Understanding Bulk API 2.0 Ingest

While processing ingest jobs, Salesforce Bulk API 2.0 automatically divides your job’s data into multiple "batches" to improve performance.

Salesforce creates a separate batch for every 10,000 records in your job data, up to a daily maximum of 150,000,000 records. If the limit is exceeded while processing your job data, the remaining data isn’t processed. The ingest job is marked as having failed.

Just as a job can fail, so can an individual batch. If Salesforce can’t process all the records in a batch within 10 minutes, the batch fails. Salesforce automatically retries failed batches up to a maximum of 10 times. If the batch still can’t be processed after 10 retries, the entire ingest job is moved to the Failed state and remaining job data isn’t processed.

If there’s a failure, create a new ingest job to process the records that weren’t processed.

To determine what records weren’t processed and what errors occurred, use the Failed Record Results and Get Job Unprocessed Record Results resources.