Our replication process uses Aurora PostgreSQL as the source and S3 as the target. We’ve often observed that when a task errors out due to issues such as “SQLite general error. Code <250>, Message <unknown error>” or “The task stopped abnormally,” and we restart it using the option ‘Tables are already loaded. Start processing changes from:’ with the date set to one day earlier, the process fails to capture the changes from the previous day. Is that an expected behaviour?
We don't have the possibility of doing a full load as the source DB gets purged, so a full load results in losing the history.
If you are starting the task from an LSN back far enough to catch the missed events then you should not be missing data. The three errors you mention should be investigated on a support case. The first could be due to file corruption or disk full, etc., there is not enough detail. Same for the second error. The third indicates that the task abended for some reason (abnormal stop).
