urls = get_urls(URLS_DB)
client = GearmanClient(['10.61.0.145'])
tasks = Taskset()
TASK_URLS_NUM = 100
# disptribute task
i = 0
while i < len(urls):
sub_urls = urls[i:i + TASK_URLS_NUM]
workload = '\n'.join(sub_urls)
t = Task1('crawl',
workload,
str(i),
timeout=TASK1_TIMEOUT,
retry_count=1)
tasks.add(t)
print "add task:%s" % t.uniq
i += TASK_URLS_NUM
# test
pass
# 0.init database for return result from worker
print "0.initialize database for results."
tmps = ["%s int" % cate for cate in CATES]
cates_str = ','.join(tmps)
tb_sql = "create table %s (word text primary key,%s,total int);" % (
RAW_WORDS_TB, cates_str)
print tb_sql
init_db(TASK1_RESULT_DB, tb_sql)
# 1.run the tasks in parallel
print "1.Preprocess tasks:"
return urls
if __name__ == "__main__":
urls = get_urls(URLS_DB)
client = GearmanClient(["10.61.0.145"])
tasks = Taskset()
TASK_URLS_NUM = 100
# disptribute task
i = 0
while i < len(urls):
sub_urls = urls[i : i + TASK_URLS_NUM]
workload = "\n".join(sub_urls)
t = Task1("crawl", workload, str(i), timeout=TASK1_TIMEOUT, retry_count=1)
tasks.add(t)
print "add task:%s" % t.uniq
i += TASK_URLS_NUM
# test
pass
# 0.init database for return result from worker
print "0.initialize database for results."
tmps = ["%s int" % cate for cate in CATES]
cates_str = ",".join(tmps)
tb_sql = "create table %s (word text primary key,%s,total int);" % (RAW_WORDS_TB, cates_str)
print tb_sql
init_db(TASK1_RESULT_DB, tb_sql)
# 1.run the tasks in parallel
print "1.Preprocess tasks:"
client.do_taskset(tasks)
