/2020
+ms.date: 04/20/2020
 ---
 
 # Alter row transformation in mapping data flow
@@ -43,11 +43,13 @@ For the alter row policies to work, the data stream must write to a database or
 
 ![Alter row sink](media/data-flow/alter-row2.png "Alter Row Sink")
 
- The default behavior is to only allow inserts. To allow updates, upserts, or deletes, check the box in the sink corresponding to that condition. If updates, upserts, or, deletes are enabled, you must specify which key columns in the sink to match on.
+The default behavior is to only allow inserts. To allow updates, upserts, or deletes, check the box in the sink corresponding to that condition. If updates, upserts, or, deletes are enabled, you must specify which key columns in the sink to match on.
 
 > [!NOTE]
 > If your inserts, updates, or upserts modify the schema of the target table in the sink, the data flow will fail. To modify the target schema in your database, choose **Recreate table** as the table action. This will drop and recreate your table with the new schema definition.
 
+The sink transformation requires either a single key or a series of keys for unique row identification in your target database. For SQL sinks, set the keys in the sink settings tab. For CosmosDB, set the partition key in the settings and also set the CosmosDB system field "id" in your sink mapping. For CosmosDB, it is mandatory to include the system column "id" for updates, upserts, and deletes.
+
 ## Data flow script
 
 ### Syntax