Java 8 JIT线程似乎陷入无限循环

我在Java 8中编写了一个服务器应用程序，并使用java 1.8.0u25运行它。

I wrote a server application in Java 8, and running it with java 1.8.0u25.

它在前几个小时工作正常，但在它实现之后5k~10k请求，VM进程的一个线程使用其中一个CPU的100％。

It works fine for the first several hours, but after it gets about 5k~10k requests, a thread of the VM process uses 100% of one of the CPUs.

所以我尝试了 jstack 让VM进程检查有问题的线程是什么，并且它显示线程（线程id是14303 = 0x37df）是C2 CompilerThread0：

So I tried jstack for the VM process to check what the problematic thread was, and it showed that the thread(thread id was 14303=0x37df) was "C2 CompilerThread0":

"C2 CompilerThread0" #6 daemon prio=9 os_prio=0 tid=0x00002aaabc12a000 nid=0x37df runnable [0x0000000000000000]
   java.lang.Thread.State: RUNNABLE

使用 jstack -m ，线程的堆栈跟踪为以下内容：

And with jstack -m, the stack trace of the thread was as the following:

----------------- 14303 -----------------
0x00002b99b67693c3  _ZN16PhaseMacroExpand27process_users_of_allocationEP8CallNode + 0x2a3
0x00002b99b676ec3b  _ZN16PhaseMacroExpand23eliminate_allocate_nodeEP12AllocateNode + 0x1cb
0x00002b99b676ee65  _ZN16PhaseMacroExpand21eliminate_macro_nodesEv + 0x1a5
0x00002b99b6772769  _ZN16PhaseMacroExpand18expand_macro_nodesEv + 0x19
0x00002b99b640b01b  _ZN7Compile8OptimizeEv + 0xa6b
0x00002b99b640c53c  _ZN7CompileC1EP5ciEnvP10C2CompilerP8ciMethodibbb + 0x13bc
0x00002b99b635f9c8  _ZN10C2Compiler14compile_methodEP5ciEnvP8ciMethodi + 0x198
0x00002b99b6414c6a  _ZN13CompileBroker25invoke_compiler_on_methodEP11CompileTask + 0xc8a
0x00002b99b6417650  _ZN13CompileBroker20compiler_thread_loopEv + 0x620
0x00002b99b69a2e8f  _ZN10JavaThread17thread_main_innerEv + 0xdf
0x00002b99b69a2fbc  _ZN10JavaThread3runEv + 0x11c
0x00002b99b6860d48  _ZL10java_startP6Thread + 0x108

每次我尝试 jstack -m 时，这个线程的堆栈跟踪都是一样的，但是方法旁边的数字（程序计数器？）at堆栈顶部 _ZN16PhaseMacroExpand27process_users_of_allocationEP8CallNode 是 0x290 ， 0x2b1 ， 0x2a3 或 0x29f 。

And every time I tried jstack -m, the stack trace of this thread was all the same, but the number beside the method(program counter?) at the top of the stack _ZN16PhaseMacroExpand27process_users_of_allocationEP8CallNode was 0x290, 0x2b1, 0x2a3, or 0x29f.

C2 CompilerThread0 看起来像是一个执行JIT编译的线程，并且堆栈跟踪似乎陷入无限循环或其他什么。

C2 CompilerThread0 looks like a thread doing JIT compilation, and the stack trace seems like it fell into an infinite loop or something.

我想知道这是否是JVM的JIT编译器的错误。如果是，我如何指定我的应用程序的哪个方法使JVM变得疯狂，以及如何解决（或解决）这个问题？我尝试了 -XX：+ PrintCompilation 选项，但它没有多大帮助，因为它没有显示哪个线程编译了哪个方法。
如果它不是JVM的问题，那会发生什么呢？

I wonder if this could be a bug of JIT compiler of JVM. If it is, how can I specify which method of my application makes the JVM crazy and how can I solve(or work around) this problem? I tried -XX:+PrintCompilation option, but it did not help much because it didn't show which thread compiled which method. If it is not a problem of JVM, what could make this thing happen?