如何在Python中获取当前的CPU和RAM使用率？

Question

在Python中获取当前系统状态(当前CPU,RAM,可用磁盘空间等)的首选方法是什么？*nix和Windows平台的奖励积分.

似乎有一些可能的方法从我的搜索中提取:

1. 使用像PSI这样的库(目前似乎没有积极开发并且在多个平台上不受支持)或像pystatgrab这样的东西(自2007年以来再没有活动似乎也不支持Windows).

2. 使用特定于平台的代码,例如os.popen("ps")在*nix系统中使用或类似代码,以及MEMORYSTATUS在Windows平台上使用ctypes.windll.kernel32(参见ActiveState上的此配方).可以将Python类与所有这些代码片段放在一起.

并不是说这些方法很糟糕,但是已经有一个支持良好的多平台方式来做同样的事情了吗？

Answer 1

psutil库将在各种平台上为您提供一些系统信息(CPU /内存使用情况):

psutil是一个模块,提供了一个接口,用于通过使用Python以可移植的方式检索有关正在运行的进程和系统利用率(CPU,内存)的信息,实现ps,top和Windows任务管理器等工具提供的许多功能.

它目前支持Linux,Windows,OSX,Sun Solaris,FreeBSD,OpenBSD和NetBSD,32位和64位架构,Python版本从2.6到3.5(Python 2.4和2.5的用户可能使用2.1.3版本).

更新:以下是一些示例用法psutil:

#!/usr/bin/env python
import psutil
# gives a single float value
psutil.cpu_percent()
# gives an object with many fields
psutil.virtual_memory()
# you can convert that object to a dictionary 
dict(psutil.virtual_memory()._asdict())

Answer 2

tqdm通过组合和可以得到实时 CPU 和 RAM 监控psutil。在运行繁重的计算/处理时它可能很方便。

它也可以在 Jupyter 中运行，无需更改任何代码：

from tqdm import tqdm
from time import sleep
import psutil

with tqdm(total=100, desc='cpu%', position=1) as cpubar, tqdm(total=100, desc='ram%', position=0) as rambar:
    while True:
        rambar.n=psutil.virtual_memory().percent
        cpubar.n=psutil.cpu_percent()
        rambar.refresh()
        cpubar.refresh()
        sleep(0.5)

使用多处理库将这些进度条放在单独的进程中很方便。

该代码片段也可以作为要点提供。

Answer 3

使用psutil库.在Ubuntu 18.04上,从1-30-2019开始,pip安装了5.5.0(最新版本).较旧的版本可能会有所不同.您可以通过在Python中执行此操作来检查您的psutil版本:

from __future__ import print_function  # for Python2
import psutil
print(psutil.__versi??on__)

获取一些内存和CPU统计信息:

from __future__ import print_function
import psutil
print(psutil.cpu_percent())
print(psutil.virtual_memory())  # physical memory usage
print('memory % used:', psutil.virtual_memory()[2])

所述virtual_memory(元组)将具有使用全系统的百分比存储器.在Ubuntu 18.04上,我似乎高估了几个百分点.

您还可以获取当前Python实例使用的内存:

import os
import psutil
pid = os.getpid()
py = psutil.Process(pid)
memoryUse = py.memory_info()[0]/2.**30  # memory use in GB...I think
print('memory use:', memoryUse)

它给出了Python脚本当前内存的使用.

psutil的pypi页面上有一些更深入的例子.

Answer 4

仅使用stdlib依赖的RAM使用的单线程:

import os
tot_m, used_m, free_m = map(int, os.popen('free -t -m').readlines()[-1].split()[1:])

Answer 5

下面的代码,没有外部库为我工作.我在Python 2.7.9上测试过

CPU使用率

import os

    CPU_Pct=str(round(float(os.popen('''grep 'cpu ' /proc/stat | awk '{usage=($2+$4)*100/($2+$4+$5)} END {print usage }' ''').readline()),2))

    #print results
    print("CPU Usage = " + CPU_Pct)

和Ram使用,总计,使用和免费

import os
mem=str(os.popen('free -t -m').readlines())
"""
Get a whole line of memory output, it will be something like below
['             total       used       free     shared    buffers     cached\n', 
'Mem:           925        591        334         14         30        355\n', 
'-/+ buffers/cache:        205        719\n', 
'Swap:           99          0         99\n', 
'Total:        1025        591        434\n']
 So, we need total memory, usage and free memory.
 We should find the index of capital T which is unique at this string
"""
T_ind=mem.index('T')
"""
Than, we can recreate the string with this information. After T we have,
"Total:        " which has 14 characters, so we can start from index of T +14
and last 4 characters are also not necessary.
We can create a new sub-string using this information
"""
mem_G=mem[T_ind+14:-4]
"""
The result will be like
1025        603        422
we need to find first index of the first space, and we can start our substring
from from 0 to this index number, this will give us the string of total memory
"""
S1_ind=mem_G.index(' ')
mem_T=mem_G[0:S1_ind]
"""
Similarly we will create a new sub-string, which will start at the second value. 
The resulting string will be like
603        422
Again, we should find the index of first space and than the 
take the Used Memory and Free memory.
"""
mem_G1=mem_G[S1_ind+8:]
S2_ind=mem_G1.index(' ')
mem_U=mem_G1[0:S2_ind]

mem_F=mem_G1[S2_ind+8:]
print 'Summary = ' + mem_G
print 'Total Memory = ' + mem_T +' MB'
print 'Used Memory = ' + mem_U +' MB'
print 'Free Memory = ' + mem_F +' MB'

Answer 6

要获得程序的逐行内存和时间分析，我建议使用memory_profiler和line_profiler。

安装：

# Time profiler
$ pip install line_profiler
# Memory profiler
$ pip install memory_profiler
# Install the dependency for a faster analysis
$ pip install psutil

共同的部分是，您可以使用相应的装饰器指定要分析的函数。

示例：我的 Python 文件main.py中有几个要分析的函数。其中之一是linearRegressionfit()。我需要使用装饰器@profile来帮助我分析关于两个方面的代码：时间和内存。

对函数定义进行以下更改

@profile
def linearRegressionfit(Xt,Yt,Xts,Yts):
    lr=LinearRegression()
    model=lr.fit(Xt,Yt)
    predict=lr.predict(Xts)
    # More Code

对于时间分析，

跑：

$ kernprof -l -v main.py

输出

Total time: 0.181071 s
File: main.py
Function: linearRegressionfit at line 35

Line #      Hits         Time  Per Hit   % Time  Line Contents
==============================================================
    35                                           @profile
    36                                           def linearRegressionfit(Xt,Yt,Xts,Yts):
    37         1         52.0     52.0      0.1      lr=LinearRegression()
    38         1      28942.0  28942.0     75.2      model=lr.fit(Xt,Yt)
    39         1       1347.0   1347.0      3.5      predict=lr.predict(Xts)
    40                                           
    41         1       4924.0   4924.0     12.8      print("train Accuracy",lr.score(Xt,Yt))
    42         1       3242.0   3242.0      8.4      print("test Accuracy",lr.score(Xts,Yts))

对于内存分析，

跑：

$ python -m memory_profiler main.py

输出

Filename: main.py

Line #    Mem usage    Increment   Line Contents
================================================
    35  125.992 MiB  125.992 MiB   @profile
    36                             def linearRegressionfit(Xt,Yt,Xts,Yts):
    37  125.992 MiB    0.000 MiB       lr=LinearRegression()
    38  130.547 MiB    4.555 MiB       model=lr.fit(Xt,Yt)
    39  130.547 MiB    0.000 MiB       predict=lr.predict(Xts)
    40                             
    41  130.547 MiB    0.000 MiB       print("train Accuracy",lr.score(Xt,Yt))
    42  130.547 MiB    0.000 MiB       print("test Accuracy",lr.score(Xts,Yts))

此外，还可以使用以下matplotlib方法绘制内存分析器结果

$ mprof run main.py
$ mprof plot

注意：经过测试

line_profiler 版本 == 3.0.2

memory_profiler 版本 == 0.57.0

psutil 版本 == 5.7.0

编辑：可以使用TAMPPA包解析分析器的结果。使用它，我们可以获得逐行所需的图

Answer 7

这是我刚才放在一起的东西,它只是窗户,但可以帮助你获得你需要做的一部分.

源自:"for sys available mem" http://msdn2.microsoft.com/en-us/library/aa455130.aspx

"单个进程信息和python脚本示例" http://www.microsoft.com/technet/scriptcenter/scripts/default.mspx?mfr=true

注意:WMI接口/进程也可用于执行我在此处未使用的类似任务,因为当前方法涵盖了我的需求,但是如果有一天需要扩展或改进它,那么可能需要调查WMI工具.

python的WMI:

http://tgolden.sc.sabren.com/python/wmi.html

代码:

'''
Monitor window processes

derived from:
>for sys available mem
http://msdn2.microsoft.com/en-us/library/aa455130.aspx

> individual process information and python script examples
http://www.microsoft.com/technet/scriptcenter/scripts/default.mspx?mfr=true

NOTE: the WMI interface/process is also available for performing similar tasks
        I'm not using it here because the current method covers my needs, but if someday it's needed
        to extend or improve this module, then may want to investigate the WMI tools available.
        WMI for python:
        http://tgolden.sc.sabren.com/python/wmi.html
'''

__revision__ = 3

import win32com.client
from ctypes import *
from ctypes.wintypes import *
import pythoncom
import pywintypes
import datetime


class MEMORYSTATUS(Structure):
    _fields_ = [
                ('dwLength', DWORD),
                ('dwMemoryLoad', DWORD),
                ('dwTotalPhys', DWORD),
                ('dwAvailPhys', DWORD),
                ('dwTotalPageFile', DWORD),
                ('dwAvailPageFile', DWORD),
                ('dwTotalVirtual', DWORD),
                ('dwAvailVirtual', DWORD),
                ]


def winmem():
    x = MEMORYSTATUS() # create the structure
    windll.kernel32.GlobalMemoryStatus(byref(x)) # from cytypes.wintypes
    return x    


class process_stats:
    '''process_stats is able to provide counters of (all?) the items available in perfmon.
    Refer to the self.supported_types keys for the currently supported 'Performance Objects'

    To add logging support for other data you can derive the necessary data from perfmon:
    ---------
    perfmon can be run from windows 'run' menu by entering 'perfmon' and enter.
    Clicking on the '+' will open the 'add counters' menu,
    From the 'Add Counters' dialog, the 'Performance object' is the self.support_types key.
    --> Where spaces are removed and symbols are entered as text (Ex. # == Number, % == Percent)
    For the items you wish to log add the proper attribute name in the list in the self.supported_types dictionary,
    keyed by the 'Performance Object' name as mentioned above.
    ---------

    NOTE: The 'NETFramework_NETCLRMemory' key does not seem to log dotnet 2.0 properly.

    Initially the python implementation was derived from:
    http://www.microsoft.com/technet/scriptcenter/scripts/default.mspx?mfr=true
    '''
    def __init__(self,process_name_list=[],perf_object_list=[],filter_list=[]):
        '''process_names_list == the list of all processes to log (if empty log all)
        perf_object_list == list of process counters to log
        filter_list == list of text to filter
        print_results == boolean, output to stdout
        '''
        pythoncom.CoInitialize() # Needed when run by the same process in a thread

        self.process_name_list = process_name_list
        self.perf_object_list = perf_object_list
        self.filter_list = filter_list

        self.win32_perf_base = 'Win32_PerfFormattedData_'

        # Define new datatypes here!
        self.supported_types = {
                                    'NETFramework_NETCLRMemory':    [
                                                                        'Name',
                                                                        'NumberTotalCommittedBytes',
                                                                        'NumberTotalReservedBytes',
                                                                        'NumberInducedGC',    
                                                                        'NumberGen0Collections',
                                                                        'NumberGen1Collections',
                                                                        'NumberGen2Collections',
                                                                        'PromotedMemoryFromGen0',
                                                                        'PromotedMemoryFromGen1',
                                                                        'PercentTimeInGC',
                                                                        'LargeObjectHeapSize'
                                                                     ],

                                    'PerfProc_Process':              [
                                                                          'Name',
                                                                          'PrivateBytes',
                                                                          'ElapsedTime',
                                                                          'IDProcess',# pid
                                                                          'Caption',
                                                                          'CreatingProcessID',
                                                                          'Description',
                                                                          'IODataBytesPersec',
                                                                          'IODataOperationsPersec',
                                                                          'IOOtherBytesPersec',
                                                                          'IOOtherOperationsPersec',
                                                                          'IOReadBytesPersec',
                                                                          'IOReadOperationsPersec',
                                                                          'IOWriteBytesPersec',
                                                                          'IOWriteOperationsPersec'     
                                                                      ]
                                }

    def get_pid_stats(self, pid):
        this_proc_dict = {}

        pythoncom.CoInitialize() # Needed when run by the same process in a thread
        if not self.perf_object_list:
            perf_object_list = self.supported_types.keys()

        for counter_type in perf_object_list:
            strComputer = "."
            objWMIService = win32com.client.Dispatch("WbemScripting.SWbemLocator")
            objSWbemServices = objWMIService.ConnectServer(strComputer,"root\cimv2")

            query_str = '''Select * from %s%s''' % (self.win32_perf_base,counter_type)
            colItems = objSWbemServices.ExecQuery(query_str) # "Select * from Win32_PerfFormattedData_PerfProc_Process")# changed from Win32_Thread        

            if len(colItems) > 0:        
                for objItem in colItems:
                    if hasattr(objItem, 'IDProcess') and pid == objItem.IDProcess:

                            for attribute in self.supported_types[counter_type]:
                                eval_str = 'objItem.%s' % (attribute)
                                this_proc_dict[attribute] = eval(eval_str)

                            this_proc_dict['TimeStamp'] = datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S.') + str(datetime.datetime.now().microsecond)[:3]
                            break

        return this_proc_dict      


    def get_stats(self):
        '''
        Show process stats for all processes in given list, if none given return all processes   
        If filter list is defined return only the items that match or contained in the list
        Returns a list of result dictionaries
        '''    
        pythoncom.CoInitialize() # Needed when run by the same process in a thread
        proc_results_list = []
        if not self.perf_object_list:
            perf_object_list = self.supported_types.keys()

        for counter_type in perf_object_list:
            strComputer = "."
            objWMIService = win32com.client.Dispatch("WbemScripting.SWbemLocator")
            objSWbemServices = objWMIService.ConnectServer(strComputer,"root\cimv2")

            query_str = '''Select * from %s%s''' % (self.win32_perf_base,counter_type)
            colItems = objSWbemServices.ExecQuery(query_str) # "Select * from Win32_PerfFormattedData_PerfProc_Process")# changed from Win32_Thread

            try:  
                if len(colItems) > 0:
                    for objItem in colItems:
                        found_flag = False
                        this_proc_dict = {}

                        if not self.process_name_list:
                            found_flag = True
                        else:
                            # Check if process name is in the process name list, allow print if it is
                            for proc_name in self.process_name_list:
                                obj_name = objItem.Name
                                if proc_name.lower() in obj_name.lower(): # will log if contains name
                                    found_flag = True
                                    break

                        if found_flag:
                            for attribute in self.supported_types[counter_type]:
                                eval_str = 'objItem.%s' % (attribute)
                                this_proc_dict[attribute] = eval(eval_str)

                            this_proc_dict['TimeStamp'] = datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S.') + str(datetime.datetime.now().microsecond)[:3]
                            proc_results_list.append(this_proc_dict)

            except pywintypes.com_error, err_msg:
                # Ignore and continue (proc_mem_logger calls this function once per second)
                continue
        return proc_results_list     


def get_sys_stats():
    ''' Returns a dictionary of the system stats'''
    pythoncom.CoInitialize() # Needed when run by the same process in a thread
    x = winmem()

    sys_dict = { 
                    'dwAvailPhys': x.dwAvailPhys,
                    'dwAvailVirtual':x.dwAvailVirtual
                }
    return sys_dict


if __name__ == '__main__':
    # This area used for testing only
    sys_dict = get_sys_stats()

    stats_processor = process_stats(process_name_list=['process2watch'],perf_object_list=[],filter_list=[])
    proc_results = stats_processor.get_stats()

    for result_dict in proc_results:
        print result_dict

    import os
    this_pid = os.getpid()
    this_proc_results = stats_processor.get_pid_stats(this_pid)

    print 'this proc results:'
    print this_proc_results

http://monkut.webfactional.com/blog/archive/2009/1/21/windows-process-memory-logging-python

Answer 8

为此我们选择使用通常的信息源，因为我们可以发现空闲内存的瞬时波动，并且觉得查询meminfo数据源很有帮助。这也帮助我们获得了更多预先解析的相关参数。

代码

import os

linux_filepath = "/proc/meminfo"
meminfo = dict(
    (i.split()[0].rstrip(":"), int(i.split()[1]))
    for i in open(linux_filepath).readlines()
)
meminfo["memory_total_gb"] = meminfo["MemTotal"] / (2 ** 20)
meminfo["memory_free_gb"] = meminfo["MemFree"] / (2 ** 20)
meminfo["memory_available_gb"] = meminfo["MemAvailable"] / (2 ** 20)

输出供参考（我们删除了所有换行符以供进一步分析）

MemTotal：1014500 kB MemFree：562680 kB MemAvailable：646364 kB 缓冲区：15144 kB Cached：210720 kB SwapCached：0 kB Active：261476 kB Inactive：12816kB Active(kB Active)()(kB Active)(88on078)(kB) ：94384 kB 非活动（文件）：108000 kB 不可控制：3652 kB Mlocked：3652 kB SwapTotal：0 kB SwapFree：0 kB Dirty：0 kB 回写：0 kB AnonPages：168160 kB kB 1350m24224 kB kB 136024234 kB 映射： SReclaimable：18044 KB SUnreclaim：16448 KB KernelStack：2672个KB PageTables：8180 KB NFS_Unstable：0 KB弹跳：0 KB WritebackTmp：0 KB CommitLimit：507248 KB Committed_AS：1038756 KB VmallocTotal：34359738367 KB VmallocUsed：0 KB VmallocChunk：0 KB HardwareCorrupted： 0 kB AnonHugePages: 88064 kB CmaTotal: 0 kB CmaFree: 0 kB HugePages_Total: 0 HugePages_Free: 0 HugePages_Rsvd: 0 HugePages_Surp: 0 Hugepagesize:2048 kB DirectMap4k：43008 kB DirectMap2M：1005568 kB

Answer 9

我觉得这些答案是为 Python 2 编写的，在任何情况下都没有人提到resource可用于 Python 3的标准包。它提供了用于获取给定进程（默认情况下调用 Python 进程）的资源限制的命令。这与获取系统作为一个整体的当前资源使用情况不同，但它可以解决一些相同的问题，例如“我想确保我只对这个脚本使用 X 多 RAM”。

Answer 10

这汇总了所有优点： psutil+os获得 Unix 和 Windows 兼容性：这使我们能够获得：

1. 中央处理器
2. 记忆
3. 磁盘

代码：

import os
import psutil  # need: pip install psutil

In [32]: psutil.virtual_memory()
Out[32]: svmem(total=6247907328, available=2502328320, percent=59.9, used=3327135744, free=167067648, active=3671199744, inactive=1662668800,     buffers=844783616, cached=1908920320, shared=123912192, slab=613048320)

In [33]: psutil.virtual_memory().percent
Out[33]: 60.0

In [34]: psutil.cpu_percent()
Out[34]: 5.5

In [35]: os.sep
Out[35]: '/'

In [36]: psutil.disk_usage(os.sep)
Out[36]: sdiskusage(total=50190790656, used=41343860736, free=6467502080, percent=86.5)

In [37]: psutil.disk_usage(os.sep).percent
Out[37]: 86.5

Answer 11

从第一响应中获取反馈并进行一些小的更改

#!/usr/bin/env python
#Execute commond on windows machine to install psutil>>>>python -m pip install psutil
import psutil

print ('                                                                   ')
print ('----------------------CPU Information summary----------------------')
print ('                                                                   ')

# gives a single float value
vcc=psutil.cpu_count()
print ('Total number of CPUs :',vcc)

vcpu=psutil.cpu_percent()
print ('Total CPUs utilized percentage :',vcpu,'%')

print ('                                                                   ')
print ('----------------------RAM Information summary----------------------')
print ('                                                                   ')
# you can convert that object to a dictionary 
#print(dict(psutil.virtual_memory()._asdict()))
# gives an object with many fields
vvm=psutil.virtual_memory()

x=dict(psutil.virtual_memory()._asdict())

def forloop():
    for i in x:
        print (i,"--",x[i]/1024/1024/1024)#Output will be printed in GBs

forloop()
print ('                                                                   ')
print ('----------------------RAM Utilization summary----------------------')
print ('                                                                   ')
# you can have the percentage of used RAM
print('Percentage of used RAM :',psutil.virtual_memory().percent,'%')
#79.2
# you can calculate percentage of available memory
print('Percentage of available RAM :',psutil.virtual_memory().available * 100 / psutil.virtual_memory().total,'%')
#20.8

Answer 12

“...当前系统状态（当前 CPU、RAM、可用磁盘空间等）”和“*nix 和 Windows 平台”可能很难实现组合。

操作系统在管理这些资源的方式上有着根本的不同。实际上，它们在核心概念上有所不同，例如定义什么算作系统和什么算作应用程序时间。

“可用磁盘空间”？什么算作“磁盘空间”？所有设备的所有分区？多引导环境中的外部分区怎么样？

我认为 Windows 和 *nix 之间没有足够明确的共识使这成为可能。事实上，在称为 Windows 的各种操作系统之间甚至可能没有达成任何共识。是否有一个适用于 XP 和 Vista 的 Windows API？