今天后台组发现新上线的应用CPU总是会占用过高。(心里暗骂,新来的运维真无聊,闲着没事看top干啥)
首先发送命令 jps -lv 查询运行的进程pid=18182,
[admin@HCX-SER04 service-8072]$ jps 19017 Jps 18827 jar
然后通过top -Hp 18182,查看哪个线程占用CPU过高。
[admin@HCX-SER04 ~]$ top -Hp 18827 top - 18:22:02 up 135 days, 4:09, 8 users, load average: 0.14, 0.16, 0.07 Tasks: 137 total, 1 running, 136 sleeping, 0 stopped, 0 zombie Cpu(s): 43.6%us, 5.2%sy, 0.0%ni, 50.2%id, 0.0%wa, 0.0%hi, 1.0%si, 0.0%st Mem: 2054152k total, 1816632k used, 237520k free, 30808k buffers Swap: 4194300k total, 0k used, 4194300k free, 415516k cached PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 18838 admin 20 0 3373m 1.1g 15m R 39.9 55.9 0:22.16 java 18871 admin 20 0 3373m 1.1g 15m S 4.7 55.9 0:04.73 java 18881 admin 20 0 3373m 1.1g 15m S 0.3 55.9 0:00.05 java 18905 admin 20 0 3373m 1.1g 15m S 0.3 55.9 0:00.05 java
将线程ID转为16进制
[admin@HCX-SER04 ~]$ python -c "print hex(18838)" 0x4996
然后使用命令jstack 18827 做线程dump
[admin@HCX-SER04 ~]$ jstack 18827 > 18827.log [admin@HCX-SER04 ~]$
后面就简单了,把线程dump文件直接less打开,找到占用高的堆栈信息(这里是十六进制的线程ID)。
(END) "C2 CompilerThread0" #8 daemon prio=9 os_prio=0 tid=0x00007fb6e410c000 nid=0x4996 waiting on condition [0x0000000000000000] java.lang.Thread.State: RUNNABLE "JDWP Event Helper Thread" #7 daemon prio=10 os_prio=0 tid=0x00007fb6e4109800 nid=0x4995 runnable [0x0000000000000000] java.lang.Thread.State: RUNNABLE
原来是C2 CompilerThread。呵呵,这下简单了原来是C2编译器导致的。凡是服务器重启后,java都会把class文件进行编译优化,没办法,告诉他们这是无法避免的,java程序就是这样,上线还是晚上吧,免的服务抖动,大惊小怪。;)