Topic
6 replies Latest Post - ‏2014-04-14T07:32:44Z by Hoozuki
SadanandaAithal
SadanandaAithal
3 Posts
ACCEPTED ANSWER

Pinned topic Heapdumps don't contain Object contents

‏2013-11-20T04:33:42Z |

Environment:

AIX 64

-Djava.runtime.version=pap6460sr12-20121025_01 (SR12)

JRE 1.6.0 AIX ppc64-64 build jvmap6460sr12-20121024_126067 (pap6460sr12-20121025_01(SR12))
VM build 20121024_126067
r9_20120914_26057
GC - 20120928_AA
JIT enabled, AOT enabled, FSD disabled, HCR disabled
 

Issue:

Of late, I see that IBM Heapdumps [ generated from a systhrow/OOME ] are very small as they do not contain Object contents.

It probably has to do with design change in Java6 SR12 or even an earlier release.

I say so because I have with me a heapdump from Java6 SR9 and I can see the contents of objects.

 

Can someone here help me with the answers to the following?

1) Is this a design change? If yes, from what version of JDK ?  And what options do I have to add to the commandline to ensure Objects are retained?

2) If this is NOT a design change, is there something that is going wrong ?

 

Thanks in advance!

 

Updated on 2013-11-20T04:34:00Z at 2013-11-20T04:34:00Z by SadanandaAithal
  • ParamS
    ParamS
    8 Posts
    ACCEPTED ANSWER

    Re: Heapdumps don't contain Object contents

    ‏2013-11-20T11:16:03Z  in response to SadanandaAithal

    Hi,

    For better understanding of the issue, can you share output of an object from the dump where you can read & can't read the content.

    Thank you, Param

     

     

    • SadanandaAithal
      SadanandaAithal
      3 Posts
      ACCEPTED ANSWER

      Re: Heapdumps don't contain Object contents

      ‏2013-11-20T13:05:38Z  in response to ParamS

      Ive attached a doc file showing the objects without contents.

      For the objects with contents I don't have a heapdump at this moment.

      Attachments

  • brataj
    brataj
    41 Posts
    ACCEPTED ANSWER

    Re: Heapdumps don't contain Object contents

    ‏2013-11-20T13:17:44Z  in response to SadanandaAithal

    I don't believe *.phd files have ever included the object contents, at least I've never seen any in years.

    You will see class names of course, and they do contain the size and location of the objects, and references between objects.

    If you have a matching core dump, you can use a tool like jdmpview to look at the object contents based on data in the *.phd file if required.

    Bernie

    • SadanandaAithal
      SadanandaAithal
      3 Posts
      ACCEPTED ANSWER

      Re: Heapdumps don't contain Object contents

      ‏2013-11-20T13:42:08Z  in response to brataj

      Thanks Param and Bernie.

      We use a script to generate heapdump based on the JDK type.

      I may have confused a heap from Oracle JVM to that of IBM's.

      So, i take it that IBM heapdumps *dont* have object contents. 

      Could we revert this behaviour with a simple command line option? [ say, modify the Xdump:request flags for heapdump alone?]

       

      • brataj
        brataj
        41 Posts
        ACCEPTED ANSWER

        Re: Heapdumps don't contain Object contents

        ‏2013-11-20T15:16:06Z  in response to SadanandaAithal

        You cannot revert to behavior that never existed.

        • Hoozuki
          Hoozuki
          8 Posts
          ACCEPTED ANSWER

          Re: Heapdumps don't contain Object contents

          ‏2014-04-14T07:32:44Z  in response to brataj

          WIMBoot. It sounds a bit like something 70-412 you'd Microsoft want to purchase before going on a cross-country hike in the snow. In actuality, passokay it's a new technological technique from Microsoft that arrived as a part of Windows 8.1 Update, and it should free up a bit more space on Windows 8 certified devices than your typical OS installation.

           

          The catch? Said devices will want to be 70-412 running Microsoft speedier flash-based storage - like a 32GB SSD, for example - since the WIMBoot passokay technology will impact performance to an undisclosed degree. And, honestly, WIMBoot is the kind of setup you'll want to run on these speedier, smaller-capacity devices; it wouldn't make much sense for a 500GB hard drive.

           

          So how, then, does WIMBoot work? In short, 70-412 it's a Microsoft tweak on the conventional means of installing Windows on a device. Instead of passokay unpacking all of the core Windows files from an installation package and dumping them on a partition - which could eat up around 9GB of space on a typical installation (sans recovery image) - the WIMBoot technique dumps this compressed "WIM" file into its own separate "Images" partition. Windows then creates pointer files within what looks like a conventional Windows installation that actually reference the WIM file on the separate partition.