Note that there are some explanatory texts on larger screens.

plurals
  1. PO
    text
    copied!<p><strong>added comment to longer GC log</strong> I think your survivor spaces are too small, note that eden is full with survivor at 56% before and eden is not full after but survivor is at 100%. This means it is forced into a completely non concurrent tenured collection, that is reported as a young gc because it was not triggered by CMS, because it has nowhere to put the extra garbage but tenured. Only the tenured collector can do stuff to tenured and hence a full GC is triggered. The remedy is making eden bigger and/or making the survivor spaces bigger. You can use <code>SurvivorRatio</code> to do this, e.g <code>-XX:SurvivorRatio=8</code> means set each survivor space to be 1/8 of eden.</p> <p>Stopped time is really "time the JVM spent at a safepoint", you can use <code>PrintSafepointStatistics</code> to determine what was going on during the safepoint. <a href="https://stackoverflow.com/questions/2850514/meaning-of-message-operations-coalesced-during-safepoint/2897599#2897599">This post</a> has a link to a list of the operations that cause the VM to reach a safepoint which, if you peruse it, will show there are many and varied ways it could sit there doing "nothing". Don't rule out OS clock shifts either as the times are recorded via a timeofday aware timestamp.</p> <p>btw pre hotspot17 those timestamps can be inaccurate. The bug is described in <a href="http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6782663" rel="nofollow noreferrer">Bug ID:6782663</a> which was fixed in hs17 which was part of 6u21 (as per <a href="http://www.oracle.com/technetwork/java/javase/6u21-156341.html" rel="nofollow noreferrer">the release notes</a>). Since you're on 6u20, it (the 30s) may not even be a real value.</p>
 

Querying!

 
Guidance

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload