Details
-
Improvement
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
3.0.0-alpha-1
-
None
-
None
Description
The Spark BulkLoad operations (and BulkLoadThin) only supports upserts, it does not support delete operations. To do so, https://github.com/apache/hbase/blob/f4f2b68238a094d7b1931dc8b7939742ccbb2b57/hbase-spark/src/main/scala/org/apache/hadoop/hbase/spark/HBaseContext.scala#L1034 would need to take the KeyValue.Type argument and be settable to KeyValue.Type.Delete.
We'd need to think about how we represented deletes in the input RDD, I guess with some container that contains either a Byte[] or a Delete marker. This would be a non-backwards compatible change, unless we override the method.