绘制Linux资源随时间变化曲线图

1. 脚本源码

执行脚本前请确保监控的Linux操作系统已安装sysstat工具（脚本中sar命令需要安装此工具）。

#!/usr/bin/env python3

import fabric
import matplotlib.dates as mdates
import matplotlib.pyplot as plt
import pandas
import re
import time
import typing


class UnixSystemMonitor(object):
    """
    Unix操作系统资源使用情况监控.

    为避免连接异常断开的情况,所有连接均在使用时临时建立.
    """

    def __init__(self, hostname: str, username: str, password: str):
        """
        :param hostname: 远端Unix服务器的主机名
        :param username: 操作系统的用户名
        :param password: 用户名对应的密码
        """
        self.hostname = hostname
        self.kwargs = {
            "host": hostname,
            "user": username,
            "connect_kwargs": {"password": password}
        }

    def get_resource_usage(self, dirname: str, nics: typing.Iterable[str] = (),
                           disks: typing.Iterable[str] = (), first: bool = False) -> None:
        """
        定期从服务器换获取资源使用情况并记录到日志文件中.
        :param dirname: 存放日志文件的目录
        :param nics: 需要监控的网卡名列表
        :param disks: 需要监控的磁盘名列表
        :param first: 如果为True,会将清空文件并写每列的含义
        :return:
        """
        date = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())

        def record_usage(path: str, keys, vals):
            with open(path, mode="w" if first else "a") as file:
                if first:
                    file.write("\t".join(["date"] + keys) + "\n")
                file.write("\t".join([date] + vals) + "\n")

        # 从远端服务器收集系统资源的使用情况
        with fabric.Connection(**self.kwargs) as conn:
            command = "sar -u -r -d -n DEV -p 1 3 | grep 'Average:'"
            resp = conn.run(command, warn=False, hide=True)
        if resp.exited != 0:
            return

        # 获取网卡使用情况
        nic_keys = re.findall(f"Average: +IFACE +(.*)\n", resp.stdout)[0].split()
        for nic in nics:
            nic_vals = re.findall(f"Average: +{nic} +(.*)\n", resp.stdout)[0].split()
            record_usage(f"{dirname}/{self.hostname}_nic_{nic}.log", nic_keys, nic_vals)

        # 获取磁盘使用率情况
        disk_keys = re.findall(f"Average: +DEV +(.*)\n", resp.stdout)[0].split()
        for disk in disks:
            disk_vals = re.findall(f"Average: +{disk} +(.*)\n", resp.stdout)[0].split()
            record_usage(f"{dirname}/{self.hostname}_disk_{disk}.log", disk_keys, disk_vals)

        # 获取内核使用情况
        cpu_keys = re.findall(f"Average: +CPU +(.*)\n", resp.stdout)[0].split()
        cpu_vals = re.findall(f"Average: +all +(.*)\n", resp.stdout)[0].split()
        record_usage(f"{dirname}/{self.hostname}_cpu.log", cpu_keys, cpu_vals)

        # 获取内存使用情况
        mem_keys = re.findall(f"Average: +(kbmemfree.*)\n", resp.stdout)[0].split()
        mem_vals = re.findall(f"Average: +([0-9]+.*)\n", resp.stdout)[0].split()
        record_usage(f"{dirname}/{self.hostname}_mem.log", mem_keys, mem_vals)

    def draw_resource_usage_charts(self, dirname: str, nics: typing.Iterable[str] = (),
                                   disks: typing.Iterable[str] = ()) -> None:
        """
        根据指定目录下的日志文件绘制资源使用情况趋势图.
        :param dirname: 存放日志文件的目录
        :param nics: 需要监控的网卡名列表
        :param disks: 需要监控的磁盘名列表
        :return:
        """
        nics, disks = tuple(nics), tuple(disks)  # 解决参数为迭代器时不可重入问题
        num_sun_plots = len(nics) + len(disks) + 2

        plt.figure(figsize=(40, 6 * num_sun_plots), dpi=100)  # 设置画布尺寸和分辨率

        index_sub_plots = 1  # 子图索引号,用于计算子图在画布的纵向位置

        def add_sub_plots(df_: pandas.DataFrame, title: str, lables_: typing.List[str], ylim=None):
            distance_to_bottom = 1 - 1 / num_sun_plots * index_sub_plots + 0.06
            ax = plt.axes([0.05, distance_to_bottom, 0.9, 0.6 / num_sun_plots])  # 设置子图位置和大小
            ax.plot(df_)
            ax.set(title=title, ylim=ylim)
            ax.legend(labels=lables_, loc="lower right")  # 设置图例
            ax.xaxis.set_major_formatter(mdates.DateFormatter("%Y-%m-%d %H:%M:%S"))
            # plt.xticks(rotation=90)  # 设置X轴坐标旋转角度

        # 添加内核使用率到画布中
        df = pandas.read_csv(f"{dirname}/{self.hostname}_cpu.log", delimiter="\t", header=0,
                             index_col="date", usecols=["date", "%idle"])
        df.index = pandas.to_datetime(df.index)
        df["%idle"] = df["%idle"].map(lambda x: 100-x)
        add_sub_plots(df, f"{self.hostname} cpu usage(%)", ["total used(%)"], (0, 100))
        index_sub_plots += 1

        # 添加内存使用率到画布中
        df = pandas.read_csv(f"{dirname}/{self.hostname}_mem.log", delimiter="\t", header=0,
                             index_col="date", usecols=["date", "%memused"])
        df.index = pandas.to_datetime(df.index)
        add_sub_plots(df, f"{self.hostname} mem usage(%)", ["total used(%)"], (0, 100))
        index_sub_plots += 1

        # 添加磁盘使用情况到画布中
        for disk in disks:
            lables = ["%util"]
            df = pandas.read_csv(f"{dirname}/{self.hostname}_disk_{disk}.log", delimiter="\t", header=0,
                                 index_col="date", usecols=["date"] + lables)
            df.index = pandas.to_datetime(df.index)
            add_sub_plots(df, f"{self.hostname} disk {disk} usage(%)", lables, (0, 100))
            index_sub_plots += 1

        # 添加网卡使用情况到画布中
        for nic in nics:
            lables = ["rxpck/s", "txpck/s", "rxkB/s", "txkB/s"]
            df = pandas.read_csv(f"{dirname}/{self.hostname}_nic_{nic}.log", delimiter="\t", header=0,
                                 index_col="date", usecols=["date"] + lables)
            df.index = pandas.to_datetime(df.index)
            add_sub_plots(df, f"{self.hostname} nic {nic} usage(%)", lables)
            index_sub_plots += 1

        plt.savefig(f"{dirname}/{self.hostname}.png")


if __name__ == '__main__':
    hostname = "..."
    username = "..."
    password = "..."
    nics, disks = list(), list()
    duration = 300
    om = UnixSystemMonitor(hostname, username=username, password=password)
    om.get_resource_usage("./log", nics=nics, disks=disks, first=True)
    start_date = time.time()
    while time.time() - start_date < duration:
        om.get_resource_usage("./log", nics=nics, disks=disks)
    om.draw_resource_usage_charts("./log", nics=nics, disks=disks)

2. 执行结果