(转)Linux下java进程CPU占用率高-分析方法

Linux下java进程CPU占用率高-分析方法

原文：http://itindex.net/detail/47420-linux-java-%E8%BF%9B%E7%A8%8B?utm_source=tuicool&utm_medium=referral

今天登陆同事的一台gateway 开始以为hive环境登陆不了了，仔细一看看了下是因为机器很卡，我每次等几秒没登陆就ctrl+c了，看了下是有个java进程cpu:340.4%  mem:14.6% 
一般解决方法是通过top命令找出消耗资源高的线程id，利用strace命令查看该线程所有系统调用 
1. 通过top命令找到可疑进程PID 
top 一下 
可以看出java进程CPU利用率一直保持100%，稳居不下，找到PID 24138 
2. 找出消耗资源最高的线程 
top -H -p  29580  可以不用第一步，直接执行命令 top -H ，就可以查看到消耗资源最高的线程 
top - 20:42:01 up 633 days,  9:30,  9 users,  load average: 6.75, 8.32, 15.86 
Tasks:  28 total,   2 running,  26 sleeping,   0 stopped,   0 zombie 
Cpu(s): 42.4%us,  4.3%sy,  0.0%ni, 53.1%id,  0.0%wa,  0.0%hi,  0.1%si,  0.1%st 
Mem:   7680000k total,  5774940k used,  1905060k free,   400792k buffers 
Swap:  2096472k total,   876580k used,  1219892k free,  1727652k cached 
  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND                                                                                                            
29679 baishou   16   0 1560m 1.1g  18m R 72.0 15.3 115:41.08 java                                                                                                               
29678 baishou   15   0 1560m 1.1g  18m R 63.3 15.3 118:44.99 java                                                                                                               
29673 baishou   15   0 1560m 1.1g  18m S  1.0 15.3   0:59.72 java                                                                                                               
29677 baishou   15   0 1560m 1.1g  18m S  1.0 15.3   1:01.34 java                                                                                                               
3. 查看这个线程所有系统调用 
strace -p 29679 

read(114, "22561043p3651_4"..., 2064) = 149
write(114, "733363201r4 B25274252*275."..., 2011) = 2011
write(114, "73336Adddddc2301!4302dQQ3300R373300c"..., 2011) = 2011
write(114, "73336000630561JdK	tb15218410101200"..., 2011) = 2011
write(114, "73336000630561K627	tb15331130101200"..., 2011) = 2011
write(114, "73336<|0630561L2320	tb15184371"..., 2011) = 2011
write(114, "73336010120012002301)1200230121200120012001200"..., 2011) = 2011
write(114, "346200120012001200
3003120F1027#2323)", 28) = 28
read(114, "22561043q3651_4"..., 2064) = 149
write(114, "733363201s4 B25274252*311."..., 2011) = 2011

发现有大量写操作，应该是由datax任务在跑。 
+++++++++++++++++++++++++++++++++++++++++++++++++++ 
++++++++++++++如果是web应用，可以继续打印线程的堆栈信息+++++++++ 
将需要的线程ID转换为16进制格式： 
printf "%x " 29679 
73ef 
最后打印线程的堆栈信息： 
jstack 29679|grep 73ef -A 30