Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-20150

Add permsize statistics for worker memory which may be very useful for the memory usage assessment

    XMLWordPrintableJSON

Details

    • Wish
    • Status: Resolved
    • Major
    • Resolution: Not A Problem
    • 2.0.2
    • None
    • Web UI
    • None

    Description

      It seems worker memory only be assigned to executor heap which is usually not enough for estimating the whole clauster memory usage,especially when memory becomes a bottleneck of the clauster.In many case,we found a executor's real memory usage was much larger than its heap size which make me have to check for every application's real memory expenditure.

      This can be improved by adding a mechanism for Non-Heap(permsize) statistics,only shown for extra memory usage which has no effect on the current worker memory allocation and statistics.The permsize can be obtained easily from executor java options.

      Attachments

        Activity

          People

            Unassigned Unassigned
            Jinhua Fu Jinhua Fu
            Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: