大橙子网站建设,新征程启航
为企业提供网站建设、域名注册、服务器等服务
创新互联不只是一家网站建设的网络公司;我们对营销、技术、服务都有自己独特见解,公司采取“创意+综合+营销”一体化的方式为您提供更专业的服务!我们经历的每一步也许不一定是最完美的,但每一步都有值得深思的意义。我们珍视每一份信任,关注我们的成都网站制作、成都做网站质量和服务品质,在得到用户满意的同时,也能得到同行业的专业认可,能够为行业创新发展助力。未来将继续专注于技术创新,服务升级,满足企业一站式全网营销推广需求,让再小的高端网站设计也能产生价值!
生产中会生成大量的系统日志、应用程序日志、安全日志等等,通过对日志的分析,可了解服务器的负载、健康状态,可分析客户的分布情况、客户的行为,甚至基于这些分析可做出预测;
一般采集流程:
日志产出-->采集-->存储-->分析-->存储-->可视化;
采集(logstash、flume(apache)、scribe(facebook));
开源实时日志分析,ELK平台:
logstash收集日志,存放到ES集群中,kibana从ES中查询数据生成图表,返回browser;
离线分析;
在线分析,一份生成日志,一份传给大数据实时处理服务;
实时处理技术:storm、spark;
分析的前提:
半结构化数据:日志是半结构化数据,是有组织的,有格式的数据,可分割成行和列,可当作表来处理,也可分析里面的数据;
文本分析:日志是文本文件,需要依赖文件io、字符串操作、正则等技术,通过这些技术能把日志中需要的数据提取出来;
例:
123.125.71.36 - - [06/Apr/2017:18:09:25 +0800] "GET / HTTP/1.1" 200 8642 "-" "Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)"
提取数据:
1、用空格分割;
方1:
方2:先空格分割,遇""[]特殊处理;
2、用正则提取;
1、
import datetime
logs = '''123.125.71.36 - - [06/Apr/2017:18:09:25 +0800]
"GET / HTTP/1.1" 200 8642 "-"
"Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)"'''
names = ('remote','','','datetime','request','status','length','','useragent')
ops = (None,None,None,lambda timestr: datetime.datetime.strptime(timestr,'%d/%b/%Y:%H:%M:%S %z'),
lambda request: dict(zip(['method','url','protocol'],request.split())),int,int,None,None)
def extract(line):
fields = []
flag = False
tmp = ''
for field in line.split():
# print(field)
if not flag and (field.startswith('[') or field.startswith('"')):
if field.endswith(']') or field.endswith('"'):
fields.append(field.strip())
else:
tmp += field[1:]
# print(tmp)
flag = True
continue
if flag:
if field.endswith(']') or field.endswith('"'):
tmp += ' ' + field[:-1]
fields.append(tmp)
flag = False
tmp = ''
else:
tmp += ' ' + field
continue
fields.append(field)
print(fields)
info = {}
for i,field in enumerate(fields):
# print(i,field)
name = names[i]
op = ops[i]
if op:
info[name] = (op(field),op)
return info
print(extract(logs))
输出:
['123.125.71.36', '-', '-', '06/Apr/2017:18:09:25 +0800', 'GET / HTTP/1.1', '200', '8642', '"-"', 'Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)']
Out[16]:
{'datetime': (datetime.datetime(2017, 4, 6, 18, 9, 25, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))),
'length': (8642, int),
'request': ({'method': 'GET', 'protocol': 'HTTP/1.1', 'url': '/'},
'status': (200, int)}
2、
((?:\d{1,3}\.){3}\d{1,3}) - - \[([/:+ \w]+)\] "(\w+) (\S+) ([/\.\w\d]+)" (\d+) (\d+) .+ "(.+)"
import datetime
import re
# logs = '''123.125.71.36 - - [06/Apr/2017:18:09:25 +0800] "GET / HTTP/1.1" 200 8642 "-" "Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)"'''
ops = {
'datetime': lambda timestr: datetime.datetime.strptime(timestr,'%d/%b/%Y:%H:%M:%S %z'),
'status': int,
'length': int
}
pattern = '''(?P
regex = re.compile(pattern)
def extract(line)->dict:
matcher = regex.match(line)
info = None
if matcher:
info = {k:ops.get(k,lambda x:x)(v) for k,v in matcher.groupdict().items()}
return info
# print(extract(logs))
def load(path:str): #装载日志文件
with open(path) as f:
for line in f:
d = extract(line)
if d:
yield d #生成器函数
else:
continue #不合格数据,pycharm中左下角TODO(view-->Status Bar)
g = load('access.log')
print(next(g))
print(next(g))
print(next(g))
# for i in g:
# print(i)
输出:
{'remote': '123.125.71.36', 'datetime': datetime.datetime(2017, 4, 6, 18, 9, 25, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 8642, 'useragent': 'Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)'}
{'remote': '112.64.118.97', 'datetime': datetime.datetime(2017, 4, 6, 19, 13, 59, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/favicon.ico', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 4101, 'useragent': 'Dalvik/2.1.0 (Linux; U; Android 5.1.1; SM-G9250 Build/LMY47X)'}
{'remote': '119.123.183.219', 'datetime': datetime.datetime(2017, 4, 6, 20, 59, 39, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/favicon.ico', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 4101, 'useragent': 'Mozilla/5.0 (Windows NT 6.3; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/49.0.2623.221 Safari/537.36 SE 2.X MetaSr 1.0'}
注:
代码若在jupyter下,注意logs中内容不能换行;
滑动窗口:
或叫时间窗口,时间窗口函数,在数据分析领域极其重要;
很多数据,如日志,都是和时间相关的,都是按时间顺序产生的,在数据分析时,要按照时间来求值;
interval,表示每一次求值的时间间隔;
width,时间窗口宽度,指一次求值的时间窗口宽度,每个时间窗口的数据不均匀;
当width > interval
有重叠;
当width = interval
数据求值没有重叠;
当width < interval
一般不采纳这种方案,会有数据缺失;
如业务数据有1000万条,要求每次漏几个,这不影响统计趋势;
c2 = c1 - delta
delta = width - interval
delta = 0时,width = interval
时序数据,运维环境中,日志、监控等产生的数据是按时间先后产生并记录下来的,与时间相关的数据,一般按时间对数据进行分析;
数据分析基本程序结构:
例:
一函数,无限的生成随机数函数,产生时间相关的数据,返回->时间+随机数;
每次取3个数据,求平均值;
import random
import datetime
# def source():
# while True:
# yield datetime.datetime.now(),random.randint(1,100)
# i = 0
# for x in source():
# print(x)
# i += 1
# if i > 100:
# break
# for _ in range(100):
# print(next(source()))
def source():
while True:
yield {'value': random.randint(1,100),'datetime':datetime.datetime.now()}
src = source()
# lst = []
# lst.append(next(src))
# lst.append(next(src))
# lst.append(next(src))
lst = [next(src) for _ in range(3)]
def handler(iterable):
values = [x['value'] for x in iterable]
return sum(values) // len(values)
print(lst)
print(handler(lst))
窗口函数:
import datetime
import re
# logs = '''123.125.71.36 - - [06/Apr/2017:18:09:25 +0800] "GET / HTTP/1.1" 200 8642 "-" "Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)"'''
ops = {
'datetime': lambda timestr: datetime.datetime.strptime(timestr,'%d/%b/%Y:%H:%M:%S %z'),
'status': int,
'length': int
}
pattern = '''(?P
regex = re.compile(pattern)
def extract(line)->dict:
matcher = regex.match(line)
info = None
if matcher:
info = {k:ops.get(k,lambda x:x)(v) for k,v in matcher.groupdict().items()}
return info
# print(extract(logs))
def load(path:str):
with open(path) as f:
for line in f:
d = extract(line)
if d:
yield d
else:
continue
# g = load('access.log')
# print(next(g))
# print(next(g))
# print(next(g))
# for i in g:
# print(i)
def window(src,handler,width:int,interval:int):
# src = {'remote': '112.64.118.97', 'datetime': datetime.datetime(2017, 4, 6, 19, 13, 59, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/favicon.ico', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 4101, 'useragent': 'Dalvik/2.1.0 (Linux; U; Android 5.1.1; SM-G9250 Build/LMY47X)'}
start = datetime.datetime.strptime('1970/01/01 01:01:01 +0800','%Y/%m/%d %H:%M:%S %z')
current = datetime.datetime.strptime('1970/01/01 01:01:02 +0800','%Y/%m/%d %H:%M:%S %z')
seconds = width - interval
delta = datetime.timedelta(seconds)
buffer = []
for x in src:
if x:
buffer.append(x)
current = x['datetime']
if (current-start).total_seconds() >= interval:
ret = handler(buffer)
# print(ret)
start = current
# tmp = []
# for i in buffer:
# if i['datetime'] > current - delta:
# tmp.append(i)
buffer = [i for i in buffer if i['datetime'] > current - delta]
def donothing_handler(iterable:list):
print(iterable)
return iterable
def handler(iterable:list):
pass #TODO
def size_handler(iterable:list):
pass #TODO
# window(load('access.log'),donothing_handler,8,5)
# window(load('access.log'),donothing_handler,10,5)
window(load('access.log'),donothing_handler,5,5)
输出:
[{'remote': '123.125.71.36', 'datetime': datetime.datetime(2017, 4, 6, 18, 9, 25, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 8642, 'useragent': 'Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)'}]
[{'remote': '112.64.118.97', 'datetime': datetime.datetime(2017, 4, 6, 19, 13, 59, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/favicon.ico', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 4101, 'useragent': 'Dalvik/2.1.0 (Linux; U; Android 5.1.1; SM-G9250 Build/LMY47X)'}]
[{'remote': '119.123.183.219', 'datetime': datetime.datetime(2017, 4, 6, 20, 59, 39, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/favicon.ico', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 4101, 'useragent': 'Mozilla/5.0 (Windows NT 6.3; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/49.0.2623.221 Safari/537.36 SE 2.X MetaSr 1.0'}]
分发:
生产者消费者模型:
对于一个监控系统,需要处理很多数据,包括日志;
要有数据的采集、分析;
被监控对象,即数据的producer生产者,数据的处理程序,即数据的consumer消费者;
传统的生产者消费者模型,生产者生产,消费者消费,这种模型有些问题,开发的代码耦合太高,如果生产规模扩大,不易扩展,生产和消费的速度难匹配;
queue队列,食堂打饭;
producer-consumer,卖包子;消费速度 >= 生产速度;解决办法:queue,作用:解耦(在程序间实现解耦(服务间解耦))、缓冲;
注:
zeromq,底层通信协议用;
大多数*mq,都是消费队列;
kafka,性能极高;
FIFO,先进先出;
LIFO,后进先出;
数据的生产是不稳定的,会造成短时间数据的潮涌,需要缓冲;
消费者消费能力不一样,有快有慢,消费者可以自己决定消费缓冲区中的数据;
单机可用queue(内建模块)构建进程内的队列,满足多个线程间的生产消费需要;
大型系统可使用第三方消息中间件,rabbitmq、rocketmq、kafka;
queue模块:
queue.Queue(maxsize=0),queue提供了一个FIFO先进先出的队列Queue,创建FIFO队列,返回Queue对象;maxsize <= 0,队列长度没有限制;
q = queue.Queue()
q.get(block=True,timeout=None),从队列中移除元素并返回这个元素,只要get过即拿走就没了;
block阻塞,timeout超时;
若block=True,是阻塞,timeout=None,就是一直阻塞,timeout有值,即阻塞到一定秒数抛Empty异常;
若blcok=False,是非阻塞,timeout将被忽略,要么成功返回一个元素,要么抛Empty异常;
q.get_nowait(),等价于q.get(block=False)或q.get(False),即要么成功返回一个元素,要么抛Empty异常;这种阻塞效果,要多线程中举例;
q.put(item,block=True,timeout=None),把一个元素加入到队列中去,
block=True,timeout=None,一直阻塞直至有空位放元素;
block=True,timeout=5,阻塞5秒抛Full异常;
block=False,timeout失效,立即返回,能塞进去就塞,不能则抛Full异常;
q.put_nowait(item),等价于q.put(item,False);
注:
Queue的长度是个近似值,不准确,因为生产消费一直在进行;
q.get(),只要get过,即拿走,数据就没了;而kafka中,拿走数据后,kafka中仍保留有,由consumer来清理;
例:
from queue import Queue
import random
q = Queue()
q.put(random.randint(1,100))
q.put(random.randint(1,100))
print(q.get())
print(q.get())
# print(q.get()) #block
print(q.get(timeout=3))
输出:
2
35
Traceback (most recent call last):
File "/home/python/magedu/projects/cmdb/queue_Queue.py", line 12, in
print(q.get(timeout=3))
File "/ane/python3.6/lib/python3.6/queue.py", line 172, in get
raise Empty
queue.Empty
分发器的实现:
生产者(数据源)生产数据,缓冲到消息队列中;
数据处理流程:数据加载-->提取-->分析(滑动窗口函数);
处理大量数据时,对于一个数据源来说,需要多个消费者处理,但如何分配数据?
需要一个分发器(调度器),把数据分发给不同的消费者处理;
每一个消费者拿到数据后,有自己的处理函数,所以要有一种注册机制;
数据加载-->提取-->分发-->分析函数1|分析函数2,一个数据通过分发器,发送给n个消费者,分析函数1|分析函数2为不同的handler,不同的窗口宽度,间隔时间;
如何分发?
一对多,副本发送(一个数据通过分发器,发送到n个消费者),用轮询;
MQ?
在生产者和消费者之间用消息队列,那么所有的消费者共用一个消息队列?(这需要解决争抢的问题);还是各自拥有一个消息队列?(较容易);
注册?
在调度器内部记录有哪些消费者,记录消费者自己的队列;
线程?
由于一条数据会被多个不同的注册过的handler处理,所以最好的方式是多线程;
注:
import threading
t = threading.Thread(target=window,args=(src,handler,width,interval)) #target,线程中运行的函数,args,这个函数运行时需要的实参用tuple
t.start()
分析功能:
分析日志很重要,通过海量数据的分析就能知道是否遭受了***,是否是爬取的高峰期,是否有盗链;
分析的逻辑放到handler中;
window仅通过时间窗口挪动取数据,不要将其的功能做的丰富全面,若需统一处理,独立出单独的函数;
注:
爬虫:baiduspider,googlebot,SEO,http,request,response;
状态码分析:
状态码中包含了很多信息;
304,服务器收到客户端提交的请求数,发现资源未变化,要求browser使用静态资源的缓存;
404,server找不到请求的资源;
304占比大,说明静态缓存效果明显;
404占比大,说明出现了错误链接,或深度嗅探网站资源;
若400,500占比突然开始增大,网站一定出问题了;
import datetime
import re
from queue import Queue
import threading
# logs = '''123.125.71.36 - - [06/Apr/2017:18:09:25 +0800] "GET / HTTP/1.1" 200 8642 "-" "Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)"'''
ops = {
'datetime': lambda timestr: datetime.datetime.strptime(timestr,'%d/%b/%Y:%H:%M:%S %z'),
'status': int,
'length': int
}
pattern = '''(?P
regex = re.compile(pattern)
def extract(line)->dict:
matcher = regex.match(line)
info = None
if matcher:
info = {k:ops.get(k,lambda x:x)(v) for k,v in matcher.groupdict().items()}
return info
# print(extract(logs))
def load(path:str):
with open(path) as f:
for line in f:
d = extract(line)
if d:
yield d
else:
continue
# g = load('access.log')
# print(next(g))
# print(next(g))
# print(next(g))
# for i in g:
# print(i)
# def window(src,handler,width:int,interval:int):
# # src = {'remote': '112.64.118.97', 'datetime': datetime.datetime(2017, 4, 6, 19, 13, 59, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/favicon.ico', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 4101, 'useragent': 'Dalvik/2.1.0 (Linux; U; Android 5.1.1; SM-G9250 Build/LMY47X)'}
# start = datetime.datetime.strptime('1970/01/01 01:01:01 +0800','%Y/%m/%d %H:%M:%S %z')
# current = datetime.datetime.strptime('1970/01/01 01:01:02 +0800','%Y/%m/%d %H:%M:%S %z')
# seconds = width - interval
# delta = datetime.timedelta(seconds)
# buffer = []
#
# for x in src:
# if x:
# buffer.append(x)
# current = x['datetime']
# if (current-start).total_seconds() >= interval:
# ret = handler(buffer)
# # print(ret)
# start = current
# # tmp = []
# # for i in buffer:
# # if i['datetime'] > current - delta:
# # tmp.append(i)
# buffer = [i for i in buffer if i['datetime'] > current - delta]
# window(load('access.log'),donothing_handler,8,5)
# window(load('access.log'),donothing_handler,10,5)
# window(load('access.log'),donothing_handler,5,5)
def window(src:Queue,handler,width:int,interval:int):
# src = {'remote': '112.64.118.97', 'datetime': datetime.datetime(2017, 4, 6, 19, 13, 59, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/favicon.ico', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 4101, 'useragent': 'Dalvik/2.1.0 (Linux; U; Android 5.1.1; SM-G9250 Build/LMY47X)'}
start = datetime.datetime.strptime('1970/01/01 00:01:01 +0800','%Y/%m/%d %H:%M:%S %z')
current = datetime.datetime.strptime('1970/01/01 01:01:02 +0800','%Y/%m/%d %H:%M:%S %z')
delta = datetime.timedelta(width-interval)
buffer = []
while True:
data = src.get()
if data:
buffer.append(data)
current = data['datetime']
if (current-start).total_seconds() >= interval:
ret = handler(buffer)
# print(ret)
start = current
buffer = [i for i in buffer if i['datetime'] > current - delta]
def donothing_handler(iterable:list):
print(iterable)
return iterable
def handler(iterable:list):
pass #TODO
def size_handler(iterable:list):
pass #TODO
def status_handler(iterable:list):
d = {}
for item in iterable:
key = item['status']
if key not in d.keys():
d[key] = 0
d[key] += 1
total = sum(d.values())
print({k:v/total*100 for k,v in d.items()}) #return
def dispatcher(src):
queues = []
threads = []
def reg(handler,width,interval):
q = Queue()
queues.append(q)
t = threading.Thread(target=window,args=(q,handler,width,interval))
threads.append(t)
def run():
for t in threads:
t.start()
for x in src:
for q in queues:
q.put(x)
return reg,run
reg,run = dispatcher(load('access.log'))
reg(status_handler,8,5)
run()
日志文件加载:
改为接受一批;
如果一批路径,迭代每一个路径;
如果路径是一个普通文件,按行读取内容(假设是日志文件);
如果路径是一个目录,就遍历路径下的所有普通文件,每一个文件按行处理,不递归处理子目录;
def openfile(path:str):
with open(path) as f:
for line in f:
d = extract(line)
if d:
yield d
else:
continue
def load(*paths):
for file in paths:
p = Path(file)
if not p.exists():
continue
if p.is_dir():
for x in p.iterdir():
if x.is_file():
# for y in openfile(str(x)):
# yield y
yield from openfile(str(x))
elif p.is_file():
# for y in openfile(str(p)):
# yield y
yield from openfile(str(p))
离线日志分析项目:
可指定文件或目录,对日志进行数据分析;
分析函数可动态注册;
数据可分发给不同的分析处理程序处理;
关键步骤:
数据源处理(处理一行行数据);
拿到数据后的处理(作为分析,一小批一小批处理,窗口函数);
分发器(生产者和消费者间作为桥梁作用);
浏览器分析:
useragent,指软件按一定的格式向远端服务器提供一个标记自己的字符串;
在http协议中,使用user-agent字段传送一这个字符串,这个值可被修改(想伪装谁都可以);
格式:([platform details]) [extensions]
例如:"Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/28.0.1500.72 Safari/537.36"
注:
chrome-->console,navigator.userAgent,将内容复制粘贴到傲游的自定义UserAgent中;
信息提取模块:
user-agents、pyyaml、ua-parser;
]$ pip install user-agents pyyaml ua-parser
例:
from user_agents import parse
u = 'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/28.0.1500.72 Safari/537.36'
ua = parse(u)
print(ua.browser)
print(ua.browser.family)
print(ua.browser.version_string)
输出:
Browser(family='Chrome', version=(28, 0, 1500), version_string='28.0.1500')
Chrome
28.0.1500
整合,完整代码:
import datetime
import re
from queue import Queue
import threading
from pathlib import Path
from user_agents import parse
from collections import defaultdict
# logs = '''123.125.71.36 - - [06/Apr/2017:18:09:25 +0800] "GET / HTTP/1.1" 200 8642 "-" "Mozilla/5.0 (compatible; Baiduspider/2.0; +http://www.baidu.com/search/spider.html)"'''
ops = {
'datetime': lambda timestr: datetime.datetime.strptime(timestr,'%d/%b/%Y:%H:%M:%S %z'),
'status': int,
'length': int,
'request': lambda request: dict(zip(('method','url','protocol'),request.split())),
'useragent': lambda useragent: parse(useragent)
}
# pattern = '''(?P
pattern = '''(?P
regex = re.compile(pattern)
def extract(line)->dict:
matcher = regex.match(line)
info = None
if matcher:
info = {k:ops.get(k,lambda x:x)(v) for k,v in matcher.groupdict().items()}
# print(info)
return info
# print(extract(logs))
# def load(path:str):
# with open(path) as f:
# for line in f:
# d = extract(line)
# if d:
# yield d
# else:
# continue
def openfile(path:str):
with open(path) as f:
for line in f:
d = extract(line)
if d:
yield d
else:
continue
def load(*paths):
for file in paths:
p = Path(file)
if not p.exists():
continue
if p.is_dir():
for x in p.iterdir():
if x.is_file():
# for y in openfile(str(x)):
# yield y
yield from openfile(str(x))
elif p.is_file():
# for y in openfile(str(p)):
# yield y
yield from openfile(str(p))
# g = load('access.log')
# print(next(g))
# print(next(g))
# print(next(g))
# for i in g:
# print(i)
# def window(src,handler,width:int,interval:int):
# # src = {'remote': '112.64.118.97', 'datetime': datetime.datetime(2017, 4, 6, 19, 13, 59, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/favicon.ico', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 4101, 'useragent': 'Dalvik/2.1.0 (Linux; U; Android 5.1.1; SM-G9250 Build/LMY47X)'}
# start = datetime.datetime.strptime('1970/01/01 01:01:01 +0800','%Y/%m/%d %H:%M:%S %z')
# current = datetime.datetime.strptime('1970/01/01 01:01:02 +0800','%Y/%m/%d %H:%M:%S %z')
# seconds = width - interval
# delta = datetime.timedelta(seconds)
# buffer = []
#
# for x in src:
# if x:
# buffer.append(x)
# current = x['datetime']
# if (current-start).total_seconds() >= interval:
# ret = handler(buffer)
# # print(ret)
# start = current
# # tmp = []
# # for i in buffer:
# # if i['datetime'] > current - delta:
# # tmp.append(i)
# buffer = [i for i in buffer if i['datetime'] > current - delta]
# window(load('access.log'),donothing_handler,8,5)
# window(load('access.log'),donothing_handler,10,5)
# window(load('access.log'),donothing_handler,5,5)
def window(src:Queue,handler,width:int,interval:int):
# src = {'remote': '112.64.118.97', 'datetime': datetime.datetime(2017, 4, 6, 19, 13, 59, tzinfo=datetime.timezone(datetime.timedelta(0, 28800))), 'method': 'GET', 'request': '/favicon.ico', 'protocol': 'HTTP/1.1', 'status': 200, 'length': 4101, 'useragent': 'Dalvik/2.1.0 (Linux; U; Android 5.1.1; SM-G9250 Build/LMY47X)'}
start = datetime.datetime.strptime('1970/01/01 00:01:01 +0800','%Y/%m/%d %H:%M:%S %z')
current = datetime.datetime.strptime('1970/01/01 01:01:02 +0800','%Y/%m/%d %H:%M:%S %z')
delta = datetime.timedelta(width-interval)
buffer = []
while True:
data = src.get()
if data:
buffer.append(data)
current = data['datetime']
if (current-start).total_seconds() >= interval:
ret = handler(buffer)
# print(ret)
start = current
buffer = [i for i in buffer if i['datetime'] > current - delta]
def donothing_handler(iterable:list):
print(iterable)
return iterable
def handler(iterable:list):
pass #TODO
def size_handler(iterable:list):
pass #TODO
def status_handler(iterable:list):
d = {}
for item in iterable:
key = item['status']
if key not in d.keys():
d[key] = 0
d[key] += 1
total = sum(d.values())
print({k:v/total*100 for k,v in d.items()}) #return
browsers = defaultdict(lambda :0)
def browser_handler(iterable:list):
# browsers = {}
for item in iterable:
ua = item['useragent']
key = (ua.browser.family,ua.browser.version_string)
# browsers[key] = browsers.get(key,0) + 1
browsers[key] += 1
return browsers
def dispatcher(src):
queues = []
threads = []
def reg(handler,width,interval):
q = Queue()
queues.append(q)
t = threading.Thread(target=window,args=(q,handler,width,interval))
threads.append(t)
def run():
for t in threads:
t.start()
for x in src:
for q in queues:
q.put(x)
return reg,run
reg,run = dispatcher(load('access.log'))
reg(status_handler,8,5)
reg(browser_handler,5,5)
run()
print(browsers)