Details
-
Bug
-
Status: Open
-
Major
-
Resolution: Unresolved
-
1.8.2
-
None
-
None
Description
While profiling a spark job using AvroMultipleOutputs, I noticed that a great deal of time is wasted by copying (hadoop) Configuration. Indeed this happens on every invocation of write: https://github.com/apache/avro/blob/master/lang/java/mapred/src/main/java/org/apache/avro/mapreduce/AvroMultipleOutputs.java#L437
After patching, I am seeing a speed-up of 2x and above in running time of the same job.