寫一個python的服務監控程序

前言:

Redhat下安裝Python2.7

rhel6.4自帶的是2.6， 發現有的機器是python2.4。 到python站點下載源碼，解壓到Redhat上。然后執行以下的命令：

# ./configure --prefix=/usr/local/python27

# make

# make install

這樣安裝之后默認不會啟用Python2.7，須要使用/usr/local/python27/bin/python2.7調用新版本號的python。

而以下的安裝方式會直接接管現有的python

# ./configure

# make

# make install

開始：

服務子進程被監控主進程創建並監控，當子進程異常關閉。主進程能夠再次啟動之。使用了python的subprocess模塊。就這個簡單的代碼，竟然互聯網上沒有現成可用的樣例。

沒辦法，我寫好了貢獻出來吧。

首先是主進程代碼：service_mgr.py

#!/usr/bin/python
#-*- coding: UTF-8 -*-
# cheungmine
# stdin、stdout和stderr分別表示子程序的標准輸入、標准輸出和標准錯誤。

# 
# 可選的值有:
#   subprocess.PIPE - 表示須要創建一個新的管道.
#   一個有效的文件描寫敘述符（事實上是個正整數）
#   一個文件對象
#   None - 不會做不論什么重定向工作。子進程的文件描寫敘述符會繼承父進程的.
# 
# stderr的值還能夠是STDOUT, 表示子進程的標准錯誤也輸出到標准輸出.
# 
# subprocess.PIPE
# 一個能夠被用於Popen的stdin、stdout和stderr 3個參數的特輸值。表示須要創建一個新的管道.
# 
# subprocess.STDOUT
# 一個能夠被用於Popen的stderr參數的特輸值，表示子程序的標准錯誤匯合到標准輸出.
################################################################################
import os
import sys
import getopt

import time
import datetime

import codecs

import optparse
import ConfigParser

import signal
import subprocess
import select

# logging
# require python2.6.6 and later
import logging  
from logging.handlers import RotatingFileHandler

## log settings: SHOULD BE CONFIGURED BY config
LOG_PATH_FILE = "./my_service_mgr.log"
LOG_MODE = 'a'
LOG_MAX_SIZE = 4*1024*1024 # 4M per file
LOG_MAX_FILES = 4          # 4 Files: my_service_mgr.log.1, printmy_service_mgrlog.2, ...  
LOG_LEVEL = logging.DEBUG  

LOG_FORMAT = "%(asctime)s %(levelname)-10s[%(filename)s:%(lineno)d(%(funcName)s)] %(message)s"  

handler = RotatingFileHandler(LOG_PATH_FILE, LOG_MODE, LOG_MAX_SIZE, LOG_MAX_FILES)
formatter = logging.Formatter(LOG_FORMAT)
handler.setFormatter(formatter)

Logger = logging.getLogger()
Logger.setLevel(LOG_LEVEL)
Logger.addHandler(handler) 

# color output
#
pid = os.getpid() 

def print_error(s):
    print '\033[31m[%d: ERROR] %s\033[31;m' % (pid, s)

def print_info(s):
    print '\033[32m[%d: INFO] %s\033[32;m' % (pid, s)

def print_warning(s):
    print '\033[33m[%d: WARNING] %s\033[33;m' % (pid, s)


def start_child_proc(command, merged):
    try:
        if command is None:
            raise OSError, "Invalid command"

        child = None

        if merged is True:
            # merge stdout and stderr
            child = subprocess.Popen(command,
                stderr=subprocess.STDOUT, # 表示子進程的標准錯誤也輸出到標准輸出
                stdout=subprocess.PIPE    # 表示須要創建一個新的管道
            )
        else:
            # DO NOT merge stdout and stderr
            child = subprocess.Popen(command,
                stderr=subprocess.PIPE,
                stdout=subprocess.PIPE)

        return child

    except subprocess.CalledProcessError:
        pass # handle errors in the called executable
    except OSError:
        pass # executable not found

    raise OSError, "Failed to run command!"


def run_forever(command):
    print_info("start child process with command: " + ' '.join(command))
    Logger.info("start child process with command: " + ' '.join(command))

    merged = False
    child = start_child_proc(command, merged)

    line = ''
    errln = ''

    failover = 0

    while True:
        while child.poll() != None:
            failover = failover + 1
            print_warning("child process shutdown with return code: " + str(child.returncode))           
            Logger.critical("child process shutdown with return code: " + str(child.returncode))

            print_warning("restart child process again, times=%d" % failover)
            Logger.info("restart child process again, times=%d" % failover)
            child = start_child_proc(command, merged)

        # read child process stdout and log it
        ch = child.stdout.read(1)
        if ch != '' and ch != '\n':
            line += ch
        if ch == '\n':
            print_info(line)
            line = ''

        if merged is not True:
            # read child process stderr and log it
            ch = child.stderr.read(1)
            if ch != '' and ch != '\n':
                errln += ch
            if ch == '\n':
                Logger.info(errln)
                print_error(errln)
                errln = ''

    Logger.exception("!!!should never run to this!!!")  


if __name__ == "__main__":
    run_forever(["python", "./testpipe.py"])

然后是子進程代碼：testpipe.py

#!/usr/bin/python
#-*- coding: UTF-8 -*-
# cheungmine
# 模擬一個woker進程,10秒掛掉
import os
import sys

import time
import random

cnt = 10

while cnt >= 0:
    time.sleep(0.5)
    sys.stdout.write("OUT: %s\n" % str(random.randint(1, 100000)))
    sys.stdout.flush()

    time.sleep(0.5)
    sys.stderr.write("ERR: %s\n" % str(random.randint(1, 100000)))
    sys.stderr.flush()

    #print str(cnt)
    #sys.stdout.flush()
    cnt = cnt - 1

sys.exit(-1)

Linux上執行非常easy：

$ python service_mgr.py

Windows上以后台進程執行：

> start pythonw service_mgr.py

代碼中須要改動：

run_forever(["python", "testpipe.py"])