Merge branch 'feature/zhangbo_flow_recall' of algorithm/rov-offline into master

write_redis/a-crontab的任务

@@ -5,7 +5,8 @@
 10 * * * * cd /root/zhangbo/rov-offline && /bin/sh alg_recsys_recall_tags_videos_task.sh > my_logs_tags/tags_$(date +\%Y-\%m-\%d_\%H).log 2>&1
 16 * * * * cd /root/zhangbo/rov-offline/write_redis && /bin/sh alg_recsys_feature_02_vidhasreturnrov_redis_task.sh > my_logs_alg_recsys_feature_02_vidhasreturnrov_redis_task/$(date +\%Y-\%m-\%d_\%H_\%M).log 2>&1
 32 * * * * cd /root/zhangbo/rov-offline/write_redis && /bin/sh alg_recsys_rank_item_realtime_1h_task.sh > my_logs_alg_recsys_rank_item_realtime_1h_task/$(date +\%Y-\%m-\%d_\%H_\%M).log 2>&1
-
 25 * * * * cd /root/zhangbo/rov-offline/write_redis && /bin/sh alg_recsys_feature_03_vidhasreturnrov1d7d_redis_task.sh > my_logs_alg_recsys_feature_03_vidhasreturnrov1d7d_redis/$(date +\%Y-\%m-\%d_\%H_\%M).log 2>&1
+22 * * * * cd /root/zhangbo/rov-offline/write_redis && /bin/sh alg_recsys_feature_04_vidvov1d3d_redis_task.sh > my_logs_alg_recsys_feature_04_vidvov1d3d_redis/$(date +\%Y-\%m-\%d_\%H_\%M).log 2>&1
+13 05 * * * cd /root/zhangbo/rov-offline/write_redis && /bin/sh alg_recsys_feature_05_vidvovdaily4filter_redis_task.sh > my_logs_alg_recsys_feature_05/$(date +\%Y-\%m-\%d_\%H_\%M).log 2>&1
 
-22 * * * * cd /root/zhangbo/rov-offline/write_redis && /bin/sh alg_recsys_feature_04_vidvov1d3d_redis_task.sh > my_logs_alg_recsys_feature_04_vidvov1d3d_redis/$(date +\%Y-\%m-\%d_\%H_\%M).log 2>&1
+40 * * * * cd /root/zhangbo/rov-offline/write_redis && /bin/sh alg_recsys_task_doudi_videos_task.sh > my_logs_alg_recsys_task_doudi_videos/$(date +\%Y-\%m-\%d_\%H_\%M).log 2>&1

write_redis/alg_recsys_feature_05_vidvovdaily4filter_redis.py

@@ -0,0 +1,196 @@
+# -*- coding: utf-8 -*-
+import os
+import sys
+import traceback
+
+root_dir = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
+if root_dir not in sys.path:
+    sys.path.append(root_dir)
+    print("******** sys.path ********")
+    print(sys.path)
+
+from multiprocessing import Process
+from odps import ODPS
+from threading import Timer
+import threading
+from my_utils import RedisHelper, execute_sql_from_odps
+from my_config import set_config
+from log import Log
+import json
+from datetime import datetime
+from queue import Queue
+from tqdm import tqdm
+import time
+
+
+
+
+config_, _ = set_config()
+log_ = Log()
+redis_helper = RedisHelper()
+
+REDIS_PREFIX = "redis:vid_vov_daily4filter:"
+EXPIRE_TIME = 48 * 3600
+
+def worker(queue, executor):
+    while True:
+        row = queue.get()
+        if row is None:  # 结束信号
+            queue.task_done()
+            break
+        executor(row)
+        queue.task_done()
+def records_process_for_list(records, executor, max_size=50, num_workers=10):
+    # 创建一个线程安全的队列
+    queue = Queue(maxsize=max_size)  # 可以调整 maxsize 以控制内存使用
+    # 设置线程池大小
+    num_workers = num_workers
+    # 启动工作线程
+    threads = []
+    for _ in range(num_workers):
+        t = threading.Thread(target=worker, args=(queue, executor))
+        t.start()
+        threads.append(t)
+    # 读取数据并放入队列
+    for row in tqdm(records):
+        queue.put(row)
+    # 发送结束信号
+    for _ in range(num_workers):
+        queue.put(None)
+    # 等待所有任务完成
+    queue.join()
+    # 等待所有工作线程结束
+    for t in threads:
+        t.join()
+def process_and_store(row):
+    table_key, json_str = row
+    key = REDIS_PREFIX + str(table_key)
+    expire_time = EXPIRE_TIME
+    redis_helper.set_data_to_redis(key, json_str, expire_time)
+
+def check_data(project, table,  date, hour, mm) -> int:
+    """检查数据是否准备好,输出数据条数"""
+    odps = ODPS(
+        access_id=config_.ODPS_CONFIG['ACCESSID'],
+        secret_access_key=config_.ODPS_CONFIG['ACCESSKEY'],
+        project=project,
+        endpoint=config_.ODPS_CONFIG['ENDPOINT'],
+        connect_timeout=3000,
+        read_timeout=500000,
+        pool_maxsize=1000,
+        pool_connections=1000
+    )
+    try:
+        t = odps.get_table(name=table)
+        log_.info(f"检查分区是否存在-【 dt={date}")
+        check_res = t.exist_partition(partition_spec=f'dt={date}')
+        if check_res:
+            sql = f'select * from {project}.{table} where dt = {date}'
+            log_.info(sql)
+            with odps.execute_sql(sql=sql).open_reader() as reader:
+                data_count = reader.count
+        else:
+            log_.info("表{}分区{}不存在".format(table, date))
+            data_count = 0
+    except Exception as e:
+        log_.error("table:{},date:{} no data. return data_count=0,报错原因是:{}".format(table, date, hour, e))
+        data_count = 0
+    return data_count
+
+def get_sql(project, table, date, hour, mm):
+    sql = '''
+    SELECT  vid
+            ,feature
+    FROM    {}.{}
+    WHERE   dt = '{}'
+    '''.format(
+        project, table, date
+    )
+    print("sql:" + sql)
+    records = execute_sql_from_odps(project=project, sql=sql)
+    video_list = []
+    with records.open_reader() as reader:
+        for record in reader:
+            key = record['vid']
+            m = dict()
+            try:
+                json_str = record['feature']
+            except Exception as e:
+                json_str = json.dumps(m)
+                log_.error(e)
+                log_.error(traceback.format_exc())
+
+            video_list.append([key, json_str])
+    return video_list
+
+
+def main():
+    try:
+        date = sys.argv[1]
+        hour = sys.argv[2]
+        # mm = sys.argv[3]
+        mm = "00"
+    except Exception as e:
+        date = datetime.now().strftime('%Y%m%d')
+        hour = datetime.now().hour
+        # mm = datetime.now().minute
+        mm = "00"
+        log_.info("没有读取到参数，采用系统时间：{}".format(e))
+    log_.info("使用时间参数-日期：{}，小时：{}".format(date, str(hour)))
+    # if hour in []:
+    #     log_.info(f"hour={hour}不执行，直接返回。")
+    #     return
+    # 1 判断上游数据表是否生产完成
+    project = "loghubods"
+    table = "alg_vid_vov_new"
+    run_flag = True
+    begin_ts = int(time.time())
+    table_data_cnt = 0
+    while run_flag:
+        if int(time.time()) - begin_ts >= 60 * 60 * 5:
+            log_.info("等待上游数据超过5小时了，认为失败退出:过了{}秒。".format(int(time.time()) - begin_ts))
+            sys.exit(1)
+        table_data_cnt = check_data(project, table, date, hour, mm)
+        if table_data_cnt == 0:
+            log_.info("上游数据{}未就绪{}，等待...".format(table, date))
+            log_.info("等待30分钟")
+            time.sleep(60 * 30)
+        else:
+            run_flag = False
+
+    log_.info("上游数据就绪，count={}，开始读取数据表".format(table_data_cnt))
+    # 2 读取数据表 处理特征
+    video_list = get_sql(project, table, date, hour, mm)
+    # 3 写入redis
+    log_.info("video的数据量:{}".format(len(video_list)))
+    records_process_for_list(video_list, process_and_store, max_size=50, num_workers=8)
+
+if __name__ == '__main__':
+    log_.info("开始执行:" + datetime.now().strftime("%Y-%m-%d %H:%M:%S"))
+    process = Process(target=main)
+    process.start()
+    # 等待子进程完成或超时
+    timeout = 3600 * 6
+    process.join(timeout=timeout)  # 设置超时为3600秒（1小时）
+    if process.is_alive():
+        print("脚本执行时间超过6小时，执行失败，经过了{}秒。".format(timeout))
+        process.terminate()  # 终止子进程
+        sys.exit(1)  # 直接退出主进程并返回状态码999
+    log_.info("完成执行:" + datetime.now().strftime("%Y-%m-%d %H:%M:%S"))
+
+
+
+# cd /root/zhangbo/rov-offline
+# python alg_recsys_feature_04_vidvov1d3d_redis.py 20240731 14
+
+"""
+    ！！！！！！！！！！！！！！
+    更改字段：table 表名
+            REDIS_PREFIX redis的key
+            EXPIRE_TIME redis的过期时间
+            两段sql 各种字段 注意分区是否有“分钟”
+            record 各种字段
+            if hour in ["00"]: 哪些小时不执行
+            process.join(timeout=3600) 任务超时时间3600
+            int(time.time()) - begin_ts >= 60*50 任务超时时间3000
+"""

write_redis/alg_recsys_feature_05_vidvovdaily4filter_redis_task.sh

@@ -0,0 +1,40 @@
+source /etc/profile
+echo $ROV_OFFLINE_ENV
+
+log_dir="my_logs_alg_recsys_feature_05"
+if [ ! -d ${log_dir} ]; then
+    mkdir ${log_dir}
+fi
+
+cur_time="`date --date='-1 day' +%Y%m%d`"  # 注意 这是前一天 不要复制错了
+cur_h="`date +%H`"
+echo "开始执行时间:{$(date "+%Y-%m-%d %H:%M:%S")}"
+if [[ $ROV_OFFLINE_ENV == 'test' ]]; then
+  /root/anaconda3/bin/python alg_recsys_feature_05_vidvovdaily4filter_redis.py $cur_time $cur_h
+elif [[ $ROV_OFFLINE_ENV == 'pro' ]]; then
+  /root/anaconda3/bin/python alg_recsys_feature_05_vidvovdaily4filter_redis.py $cur_time $cur_h
+  if [ $? -ne 0 ]; then
+    msg="写redis的任务，视频特征排序使用，vov使用的公式1d3d，alg_recsys_feature_05_vidvovdaily4filter_redis.py:${cur_time}-${cur_h}-something-is-wrong."
+    /root/anaconda3/bin/python utils_monitor.py ${msg}
+  fi
+fi
+echo "结束执行时间:{$(date "+%Y-%m-%d %H:%M:%S")}"
+echo "all done"
+
+day="$(date -d '10 days ago' +%Y-%m-%d)"
+base_path=${log_dir}
+pattern="${base_path}/${day}*"
+if compgen -G "$pattern" > /dev/null; then
+    echo "找到匹配的文件，正在删除..."
+    # 使用find命令查找并删除匹配的文件
+    find "$base_path" -name "${day}*" -exec rm -rf {} \;
+    echo "文件已被删除。"
+else
+    echo "没有找到匹配的文件，无需删除。"
+fi
+
+
+# sh alg_recsys_feature_05_vidvovdaily4filter_redis_task.sh
+
+
+#my_logs_alg_recsys_feature_05

write_redis/alg_recsys_task_doudi_videos.py

@@ -0,0 +1,177 @@
+# -*- coding: utf-8 -*-
+import os
+import sys
+import traceback
+
+root_dir = os.path.dirname(os.path.dirname(os.path.realpath(__file__)))
+if root_dir not in sys.path:
+    sys.path.append(root_dir)
+    print("******** sys.path ********")
+    print(sys.path)
+
+from multiprocessing import Process
+from odps import ODPS
+from threading import Timer
+import threading
+from my_utils import RedisHelper, execute_sql_from_odps, request_post
+from my_config import set_config
+from log import Log
+import json
+from datetime import datetime
+from queue import Queue
+from tqdm import tqdm
+import time
+
+
+DOUDI_URL = "http://recommend-common-internal.piaoquantv.com/longvideoapi/openapi/recommend/updateFallBackVideoListV2"
+
+config_, _ = set_config()
+log_ = Log()
+
+def check_data(project, table,  date, hour, mm) -> int:
+    """检查数据是否准备好,输出数据条数"""
+    odps = ODPS(
+        access_id=config_.ODPS_CONFIG['ACCESSID'],
+        secret_access_key=config_.ODPS_CONFIG['ACCESSKEY'],
+        project=project,
+        endpoint=config_.ODPS_CONFIG['ENDPOINT'],
+        connect_timeout=3000,
+        read_timeout=500000,
+        pool_maxsize=1000,
+        pool_connections=1000
+    )
+    try:
+        t = odps.get_table(name=table)
+        log_.info(f"检查分区是否存在-dt={date}.{hour}")
+        check_res = t.exist_partition(partition_spec=f'dt={date},hour={hour},minute={hour}0000')
+        if check_res:
+            sql = f'select * from {project}.{table} where dt = \"{date}\" and hour = \"{hour}\" and minute = \"{hour}0000\"'
+            log_.info(sql)
+            with odps.execute_sql(sql=sql).open_reader() as reader:
+                data_count = reader.count
+        else:
+            log_.info("表{}分区{}不存在".format(table, date + hour))
+            data_count = 0
+    except Exception as e:
+        log_.error("table:{},date:{}.{} no data. return data_count=0,报错原因是:{}".format(table, date, hour, e))
+        data_count = 0
+    return data_count
+
+def get_sql(project, table, date, hour, mm):
+    sql = '''
+    SELECT  vid
+        ,AVG(rank) AS score
+        ,SUM(1) as cnt
+FROM    (
+            SELECT  a.dt
+                    ,a.hour
+                    ,a.region_provience
+                    ,t1.videoid AS vid
+                    ,t1.pos + 1 AS rank -- pos 从 0 开始，所以 +1 作为 rank
+            FROM    {}.{} a
+            LATERAL VIEW POSEXPLODE(SPLIT(a.videoid_arr,",")) t1 AS
+                    pos
+                    ,videoid
+            WHERE   a.dt = "{}"
+            AND     a.hour = "{}"
+            AND     a.region_provience IN ("上海","中国","云南","内蒙古","北京","吉林","四川","天津","宁夏","安徽","山东","山西","广东","广西","新疆","江苏","江西","河北","河南","浙江","海南","湖北","湖南","甘肃","福建","西藏","贵州","辽宁","重庆","陕西","青海","黑龙江")
+        ) 
+GROUP BY vid
+ORDER BY AVG(rank)
+    '''.format(
+        project, table, date, hour
+    )
+    print("sql:" + sql)
+    records = execute_sql_from_odps(project=project, sql=sql)
+    video_list = []
+    with records.open_reader() as reader:
+        for record in reader:
+            key = record['vid']
+            score = record['score']
+            video_list.append({'videoId': key, 'rovScore': float(score)})
+    return video_list
+
+
+def main():
+    try:
+        date = sys.argv[1]
+        hour = sys.argv[2]
+        # mm = sys.argv[3]
+        mm = "00"
+    except Exception as e:
+        date = datetime.now().strftime('%Y%m%d')
+        hour = datetime.now().hour
+        # mm = datetime.now().minute
+        mm = "00"
+        log_.info("没有读取到参数，采用系统时间：{}".format(e))
+    log_.info("使用时间参数-日期：{}，小时：{}".format(date, str(hour)))
+    if hour in ["23", "00", "01", "02", "03", "04", "05"]:
+        log_.info(f"hour={hour}不执行，直接返回。")
+        return
+    # 1 判断上游数据表是否生产完成
+    project = "loghubods"
+    table = "alg_recsys_recall_region_1h_v1"
+    run_flag = True
+    begin_ts = int(time.time())
+    table_data_cnt = 0
+    while run_flag:
+        if int(time.time()) - begin_ts >= 60 * 40:
+            log_.info("等待上游数据超过40分钟了，认为失败退出:过了{}秒。".format(int(time.time()) - begin_ts))
+            sys.exit(1)
+        table_data_cnt = check_data(project, table, date, hour, mm)
+        if table_data_cnt == 0:
+            log_.info("上游数据{}未就绪{}，等待...".format(table, date))
+            log_.info("等待10分钟")
+            time.sleep(60 * 10)
+        else:
+            run_flag = False
+
+    log_.info("上游数据就绪，count={}，开始读取数据表".format(table_data_cnt))
+    # 2 读取数据表 处理特征
+    video_list = get_sql(project, table, date, hour, mm)
+    # 3 通过url请求，写入后端。
+    video_list = video_list[:2000]
+    log_.info("video的数据量:{}".format(len(video_list)))
+    log_.info("10个videos:{}".format(video_list[:10]))
+    result = request_post(request_url=DOUDI_URL, request_data={'videos': video_list})
+    log_.info("请求结果result={}".format(str(result)))
+    if result is None:
+        msg = "请求失败1"
+        log_.error(msg)
+        sys.exit(1)
+    elif result['code'] == 0:
+        msg = "请求成功"
+        log_.info(msg)
+    else:
+        msg = "请求失败2"
+        log_.error(msg)
+        sys.exit(1)
+
+
+if __name__ == '__main__':
+    log_.info("开始执行:" + datetime.now().strftime("%Y-%m-%d %H:%M:%S"))
+    process = Process(target=main)
+    process.start()
+    # 等待子进程完成或超时
+    timeout = 3600 * 1
+    process.join(timeout=timeout)  # 设置超时为3600秒（1小时）
+    if process.is_alive():
+        print("脚本执行时间超过1小时，执行失败，经过了{}秒。".format(timeout))
+        process.terminate()  # 终止子进程
+        sys.exit(1)  # 直接退出主进程并返回状态码999
+    log_.info("完成执行:" + datetime.now().strftime("%Y-%m-%d %H:%M:%S"))
+
+
+
+# cd /root/zhangbo/rov-offline
+# python alg_recsys_task_doudi_videos.py 20240731 14
+
+"""
+    ！！！！！！！！！！！！！！ 通过url，给后端传送兜底视频。每小时执行一次，方案是从loghubods.alg_recsys_recall_region_1h_v1 的单路召回源中挑选视频。
+    更改字段：table 表名
+            两段sql 各种字段 注意分区是否有“分钟”
+            record 各种字段
+            if hour in ["00"]: 哪些小时不执行
+            process.join(timeout=3600) 任务超时时间3600
+            int(time.time()) - begin_ts >= 60*50 任务超时时间3000
+"""

write_redis/alg_recsys_task_doudi_videos_task.sh

@@ -0,0 +1,40 @@
+source /etc/profile
+echo $ROV_OFFLINE_ENV
+
+log_dir="my_logs_alg_recsys_task_doudi_videos"
+if [ ! -d ${log_dir} ]; then
+    mkdir ${log_dir}
+fi
+
+cur_time="`date +%Y%m%d`"
+cur_h="`date +%H`"
+echo "开始执行时间:{$(date "+%Y-%m-%d %H:%M:%S")}"
+if [[ $ROV_OFFLINE_ENV == 'test' ]]; then
+  /root/anaconda3/bin/python alg_recsys_task_doudi_videos.py $cur_time $cur_h
+elif [[ $ROV_OFFLINE_ENV == 'pro' ]]; then
+  /root/anaconda3/bin/python alg_recsys_task_doudi_videos.py $cur_time $cur_h
+  if [ $? -ne 0 ]; then
+    msg="写redis的任务，给前端的兜底视频刷新，alg_recsys_task_doudi_videos.py:${cur_time}-${cur_h}-something-is-wrong."
+    /root/anaconda3/bin/python utils_monitor.py ${msg}
+  fi
+fi
+echo "结束执行时间:{$(date "+%Y-%m-%d %H:%M:%S")}"
+echo "all done"
+
+day="$(date -d '10 days ago' +%Y-%m-%d)"
+base_path=${log_dir}
+pattern="${base_path}/${day}*"
+if compgen -G "$pattern" > /dev/null; then
+    echo "找到匹配的文件，正在删除..."
+    # 使用find命令查找并删除匹配的文件
+    find "$base_path" -name "${day}*" -exec rm -rf {} \;
+    echo "文件已被删除。"
+else
+    echo "没有找到匹配的文件，无需删除。"
+fi
+
+
+# sh alg_recsys_task_doudi_videos_task.sh
+
+
+# my_logs_alg_recsys_task_doudi_videos