当从不同设备完成许多连接时，pyodbc cursor.execute 冻结

Question

几乎 10 raspberry pi 使用与下面提到的相同的脚本并尝试每分钟（几乎同时）将数据发送到数据库。

当只有一台设备与数据库通信时几乎没有问题，但当我开始添加更多设备时问题就开始了 raspberry pi。


Code:

import pyodbc 

def update_database():
        try:            
            mydb = pyodbc.connect('DRIVER=FreeTDS;SERVER=192.xxx.xxx.xxx;PORT=xxxx;DATABASE=xxxxxxx;UID=xxxxxxx;PWD=xxxxxxx;TDS_Version=4.2;')
            mycursor = mydb.cursor()

            if mycursor.tables(table='DBDBDBDB', tableType='TABLE').fetchone():
                print("DB exists")
            else:
                print("DB absent")
                DB_Connection_status=0


            ycursor = mydb.cursor()            
            sql = "INSERT INTO DBDBDBDB (docdt, timestamp, cocd, param1, param2, param3, param4, param5, param6, param7, param8) \
                    VALUES (?,?,?,?,?,?,?,?,?,?,?)"


            DB_Connection_status=1
            systemlog("DB Connected")
        except:
            DB_Connection_status=0
            print ("Connection with database: failed")
            systemlog("DB Not Connected")



        for index_y in range (0,6):
            #few lines of code here

            for index_x in range (0,60): 
                #few lines of code here

                if(DB_Connection_status==1):

                    val = (formatted_date, formatted_time, COCD, str(var1), \
                           str(var2), str(var3),\
                           'A','1', str( var4[index_y][index_x]), \
                           str(var5[index_y][toggle_sw_num-1]),0)
                    try:
                        mycursor.execute(sql, val)
                    except:
                        systemlog("DB record skipped")


        if(DB_Connection_status==1):
            try:
                mydb.commit()
                print(mycursor.rowcount, "record inserted.")
                systemlog("Record inserted")
            except:
                systemlog("DB commit skipped")

while 1:
    #few lines of code here
    if(system_minute!=oldsystem_minute):
        oldsystem_minute=system_minute
        #few lines of code here
        update_database()

最初在我为游标执行和提交添加错误处理之前抛出如下所示的错误

Traceback (most recent call last):
  File "mod.py", line 461, in <module>
  File "mod.py", line 213, in update_database
pyodbc.ProgrammingError: ('42000', '[42000] [FreeTDS][SQL Server]Transaction (Process ID 52) was deadlocked on lock resources with another process and has been chosen as the deadlock victim. Rerun the transaction. (1205) (SQLExecDirectW)')

但是这个错误处理只是为了避免代码崩溃。

代码顺序有问题吗？例如，每次执行游标时我都应该调用 commit/close 吗？有什么帮助吗？

OS: raspbian 运行 raspberry pi, Python: 2.7, 数据库：MSSQL

谢谢

Answer 1

should I call commit/close every time on cursor execute?

是的。 commit()。插入的行和索引上的锁在事务期间保持。 commit() 将释放该会话的所有锁。

如果在每行之后提交是一个性能问题，您可以在单个语句中打开delayed durability, or, instead of a nested loop with single-row inserts, build a JSON document with all 360 rows, send that to SQL Server and have SQL Server parse and INSERT它。

当从不同设备完成许多连接时，pyodbc cursor.execute 冻结

pyodbc cursor.execute freezes when many connection are done from different devices

python

linux

sql-server

pyodbc

raspberry-pi3