问题描述
我正在研究VisualVM中的hprof文件
服务器正在运行JDK 1.4.2_30，并具有1 GB的堆，NewSize为200 Mb.
hprof显示56000个int[]实例占用了71％的堆，而在VisualVM中查看时，这些56K数组都没有引用
根据我们的说法，如果没有引用"，则应该将其收集为垃圾.所以问题是:
a)有什么方法可以找到这些参考文献?
b)这是不正确的快照-即采取堆转储的操作是否进行了某种GC的操作?
c)我们应该在VisualVM中查看保留大小"对象吗?
作为更新-我们仍然不知道这些int[]是什么，但是下一个最大的对象是Pool中的Weblogic内部EJB引用，我们发现其中之一具有不正确的设置将内存使用率从早期的频繁Full GC降低到30％
免责声明:未来的猜测，我没有一个明确的答案，但我可能会给出一些有用的提示. /p>
我最近也遇到过类似的情况(尽管使用了最新的Java版本).
大量未引用的int[] 似乎被映射为某种类型的缓冲区的原因(我不是NIO专家，但似乎与此有关).
得出此结论的原因是，存在一个非常相似的数量的byte[]，它们的内存大小几乎相同，并且这些byte[]被某些*Channel实例引用.
所以我的猜测是，那些int[]实际上是 synthetic 实例，实际上与byte[]共享了它们的内存(正确引用了 ).简而言之:它们是红色鲱鱼.
所有这些都是很多猜测，而我从来没有跟进，因为当时的实际问题完全无关紧要.
I'm investigating an hprof file in VisualVM
The server is running JDK 1.4.2_30 and has 1 GB heap with NewSize of 200 Mb.
The hprof shows 71% of the heap occupied by 56000 instances of int[] and none of these 56K arrays have references when viewed in VisualVM
As per us, this should have been garbage collected if there were "no references".So the questions are:
a) is there any way to find out these references?
b) Is this an incorrect snapshot - i.e. did the action of taking the heap dump do some sort of GC ?
c) Should we be looking at the "retained sizes" objects in VisualVM?
As an update - we still dont know what these int[] are, but the next largest objects were Weblogic internal EJB references in Pool, and we discovered an incorrect setting in one of those which has brought down memory util to 30% from earlier frequent Full GCs
Disclaimer: guesswork ahead, I don't have a clear answer, but I might be able to give some useful hints.
I've seen a similar situation lately ('though with a much more current Java version).
The reason for the big number of unreferenced int[] seems to have been mapped buffers of some kind (I'm not a NIO expert, but it seems to have been related to this).
The reason for this conclusion is that there was a very similar number of byte[] around with pretty much the same size in memory and those byte[] were referenced by some *Channel instances.
So my guess is (was) that those int[] are actually synthetic instances that actually share their memory with the byte[] (which are properly referenced). In short: they are red herrings.
All of this is a lot of guesswork and I never followed up with it, because the actual problem at that time turned out to be something entirely unrelated.
这篇关于具有“无参考"的对象被称为“无参考".在hprof中的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！