sql - 如何在三个不同的表中并行插入数据

标签 sql sql-server sql-server-2008 tsql sql-server-2012

我有一个存储过程,它将插入大量记录,现在是否有可能将数据并行插入到3个中;

  • 第一个插入 100 万条记录。
  • 第二个插入 150 万条记录。
  • 第三个插入 500k 条记录

据我所知 - 过程插入正在相继发生。

那么如何实现并行加载呢?

最佳答案

语句在 T-SQL 批处理中同步执行。要从存储过程异步并行执行多个语句,您需要使用多个并发数据库连接。请注意,异步执行的棘手部分不仅是确定所有任务何时完成,还要确定它们是成功还是失败。

方法一:SSIS包

创建一个 SSIS 包来并行执行 3 个 SQL 语句。在 SQL 2012 及更高版本中,使用 SSIS 目录存储过程运行包。在 SQL 2012 之前的版本中,您需要为包创建 SQL 代理作业并使用 sp_start_job 启动。

您需要检查 SSIS 执行状态或 SQL 代理作业状态以确定完成情况以及成功/失败结果。

方法2:Powershell和SQL代理

执行 SQL 代理作业,该作业运行 Powershell 脚本,该脚本使用 Powershell 后台作业(Start-Job 命令)并行执行查询。该脚本可以返回退出代码,零表示成功,非零表示失败,以便 SQL 代理可以确定它是否成功。检查 SQL 代理作业状态以确定完成情况以及成功/失败结果。

方法 3:多个 SQL 代理作业

同时执行多个 SQL 代理作业,每个作业都有一个包含导入脚本的 T-SQL 作业步骤。检查每个作业的 SQL Agent 作业状态以确定完成情况以及成功/失败结果。

方法 4:服务代理 使用队列激活过程并行执行导入脚本。如果您以前没有使用过 Service Broker,那么这可能会很迟钝,并且遵循经过审查的模式很重要。我提供了一个示例来帮助您入门(对于 SQL 2012 之前的版本,将 THROW 替换为 RAISERROR)。数据库必须启用 Service Broker,默认情况下启用,但在还原或附加后关闭。

USE YourDatabase;
Go

--create proc that will be automatically executed (activated) when requests are waiting
CREATE PROC dbo.ExecuteTSqlTask
AS
SET NOCOUNT ON;

DECLARE
      @TSqlJobConversationHandle uniqueidentifier = NEWID()
    , @TSqlExecutionRequestMessage xml
    , @TSqlExecutionResultMessage xml
    , @TSqlExecutionResult varchar(10)
    , @TSqlExecutionResultDetails nvarchar(MAX)
    , @TSqlScript nvarchar(MAX)
    , @TSqlTaskName sysname
    , @RowsAffected int
    , @message_type_name sysname;

WHILE 1 = 1
BEGIN

    --get the next task to execute
    WAITFOR (
        RECEIVE TOP (1)
              @TSqlJobConversationHandle = conversation_handle
            , @TSqlExecutionRequestMessage = CAST(message_body AS xml)
            , @message_type_name = message_type_name
        FROM dbo.TSqlExecutionQueue
        ), TIMEOUT 1000;

    IF @@ROWCOUNT = 0
    BEGIN
        --no work to do - exit
        BREAK;
    END;

    IF @message_type_name = N'TSqlExecutionRequest'
    BEGIN

        --get task name and script
        SELECT
              @TSqlTaskName = @TSqlExecutionRequestMessage.value('(/TSqlTaskName)[1]', 'sysname')
            , @TSqlScript = @TSqlExecutionRequestMessage.value('(/TSqlScript)[1]', 'nvarchar(MAX)');

        --execute script
        BEGIN TRY
            EXEC sp_executesql @TSqlScript;
            SET @RowsAffected = @@ROWCOUNT;
            SET @TSqlExecutionResult = 'Completed';
            SET @TSqlExecutionResultDetails = CAST(@RowsAffected as varchar(10)) + ' rows affected';
        END TRY
        BEGIN CATCH
            SET @TSqlExecutionResult = 'Erred';
            SET @TSqlExecutionResultDetails = 
                  'Msg ' + CAST(ERROR_NUMBER() AS varchar(10))
                + ', Level ' + CAST(ERROR_SEVERITY() AS varchar(2))
                + ', State ' + CAST(ERROR_STATE() AS varchar(10))
                + ', Line ' + CAST(ERROR_LINE() AS varchar(10))
                + ': ' + ERROR_MESSAGE();
        END CATCH;

        --send execution result back to initiator
        SET @TSqlExecutionResultMessage = '<TSqlTaskName /><TSqlExecutionResult /><TSqlExecutionResultDetails />';
        SET @TSqlExecutionResultMessage.modify('insert text {sql:variable("@TSqlTaskName")} into (/TSqlTaskName)[1] ');
        SET @TSqlExecutionResultMessage.modify('insert text {sql:variable("@TSqlExecutionResult")} into (/TSqlExecutionResult)[1] ');
        SET @TSqlExecutionResultMessage.modify('insert text {sql:variable("@TSqlExecutionResultDetails")} into (/TSqlExecutionResultDetails)[1] ');
        SEND ON CONVERSATION @TSqlJobConversationHandle
            MESSAGE TYPE TSqlExecutionResult
            (@TSqlExecutionResultMessage);

    END
    ELSE
    BEGIN
        IF @message_type_name = N'TSqlJobComplete'
        BEGIN
            --service has ended conversation so we're not going to get any more execution requests
            END CONVERSATION @TSqlJobConversationHandle;
        END
        ELSE
        BEGIN
            END CONVERSATION @TSqlJobConversationHandle WITH ERROR = 1 DESCRIPTION = 'Unexpected message type received by ExecuteTSqlTask';
            RAISERROR('Unexpected message type received (%s) by ExecuteTSqlTask', 16, 1, @message_type_name);
        END;
    END;
END;
GO

CREATE QUEUE dbo.TSqlResultQueue;
CREATE QUEUE dbo.TSqlExecutionQueue
    WITH STATUS=ON,
    ACTIVATION (
          STATUS = ON
        , PROCEDURE_NAME = dbo.ExecuteTSqlTask
        , MAX_QUEUE_READERS = 3 --max number of concurrent activated proc instances 
        , EXECUTE AS OWNER
        );
CREATE MESSAGE TYPE TSqlExecutionRequest VALIDATION = WELL_FORMED_XML;
CREATE MESSAGE TYPE TSqlExecutionResult VALIDATION = WELL_FORMED_XML;
CREATE MESSAGE TYPE TSqlJobComplete VALIDATION = WELL_FORMED_XML;
CREATE CONTRACT TSqlExecutionContract (
      TSqlExecutionRequest SENT BY INITIATOR
    , TSqlJobComplete SENT BY INITIATOR
    , TSqlExecutionResult SENT BY TARGET
    );
CREATE SERVICE TSqlJobService ON QUEUE dbo.TSqlResultQueue ([TSqlExecutionContract]);
CREATE SERVICE TSqlExecutorService ON QUEUE dbo.TSqlExecutionQueue ([TSqlExecutionContract]);
GO

CREATE PROC dbo.ExecuteParallelImportScripts
AS
SET NOCOUNT ON;

DECLARE
      @TSqlJobConversationHandle uniqueidentifier
    , @TSqlExecutionRequestMessage xml
    , @TSqlExecutionResultMessage xml
    , @TSqlExecutionResult varchar(10)
    , @TSqlExecutionResultDetails nvarchar(MAX)
    , @TSqlTaskName sysname
    , @CompletedCount int = 0
    , @ErredCount int = 0
    , @message_type_name sysname;

DECLARE @TsqlTask TABLE(
      TSqlTaskName sysname NOT NULL PRIMARY KEY 
    , TSqlScript nvarchar(MAX) NOT NULL
    );

BEGIN TRY

    --insert a row for each import task
    INSERT INTO @TsqlTask(TSqlTaskName, TSqlScript) 
        VALUES(N'ImportScript1', N'INSERT INTO dbo.Table1 SELECT * FROM dbo.Table1Staging;');
    INSERT INTO @TsqlTask(TSqlTaskName, TSqlScript) 
        VALUES(N'ImportScript2', N'INSERT INTO dbo.Table2 SELECT * FROM dbo.Table2Staging;');
    INSERT INTO @TsqlTask(TSqlTaskName, TSqlScript) 
        VALUES(N'ImportScript3', N'INSERT INTO dbo.Table3 SELECT * FROM dbo.Table3Staging;');

    --start a conversation for this import process
    BEGIN DIALOG CONVERSATION @TsqlJobConversationHandle
        FROM SERVICE TSqlJobService
        TO SERVICE 'TSqlExecutorService', 'CURRENT DATABASE'
        ON CONTRACT TSqlExecutionContract
        WITH ENCRYPTION = OFF;

    --send import tasks to executor service for parallel execution
    DECLARE JobTasks CURSOR LOCAL FAST_FORWARD FOR
        SELECT (SELECT TSqlTaskName, TSqlScript
            FROM @TsqlTask AS task 
            WHERE task.TSqlTaskName = job.TSqlTaskName
            FOR XML PATH(''), TYPE) AS TSqlExecutionRequest
        FROM @TsqlTask AS job;
    OPEN JobTasks;
    WHILE 1 = 1
    BEGIN
        FETCH NEXT FROM JobTasks INTO @TSqlExecutionRequestMessage;
        IF @@FETCH_STATUS = -1 BREAK;
        SEND ON CONVERSATION @TSqlJobConversationHandle
            MESSAGE TYPE TSqlExecutionRequest
            (@TSqlExecutionRequestMessage);
    END;
    CLOSE JobTasks;
    DEALLOCATE JobTasks;

    --get each parallel task execution result until all are complete
    WHILE 1 = 1
    BEGIN

        --get next task result
        WAITFOR (
            RECEIVE TOP (1)
                  @TSqlExecutionResultMessage = CAST(message_body AS xml)
                , @message_type_name = message_type_name
            FROM dbo.TSqlResultQueue
            WHERE conversation_handle = @TSqlJobConversationHandle
            ), TIMEOUT 1000;

        IF @@ROWCOUNT <> 0
        BEGIN

            IF @message_type_name = N'TSqlExecutionResult'
            BEGIN

                --get result of import script execution
                SELECT
                      @TSqlTaskName = @TSqlExecutionResultMessage.value('(/TSqlTaskName)[1]', 'sysname')
                    , @TSqlExecutionResult = @TSqlExecutionResultMessage.value('(/TSqlExecutionResult)[1]', 'varchar(10)')
                    , @TSqlExecutionResultDetails = COALESCE(@TSqlExecutionResultMessage.value('(/TSqlExecutionResultDetails)[1]', 'nvarchar(MAX)'), N'');
                RAISERROR('Import task %s %s: %s', 0, 0, @TSqlTaskName, @TSqlExecutionResult, @TSqlExecutionResultDetails) WITH NOWAIT;
                IF @TSqlExecutionResult = 'Completed'
                BEGIN
                    SET @CompletedCount += 1;
                END
                ELSE
                BEGIN
                    SET @ErredCount += 1;
                END;

                --remove task from tracking table after completion
                DELETE FROM @TSqlTask
                WHERE TSqlTaskName = @TSqlTaskName;

                IF NOT EXISTS(SELECT 1 FROM @TsqlTask)
                BEGIN
                    --all tasks are done - send TSqlJobComplete message to instruct executor service to end conversation
                    SEND ON CONVERSATION @TSqlJobConversationHandle
                        MESSAGE TYPE TSqlJobComplete;
                END
            END
            ELSE
            BEGIN
                IF @message_type_name = N'http://schemas.microsoft.com/SQL/ServiceBroker/EndDialog'
                BEGIN
                    --executor service has ended conversation so we're done
                    END CONVERSATION @TSqlJobConversationHandle;
                    BREAK;
                END
                ELSE
                BEGIN
                    END CONVERSATION @TSqlJobConversationHandle WITH ERROR = 1 DESCRIPTION = 'Unexpected message type received by ExecuteParallelInserts';
                    RAISERROR('Unexpected message type received (%s) by ExecuteParallelInserts', 16, 1, @message_type_name);
                END;
            END
        END;
    END;
    RAISERROR('Import processing completed. CompletedCount=%d, ErredCount=%d.', 0, 0, @CompletedCount, @ErredCount);
END TRY
BEGIN CATCH
    THROW;
END CATCH;
GO

--execute import scripts in parallel
EXEC dbo.ExecuteParallelImportScripts;
GO

关于sql - 如何在三个不同的表中并行插入数据,我们在Stack Overflow上找到一个类似的问题: https://stackoverflow.com/questions/39632229/

相关文章:

SQL 从数据库获取 X 个帐户,这可能是可变的行数

sql - 在 SQL 表的每一行中查找最后一个 NOT NULL 列的值

sql - Access 创建表vba

php - 如何在 MYSQL 中正确选择最高分?

java - ClassNotFoundException 和 NPE 同时出现

c# - 数据绑定(bind)到 asp.net 中的 html 标记

sql-server-2008 - 删除多行上的触发器

mysql - 一张表中两个表的信息

json - 用于 json 自动的 sql 服务器。如何获得所有结果

SQL查询来比较彼此之间的行子集