如何利用ansible callback插件对执行结果进行解析

最近在写一个批量巡检工具，利用ansible将脚本推到各个机器上执行，然后将执行的结果以json格式返回来。

如下所示：

# ansible node2 -m script -a /root/python/health_check.py

node2 | SUCCESS => {
    "changed": true,
    "rc": ,
    "stderr": "Shared connection to 192.168.244.20 closed.\r\n",
    "stdout": "{'cpu_iowait': '0.00', 'swap_out': 0, 'cpu_usr': '0.00', 'cpu_idle': '100.00', 'swap_total': '1999', 'swap_used': '78'
, 'load_average_5': '0.11', 'mem_util': '92.0', 'uptime': '', 'load_average_1': '0.03', 'cpu_sys': '0.00', 'mem_total': '', 'swap_in': , 'load_average_15': '0.06', 'disk': ['Filesystem      Size  Used Avail Use% Mounted on\\n', '/dev/sda3        18G  8.6G  8.1G  52% /\\n', 'tmpfs           238M     0  238M   0% /dev/shm\\n', '/dev/sda1       190M   27M  154M  15% /boot\\n'], 'numa': ''}\r\n",     
    "stdout_lines": [
        "{'cpu_iowait': '0.00', 'swap_out': 0, 'cpu_usr': '0.00', 'cpu_idle': '100.00', 'swap_total': '1999', 'swap_used': '78', 'loa
d_average_5': '0.11', 'mem_util': '92.0', 'uptime': '', 'load_average_1': '0.03', 'cpu_sys': '0.00', 'mem_total': '', 'swap_in': 0, 'load_average_15': '0.06', 'disk': ['Filesystem      Size  Used Avail Use% Mounted on\\n', '/dev/sda3        18G  .6G  .1G  % /\\n', 'tmpfs           238M       238M   % /dev/shm\\n', '/dev/sda1       190M   27M  154M  % /boot\\n'], 'numa': ''}"    ]
}

然后将结果重定向到一个文本文件中，再通过另外一个脚本，对该文本文件进行解析汇总，最后实现的结果如下：

ip              uptime          cpu_usr         cpu_sys         cpu_iowait      cpu_idle        load_average_1  load_average_5  ...
192.168.244.30                                                                           0.02            0.08            ...
192.168.244.20                                                                                         0.01            ...           

但总感觉这种方式有点low，对返回结果进行解析，这似乎是一个比较普遍的需求吧？

没道理，官方会对这种需求视而不见的，其实，官方提供了一个callback插件，来实现回调功能，里面定义了若干场景，譬如主机不可达，执行任务失败，执行任务成功等，分别对应不同的方法，这样就可以实现在不同的场景触发不同的操作，譬如，如果执行playbook失败了就发送邮件等，执行成功了将返回的结果保存到数据库中。

官方给了一个样例，具体可见：https://github.com/ansible/ansible/blob/devel/lib/ansible/plugins/callback/log_plays.py

基于上面这个样例，自己进行了定制性开发。本来想在callback插件中实现所有功能，但callback插件调试相当麻烦，不允许使用print函数，而且如果出现问题了，譬如列表下标越界，也只是在执行ansible时给出报错信息，并没有指出具体的报错行数。

最后，放弃了自己ALL IN ONE的想法，只是将返回的结果解析后保存到sqlite3数据库中，后续再基于数据库中的数据进行汇总。

代码如下：

from __future__ import (absolute_import, division, print_function)
__metaclass__ = type

import os
import time
import json
import sqlite3
from ansible.module_utils._text import to_bytes
from ansible.plugins.callback import CallbackBase

class CallbackModule(CallbackBase):
    """
    logs playbook results, per host, in /var/log/ansible/hosts
    """
    CALLBACK_VERSION = 2.0
    CALLBACK_TYPE = 'notification'
    CALLBACK_NAME = 'performance_check'
    CALLBACK_NEEDS_WHITELIST = False

    def __init__(self):
        super(CallbackModule, self).__init__()

    def runner_on_failed(self, host, res, ignore_errors=False):
        pass

    def runner_on_ok(self, host, res):
        performance_data=PerformanceData()
        create_table_sql = 'CREATE TABLE performance_data(ip varchar(20) primary key, uptime varchar(20),cpu_usr DECIMAL,cpu_sys DECI
MAL, cpu_iowait DECIMAL,cpu_idle DECIMAL,load_average_1 DECIMAL,load_average_5 DECIMAL,load_average_15 DECIMAL, mem_total INTEGER,mem_util DECIMAL,swap_total INTEGER,swap_used INTEGER,swap_in INTEGER,swap_out INTEGER,
numa TINYINT)'        
        insert_sql = 'insert into performance_data values (?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?)'
        insert_value = str_to_json(host,res)
        performance_data.create_table(create_table_sql)
        performance_data.insert_command(insert_sql,insert_value)
        performance_data.quit()

    def runner_on_skipped(self, host, item=None):
        #self.log(host, 'SKIPPED', '...')
        pass
    def runner_on_unreachable(self, host, res):
        #self.log(host, 'UNREACHABLE', res)
        pass
    def runner_on_async_failed(self, host, res, jid):
        #self.log(host, 'ASYNC_FAILED', res)
        pass
    def playbook_on_import_for_host(self, host, imported_file):
        pass

    def playbook_on_not_import_for_host(self, host, missing_file):
        pass

class PerformanceData():
    def __init__(self):
        self.conn = sqlite3.connect("/tmp/data.db")
        self.cursor = self.conn.cursor()

    def create_table(self,create_table_sql):
        self.cursor.execute(create_table_sql)

    def insert_command(self,insert_sql,insert_value):
        self.cursor.execute(insert_sql,insert_value)

    def query(self,query_sql):
        self.cursor.execute(query_sql)
        results=self.cursor.fetchall()
        return results

    def quit(self):
        self.conn.commit()
        self.conn.close()

def str_to_json(host,res):
            result= res["stdout"].strip(" ").replace("'",'"').strip('\n').strip('"')
            results= '{"'+host+'":'+result+'}'
            result_with_host = json.loads(results)
            value=result_with_host[host]
            return (host,value['uptime'],float(value['cpu_usr']),float(value['cpu_sys']),float(value['cpu_iowait']),
               float(value['cpu_idle']), float(value['load_average_1']), float(value['load_average_5']), float(value['load_average_15
']),               int(value['mem_total']), float(value['mem_util']),int(value['swap_total']),int(value['swap_used']),int(value['swap_in'
]),               int(value['swap_out']), int(value['numa'])
               )

这里一并附上，上述解析文本的脚本，似乎更能实现我ALL IN ONE的想法，哈哈~

#coding: utf8
import re,json,sqlite3
def get_ip_success():
    with open(r'C:\Users\Administrator\Desktop\2.txt') as f:
        ip_unreachable = []
        ip_failed = []
        ip_success=[]
        line_num=0
        for line in f.readlines():
            if re.search('UNREACHABLE', line):
                ip=line.split()[0]
                ip_unreachable.append(ip)
                flag=0
            elif re.search('FAILED',line):
                ip = line.split()[0]
                ip_failed.append(ip)
                flag=0
            elif re.search('SUCCESS',line):
                ip = line.split()[0]
                flag=1
                line_num=1
            elif flag == 1 and line_num == 7:
                line= line.strip(" ").replace("'",'"').strip('\n').strip('"')
                stdout_lines= '{"'+ip+'":'+line+'}'
                stdout_lines_with_ip = json.loads(stdout_lines)
                ip_success.append(stdout_lines_with_ip)
            line_num =line_num + 1
    return ip_success

def os_status_generator(ip_success):
    for os_status in ip_success:
        for key,value in os_status.iteritems():
            yield (key,value['uptime'],float(value['cpu_usr']),float(value['cpu_sys']),float(value['cpu_iowait']),
               float(value['cpu_idle']), float(value['load_average_1']), float(value['load_average_5']), float(value['load_average_15']),
               int(value['mem_total']), float(value['mem_util']),int(value['swap_total']),int(value['swap_used']),int(value['swap_in']),
               int(value['swap_out']), int(value['numa'])
               )

class OsStatus():
    def __init__(self,ip_success):
        try:
            self.conn = sqlite3.connect(":memory:")
            self.cursor = self.conn.cursor()
            self.cursor.execute('''CREATE TABLE os_status
                     (ip varchar(20) primary key, uptime varchar(20),cpu_usr DECIMAL,cpu_sys DECIMAL,cpu_iowait DECIMAL,cpu_idle DECIMAL,
                      load_average_1 DECIMAL,load_average_5 DECIMAL,load_average_15 DECIMAL,mem_total INTEGER,mem_util DECIMAL,
                      swap_total INTEGER,swap_used INTEGER,swap_in INTEGER,swap_out INTEGER,numa TINYINT)''')
            self.cursor.executemany("insert into os_status values (?,?,?,?,?,?,?,?,?,?,?,?,?,?,?,?)",os_status_generator(ip_success) )
        except Exception as e:
            print e;

    def query(self,sql):
        self.cursor.execute(sql)
        results=self.cursor.fetchall()
        column_size=len(results[0])
        column_name= [column[0] for column in self.cursor.description]
        for i in range(column_size):
            print column_name[i].ljust(15),
        print
        for each_result in results:
            for i in range(column_size):
                print str(each_result[i]).ljust(15),
            print

    def quit(self):
        try:
            self.cursor.close()
            self.conn.close()
        except Exception as e:
            print e;

ip_success=get_ip_success()
os_status=OsStatus(ip_success)
sql = "select * from os_status"
os_status.query(sql)

最后，再提一下ansible中如何开启callback插件功能，默认是关闭的。

开启两个选项：

callback_plugins   = /usr/share/ansible/plugins/callback
bin_ansible_callbacks = True

这两个是必需的，另外一个选项是

callback_whitelist = performance_check

其中，performance_check对应的是上面callback插件中定义的“CALLBACK_NAME”，

另一个相关参数是“CALLBACK_NEEDS_WHITELIST”，如果设置为False，则无需设置callback_whitelist选项，反之，则必须在callback_whitelist选项中指定“CALLBACK_NAME”。