To Avoid retries on scrolls that already have downstream records (see NXP-32166), we rely on the presence of processingStartTime in the BulkStatus of the command to detect unwanted retries.
This forces us to set the processing start time a little early.
We would like to use a new flushed flag instead that would be defined in the avro schema of the BulkStatus.
- is related to
-
NXP-32166 Avoid retries on scroll that has already downstream records
- Resolved