Uploaded image for project: 'Parquet'
  1. Parquet
  2. PARQUET-353

Compressors not getting recycled while writing parquet files, causing memory leak

    XMLWordPrintableJSON

Details

    • Bug
    • Status: Resolved
    • Major
    • Resolution: Fixed
    • 1.6.0, 1.7.0, 1.8.0
    • 1.9.0, 1.8.2
    • parquet-mr
    • None

    Description

      Compressors are not getting recycled while writing parquet files. This is causing native/physical memory leak in my spark app which is parquet write intensive since its creating new compressors everytime i write parquet files.

      The actual code issue is that we are creating 'codecFactory' in 'getRecordWriter' method of ParquetOutputFormat.java but not calling codecFactory.release() which is responsible for recycling compressors.

      Attachments

        Issue Links

          Activity

            People

              nitin2goyal Nitin Goyal
              nitin2goyal Nitin Goyal
              Votes:
              1 Vote for this issue
              Watchers:
              4 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: