def push_queue_items():
count_news_seed_queue_sql = """select count(*) from hainiu_queue where type=3 and fail_times=0;"""
insert_news_seed_internally_queue_items_sql = """
insert into hainiu_queue (type,action,params) values(3,%s,%s);
"""
count_news_seed_internally_sql = """select count(*) from hainiu_web_seed_internally where status=0 for update;"""
selec_news_seed_internally_sql = """select a_url,param,id from hainiu_web_seed_internally where status=0 limit %s,%s;"""
update_news_seed_internally_sql = """update hainiu_web_seed_internally set status=1 where id in (%s);"""
rl = LogUtil().get_base_logger()
try:
d = DBUtil(config._HAINIU_DB)
queue_total = d.read_one(count_news_seed_queue_sql)[0]
if queue_total != 0:
rl.info('last download_page queue not finish,last queue %s unFinish' % (queue_total))
return
starttime = time.clock()
d = DBUtil(config._HAINIU_DB)
total = long(d.read_one(count_news_seed_internally_sql)[0])
page_size = 1000
page = total / page_size
for i in range(0, page + 1):
sql = selec_news_seed_internally_sql % (0, page_size)
list = d.read_tuple(sql)
values = []
id_values = []
for l in list:
url = l[0]
url = url if url is not None else ''
param = l[1]
param = param if param is not None else ''
values.append((url,param))
id = l[2]
id_values.append(str(id))
if id_values.__len__() != 0:
random.shuffle(values)
d.executemany_no_commit(insert_news_seed_internally_queue_items_sql,values)
ids = ','.join(id_values)
sql = update_news_seed_internally_sql % (ids)
d.execute(sql)
endtime = time.clock()
worksec = int(round((endtime - starttime)))
rl.info('push seed_internally queue finish,total items %s,action time %s\'s' % (total,worksec))
except:
rl.exception()
rl.error(sql)
d.rollback()
finally:
d.close()
def push_queue_items():
count_news_seed_sql = """select count(*) from hainiu_web_seed where status=0;"""
select_news_seed_sql = """select url,category,last_crawl_time from hainiu_web_seed where status=0 limit %s,%s;"""
insert_news_seed_queue_items_sql = """insert into hainiu_queue (type,action,params) values(1,%s,%s);"""
count_news_seed_queue_sql = """select count(*) from hainiu_queue where type=1 and fail_times=0;"""
rl = LogUtil().get_base_logger()
try:
d = DBUtil(config._HAINIU_DB)
queue_total = d.read_one(count_news_seed_queue_sql)[0]
if queue_total != 0:
rl.info('last news_find queue not finish,last queue %s unFinish' %
(queue_total))
return
starttime = time.clock()
total = long(d.read_one(count_news_seed_sql)[0])
page_size = 1000
page = total / page_size
for i in range(0, page + 1):
sql = select_news_seed_sql % (i * page_size, page_size)
list = d.read_tuple(sql)
values = []
for l in list:
url = l[0]
publisher = get_tld(url)
publisher = publisher[0:publisher.index((
'.'))] if publisher.__contains__('.') else publisher
param = {}
param['category'] = l[1]
param['publisher'] = publisher
param = json.dumps(param, ensure_ascii=False)
values.append((url, param))
if values.__len__() != 0:
random.shuffle(values)
d.executemany(insert_news_seed_queue_items_sql, values)
endtime = time.clock()
worksec = int(round((endtime - starttime)))
rl.info(
'push news_find queue finish,total items %s,action time %s\'s' %
(total, worksec))
except:
rl.exception()
rl.error(sql)
d.rollback()
finally:
d.close()
def send_sms(self, content, phone=config._ALERT_PHONE):
"""send alter sms for phone with content
"""
l = LogUtil().get_base_logger()
try:
send_url = 'http://send.sms.hainiu.com:8080/s?command=cralwer&phone=%s&' % (
phone)
send_url += urllib.urlencode(
{'content': content.decode('utf-8').encode('gbk')})
r = urllib2.urlopen(send_url).read()
if '0-OK' != r:
l.error("短信发送失败,短信服务器返回状态为:%s,手机号:%s,内容:%s" %
(r, phone, content))
return False
except:
l.exception()
return False
return True
def push_queue_items():
inert_sql = """
insert into hainiu_queue (type,params,action) values(1,%s,%s);
"""
count_sql = """
select count(1) from hainiu_queue where type=1;
"""
select_sql = """
select id from hainiu_queue where type=1 limit %s,%s;
"""
rl = LogUtil().get_base_logger()
try:
d = DBUtil(config._HAINIU_DB)
sql = inert_sql
insert_list = [("aaa", "bbb"), ("dffddf", "awwee")]
d.executemany(sql, insert_list)
sql = count_sql
queue_total = d.read_one(sql)[0]
print "queue_total", queue_total
page_size = 10
page = (queue_total / page_size) + 1
print "page", page
for i in range(0, page):
sql = select_sql % (i * page_size, page_size)
select_list = d.read_tuple(sql)
print "page", i
for record in select_list:
id = record[0]
print id
except:
rl.exception()
rl.error(sql)
d.rollback()
finally:
d.close()
class KafkaUtil:
__kafka_connect_cache = {}
__lock = threading.Lock()
def __init__(self,kafka_conf):
host_list = [host for host in kafka_conf['HOST'].split(',')]
random.shuffle(host_list)
host_str = ','.join(host_list)
self.cache_key = '_'.join((host_str,kafka_conf['TOPIC']))
self.host = host_str
self.topic = kafka_conf['TOPIC']
self.rl = LogUtil().get_logger('consumer', 'consumer_kafka')
def push_message(self,message):
self.__lock.acquire()
u = Util()
producer = u.get_dict_value(self.__kafka_connect_cache,self.cache_key)
if producer is None:
client = KafkaClient(hosts=self.host)
topic = client.topics[self.topic]
producer = topic.get_producer()
self.__kafka_connect_cache[self.cache_key] = producer
is_success = True
try:
producer.produce(message)
except:
is_success = False
del self.__kafka_connect_cache[self.cache_key]
self.rl.error('kafka push error cacheKey is %s' % (self.cache_key))
self.rl.exception()
self.__lock.release()
return is_success
class HainiuConsumer(ConsumerAction):
def __init__(self, id, ac, params):
'''
初始化队列的消费者
:param id: 消息的ID，也就是数据库表里的ID
:param ac: 消息的动作信息，也就是数据库表里的action字段
:param params: 消息的动作的附加参数
'''
super(self.__class__, self).__init__()
self.id = id
self.ac = ac
self.params = params
self.rl = LogUtil().get_logger('consumer', 'consumer' + queue_name)
def action(self):
'''
处理拿到的消息
:return:消费动作的处理结果，用于消费者线程的日志打印和传递处理成功和失败方法所需要的数据
'''
is_success = True
try:
print self.ac, self.params
# 1/0
except:
is_success = False
self.rl.exception()
return super(self.__class__, self).result(is_success, [self.id])
def success_action(self, values):
'''
消息动作处理成功之后，从队列中间件删除该消息，表示这个消息最终处理完成
:param values: 消息动作处理之后的结果
'''
delete_sql = """
delete from hainiu_queue where id=%s
"""
try:
d = DBUtil(config._HAINIU_DB)
id = values[0]
sql = delete_sql % id
d.execute(sql)
except:
self.rl.exception()
self.rl.error(sql)
d.rollback()
finally:
d.close()
def fail_action(self, values):
'''
消息动作处理失败之后，更改队列中间件中该消息的失败次数并记录执行机器的IP
如果达到该机器的最大尝试失败次数，则更改队列中间件中该消息的状态为未处理，目的让其它机器再次尝试去处理该消息
:param values: 消息动作处理之后的结果
'''
update_sql = """
update hainiu_queue set fail_times=fail_times+1,fail_ip='%s' where id=%s;
"""
update_sql_1 = """
update hainiu_queue set type=1 where id=%s
"""
try:
d = DBUtil(config._HAINIU_DB)
id = values[0]
u = Util()
ip = u.get_local_ip()
sql = update_sql % (ip, id)
d.execute_no_commit(sql)
if (self.try_num == Consumer._WORK_TRY_NUM):
sql = update_sql_1 % id
d.execute_no_commit(sql)
d.commit()
except:
self.rl.exception()
self.rl.error(sql)
d.rollback()
finally:
d.close()
class HainiuProducer(ProducerAction):
def __init__(self, limit, fail_times):
'''
初始化队列的发者
:param limit: 每次从队列中取多少条记录
:param fail_times: 限定取记录的失败次数条件
'''
super(self.__class__, self).__init__()
self.limit = limit
self.fail_times = fail_times
self.rl = LogUtil().get_logger('producer', 'producer' + queue_name)
def queue_items(self):
'''
从队列中取出要处理的消息，并封装成消费者动作，然后更新队列的状态
:return: 封装好的消费者动作列表
'''
# 会限制本机处理失败之后就不再进行获取的获取，通过机器IP来限制
# select_queue_sql = """
# select id,action,params from hainiu_queue where type=1 and fail_ip <>'%s' and fail_times<=%s
# limit 0,%s for update;
# """
select_queue_sql = """
select id,action,params from hainiu_queue where type=1 and fail_times<=%s
limit 0,%s for update;
"""
update_queue_sql = """
update hainiu_queue set type=0 where id in (%s);
"""
return_list = []
try:
d = DBUtil(config._HAINIU_DB)
# u = Util()
# ip = u.get_local_ip()
# sql = select_queue_sql % (self.fail_times,ip,self.limit)
sql = select_queue_sql % (self.fail_times, self.limit)
select_dict = d.read_dict(sql)
if len(select_dict) == 0:
return return_list
query_ids = []
for record in select_dict:
id = record["id"]
action = record["action"]
params = record["params"]
query_ids.append(str(id))
c = HainiuConsumer(id, action, params)
return_list.append(c)
ids = ",".join(query_ids)
sql = update_queue_sql % ids
d.execute(sql)
except:
self.rl.exception()
self.rl.error(sql)
d.rollback()
finally:
d.close()
return return_list
class NewsFindConsumer(ConsumerAction):
def __init__(self, url, param ,queue_id):
ConsumerAction.__init__(self)
self.url = url[:-1] if url.endswith('/') else url
self.param = param
self.queue_id = queue_id
self.rl = LogUtil().get_logger('consumer', 'consumer' + queue_name)
def action(self):
is_success = True
t = TimeUtil()
u = Util()
hu = HtmlUtil()
r = RequestUtil()
in_values = []
ex_values = []
a_href = ''
main_md5 = u.get_md5(self.url)
now_time = datetime.now()
update_time = int(time.mktime(now_time.timetuple()))
create_time = update_time
create_day = int(t.now_day().replace('-', ''))
create_hour = int(t.now_hour())
try:
html = r.http_get_phandomjs(self.url)
domain = get_tld(self.url)
soup = BeautifulSoup(html, 'lxml')
a_docs = soup.find_all("a")
a_set = set()
a_param = {}
out_json_srt = ''
status = 0
host = hu.get_url_host(self.url)
for a in a_docs:
a_href = self.get_format_url(a,host)
a_title = a.get_text().strip()
if a_href == '' or a_title == '':
continue
if a_set.__contains__(a_href):
continue
a_set.add(a_href)
req = urllib2.Request(url=a_href)
a_host = req.get_host() if req.get_host() is not None else ''
a_md5 = u.get_md5(a_href)
if a_title != '':
a_param['title'] = a_title
out_json_srt = json.dumps(a_param,ensure_ascii=False)
a_xpath = hu.get_dom_parent_xpath_js(a)
insert_values = (main_md5,domain,host,a_md5,a_host,a_xpath,create_time,create_day,create_hour,update_time,status,
MySQLdb.escape_string(self.url),
MySQLdb.escape_string(a_href),
MySQLdb.escape_string(a_title),
out_json_srt)
if a_host.__contains__(domain):
in_values.append(insert_values)
else:
ex_values.append(insert_values)
in_table = 'hainiu_web_seed_internally'
ex_table = 'hainiu_web_seed_externally'
insert_sql = """
insert into <table> (md5,domain,host,a_md5,a_host,a_xpath,create_time,create_day,create_hour,update_time,status,url,a_url,a_title,param)
values (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s) ON DUPLICATE KEY UPDATE update_time=update_time;
"""
try:
d = DBUtil(config._HAINIU_DB)
d.execute_no_commit("set NAMES utf8mb4;")
if in_values.__len__() != 0:
sql = insert_sql.replace('<table>',in_table)
d.executemany_no_commit(sql,in_values)
if ex_values.__len__() != 0:
sql = insert_sql.replace('<table>',ex_table)
d.executemany_no_commit(sql,ex_values)
d.commit()
except:
is_success = False
self.rl.exception()
self.rl.error(sql)
d.rollback()
finally:
d.close()
except:
is_success = False
self.rl.exception()
finally:
r.close_phandomjs()
return super(self.__class__, self).result(is_success, [main_md5,self.url,a_href,in_values.__len__(),ex_values.__len__(),self.queue_id])
def success_action(self, values):
delete_sql = """
delete from hainiu_queue where id=%s;
"""
update_hainiu_news_seed_sql = """
update hainiu_web_seed set last_crawl_internally=%s,last_crawl_externally=%s,last_crawl_time=now() where md5="%s";"""
try:
d = DBUtil(config._HAINIU_DB)
id = values[5]
sql = delete_sql % id
d.execute_no_commit(sql)
sql = update_hainiu_news_seed_sql % (values[3],values[4],values[0])
d.execute_no_commit(sql)
d.commit()
except:
self.rl.exception()
self.rl.error(sql)
d.rollback()
d.commit()
finally:
d.close()
def fail_action(self, values):
update_sql = """
update hainiu_queue set fail_times=fail_times+1,fail_ip='%s' where id=%s;
"""
update_sql_1 = """
update hainiu_queue set type=1 where id=%s;
"""
update_hainiu_news_seed_sql = """
update hainiu_web_seed set fail_times=fail_times+1,fail_ip="%s" where md5="%s";
"""
try:
d = DBUtil(config._HAINIU_DB)
id = values[5]
u = Util()
ip = u.get_local_ip()
sql = update_sql % (ip, id)
d.execute_no_commit(sql)
main_md5 = values[0]
sql = update_hainiu_news_seed_sql % (ip, main_md5)
d.execute_no_commit(sql)
if (self.try_num == Consumer._WORK_TRY_NUM):
sql = update_sql_1 % (id)
d.execute_no_commit(sql)
d.commit()
except:
self.rl.exception()
self.rl.error(sql)
d.rollback()
d.commit()
finally:
d.close()
def get_format_url(self, a_doc, host):
a_href = a_doc.get('href')
try:
if a_href is not None and a_href.__len__() > 0:
a_href = str(a_href).strip()
a_href = a_href[:a_href.index('#')] if a_href.__contains__('#') else a_href
# a_href = a_href.encode('utf8')
# a_href = urllib.quote(a_href,safe='.:/?&=')
if a_href.startswith('//'):
url = 'https:' + a_href if self.url.startswith('https:') else 'http:' + a_href
url = mx.URL.URL(str(url))
a_href = url.url
elif a_href.startswith('/'):
url = 'https://' + host + a_href if self.url.startswith('https:') else 'http://' + host + a_href
url = mx.URL.URL(str(url))
a_href = url.url
elif a_href.startswith('./') or a_href.startswith('../'):
url = mx.URL.URL(str(self.url) + '/' + a_href)
a_href = url.url
elif not a_href.startswith('javascript') and not a_href.startswith('mailto') and not a_href.startswith('http') and a_href != '':
url = 'https://' + host + '/' + a_href if self.url.startswith('https:') else 'http://' + host + '/' + a_href
url = mx.URL.URL(str(url))
a_href = url.url
a_href = a_href[:-1] if a_href.endswith('/') else a_href
# a_href = a_href.lower()
get_tld(a_href)
except:
return ''
if not a_href.startswith('http'):
return ''
if a_href.__contains__('?'):
a_params_str = a_href[a_href.index('?') + 1:]
a_params = a_params_str.split('&')
a_params.sort()
a_params_str = '&'.join(a_params)
a_href = a_href[:a_href.index('?') + 1] + a_params_str
return a_href
#-*- encoding: utf-8 -*-
'''
log_demo.py
Created on 21-1-30 上午11:23
Copyright (c) 21-1-30, 海牛学院版权所有.
@author: 潘牛
'''
from commons.util.log_util import LogUtil
logger1 = LogUtil().get_logger("log_name", "log_file")
logger2 = LogUtil().get_logger("log_name", "log_file")
# 两个对象指向同一内存地址
print logger1 is logger2
logger1.info("测试 info 级别")
logger1.error("测试 error 级别")
try:
1 / 0
except Exception, e:
logger1.exception(e)
class DownLoadConsumer(ConsumerAction):
def __init__(self, url, param, queue_id, pro_flag):
ConsumerAction.__init__(self)
self.url = url[:-1] if url.endswith('/') else url
self.param = param
self.queue_id = queue_id
self.pro_flag = pro_flag
self.rl = LogUtil().get_logger('consumer', 'consumer' + queue_name)
def action(self):
is_success = True
t = TimeUtil()
f = FileUtil()
u = Util()
hu = HtmlUtil()
r = RequestUtil()
values = []
md5 = u.get_md5(self.url)
now_time = datetime.now()
update_time = int(time.mktime(now_time.timetuple()))
create_time = update_time
create_day = int(t.now_day().replace('-', ''))
create_hour = int(t.now_hour())
now_minute = int(t.now_min())
for i in xrange(60, -5, -5):
if now_minute >= i:
now_minute = i
break
now_minute = t.now_time(format='%Y%m%d%H') + (
'0%s' % (str(now_minute)) if now_minute < 10 else str(now_minute))
values.append(MySQLdb.escape_string(self.url))
values.append(md5)
values.append(create_time)
values.append(create_day)
values.append(create_hour)
values.append('')
values.append(MySQLdb.escape_string(self.param))
values.append(update_time)
try:
html = r.http_get_phandomjs(self.url)
domain = get_tld(self.url)
values[5] = domain
soup = BeautifulSoup(html, 'lxml')
title_doc = soup.find('title')
title = title_doc.contents[0] if title_doc is not None and len(
title_doc.contents) == 1 else ''
host = hu.get_url_host(self.url)
values.append(host)
values.append(MySQLdb.escape_string(title))
# k = KafkaUtil(config._KAFKA_CONFIG)
html = html.replace(content._SEQ1,
'').replace(content._SEQ2, content._SEQ4)
# push_str = content._SEQ3.join(('%s','%s')) % (self.url,html)
# push_str = content._SEQ3.join(('%s','%s')) % (u.get_md5(push_str),push_str)
# push_str = bytes(push_str)
# is_success = k.push_message(push_str)
is_success = True
if is_success:
self.save_file(create_time, f, now_minute, u, self.url, html)
else:
values.append('')
values.append('')
self.rl.error("kafka push error")
except:
is_success = False
values.append('')
values.append('')
self.rl.exception()
finally:
r.close_phandomjs()
try:
if is_success:
values.append(1)
insert_web_page_sql = """
insert into hainiu_web_page (url,md5,create_time,create_day,create_hour,domain,param,update_time,host,
title,status) values ("%s","%s",%s,%s,%s,"%s","%s",%s,"%s","%s",%s) on DUPLICATE KEY UPDATE update_time=values(update_time);
"""
else:
ip = u.get_local_ip()
values.append(ip)
values.append(2)
insert_web_page_sql = """
insert into hainiu_web_page (url,md5,create_time,create_day,create_hour,domain,param,update_time,host,
title,fail_ip,status) values ("%s","%s",%s,%s,%s,"%s","%s",%s,"%s","%s","%s",%s)
on DUPLICATE KEY UPDATE fail_times=fail_times+1,fail_ip=values(fail_ip);
"""
d = DBUtil(config._HAINIU_DB)
sql = insert_web_page_sql % tuple(values)
d.execute(sql)
except:
is_success = False
self.rl.exception()
self.rl.error(sql)
d.rollback()
d.commit()
finally:
d.close()
return super(self.__class__,
self).result(is_success,
[md5, self.url, update_time, self.queue_id])
def success_action(self, values):
delete_sql = """
delete from hainiu_queue where id=%s;
"""
update_hainiu_news_internally_sql = """
update hainiu_web_seed_internally set update_time=%s where md5="%s";
"""
try:
d = DBUtil(config._HAINIU_DB)
id = values[3]
sql = delete_sql % id
d.execute_no_commit(sql)
sql = update_hainiu_news_internally_sql % (values[2], values[0])
d.execute_no_commit(sql)
d.commit()
except:
self.rl.exception()
self.rl.error(sql)
d.rollback()
d.commit()
finally:
d.close()
def fail_action(self, values):
update_sql = """
update hainiu_queue set fail_times=fail_times+1,fail_ip='%s' where id=%s;
"""
update_sql_1 = """
update hainiu_queue set type=3 where id=%s;
"""
update_hainiu_news_internally_sql = """
update hainiu_web_seed_internally set fail_times=fail_times+1,fail_ip="%s",update_time=%s where md5="%s";
"""
try:
d = DBUtil(config._HAINIU_DB)
id = values[3]
u = Util()
ip = u.get_local_ip()
sql = update_sql % (ip, id)
d.execute_no_commit(sql)
sql = update_hainiu_news_internally_sql % (ip, values[2],
values[0])
d.execute_no_commit(sql)
if (self.try_num == Consumer._WORK_TRY_NUM):
sql = update_sql_1 % (id)
d.execute_no_commit(sql)
d.commit()
except:
self.rl.exception()
self.rl.error(sql)
d.rollback()
d.commit()
finally:
d.close()
def save_file(self, create_time, fi, now_minute, u, url, html):
now_file_name = '%s_%s_%s' % (self.consumer_thread_name, self.pro_flag,
now_minute)
last_file_name = u.get_dict_value(html_file_path_cache,
self.consumer_thread_name)
html_file_path_cache[self.consumer_thread_name] = now_file_name
tmp_path = config._LOCAL_DATA_DIR % (
'%s/%s_%s' % ('tmp', self.consumer_thread_name, self.pro_flag))
start_char = content._SEQ2
if last_file_name is None or now_file_name != last_file_name:
start_char = ''
if os.path.exists(tmp_path) and os.path.getsize(tmp_path) > 0:
done_path = config._LOCAL_DATA_DIR % (
'%s/%s_%s' % ('done', now_file_name, create_time))
shutil.move(tmp_path, done_path)
html = html.replace(content._SEQ1, '').replace(content._SEQ2,
content._SEQ4)
record_str = content._SEQ3.join(('%s', '%s')) % (url, html)
record_str = content._SEQ3.join(
('%s', '%s')) % (u.get_md5(record_str), record_str)
html_record_format_str = start_char + record_str
fi.write_file_content_pattern(tmp_path,
html_record_format_str,
pattern='a')
