增加日志记录文件1.txt

增加数据库连接失败时的处理逻辑

service_status_exporter/service_monitor.py

@@ -13,8 +13,20 @@ import time
 from time import sleep
 from prometheus_client import start_http_server, Gauge
 import argparse
-import pymysql
+from pymysql import connect
 import requests
+import logging
+import os
+current_file_path = os.path.abspath(__file__)
+current_dir_path = os.path.dirname(current_file_path)
+
+logging.basicConfig(
+    filename=f'{current_dir_path}/1.log',
+    # encoding='utf-8',
+    format='%(asctime)s - %(name)s - %(levelname)s - %(module)s:%(message)s',
+    datefmt='%Y-%m-%d %H:%M:%S',
+    level=logging.INFO
+)
 
 
 def service_status():
@@ -29,7 +41,7 @@ def service_status():
 
     # 连接到数据库
     try:
-        connection = pymysql.connect(**config)
+        connection = connect(**config)
         if connection.open:
             # 创建一个游标对象
             cursor = connection.cursor()
@@ -51,7 +63,8 @@ def service_status():
             for record in records:
                 try:
                     status = requests.get(record[2]).status_code
-                except Exception:
+                except Exception as e:
+                    logging.error(f"访问{record[2]}失败，错误为：{e}")
                     # 当网页无法访问时，将状态码设置为777
                     status = 777
                 result.append({
@@ -63,17 +76,14 @@ def service_status():
                     'owner_email': record[4],
                     'chk_time': time.strftime('%Y-%m-%d %H:%M:%S', time.localtime())
                 })
-
+            # 关闭游标和连接
+            cursor.close()
+            connection.close()
             return result
 
     except Exception as e:
-        print(f"数据库连接错误: {e}")
-
-    finally:
-        # 关闭游标和连接
-        if connection.open:
-            cursor.close()
-            connection.close()
+        logging.error(f"数据库连接错误: {e}")
+        return False
 
 
 if __name__ == '__main__':
@@ -84,7 +94,7 @@ if __name__ == '__main__':
 
     # 解析命令行参数
     parser = argparse.ArgumentParser(description='这是一个服务状态检查程序，说明如下：')
-    parser.add_argument('--port', '-p', default='8761', help='监听端口，8761')
+    parser.add_argument('--port', '-p', default='18000', help='监听端口，8761')
     parser.add_argument('--db_ip', default='114.116.11.225', help='指定mysql的ip，默认为114.116.11.225')
     parser.add_argument('--db_port', default='3306', help='指定mysql端口，默认为3306')
     parser.add_argument('--db_user', default='monitor', help='指定mysql的用户名，monitor')
@@ -102,11 +112,13 @@ if __name__ == '__main__':
     while True:
         # 更新指标的值
         res = service_status()
-        metric.clear()
-        for info in res:
-            metric.labels(project=info['project'], service_name=info['service_name'],
-                          url=info['url'], owner_name=info['owner_name'],
-                          owner_email=info['owner_email'], chk_time=info['chk_time']).set(info['status'])
-
+        if res:
+            metric.clear()
+            for info in res:
+                metric.labels(project=info['project'], service_name=info['service_name'],
+                              url=info['url'], owner_name=info['owner_name'],
+                              owner_email=info['owner_email'], chk_time=info['chk_time']).set(info['status'])
+        else:
+            logging.error("数据获取失败")
         # 每隔30秒更新一次指标的值
-        sleep(30)
+        sleep(60)