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()
def queue_items(self):
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);
"""
list = []
try:
d = DBUtil(config._HAINIU_DB)
sql = select_queue_sql % (self.fail_times,self.limit)
tuple = d.read_tuple(sql)
if len(tuple) == 0:
return list
queue_ids = ''
for t in tuple:
queue_id = t[0]
url = t[1]
param = '' if t[2] is None else t[2]
queue_ids += str(queue_id) + ','
c = NewsFindConsumer(url, param, queue_id)
list.append(c)
queue_ids = queue_ids[:-1]
d.execute(update_queue_sql % (queue_ids))
except:
self.rl.exception()
d.rollback()
d.commit()
finally:
d.close()
return list
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 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
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 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 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 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()
def create_seed():
url = "https://www.autohome.com.cn/all"
catetory = "汽车"
sql = """
insert into hainiu_web_seed (url,md5,domain,host,category,status) values
('%s','%s','%s','%s','%s',0);
"""
hu = HtmlUtil()
domain = get_tld(url)
host = hu.get_url_host(url)
u = Util()
md5 = u.get_md5(url)
rl = LogUtil().get_base_logger()
try:
d = DBUtil(config._HAINIU_DB)
sql = sql % (url, md5, domain, host, catetory)
d.execute(sql)
except:
rl.exception()
d.rollback()
finally:
d.close()
def create_seed():
sql = """
insert into web_seed (url,md5,domain,host,category,status) values
('%s','%s','%s','%s','%s',0);
"""
url = "https://news.sina.com.cn/"
catetory = "新闻"
hu = HtmlUtil()
domain = get_tld(url)
host = hu.get_url_host(url)
u = Util()
md5 = u.get_md5(url)
rl = LogUtil().get_base_logger()
try:
d = DBUtil(config._ZZ_DB)
sql = sql % (url, md5, domain, host, catetory)
d.execute(sql)
except:
rl.exception()
d.rollback()
finally:
d.close()
def redis2Hdfs():
select_xpath_rule_sql = """select host,xpath,type from stream_extract_xpath_rule where host='%s' and status=0"""
rl = LogUtil().get_base_logger()
try:
d = DBUtil(_ZZ_DB)
start = 0
is_finish = True
host_set = set()
f = FileUtil()
t = TimeUtil()
time_str = t.now_time(format='%Y%m%d%H%M%S')
#local_xpath_file_path = '/user/zengqingyong17/spark/xpath_cache_file' + time_str
local_xpath_file_path = 'E:/python_workspaces/data/xpath/xpath_file' + time_str
starttime = time.clock()
r = redis.Redis('nn1.hadoop', '6379', db=6)
while is_finish:
values = set()
rs = r.scan(start, "total_z:*", 10)
# 新游标
start = rs[0]
if start ==0:
is_finish = False
# print rs
for i in rs[1]:
host = i.split(":")[1]
total_key = i
txpath_key = 'txpath_z:%s' % host
fxpath_key = 'fxpath_z:%s' % host
total = r.get(total_key)
# 降序排序获得次数(0,1)
txpath = r.zrevrange(txpath_key, 0, 1)
row_format = "%s\t%s\t%s\t%s"
if txpath:
txpath_num = int(r.zscore(txpath_key, txpath[0]))
if txpath.__len__() == 2:
# 返回txpath_key 中txpath[1]的数值
txpath_num_1 = int(r.zscore(txpath_key, txpath[1]))
txpath_num_1 = txpath_num_1 if txpath_num_1 is not None else 0
if txpath_num / float(total) >= 0.8:
values.add(row_format % (host, txpath[0], 'true', '0'))
host_set.add(host)
else:
if txpath_num >= 100:
values.add(row_format % (host, txpath[0], 'true', '0'))
host_set.add(host)
if txpath_num_1 is not None and txpath_num_1 >= 100:
values.add(row_format % (host, txpath[1], 'true', '0'))
host_set.add(host)
# 获得fxpath_key的全部值
fxpath = r.smembers(fxpath_key)
if fxpath:
# print 'fxpath:%s' % fxpath
for fx in fxpath:
values.add(row_format % (host, fx, 'false', '1'))
host_set.add(host)
sql = select_xpath_rule_sql % host
list_rule = d.read_tuple(sql)
for rule in list_rule:
type = rule[2]
if type == 0:
values.add(row_format % (rule[0], rule[1], 'true', '2'))
host_set.add(host)
elif type == 1:
values.add(row_format % (rule[0], rule[1], 'false', '3'))
host_set.add(host)
f.write_file_line_pattern(local_xpath_file_path, values, "a")
#上传到HDFS的XPATH配置文件目录
# c.upload("/user/qingniu/xpath_cache_file/", local_xpath_file_path)
endtime = time.clock()
worksec = int(round((endtime - starttime)))
rl.info('total host %s,action time %s\'s' % (host_set.__len__(), worksec))
except:
rl.exception()
d.rollback()
finally:
d.close()
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 xpath_config_file():
select_xpath_rule_sql = """select host,xpath,type from stream_extract_xpath_rule where host='%s' and status=0"""
rl = LogUtil().get_base_logger()
try:
# _HAINIU_DB = {'HOST': '192.168.137.190', 'USER': '******', 'PASSWD': '12345678', 'DB': 'hainiucrawler',
# 'CHARSET': 'utf8', 'PORT': 3306}
d = DBUtil(config._HAINIU_DB)
# d = DBUtil(_HAINIU_DB)
r = redis.Redis('nn1.hadoop', 6379, db=6)
# r = redis.Redis('redis.hadoop', 6379, db=6)
f = FileUtil()
t = TimeUtil()
c = Client("http://nn1.hadoop:50070")
time_str = t.now_time(format='%Y%m%d%H%M%S')
# local_xpath_file_path = '/Users/leohe/Data/input/xpath_cache_file/xpath_file' + time_str
local_xpath_file_path = '/home/qingniu/xpath_cache_file/xpath_file' + time_str
start_cursor = 0
is_finish = True
starttime = time.clock()
host_set = set()
while is_finish:
values = set()
limit = r.scan(start_cursor, 'total:*', 10)
if limit[0] == 0:
is_finish = False
start_cursor = limit[0]
for h in limit[1]:
host = h.split(":")[1]
total_key = h
txpath_key = 'txpath:%s' % host
fxpath_key = 'fxpath:%s' % host
total = r.get(total_key)
txpath = r.zrevrange(txpath_key, 0, 1)
row_format = "%s\t%s\t%s\t%s"
if txpath:
# print 'txpath:%s' % txpath
txpath_num = int(r.zscore(txpath_key, txpath[0]))
if txpath.__len__() == 2:
txpath_num_1 = int(r.zscore(txpath_key, txpath[1]))
txpath_num_1 = txpath_num_1 if txpath_num_1 is not None else 0
# print 'txpath_max_num:%s' % txpath_num
if txpath_num / float(total) >= 0.8:
values.add(row_format % (host, txpath[0], 'true', '0'))
host_set.add(host)
else:
if txpath_num >= 1:
values.add(row_format %
(host, txpath[0], 'true', '0'))
host_set.add(host)
if txpath_num_1 is not None and txpath_num_1 >= 1:
values.add(row_format %
(host, txpath[1], 'true', '0'))
host_set.add(host)
fxpath = r.smembers(fxpath_key)
if fxpath:
# print 'fxpath:%s' % fxpath
for fx in fxpath:
values.add(row_format % (host, fx, 'false', '0'))
host_set.add(host)
sql = select_xpath_rule_sql % host
list_rule = d.read_tuple(sql)
for rule in list_rule:
type = rule[2]
if type == 0:
values.add(row_format %
(rule[0], rule[1], 'true', '2'))
host_set.add(host)
elif type == 1:
values.add(row_format %
(rule[0], rule[1], 'false', '3'))
host_set.add(host)
f.write_file_line_pattern(local_xpath_file_path, values, "a")
#上传到HDFS的XPATH配置文件目录
c.upload("/user/qingniu/xpath_cache_file/", local_xpath_file_path)
endtime = time.clock()
worksec = int(round((endtime - starttime)))
rl.info('total host %s,action time %s\'s' %
(host_set.__len__(), worksec))
except:
rl.exception()
d.rollback()
finally:
d.close()
sql = """
insert into hainiu_queue (type,action,params) values (%s, %s, %s);
"""
try:
params = [1, 'www.hainiubl.com', "ff"]
db_util.execute_no_commit(sql, params)
time.sleep(5)
1/0
params = [1, 'www.hainiubl.com', "gg"]
db_util.execute_no_commit(sql, params)
time.sleep(5)
db_util.commit()
except Exception, e:
print e
db_util.rollback()
finally:
db_util.close()
# -------------------------------------
# 悲观锁 + 事务
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])
