关于linux的进程中的各个线程cpu占用情况的分析和查看
我们常常会在新开的服搭建一个游戏的server,有时候要进行压力測试,那么怎样来看呢,一般我们会通过top命令查看各个进程的cpu和内存占用情况,获得到了我们的进程id,然后我们或许会通过pstack命令查看里边的各个线程id以及相应的线程如今正在做什么事情,分析多组数据就能够获得哪些线程里有慢操作影响了server的性能,从而得到解决方式。
比方这种以组数据:
[root@AY130816144542124256Z bin]# pstack 30222
Thread 9 (Thread 0x7f729adc1700 (LWP 30251)):
#0 0x00007f72a429b720 in sem_wait () from /lib64/libpthread.so.0
#1 0x0000000000ac5eb6 in Semaphore::down() ()
#2 0x0000000000ac5cac in Queue::get() ()
#3 0x00000000009a583f in DBManager::processUpdate(Queue*) ()
#4 0x00000000009a4bfb in dbUpdateThread(void*) ()
#5 0x00007f72a4295851 in start_thread () from /lib64/libpthread.so.0
#6 0x00007f72a459267d in clone () from /lib64/libc.so.6
Thread 8 (Thread 0x7f727c579700 (LWP 30252)):
#0 0x00007f72a429b720 in sem_wait () from /lib64/libpthread.so.0
#1 0x0000000000ac5eb6 in Semaphore::down() ()
#2 0x0000000000ac5cac in Queue::get() ()
#3 0x00000000009a5799 in DBManager::processQuery(Queue*) ()
#4 0x00000000009a4c3a in dbQueryThread(void*) ()
#5 0x00007f72a4295851 in start_thread () from /lib64/libpthread.so.0
#6 0x00007f72a459267d in clone () from /lib64/libc.so.6
Thread 7 (Thread 0x7f7257fff700 (LWP 30253)):
#0 0x00007f72a42997bb in pthread_cond_timedwait@@GLIBC_2.3.2 () from /lib64/libpthread.so.0
#1 0x00007f72a549ee08 in utils::thread::condition_impl::timed_wait(int) () from /usr/local/utils-0.0.1/lib/libutils.so.0.0.1
#2 0x00007f72a549ebd3 in utils::thread::Condition::timed_wait(int) () from /usr/local/utils-0.0.1/lib/libutils.so.0.0.1
#3 0x00000000009d5f57 in utils::MessageQueue<FightInfo*>::pop() ()
#4 0x00000000009d5557 in FightReport::svc() ()
#5 0x00007f72a5494b45 in utils::Task_Base::thread_proc(void*) () from /usr/local/utils-0.0.1/lib/libutils.so.0.0.1
#6 0x00007f72a4295851 in start_thread () from /lib64/libpthread.so.0
#7 0x00007f72a459267d in clone () from /lib64/libc.so.6
Thread 6 (Thread 0x7f72397b7700 (LWP 30254)):
#0 0x00007f72a4588fc3 in poll () from /lib64/libc.so.6
#1 0x00007f72a0fbded4 in __libc_res_nsend () from /lib64/libresolv.so.2
#2 0x00007f72a0fba76a in __libc_res_nquery () from /lib64/libresolv.so.2
#3 0x00007f72a0fbad29 in __libc_res_nquerydomain () from /lib64/libresolv.so.2
#4 0x00007f72a0fbb9cf in __libc_res_nsearch () from /lib64/libresolv.so.2
#5 0x00007f729adc37a7 in _nss_dns_gethostbyname4_r () from /lib64/libnss_dns.so.2
#6 0x00007f72a457a4c3 in gaih_inet () from /lib64/libc.so.6
#7 0x00007f72a457cb20 in getaddrinfo () from /lib64/libc.so.6
#8 0x00007f72a56fc782 in Curl_getaddrinfo_ex () from /usr/lib64/libcurl.so.4
#9 0x00007f72a56f1d42 in Curl_getaddrinfo () from /usr/lib64/libcurl.so.4
#10 0x00007f72a56c9e77 in Curl_resolv () from /usr/lib64/libcurl.so.4
#11 0x00007f72a56ca138 in Curl_resolv_timeout () from /usr/lib64/libcurl.so.4
#12 0x00007f72a56d8d88 in ?
? () from /usr/lib64/libcurl.so.4
#13 0x00007f72a56ddb79 in ?
?
() from /usr/lib64/libcurl.so.4
#14 0x00007f72a56de76e in Curl_connect () from /usr/lib64/libcurl.so.4
#15 0x00007f72a56e69b0 in Curl_perform () from /usr/lib64/libcurl.so.4
#16 0x0000000000ae6e3d in HttpClient::svc() ()
#17 0x00007f72a5494b45 in utils::Task_Base::thread_proc(void*) () from /usr/local/utils-0.0.1/lib/libutils.so.0.0.1
#18 0x00007f72a4295851 in start_thread () from /lib64/libpthread.so.0
#19 0x00007f72a459267d in clone () from /lib64/libc.so.6
Thread 5 (Thread 0x7f721af6f700 (LWP 30255)):
#0 0x00007f72a455691d in nanosleep () from /lib64/libc.so.6
#1 0x000000000098cb8a in Sleep(unsigned long) ()
#2 0x000000000098b87d in DynResource::svc() ()
#3 0x00007f72a5494b45 in utils::Task_Base::thread_proc(void*) () from /usr/local/utils-0.0.1/lib/libutils.so.0.0.1
#4 0x00007f72a4295851 in start_thread () from /lib64/libpthread.so.0
#5 0x00007f72a459267d in clone () from /lib64/libc.so.6
Thread 4 (Thread 0x7f71fc727700 (LWP 30256)):
#0 0x00007f72a455691d in nanosleep () from /lib64/libc.so.6
#1 0x000000000098cb8a in Sleep(unsigned long) ()
#2 0x0000000000a61516 in PlayerOpLogThread::svc() ()
#3 0x00007f72a5494b45 in utils::Task_Base::thread_proc(void*) () from /usr/local/utils-0.0.1/lib/libutils.so.0.0.1
#4 0x00007f72a4295851 in start_thread () from /lib64/libpthread.so.0
#5 0x00007f72a459267d in clone () from /lib64/libc.so.6
Thread 3 (Thread 0x7f71ddedf700 (LWP 30257)):
#0 0x00007f72a4592c73 in epoll_wait () from /lib64/libc.so.6
#1 0x00007f72a51f334f in Epoll_Reactor::run_reactor_event_loop() () from /usr/local/net_manager-0.0.2/lib/libnet_manager.so.0.0.2
#2 0x00007f72a51f2523 in Net_Thread::svc() () from /usr/local/net_manager-0.0.2/lib/libnet_manager.so.0.0.2
#3 0x00007f72a5494b45 in utils::Task_Base::thread_proc(void*) () from /usr/local/utils-0.0.1/lib/libutils.so.0.0.1
#4 0x00007f72a4295851 in start_thread () from /lib64/libpthread.so.0
#5 0x00007f72a459267d in clone () from /lib64/libc.so.6
Thread 2 (Thread 0x7f71bf697700 (LWP 30258)):
#0 0x00007f72a4592c73 in epoll_wait () from /lib64/libc.so.6
#1 0x00007f72a51f334f in Epoll_Reactor::run_reactor_event_loop() () from /usr/local/net_manager-0.0.2/lib/libnet_manager.so.0.0.2
#2 0x00007f72a51f2523 in Net_Thread::svc() () from /usr/local/net_manager-0.0.2/lib/libnet_manager.so.0.0.2
#3 0x00007f72a5494b45 in utils::Task_Base::thread_proc(void*) () from /usr/local/utils-0.0.1/lib/libutils.so.0.0.1
#4 0x00007f72a4295851 in start_thread () from /lib64/libpthread.so.0
#5 0x00007f72a459267d in clone () from /lib64/libc.so.6
Thread 1 (Thread 0x7f72a60ae7e0 (LWP 30222)):
#0 0x00007f72a4584c95 in _xstat () from /lib64/libc.so.6
#1 0x00007f72a45483e0 in __tzfile_read () from /lib64/libc.so.6
#2 0x00007f72a4547864 in tzset_internal () from /lib64/libc.so.6
#3 0x00007f72a4547b20 in tzset () from /lib64/libc.so.6
#4 0x00007f72a4546699 in timelocal () from /lib64/libc.so.6
#5 0x0000000000b0b08d in Achieve::GetRemainTime(AchieveTemplate*) ()
#6 0x0000000000b115ca in Achieve::update() ()
#7 0x0000000000a197ce in Player::update() ()
#8 0x0000000000b1b272 in PlayerMng::Tick() ()
#9 0x0000000000a73105 in GameServer::FrameTick(unsigned int) ()
#10 0x0000000000a6ff80 in GameServer::run() ()
#11 0x0000000000a773a1 in main ()
[root@AY130816144542124256Z gameserver]# ps -eLo pid,lwp,pcpu | grep 30222
30222 30222 31.4
30222 30251 0.0
30222 30252 0.0
30222 30253 0.0
30222 30254 0.0
30222 30255 0.0
30222 30256 1.2
30222 30257 1.2
30222 30258 1.0
多组数据显示表明我们的主逻辑线程的确占用的cpu非常高,发现事实上在Achieve::update() 的时候做了太多没实用的推断,而且能够降低循环进入的次数的。所以我在这里进行了优化
后来用了一下iptraf命令查看了网络流量:
第一项:IP流量监控
第二项:常规查看网卡流量状态。仅仅查看各网卡的总流量
第三项:具体查看网卡流量状态。比方按TCP,UDP,ARP等协议查看
选all interfaces,查看全部网卡接口
界面分上下两部分,上部分可具体显示哪个与之相连的IP。发了多少包,即时流量是多少,下部分,能够显示udp等信息。
显示连接本地server的全部连接流量情况
另外不得不提到另外一个命令sar:
sar有非常多用途。假设要来监控网络流量。使用以下的命令行方式:
sar -n DEV interval count
当中。interval是统计时间间隔。以秒为单位;count是总共统计几次。假设为0就不断的统计直到 Ctrl+C 打断,否则运行count次就退出。
比方:sar -n DEV 1 4
比方:sar -n DEV 10 0
IFACE:LAN接口
rxpck/s:每秒钟接收的数据包
txpck/s:每秒钟发送的数据包
rxbyt/s:每秒钟接收的字节数
txbyt/s:每秒钟发送的字节数
sar -n DEV -u 1 10
07:22:01 PM IFACE rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s
07:22:02 PM lo 0.00 0.00 0.00 0.00 0.00 0.00 0.00
07:22:02 PM eth0 2.47 0.00 0.11 0.00 0.00 0.00 0.00
07:22:02 PM eth1 80.25 22.22 3.53 3.52 0.00 0.00 0.00
07:22:02 PM CPU %user %nice %system %iowait %steal %idle
07:22:03 PM all 1.85 0.00 1.06 0.00 0.00 97.09
07:22:02 PM IFACE rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s
07:22:03 PM lo 0.00 0.00 0.00 0.00 0.00 0.00 0.00
07:22:03 PM eth0 0.00 0.00 0.00 0.00 0.00 0.00 0.00
07:22:03 PM eth1 66.67 19.75 2.95 3.22 0.00 0.00 0.00
07:22:03 PM CPU %user %nice %system %iowait %steal %idle
07:22:04 PM all 2.12 0.00 1.06 0.00 0.00 96.82
07:22:03 PM IFACE rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s
07:22:04 PM lo 0.00 0.00 0.00 0.00 0.00 0.00 0.00
07:22:04 PM eth0 7.32 4.88 0.65 0.38 0.00 0.00 0.00
07:22:04 PM eth1 85.37 28.05 4.18 3.88 0.00 0.00 0.00
07:22:04 PM CPU %user %nice %system %iowait %steal %idle
07:22:05 PM all 3.96 0.00 2.64 0.00 0.00 93.40
07:22:04 PM IFACE rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s
07:22:05 PM lo 412.50 412.50 104.44 104.44 0.00 0.00 0.00
07:22:05 PM eth0 13.75 12.50 1.32 1.26 0.00 0.00 0.00
07:22:05 PM eth1 903.75 707.50 174.12 62.42 0.00 0.00 0.00
07:22:05 PM CPU %user %nice %system %iowait %steal %idle
07:22:06 PM all 5.25 0.00 4.46 2.10 0.00 88.19
07:22:05 PM IFACE rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s
07:22:06 PM lo 849.41 849.41 154.54 154.54 0.00 0.00 0.00
07:22:06 PM eth0 0.00 0.00 0.00 0.00 0.00 0.00 0.00
07:22:06 PM eth1 1287.06 1337.65 84.48 461.76 0.00 0.00 0.00
07:22:06 PM CPU %user %nice %system %iowait %steal %idle
07:22:07 PM all 1.85 0.00 1.06 0.00 0.00 97.09
07:22:06 PM IFACE rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s
07:22:07 PM lo 0.00 0.00 0.00 0.00 0.00 0.00 0.00
07:22:07 PM eth0 4.94 0.00 0.22 0.00 0.00 0.00 0.00
07:22:07 PM eth1 97.53 19.75 4.32 3.18 0.00 0.00 0.00
07:22:07 PM CPU %user %nice %system %iowait %steal %idle
07:22:08 PM all 1.58 0.00 1.32 0.00 0.00 97.11
07:22:07 PM IFACE rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s
07:22:08 PM lo 4.94 4.94 0.31 0.31 0.00 0.00 0.00
07:22:08 PM eth0 1.23 0.00 0.06 0.00 0.00 0.00 0.00
07:22:08 PM eth1 67.90 24.69 3.02 3.94 0.00 0.00 0.00
Average: CPU %user %nice %system %iowait %steal %idle
Average: all 3.63 0.00 2.55 0.84 0.00 92.98
Average: IFACE rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s
Average: lo 212.61 212.61 50.54 50.54 0.00 0.00 0.00
Average: eth0 42.79 41.70 3.27 41.84 0.00 0.00 0.00
Average: eth1 492.48 432.48 53.63 110.40 0.00 0.00 0.00
关于linux的进程中的各个线程cpu占用情况的分析和查看的更多相关文章
- 查看java线程cpu占用情况的脚本
#!/bin/bash [ $# -ne ] && exit jstack $ >/tmp/jstack.log -o THREAD,tid,time|sort -k2nr| s ...
- 找出进程中各线程cpu消耗情况
以root用户执行以下命令,以PID 5423举例: 1,根据top命令,找到占用CPU高的进程,找到PID PID USER PR NI VIRT RES SHR S %CPU ...
- 命令行分析java线程CPU占用
1.使用top命令找出占用cpu最高的JAVA进程pid号 2. 找出占用cpu最高的线程: top -Hp -n 1 3. 打印占CPU最高JAVA进程pid的堆栈信息 jstack pid &g ...
- Linux多线程编程的时候怎么查看一个进程中的某个线程是否存活
pthread_kill: 别被名字吓到,pthread_kill可不是kill,而是向线程发送signal.还记得signal吗,大部分signal的默认动作是终止进程的运行,所以,我们才要用sig ...
- Linux获取进程中变量
列出所有进程 #include <linux/kernel.h> #include <linux/module.h> #include <linux/init.h> ...
- linux下top命令查看cpu占用情况
可以通过 top 命令来查看 CPU 使用状况.运行 top 命令后,CPU 使用状态会以全屏的方式显示,并且会处在对话的模式 -- 用基于 top 的命令,可以控制显示方式等等.退出 top 的命令 ...
- systemd --user进程CPU占用高问题分析
1.问题由来 近期发现堡垒机环境有如下问题,systemd占用大量cpu: 原文链接:https://www.cnblogs.com/yaohong/p/16046670.html 2.问题定位 2. ...
- 查询进程内存,cpu占用情况。僵尸进程
查使用内存最多的5个进程:ps aux | head -1 && ps aux | grep -v USER | sort -nr -k 4 | head -5 查使用CPU最多的5个 ...
- 【Linux 线程】同一个进程中的线程共享哪些资源
进程是具有一定独立功能的程序关于某个数据集合上的一次运行活动,进程是系统进行资源分配和调度的一个独立单位. 线程是进程的一个实体,是CPU调度和分派的基本单位,它是比进程更小的能独立运行的基本单位.线 ...
随机推荐
- hdu 5875(单调栈)
Function Time Limit: 7000/3500 MS (Java/Others) Memory Limit: 262144/262144 K (Java/Others)Total ...
- python模拟QQ聊天室(tcp加多线程)
python模拟QQ聊天室(tcp加多线程) 服务器代码: from socket import * from threading import * s = socket(AF_INET,SOCK_S ...
- 修改input中的placeholder属性的颜色
input::-webkit-input-placeholder{ color:#e8e8e8; } input::-moz-placeholder{ /* Mozilla Firefox 19+ * ...
- c# winform读取及发送串口信号
请参考C#的API文档:https://docs.microsoft.com/zh-cn/dotnet/api/system.io.ports.serialport?redirectedfrom=MS ...
- 验证码无法显示:Could not initialize class sun.awt.X11GraphicsEnvironment 解决方案
一.原因现象:图下图 二.原因导致: 经过Google发现很多人也出现同样的问题.从了解了X11GraphicEnvironment这个类的功能入手,一个Java服务器来处理图片的API基本上是需要运 ...
- MYSQL注入天书之开天辟地
MYSQL注入天书 在线版本:xianzhi.aliyun.com 第一篇地址:https://xianzhi.aliyun.com/forum/read/314.html第二篇地址:https:// ...
- Javascript 中的神器
Promise in js 回调函数真正的问题在于他剥夺了我们使用 return 和 throw 这些关键字的能力.而 Promise 很好地解决了这一切. 2015 年 6 月,ECMAScript ...
- hdu 2433 Travel(还不会)
Problem Description One day, Tom traveled to a country named BGM. BGM is a small country, but ...
- thunk 函数
function* f() { console.log(1); for (var i = 0; true; i++) { console.log('come in'); var reset = yie ...
- 【BZOJ 4229】 4229: 选择 (线段树+树链剖分)
4229: 选择 Time Limit: 10 Sec Memory Limit: 128 MBSubmit: 67 Solved: 41 Description 现在,我想知道自己是否还有选择. ...