Java获取Linux的fullgc,linux里查看最耗CPU的线程(full gc)

1、top后按c查看最耗cpu的进程，得到pid

2、top -Hp pid 查看该进程里的线程资源使用情况，找到最耗资源的线程的pid

3、jstack pid来查看进程的各个线程栈，注意这里的pid是第一步中进程的pid，不是第二步得到的线程id

4、将第二步得到的pid转成16进制之后在线程栈信息里查找nid等于pid16进制的，就找到最耗资源的线程的栈信息

话不多说了，先来看代码吧

public class Test{ public static void main(String args[]){ for(int i=0;i<10;i++){ new Thread(){ public void run(){ try{ Thread.sleep(100000); }catch(Exception e){} } }.start(); } Thread t=new Thread(){ public void run(){ int i=0; while(true){ i=(i++)/100; } } }; t.setName("Busiest Thread"); t.start(); } } 这个例子里新创建了11个线程，其中10个线程没干什么事，主要是sleep，另外有一个线程在循环里一直跑着，可以想象这个线程是这个进程里最耗cpu的线程了，那怎么把这个线程给抓出来呢？

首先我们可以通过top -Hp 来看这个进程里所有线程的cpu消耗情况，得到类似下面的数据

$ top -Hp 18207 top - 19:11:43 up 573 days, 2:43, 2 users, load average: 3.03, 3.03, 3.02 Tasks: 44 total, 1 running, 43 sleeping, 0 stopped, 0 zombie Cpu(s): 18.8%us, 0.0%sy, 0.0%ni, 81.1%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st Mem: 99191752k total, 98683576k used, 508176k free, 128248k buffers Swap: 1999864k total, 191064k used, 1808800k free, 17413760k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 18250 admin 20 0 26.1g 28m 10m R 99.9 0.0 0:19.50 java Test 18207 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18208 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.09 java Test 18209 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18210 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18211 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18212 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18213 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18214 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18215 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18216 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18217 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18218 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18219 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18220 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18221 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18222 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18223 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18224 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18225 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18226 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 18227 admin 20 0 26.1g 28m 10m S 0.0 0.0 0:00.00 java Test 拿到这个结果之后，我们可以看到cpu最高的线程是pid为18250的线程，占了99.8%：

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 18250 admin 20 0 26.1g 28m 10m R 99.9 0.0 0:19.50 java Test 接着我们可以通过jstack 的输出来看各个线程栈:

$ jstack 18207 2016-03-30 19:12:23 Full thread dump OpenJDK 64-Bit Server VM (25.66-b60 mixed mode):

"Attach Listener" #30 daemon prio=9 os_prio=0 tid=0x00007fb90be13000 nid=0x47d7 waiting on condition [0x0000000000000000] java.lang.Thread.State: RUNNABLE

"DestroyJavaVM" #29 prio=5 os_prio=0 tid=0x00007fb96245b800 nid=0x4720 waiting on condition [0x0000000000000000] java.lang.Thread.State: RUNNABLE

"Busiest Thread" #28 prio=5 os_prio=0 tid=0x00007fb91498d000 nid=0x474a runnable [0x00007fb9065fe000] java.lang.Thread.State: RUNNABLE at Test$2.run(Test.java:18)

"Thread-9" #27 prio=5 os_prio=0 tid=0x00007fb91498c800 nid=0x4749 waiting on condition [0x00007fb906bfe000] java.lang.Thread.State: TIMED_WAITING (sleeping) at java.lang.Thread.sleep(Native Method) at Test$1.run(Test.java:9)

"Thread-8" #26 prio=5 os_prio=0 tid=0x00007fb91498b800 nid=0x4748 waiting on condition [0x00007fb906ffe000] java.lang.Thread.State: TIMED_WAITING (sleeping) at java.lang.Thread.sleep(Native Method) at Test$1.run(Test.java:9)

"Thread-7" #25 prio=5 os_prio=0 tid=0x00007fb91498b000 nid=0x4747 waiting on condition [0x00007fb9073fe000] java.lang.Thread.State: TIMED_WAITING (sleeping) at java.lang.Thread.sleep(Native Method) at Test$1.run(Test.java:9)

"Thread-6" #24 prio=5 os_prio=0 tid=0x00007fb91498a000 nid=0x4746 waiting on condition [0x00007fb9077fe000] java.lang.Thread.State: TIMED_WAITING (sleeping) at java.lang.Thread.sleep(Native Method) at Test$1.run(Test.java:9) ... 上面的线程栈我们注意到nid的值其实就是线程ID，它是十六进制的，我们将消耗cpu最高的线程18250，转成十六进制0X47A，然后从上面的线程栈里找到nid=0X47A的线程，其栈为：

"Busiest Thread" #28 prio=5 os_prio=0 tid=0x00007fb91498d000 nid=0x474a runnable [0x00007fb9065fe000] java.lang.Thread.State: RUNNABLE at Test$2.run(Test.java:18) 即将最耗cpu的线程找出来了，是Businest Thread

fullgc cpu飙高的部分排查思路学习：

a) top -Hp pid找出最耗cpu的线程Id 看看是业务线程还是gc线程 b) 查询gc日志[Full GC [PSYoungGen: 年轻代回收前大小->年轻代回收后大小(年轻代内存总大小)] [ParOldGen: 老年代回收前大小->老年代回收后大小(老年代总空间)] 回收前堆大小->回收后堆大小堆的总大小) 看看gc后数据是否可以gc掉 (日志位于tomcat等容器的日志目录下如gc.log)

5.617(时间戳): [GC(Young GC) 5.617(时间戳): [ParNew(使用ParNew作为年轻代的垃圾回收期): 43296K(年轻代垃圾回收前的大小)->7006K(年轻代垃圾回收以后的大小)(47808K)(年轻代的总大小), 0.0136826 secs(回收时间)] 44992K(堆区垃圾回收前的大小)->8702K(堆区垃圾回收后的大小)(252608K)(堆区总大小), 0.0137904 secs(回收时间)] [Times: user=0.03(Young GC用户耗时) sys=0.00(Young GC系统耗时), real=0.02 secs(Young GC实际耗时)

c) gc不掉的话 看看哪些对象占用内存比较多 jmap -dump:format=b,file=heap.hprof pid 用mat工具分析或者jmap -histo pid(更简单)然后再看看这些对象是怎么出来的 比如导出excel或者sql里查太多数据了 d) 可以gc掉的话可以考虑下jvm内存是否需要调整