This discussion is archived
1 Reply Latest reply: Aug 9, 2011 9:28 AM by 802316 RSS

survivor spaces

user50021 Newbie
Currently Being Moderated

we have a problem with heap sizing, in fact much more objects as expected survive enough young gcs to end up in the old heap where the finally become garbage. Since we have to use CMS we get a very fragmented old heap after a few hours and finally we see "concurent mode failure". To find the root cause we have to understand the aging of our objects. Is there a way to figure out which objects filling up the survivor space, e.g. a kind of "selective heap dump"? Thanks.

  • 1. Re: survivor spaces
    802316 Pro
    Currently Being Moderated
    I would have a look at the live objects in the heap.

    I would start with the following to get the most objects first.
    jmap -histo:live {pid}
    Another thing I would do is increase the young generation. I tend to start with 8 GB new size and see how the system behaves and reduce the young generation if its not needed.


  • Correct Answers - 10 points
  • Helpful Answers - 5 points