<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    Hi Ramki,<br>
    <br>
    thanks for your answere.<br>
    I have compared the gc-log with the jconsole chart and for every
    jump of the eden space to zero there is a (young) gc-entry in the
    log. So the jconsole chart is right and the scavenger period
    suddenly falls (even I am almost sure the application does not
    allocate more objects than before).<br>
    Thank you for the hint with the TLAB size. I will trying this now.<br>
    <br>
    Flo<br>
    <br>
    Am 12.12.2011 20:40, schrieb Srinivas Ramakrishna:
    <blockquote
cite="mid:CABzyjy=vtm2QFAqhcNdC85SKogXRZAnuHSC7AvA5Rfc-qQYCKA@mail.gmail.com"
      type="cite"><br>
      As I said, instead of relying on jconsole's asynchronous heap size
      samples, look at the GC log to see<br>
      if the inter-scavenge period suddenly falls (i.e. the application
      is presumably allocating at a much faster rate<br>
      than is normal). If the GC log confirns that, then see if turning
      off  TLAB resizing (and fixing it to a reasonable<br>
      value) makes any difference.<br>
      <br>
      Your existing GC logs should be sufficient to answer the first
      part of the question above.<br>
      <br>
      best.<br>
      -- ramki<br>
      <br>
      <div class="gmail_quote">On Mon, Dec 12, 2011 at 11:08 AM, Florian
        Binder <span dir="ltr"><<a moz-do-not-send="true"
            href="mailto:java@java4.info">java@java4.info</a>></span>
        wrote:<br>
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <div bgcolor="#FFFFFF" text="#000000"> Tony,<br>
            <br>
            As you have seen at the jconsole-screenshot the app does
            make very few garbage. It is optimized to make no garbage.
            But suddenly there seems to be every 90 seconds about 4 gb
            of (new) garbage (even the load of the app does not change).
            Therefore I thought it might be the g1-collector. I will try
            to reproduce this senario an have a closure look on it.<br>
            <br>
            Flo<br>
            <br>
            Am 12.12.2011 15:46, schrieb Tony Printezis:
            <div>
              <div class="h5">
                <blockquote type="cite"> Florian,<br>
                  <br>
                  Not sure what you mean by "the same GC". It is because
                  the space transition looks like:
                  16394M->12914M(20480M)? This is normal. Heap
                  occupancy reaches 16394M, GC happens, heap occupancy
                  goes down to 12914M. Next time it happens, the heap
                  will have new objects. This is how GCs work.<br>
                  <br>
                  Tony<br>
                  <br>
                  On 12/09/2011 09:08 PM, Florian Binder wrote:
                  <blockquote type="cite"> In the gc log it seems that
                    there is always running the same garbage collection.<br>
                    For example:<br>
                    $ zcat out_err.log.gz | grep 20480M | tail<br>
                    2011-11-28T19:13:19.482+0100: [GC cleanup
                    14515M->14515M(20480M), 0.1370060 secs]<br>
                       [ 16394M->12914M(20480M)]<br>
                       [ 16394M->12914M(20480M)]<br>
                    2011-11-28T19:17:12.509+0100: [GC cleanup
                    15582M->15582M(20480M), 0.1387230 secs]<br>
                       [ 16394M->12914M(20480M)]<br>
                       [ 16394M->12914M(20480M)]<br>
                       [ 16394M->12914M(20480M)]<br>
                    2011-11-28T19:21:06.089+0100: [GC cleanup
                    12978M->12978M(20480M), 0.1344170 secs]<br>
                       [ 16394M->12914M(20480M)]<br>
                       [ 16394M->12914M(20480M)]<br>
                    <br>
                    Therefore I assume this might be a bug ;-)<br>
                    You can download the whole log at: <a
                      moz-do-not-send="true"
                      href="http://java4.info/g1/out_err.log.gz"
                      target="_blank">http://java4.info/g1/out_err.log.gz</a><br>
                    <br>
                    I don't think that we still have the logs of that
                    very long gc, but I will have a look for it at
                    monday.<br>
                    Furthermore I do not think that we have logged there
                    much details of the gc. But I know that this happend
                    on a very special server, which contains more the
                    30gb of references (yes, just references to other
                    objects). If we run it with CompressedOops we reduce
                    the memory useage to nearly 50%.<br>
                    <br>
                    Flo<br>
                    <br>
                    <br>
                    <br>
                    Am 10.12.2011 01:34, schrieb Srinivas Ramakrishna:
                    <blockquote type="cite"><br>
                      <br>
                      A couple of things caught my eye....<br>
                      <br>
                      <div class="gmail_quote">
                        <blockquote class="gmail_quote" style="margin:0
                          0 0 .8ex;border-left:1px #ccc
                          solid;padding-left:1ex">
                          <div> On 12/05/2011 02:19 PM, Florian Binder
                            wrote:<br>
                            >> I assume you know when the data
                            will be dropped and reinitialized,<br>
                            >> right? Can you do a Full GC (with
                            System.gc()) after you re-initialize<br>
                            >> the heap. This typically helps a
                            lot with CMS.<br>
                            > Yes, this is exactly what we are trying
                            at this time ;-)<br>
                            > Either this Full-GC pauses are very
                            long (60-300 seconds) they are much<br>
                            > shorter than after a CMS-failure (when
                            they take sometimes more than<br>
                            > 4000 seconds ;-)) and we can define
                            when they occure.<br>
                          </div>
                        </blockquote>
                        <div><br>
                          A full GC that takes an hour is definitely a
                          bug. Have you logged that bug?<br>
                          Or at least share the GC log? What's the
                          version of the JDK that this behaviour<br>
                          was seen with?<br>
                           <br>
                        </div>
                        <div>... <br>
                        </div>
                        <blockquote class="gmail_quote"
                          style="margin:0pt 0pt 0pt
                          0.8ex;border-left:1px solid
                          rgb(204,204,204);padding-left:1ex">
                          <div> > Furthermore this should not
                            increase the total-heap-space:<br>
                            > <a moz-do-not-send="true"
                              href="http://java4.info/g1/all.png"
                              target="_blank">http://java4.info/g1/all.png</a><br>
                          </div>
                        </blockquote>
                        <div><br>
                           Remember that jconsole asynchronously samples
                          the heap, whose size is<br>
                          read "with possible glitches". Rather, you
                          should probably rely on the GC log in order to
                          assess<br>
                          the heap size after each GC event, rather than
                          the asynchronous samples from<br>
                          jconsole. I myself have not had the chance to
                          look at yr GC logs to see<br>
                          what that indicated wrt the size of Eden and
                          of the Heap.<br>
                          <br>
                          -- ramki<br>
                        </div>
                      </div>
                    </blockquote>
                    <br>
                    <br>
                    <fieldset></fieldset>
                    <br>
                    <pre>_______________________________________________
hotspot-gc-use mailing list
<a moz-do-not-send="true" href="mailto:hotspot-gc-use@openjdk.java.net" target="_blank">hotspot-gc-use@openjdk.java.net</a>
<a moz-do-not-send="true" href="http://mail.openjdk.java.net/mailman/listinfo/hotspot-gc-use" target="_blank">http://mail.openjdk.java.net/mailman/listinfo/hotspot-gc-use</a>
</pre>
                  </blockquote>
                </blockquote>
                <br>
              </div>
            </div>
          </div>
        </blockquote>
      </div>
      <br>
    </blockquote>
    <br>
  </body>
</html>