Realtime Compute for Apache Flink aggregates data by using window functions based on the time attribute. To use window functions based on the event time of a job, you must define a watermark when you declare a source table.

A watermark is used to measure the progress of Event time. It is a hidden data attribute. A watermark is defined in the DDL statement of the source table. Flink provides the following statement to define a watermark: 
WATERMARK [watermarkName] FOR <rowtime_field> AS withOffset(<rowtime_field>, offset)
Note For more information about the time attributes of Realtime Compute for Apache Flink, see Time attributes.
Parameter Required Description
watermarkName No The name of the watermark.
<rowtime_field> Yes The column used to generate the watermark. <rowtime_field> is identified as the event time column and must be a column of the TIMESTAMP type defined in the table. You can use <rowtime_field> to define a window in the job code.
withOffset Yes The policy to generate a watermark. In this example, the watermark value is generated by using the following formula: <rowtime_field> - offset. The first parameter in withOffset must be <rowtime_field>.
offset Yes The offset between the watermark value and event time, in milliseconds.
A specific field in a data record indicates the time when the record was generated. For example, a table contains a rowtime field whose data type is TIMESTAMP, and one of its values is 1501750584000 (2017-08-03 08:56:24.000). If you want to define a watermark based on the rowtime field and configure a 4-second offset, add the following definition: 
WATERMARK FOR rowtime AS withOffset(rowtime, 4000)

In this example, the watermark time of the data record is 1501750584000 - 4000 = 1501750580000 (2017-08-03 08:56:20.000). This means that all data whose timestamp is earlier than 1501750580000 (2017-08-03 08:56:20.000) has arrived.

Note
  • If you use an event time-based watermark, the rowtime field must be of the TIMESTAMP type. Realtime Compute for Apache Flink supports 13-digit UNIX timestamps in milliseconds. If the rowtime field is of another type or the UNIX timestamp is not 13 digits in length, we recommend that you use a computed column to convert the time. For more information, see Computed column.
  • The event time and processing time can only be declared in the source table. For more information, see Event time and Processing time.

Summary

  • A watermark indicates that all the events whose timestamp t' is earlier than the watermark time t (t'< t) have occurred. After the watermark time t takes effect, all subsequently received data records whose event time is earlier than t are discarded. Realtime Compute for Apache Flink will allow you to change the configuration and update the subsequent data.
  • Watermarks are important for data streams that arrive out of order because the watermarks help ensure that the computing in a window is correct even if some events arrive late.
  • If an operator has multiple input data streams for parallel processing, the event time of the data stream with the shortest time is used as the event time of the operator.