[DRILL-473] Memory leak reported in ParquetRecordReader when running a hash aggregation test

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.4.0
• Assignee: Aman Sinha
• Reporter: Aman Sinha
• Created: 29/Mar/14 23:17
• Updated: 31/Jul/14 06:33
• Resolved: 09/Jun/14 05:33

Description

This memory leak was detected when working on Drill-443. To reproduce, remove the @Ignore label of testQ8() and testQ8_1() in TestHashAggr.java. See stack below.

Total 2 allocation(s) of byte size(s): 40000, 40000, at stack location:
org.apache.drill.exec.memory.TopLevelAllocator$ChildAllocator.buffer(TopLevelAllocator.java:96)
org.apache.drill.exec.memory.TopLevelAllocator$ChildAllocator.buffer(TopLevelAllocator.java:101)
org.apache.drill.exec.vector.BigIntVector.allocateNew(BigIntVector.java:103)
org.apache.drill.exec.vector.AllocationHelper.allocate(AllocationHelper.java:29)
org.apache.drill.exec.store.VectorHolder.allocateNew(VectorHolder.java:90)
org.apache.drill.exec.store.VectorHolder.reset(VectorHolder.java:73)
org.apache.drill.exec.store.parquet.ParquetRecordReader.resetBatch(ParquetRecordReader.java:275)
org.apache.drill.exec.store.parquet.ParquetRecordReader.next(ParquetRecordReader.java:333)
org.apache.drill.exec.physical.impl.ScanBatch.next(ScanBatch.java:94)
org.apache.drill.exec.physical.impl.validate.IteratorValidatorBatchIterator.next(IteratorValidatorBatchIterator.java:107)
org.apache.drill.exec.record.AbstractSingleRecordBatch.next(AbstractSingleRecordBatch.java:42)
org.apache.drill.exec.physical.impl.validate.IteratorValidatorBatchIterator.next(IteratorValidatorBatchIterator.java:107)
org.apache.drill.exec.physical.impl.aggregate.HashAggBatch.next(HashAggBatch.java:105)