We have just switched from deprecated batch mode collection and enrichment to the new streaming versions.
Everything was going fine till last week, when we pushed a datatype error to our production JS tracker and started trying to put a string in a numeric field in an unstruct schema. So now I have an enriched/bad S3 bucket full of events that I want to edit and put back into ETL.
A while back I did this with the batch mode collection and enrichment: pulled events out of enriched/bad, cleaned them up, generated fresh input files, placed them in the raw input bucket. All that worked fine.
Can I do a similar thing now? If I take the payload from an enriched/bad record and correct the defects, can I put it in a bucket where it will get picked up by emr-etl-runner?