Oplog truncation can fall behind (currently because of SERVER-123045 but potentially for other reasons). Multiple triggers have been identified for the known bug but there may be other bugs and/or other triggers. Monitoring will help confirm that the fix is correct.

On a per-cluster basis, monitor time spent truncating oplog, and amount of oplog present in excess of the configured retention time.

On a global basis, monitor top five and worst case excess oplog retention.