How do I provide resiliency in AWS-DMS?

I'm using AWS-DMS for replicating from PostgreSQL (as source endpoint) to S3 (as target endpoint). I want to provide resiliency for the process.

Particularly, if the DMS goes down due to some reason, how do I ask it to start from the last point where it had read?

This translates to :

1. If DMS goes down during the full load, it should perform full load again and then move to CDC (change data capture).
2. If DMS goes down during CDC, it should resume from the point where it had gone down.

This poses some more challenges, like the write-ahead-logs corresponding to the point where it had gone down may no longer be present, and so it needs to do a full load again followed by CDC.

But, being optimistic, I'd like a solution to the scenario when the log corresponding to the point of time when the DMS went down is still persisted.

Thanks.