The RangeDeleter not only deletes the necessary documents, but also waits for a majority of members to be caught up. In 3.4, this happens all the time, whereas in master, this happens only if any documents were deleted.
In 3.4, the w:majority wait happens after the (apparently final) "rangeDeleter deleted X documents" message, whereas in master, it happens before the final "Deleted X documents in DB.COLL range ..." message.
This ticket is to request log message(s) (at default level) to indicate when the RangeDeleter is spending time waiting for replication.
- is related to
SERVER-29812 RangeDeleter unnecessarily waits for 'majority' write concern