Connect CDC uses the following rules to start a request at its scheduled time.
-
The kernel needs to be active for the model.
-
If the kernel is not running, the scheduled request starts are missed during that time.
-
When the kernel runs at a later time, it silently ignores the scheduled start requests that were missed.
-
If the schedule contains a request duration, the request stops on schedule even if the request has not “completed” its work.
-
For instance, the kernel stops a copy request at its scheduled time even if it has not completed copying all tables in the request.
-
If a request is not active at its scheduled stop time (due to the request running to completion before its scheduled stop time or a user manually stopping the request), the scheduled stop request results in no action in the kernel.
-
If a bi-directional request has a schedule associated with it, it is started on all its source hosts at the same time, provided the kernel is running on all these hosts.
-
You may manually start or stop the request that has a schedule associated with it
-
You may schedule multiple requests to run at the same time as long as the requests belong to different distributions and do not share tables with other running requests.
-
Using the start date and time and time zone, Connect CDC applies date arithmetic to handle changes due to daylight saving time and variation in lengths of months. For example, it is possible to run a request on the same day every week or at the end of every month.
-
Connect CDC does not support scheduling a request in the duration when the time zone switches over to or from daylight saving. For instance, if a request is scheduled to run every day at 2 AM, then its behavior cannot be predicted on the day when the time is changed for daylight saving time adjustment.