基于SQL Server的简单数据同步方案
软件系统中经常需要进行数据同步,如 C/S 程序为了支持离线应用和服务端之间双向同步数据、数据库集群中主服务器向从服务器同步数据、业务子系统之间同步共用的数据等。
不同需求下的数据同步方法:
a)在 C/S 客户端只有数据查询,而且同步的数据比较少时
1、可以在每次同步时先清空客户端(目标)数据库表(truncate(不会记录删除日志且不会触发触发器)/delete(会记录行删除日志))的数据,然后直接从服务器写入最新的所有数据到客户端客户端数据库表。
优点:程序简单,不需要区分哪些是新增的数据哪些是修改的数据。
缺点:当数据量很大时,select 全表扫描会产生性能问题。
2、可以在每次同步时先清空客户端(目标)数据库表(truncate(不会记录删除日志且不会触发触发器)/delete(会记录行删除日志))的数据,然后把服务器端大量的数据按分页的方式写到客户端数据库表。
public class SynchronizeSqlStr { public static readonly string TruncateSql = null; public static readonly string[] GetRecordCountStr = null; public static readonly string[] AllSynchronizeSqlStr = null; static SynchronizeSqlStr() { TruncateSql = @"TRUNCATE TABLE T_Target;"; GetRecordCountStr = new string[] { "SELECT COUNT(1) AS Co FROM T_Source;", }; AllSynchronizeSqlStr = new string[] //注意:这里每次都要排序,会影响查询效率,如果改成 ID 范围会更好!! { @"INSERT INTO dbo.T_Target(TId, TName, DeptId) SELECT S.TId, S.TName, S.DeptId FROM (SELECT ROW_NUMBER() OVER (ORDER BY TId) AS ROWNUM, TId, TName, DeptId FROM dbo.T_Source) AS S WHERE S.ROWNUM >= {0} AND S.ROWNUM < {1};", }; } }
public class HelloJob : IJob { private static readonly NLog.Logger log = NLog.LogManager.GetCurrentClassLogger(); private static readonly string connStr = ConfigurationManager.AppSettings["ConnectionString"].ToString(); int pageSize = 1000; // 默认每次同步1000条记录 public virtual Task Execute(IJobExecutionContext context) { try { DbHelperSQL.connectionString = DESEncrypt.Decrypt(connStr); // 步骤一:清理客户端的历史数据(注意:照片表采取增量同步方式,且不删除历史记录) DbHelperSQL.ExecuteSqlTran(SynchronizeSqlStr.TruncateSql); // 步骤二:依次同步服务器的数据到客户端 for (int i = 0; i < SynchronizeSqlStr.AllSynchronizeSqlStr.Length; i++) { try { log.Info(string.Format("准备异步执行脚本:" + i.ToString() + " - {0}", DateTime.Now)); Thread thread = new Thread(SynchronizeDataByPager); thread.IsBackground = true; thread.Start(SynchronizeSqlStr.GetRecordCountStr[i] + "~" + SynchronizeSqlStr.AllSynchronizeSqlStr[i]); } catch (Exception e) { log.Error("异步执行脚本:" + i.ToString() + " 时出错:\r\n" + e.ToString() + " " + DateTime.Now + "\r\n"); break; } Thread.Sleep(200); } log.Info(string.Format("Synchronize 完成! - {0}", DateTime.Now)); return Task.FromResult(true); } catch (Exception ex) { log.Error("执行Execute方法时出错:" + ex.ToString() + " " + DateTime.Now + "\r\n"); return Task.FromResult(false); } } /// <summary> /// 按记录总条数进行分页,然后每次同步一页的数据,防止数据库操作超时 /// </summary> private void SynchronizeDataByPager(object querySqlAndUpdateSql) { try { string sql = Convert.ToString(querySqlAndUpdateSql); if (string.IsNullOrEmpty(sql)) return; string querySql = sql.Split(new char[] { '~' })[0]; string updateSql = sql.Split(new char[] { '~' })[1]; DataSet ds = DbHelperSQL.Query(querySql); if (ds == null || ds.Tables == null || ds.Tables.Count < 1 || ds.Tables[0].Rows == null || ds.Tables[0].Rows.Count < 1) return; int recordCount = 0; int.TryParse(ds.Tables[0].Rows[0]["Co"].ToString(), out recordCount); if (recordCount < 1) return; int pageCount = recordCount / pageSize + 1; for (int i = 0; i < pageCount; i++) { string eachPageSql = string.Format(updateSql, i * pageSize, i * pageSize + pageSize); DbHelperSQL.ExecuteSqlTran(eachPageSql); Thread.Sleep(TimeSpan.FromMilliseconds(100)); } } catch (Exception ex) { log.Error("执行分页同步操作时出错:" + ex.ToString() + " " + DateTime.Now + "\r\n"); } } }
b)在 C/S 客户端既有数据查询也有数据修改,而且的数据比较多时(如:照片表可能比较大)
1.如果源数据库中的表设计规范,即表中包含 ID(自增长,当数据重复时这个字段很有用)、CreateTime、UpdateTime、IsDelete 字段,而且没有物理删除,这时在每次同步时可以先取目的表中最新的 UpdateTime或数据库的Timespan 值,保存到配置文件中,然后从源表中筛选大于或等于 UpdateTime 的记录,然后通过比对 Id 来区分哪些是新增记录,哪些是修改记录(从源表中筛选大于目标表最大 Id 的记录,即为新增的记录,其他的则是修改的记录),然后分别新增或修改到目的表中。
1.1 利用存储过程来实现
CREATE PROCEDURE [DBO].[P_SynchronizeData] AS BEGIN SET NOCOUNT ON; DECLARE @Max_UpdateTime NVARCHAR(100); DECLARE @SQL1 NVARCHAR(1000); DECLARE @SQL2 NVARCHAR(1000); SELECT @Max_UpdateTime = MAX(UpdateTime) FROM T_Target; IF (@Max_UpdateTime IS NULL) BEGIN INSERT INTO T_Target(Id,Name) SELECT * FROM OPENQUERY(HTIMSDB,'SELECT Id,Name FROM T_Source '); END ELSE BEGIN --新增的记录 SET @SQL1 = 'SELECT Id,Name FROM T_Source WHERE UpdateTime >'''+ @Max_UpdateTime + ''''; SET @SQL2 = 'INSERT INTO T_Target(Id,Name) VALUES(@SQL1)' EXEC(@SQL2) --修改的记录(修改时可以先进行物理删除然后进行新增
--删除的记录 END END
1.2 利用 c# 程序来实现
// 同步 HT_POWER 表的记录 // 步骤一:先把源表的数据同步到目的表,不管RECORD_COUNTER字段的值(注意:这里把 RECORD_COUNTER 赋值为 0,是为了后面方便更新其值) //门禁库中全部的权限记录 DataSet sourceData = SQLHelper.Query(_connStr, "SELECT ControllerID + '00' + cast(DoorNo as varchar(35)) + StaffID as sourceId FROM [SKEP_DAS].[dbo].[DAS_DoorPoint];"); List<string> sourceDataIds = new List<string>(); if (!IsNullDataSet(sourceData)) { int count = sourceData.Tables[0].Rows.Count; for (int i = 0; i < count; i++) { try { sourceDataIds.Add(sourceData.Tables[0].Rows[i]["sourceId"].ToString()); } catch (Exception e) { } } } //门禁库中最近更新过的的记录集合 object lastUpdateTime = SQLHelper.ExecuteScalar(_connStr, CommandType.Text, "SELECT max(InsertDateTime) as InsertDateTime FROM [HT_ACCESS].[dbo].[HT_POWER];"); string lastUpdateTimeStr = "1900-01-01 00:00:00"; if (lastUpdateTime != null) lastUpdateTimeStr = Convert.ToDateTime(lastUpdateTime).ToString("yyyy-MM-dd HH:mm:ss"); DataSet sourceUpdateData = SQLHelper.Query(_connStr, "SELECT ControllerID + '00' + cast(DoorNo as varchar(35)) + StaffID as sourceId FROM [SKEP_DAS].[dbo].[DAS_DoorPoint] where InsertDateTime > '" + lastUpdateTimeStr + @"'"); List<string> sourceUpdateDataIds = new List<string>(); if (!IsNullDataSet(sourceUpdateData)) //IsNullDataSet方法见文末 { int count = sourceUpdateData.Tables[0].Rows.Count; for (int i = 0; i < count; i++) { try { sourceUpdateDataIds.Add(sourceUpdateData.Tables[0].Rows[i]["sourceId"].ToString()); } catch (Exception e) { } } } //中间库中全部(未删除)的权限记录 DataSet targetData = SQLHelper.Query(_connStr, "SELECT DOOR_NO+EMP_NO as targetId FROM [HT_ACCESS].[dbo].[HT_POWER] WHERE [DELETE_FLAG] = 0;"); List<string> targetDataIds = new List<string>(); if (!IsNullDataSet(targetData)) { int count = targetData.Tables[0].Rows.Count; for (int i = 0; i < count; i++) { try { targetDataIds.Add(targetData.Tables[0].Rows[i]["targetId"].ToString()); } catch (Exception e) { } } } try { //1.已删除的记录(这里不删只修改DELETE_FLAG=1) List<string> deleteIds = targetDataIds.Where(t => !sourceDataIds.Contains(t)).ToList(); foreach (var id in deleteIds) //依次添加防止RECORD_COUNTER重复 { string clearOldDatas = @"UPDATE [HT_ACCESS].[DBO].HT_POWER SET DELETE_FLAG = 1, RECORD_COUNTER = (SELECT ISNULL(MAX(RECORD_COUNTER),0) + 1 FROM [HT_ACCESS].[dbo].[HT_POWER]) where DOOR_NO+EMP_NO = '" + id + @"' AND [DELETE_FLAG] = 0;"; SQLHelper.ExecuteNonQuery(_connStr, CommandType.Text, clearOldDatas); } //2.新增的记录 List<string> addIds = sourceUpdateDataIds.Where(t => !targetDataIds.Contains(t)).ToList(); foreach (var id in addIds) { string addSql = @"INSERT INTO [HT_ACCESS].[DBO].HT_POWER (POWER_ID,DOOR_NO,EMP_NO,POWER_ZONE_NO,IS_HOLIDAY_ENABLED,InsertDateTime,RECORD_COUNTER) SELECT DoorPointID, ControllerID + '00' + cast(DoorNo as varchar(35)) as [DOOR_ID], StaffID,DoorTztNo,IsHolidayEnabled,InsertDateTime,(SELECT ISNULL(MAX(RECORD_COUNTER),0) + 1 FROM [HT_ACCESS].[dbo].[HT_POWER]) as RECORD_COUNTER FROM [SKEP_DAS].[dbo].[DAS_DoorPoint] where ControllerID + '00' + cast(DoorNo as varchar(35)) + StaffID = '" + id + @"'"; SQLHelper.ExecuteNonQuery(_connStr, CommandType.Text, addSql); } //3.需要修改的记录(由于删除操作会影响RECORD_COUNTER的赋值,所以把删除语句放在insert语句后面执行!否则先执行deleteSql的话,RECORD_COUNTER的值总会少1) List<string> updateIds = sourceUpdateDataIds.Where(t => targetDataIds.Contains(t)).ToList(); //3.1删除旧记录 foreach (var id in updateIds) { string clearNeedUpdateData = @"DELETE FROM [HT_ACCESS].[DBO].HT_POWER where DOOR_NO+EMP_NO = '" + id + @"' AND ISNULL(DOOR_NO,'')<>'' AND ISNULL(EMP_NO,'')<>'';"; SQLHelper.ExecuteNonQuery(_connStr, CommandType.Text, clearNeedUpdateData); } //3.2插入最新的记录 foreach (var id in updateIds) { string insertUpdateData = @"INSERT INTO [HT_ACCESS].[DBO].HT_POWER(POWER_ID,DOOR_NO,EMP_NO,POWER_ZONE_NO,IS_HOLIDAY_ENABLED,InsertDateTime,RECORD_COUNTER) SELECT DoorPointID,ControllerID + '00' + cast(DoorNo as varchar(35)) as [DOOR_ID], StaffID,DoorTztNo,IsHolidayEnabled,InsertDateTime,(SELECT ISNULL(MAX(RECORD_COUNTER),0) + 1 FROM [HT_ACCESS].[dbo].[HT_POWER]) as RECORD_COUNTER FROM [SKEP_DAS].[dbo].[DAS_DoorPoint] where ControllerID + '00' + cast(DoorNo as varchar(35)) + StaffID = '" + id + @"';"; SQLHelper.ExecuteNonQuery(_connStr, CommandType.Text, insertUpdateData); } } catch (Exception e) { logger.LogError("同步 HT_POWER 表的记录时出错:\r\n" + e.ToString() + " " + DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss:fff")); }
2.如果源数据库表设计粗糙,没有 UpdateTime 字段,而且还可能存在物理删除,这时需分两种情况:
2.1如果不允许在源数据库中创建触发器,由于无法区分新增还是修改而且无法直接找到被删除的记录,则只能进行全表比对或者按 1. 中的方法按分页的方式进行同步了。
2.2 如果允许在源数据库中创建触发器和表,则可以在源表中创建相应的触发器来监控增删改操作,然后把操作的表名、类型、主键等关键字段值保存到新建的中间表中,最后创建对应的目的表(包含 Id、CreateTime、UpdateTime、IsDelete、RecordCount 字段,这样如果需要同步这里的数据会简单很多),通过后台程序定时同步数据到目的表中。
问题:当提交批量sql语句,包括增删改时,触发器无法区分操作顺序,顺序错乱的话会导致数据同步不一致!
如下所示:
创建相关的数据库和表:
USE [TestDb] GO /****** Object: StoredProcedure [dbo].[P_TriggerCommon] Script Date: 2019/7/5 13:59:42 ******/ SET ANSI_NULLS ON GO SET QUOTED_IDENTIFIER ON GO -- ============================================= -- Author: wzl -- Create date: 2019/03/01 -- Description: 把各触发器中的公共操作部分提取出来(用于基础信息库) -- ============================================= CREATE PROCEDURE [dbo].[P_TriggerCommon] @SOURCE_TABLE_NAME nvarchar(200), @SOURCE_PRIMARYKEY_NAME nvarchar(200), @SOURCE_PRIMARYKEY_VALUE nvarchar(200), @TYPE SMALLINT AS BEGIN SET NOCOUNT ON; if(not exists(select 1 from [TestDb].[DBO].[OperateData] where [TABLENAME] = @SOURCE_TABLE_NAME and [TYPE] = @TYPE and [PrimaryFeildValue] = @SOURCE_PRIMARYKEY_VALUE)) begin INSERT INTO [TestDb].[DBO].[OperateData]([TABLENAME],[TYPE],[PrimaryFeild],[PrimaryFeildValue],[CreateTime]) VALUES(@SOURCE_TABLE_NAME, @TYPE, @SOURCE_PRIMARYKEY_NAME, @SOURCE_PRIMARYKEY_VALUE, GETDATE()); end else begin UPDATE [TestDb].[DBO].[OperateData] SET [CreateTime] = GETDATE() where [TABLENAME] = @SOURCE_TABLE_NAME and [TYPE] = @TYPE and [PrimaryFeild] = @SOURCE_PRIMARYKEY_NAME and [PrimaryFeildValue] = @SOURCE_PRIMARYKEY_VALUE; end END GO /****** Object: Table [dbo].[OperateData] Script Date: 2019/7/5 13:59:42 ******/ SET ANSI_NULLS ON GO SET QUOTED_IDENTIFIER ON GO CREATE TABLE [dbo].[OperateData]( [Id] [int] IDENTITY(1,1) NOT NULL, [TableName] [nvarchar](200) NOT NULL, [Type] [smallint] NOT NULL, [PrimaryFeild] [nvarchar](200) NOT NULL, [PrimaryFeildValue] [nvarchar](200) NOT NULL, [CreateTime] [datetime] NULL, CONSTRAINT [PK_OperateData] PRIMARY KEY CLUSTERED ( [Id] ASC )WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY] ) ON [PRIMARY] GO /****** Object: Table [dbo].[Student] Script Date: 2019/7/5 13:59:42 ******/ SET ANSI_NULLS ON GO SET QUOTED_IDENTIFIER ON GO CREATE TABLE [dbo].[Student]( [Id] [int] IDENTITY(1,1) NOT NULL, [Number] [nvarchar](200) NOT NULL, [Name] [nvarchar](200) NULL, [Sex] [bit] NULL, [Age] [smallint] NULL, [Version] [timestamp] NULL, CONSTRAINT [PK_Student] PRIMARY KEY CLUSTERED ( [Id] ASC )WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY] ) ON [PRIMARY] GO /****** Object: Table [dbo].[Student_Log] Script Date: 2019/7/5 13:59:42 ******/ SET ANSI_NULLS ON GO SET QUOTED_IDENTIFIER ON GO CREATE TABLE [dbo].[Student_Log]( [Id] [int] NOT NULL, [Number] [nvarchar](200) NOT NULL, [Name] [nvarchar](200) NULL, [Sex] [bit] NULL, [Age] [smallint] NULL, [Version] [timestamp] NULL, [Source] [nvarchar](50) NULL ) ON [PRIMARY] GO
在需要同步数据的表上创建触发器:
USE [TestDb] GO /****** Object: Trigger [dbo].[T_Student] Script Date: 2019/7/5 14:00:28 ******/ SET ANSI_NULLS ON GO SET QUOTED_IDENTIFIER ON GO CREATE TRIGGER [dbo].[T_Student] ON [dbo].[Student] AFTER INSERT,DELETE,UPDATE AS BEGIN --触发器对应的源表名 DECLARE @SOURCE_TABLE_NAME nvarchar(200); SET @SOURCE_TABLE_NAME = '[TestDb].[DBO].[Student]'; --源表中的主键名称 DECLARE @SOURCE_PRIMARYKEY_NAME nvarchar(200); SET @SOURCE_PRIMARYKEY_NAME = 'Number'; --源表中的主键的值 DECLARE @SOURCE_PRIMARYKEY_VALUE nvarchar(200); SET @SOURCE_PRIMARYKEY_VALUE = ''; --触发类型(1 新增、2 删除、3 修改) DECLARE @TYPE SMALLINT; --DML触发器使用了deleted和inserted表,它们保存了被用户修改的行的新值和原来的值 --新增(只有inserted中有记录,说明发生在[DAS_TenantTimezoneTable]表上的所有操作都是新增) IF(EXISTS(SELECT 1 FROM INSERTED) AND NOT EXISTS(SELECT 1 FROM DELETED)) BEGIN SET @TYPE = 1; --test insert into dbo.Student_log ([Id],[Number],[Name],[Sex],[Age],[Source]) select [Id],[Number],[Name],[Sex],[Age],'INSERTED' from INSERTED declare MyCursor cursor for SELECT Number FROM INSERTED; open MyCursor; fetch next from MyCursor into @SOURCE_PRIMARYKEY_VALUE; while @@FETCH_STATUS = 0 begin EXEC DBO.P_TriggerCommon @SOURCE_TABLE_NAME, @SOURCE_PRIMARYKEY_NAME, @SOURCE_PRIMARYKEY_VALUE, @TYPE; fetch next from MyCursor into @SOURCE_PRIMARYKEY_VALUE; end close MyCursor; deallocate MyCursor; END --删除(只有deleted中有记录,说明发生在[DAS_TenantTimezoneTable]表上的所有操作都是删除) ELSE IF(EXISTS(SELECT 1 FROM DELETED) AND NOT EXISTS(SELECT 1 FROM INSERTED)) BEGIN SET @TYPE = 2; --test insert into dbo.Student_log ([Id],[Number],[Name],[Sex],[Age],[Source]) select [Id],[Number],[Name],[Sex],[Age],'DELETED' from DELETED declare MyCursor cursor for SELECT Number FROM DELETED; open MyCursor; fetch next from MyCursor into @SOURCE_PRIMARYKEY_VALUE; while @@FETCH_STATUS = 0 begin EXEC DBO.P_TriggerCommon @SOURCE_TABLE_NAME, @SOURCE_PRIMARYKEY_NAME, @SOURCE_PRIMARYKEY_VALUE, @TYPE; fetch next from MyCursor into @SOURCE_PRIMARYKEY_VALUE; end close MyCursor; deallocate MyCursor; END --新增、删除、修改(inserted和deleted中都有记录,说明发生在[DAS_TenantTimezoneTable]表上的操作可能包含:新增或删除或修改) ELSE IF(EXISTS(SELECT 1 FROM INSERTED) AND EXISTS(SELECT 1 FROM DELETED)) BEGIN --新增的记录 SET @TYPE = 1; --test insert into dbo.Student_log ([Id],[Number],[Name],[Sex],[Age],[Source]) select [Id],[Number],[Name],[Sex],[Age],'INSERTED' from INSERTED --test insert into dbo.Student_log ([Id],[Number],[Name],[Sex],[Age],[Source]) select [Id],[Number],[Name],[Sex],[Age],'DELETED' from DELETED declare MyCursor cursor for select Number from inserted I where Number not in (select Number from deleted); open MyCursor; fetch next from MyCursor into @SOURCE_PRIMARYKEY_VALUE; while @@FETCH_STATUS = 0 begin EXEC DBO.P_TriggerCommon @SOURCE_TABLE_NAME, @SOURCE_PRIMARYKEY_NAME, @SOURCE_PRIMARYKEY_VALUE, @TYPE; fetch next from MyCursor into @SOURCE_PRIMARYKEY_VALUE; end close MyCursor; deallocate MyCursor; --删除的记录 SET @TYPE = 2; declare MyCursor cursor for select Number from deleted I where Number not in (select Number from inserted); open MyCursor; fetch next from MyCursor into @SOURCE_PRIMARYKEY_VALUE; while @@FETCH_STATUS = 0 begin EXEC DBO.P_TriggerCommon @SOURCE_TABLE_NAME, @SOURCE_PRIMARYKEY_NAME, @SOURCE_PRIMARYKEY_VALUE, @TYPE; fetch next from MyCursor into @SOURCE_PRIMARYKEY_VALUE; end close MyCursor; deallocate MyCursor; --修改的记录 SET @TYPE = 3; declare MyCursor cursor for select Number from inserted I where Number in (select Number from deleted); open MyCursor; fetch next from MyCursor into @SOURCE_PRIMARYKEY_VALUE; while @@FETCH_STATUS = 0 begin EXEC DBO.P_TriggerCommon @SOURCE_TABLE_NAME, @SOURCE_PRIMARYKEY_NAME, @SOURCE_PRIMARYKEY_VALUE, @TYPE; fetch next from MyCursor into @SOURCE_PRIMARYKEY_VALUE; end close MyCursor; deallocate MyCursor; END END GO
测试脚本:
--批量执行同类型的语句 INSERT INTO [dbo].[Student] ([Number],[Name],[Sex],[Age]) VALUES ('091001','张三',1,20); INSERT INTO [dbo].[Student] ([Number],[Name],[Sex],[Age]) VALUES ('091002','李四',1,18); --批量执行不同类型的语句 UPDATE Student SET Name = Name + '1' WHERE Number = '091001'; DELETE FROM Student WHERE Number = '091002'; INSERT INTO [dbo].[Student] ([Number],[Name],[Sex],[Age]) VALUES ('091003','王五',1,21); --查看结果 select * FROM [TestDb].[dbo].[Student] select * FROM [TestDb].[dbo].Student_Log select * FROM [TestDb].[dbo].OperateData delete FROM [TestDb].[dbo].[Student] delete FROM [TestDb].[dbo].Student_Log delete FROM [TestDb].[dbo].OperateData dbcc CHECKIDENT('Student',reseed,0) --dbcc CHECKIDENT('Student_Log',reseed,0) dbcc CHECKIDENT('OperateData',reseed,0)
C#后台同步程序(根据操作跟踪表的数据来同步数据):
public class SynchronizeSqlStr { public static readonly string TriggerTrackSql = "SELECT [TableName],[Type],[PrimaryKey] FROM [SKEP_DAS].[dbo].[SKEP_SYNC]"; public static readonly string checkHasDataSql = "SELECT 1 FROM [HT_ACCESS].[dbo].[HT_PHOTO]"; public static readonly string initDataSql = ""; public static readonly string updateRecordCounterSql = "";
public static Dictionary<string, string> TableMappings = new Dictionary<string, string>(); public static readonly string InsertSQLs = @"INSERT INTO {0} ({1}) SELECT {2} FROM {3} WHERE {4} = '{5}';"; public static readonly string DeleteSQLs = "DELETE FROM {0} WHERE {1} = '{2}';"; public static readonly string UpdateSQLs = ""; // DeleteSQLs + "\r\n" + InsertSQLs; public static Dictionary<string, string[]> BaseParams = new Dictionary<string, string[]>(); static SynchronizeSqlStr() { int RecordDay = 1; int.TryParse(ConfigurationManager.AppSettings["RecordDay"].ToString(), out RecordDay); string getRecordDay = DateTime.Now.AddDays(-1 * RecordDay).ToShortDateString(); // 首次启动同步程序时,先初始化已有的数据 initDataSql = @"INSERT INTO [HT_ACCESS].[DBO].HT_PHOTO(EMP_NO,DEPT_NO,EMP_NAME,PHOTO_CONTENT,IS_QUALIFIED) SELECT StaffID,TenantID,StaffDisplayName,StaffPhotoImg,1 FROM [SKEP_DAS].[dbo].[DAS_Staff]; // RECORD_COUNTER 用于标记最新被修改过的数据,在初始化数据时由于 ID 是从1开始自增长的,所以可以用 ID 的值来初始 RECORD_COUNTER,(而首次用 "SELECT MAX(RECORD_COUNTER) + 1 FROM xxx" 插入的是 NULL) updateRecordCounterSql = @"UPDATE HT_ACCESS.DBO.HT_PHOTO SET RECORD_COUNTER = ID;"; // 门禁库表和中间库表的映射关系 TableMappings.Add("DAS_Staff", "HT_PHOTO"); BaseParams.Add("HT_PHOTO", new string[] { "[HT_ACCESS].[DBO].HT_PHOTO", "EMP_NO,DEPT_NO,EMP_NAME,PHOTO_CONTENT", "StaffID,TenantID,StaffDisplayName,StaffPhotoImg", "EMP_NO", "StaffID" }); } }
C#同步程序逻辑:
using HT.IMS.DataSync; using IMS.BaseFramework.Logging; using IMS.DBHelper; using Quartz; using System; using System.Configuration; using System.Data; using System.Text; using System.Threading.Tasks; namespace SynchronizeTask { public class HelloJob : IJob { private static LoggerAdapter<Program> logger = new LoggerAdapter<Program>(); private static string _connStr = ConfigurationManager.ConnectionStrings["SQLConnectionStr"].ToString(); string dtNow = "1900-01-01"; private static readonly object _syncRoot = new object(); // 注意必须是 static 类型,否则每个实例都有自己的 _syncRoot,达不到全局互斥效果 public virtual Task Execute(IJobExecutionContext context) { try { #region 初始化数据库表中的数据 dtNow = DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss:fff"); DataSet checkData = SQLHelper.Query(_connStr, SynchronizeSqlStr.checkHasDataSql); try { if (IsNullDataSet(checkData)) { // 由于各主键ID是自增长的,因此为了简便,在初始化数据后,直接让 RECORD_COUNTER = ID(注意:updateRecordCounterSql 里排除里 RECORD 表,因为它只有一条记录) string initDataSql = Convert.ToString(SynchronizeSqlStr.initDataSql + " " + SynchronizeSqlStr.updateRecordCounterSql); // 这里不能用异步操作,因为后面的操作要在它执行成功后才能执行 SQLHelper.ExecuteNonQuery(_connStr, CommandType.Text, initDataSql); } } catch (Exception e) { logger.LogError("初始化数据库表中的数据时出错:\r\n" + e.ToString() + " " + DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss:fff")); return Task.FromResult(false); } #endregion #region 同步表的数据(根据门禁触发器的记录) // 获取门禁数据库服务器上的触发器生成的待同步的数据 DataSet trackData = SQLHelper.Query(_connStr, SynchronizeSqlStr.TriggerTrackSql); if (IsNullDataSet(trackData)) { return Task.FromResult(true); } DataTable dt = trackData.Tables[0]; int rowCount = trackData.Tables[0].Rows.Count; StringBuilder dataSyncSqls = new StringBuilder(""); StringBuilder clearSqls = new StringBuilder(""); //这样写有个问题,如果同一个表有多条记录,由于语句没有执行,则取maxrecordCount都是同一个值 //改成每次拼接一条语句就执行一次 for (int i = 0; i < rowCount; i++) { try { // 提取表名(由于表名在各个数据库中都保持一样,这里提取表名后方便后面获取 Dictionary 中的参数) string tableName = dt.Rows[i]["TableName"].ToString(); string tableCache = tableName; tableName = tableName.Substring(tableName.LastIndexOf('[') + 1).TrimEnd(new char[] { ']' }); tableName = SynchronizeSqlStr.TableMappings[tableName]; string[] tempParams = SynchronizeSqlStr.BaseParams[tableName]; string getMaxRecordCountSql = string.Format("SELECT MAX([RECORD_COUNTER]) FROM {0}", tempParams[0]); object maxRecordCount = SQLHelper.ExecuteScalar(_connStr, CommandType.Text, getMaxRecordCountSql); string maxCount = Convert.ToString(maxRecordCount); if (string.IsNullOrWhiteSpace(maxCount)) { maxCount = "1"; } else { maxCount = Convert.ToString(Convert.ToInt32(maxCount) + 1); } if (dt.Rows[i]["Type"].ToString() == "1") // 新增 { dataSyncSqls.Append( string.Format(SynchronizeSqlStr.InsertSQLs, tempParams[0], tempParams[1], tempParams[2], dt.Rows[i]["TableName"].ToString(), tempParams[4], dt.Rows[i]["PrimaryKey"].ToString()) + "\r\n" + string.Format("UPDATE {0} SET [MODIFY_TIME] = '" + dtNow + "', [RECORD_COUNTER] = " + maxCount + " WHERE {1} = '{2}' AND DELETE_FLAT<>1", tempParams[0], tempParams[3], dt.Rows[i]["PrimaryKey"].ToString()) + ";" ); } else if (dt.Rows[i]["Type"].ToString() == "2") // 删除(由于是软删除,所以和修改处理方式类似) { // dataSyncSqls.Append(string.Format(SynchronizeSqlStr.DeleteSQLs, tempParams[0], tempParams[3], dt.Rows[i]["PrimaryKey"].ToString())); dataSyncSqls.Append( string.Format("UPDATE {0} SET [DELETE_FLAG] = 1, [MODIFY_TIME] = '" + dtNow + "', [RECORD_COUNTER] = " + maxCount + " WHERE {1} = '{2}' AND DELETE_FLAT<>1", tempParams[0], tempParams[3], dt.Rows[i]["PrimaryKey"].ToString()) + ";" ); } else if (dt.Rows[i]["Type"].ToString() == "3") // 修改 { //由于找删除是把照片改成默认的照片,而12所开发人员要求把删除的照片标记为删除,这里对照片的修改做特殊处理 if (tempParams[0] == "[HT_ACCESS].[DBO].HT_PHOTO") { string getPhotoContenSql = string.Format("SELECT TOP 1 [StaffPhoto] FROM [SKEP_DASPHOTO].[dbo].[DAS_StaffPhoto] WHERE [PhotoID] = '{0}' AND DELETE_FLAG<>1", dt.Rows[i]["PrimaryKey"].ToString()); object photoContent = SQLHelper.ExecuteScalar(_connStr, CommandType.Text, getPhotoContenSql); byte[] pc = null; StringBuilder sb = new StringBuilder("0x"); try { pc = ((byte[])photoContent); foreach (var b in pc) { sb.Append(b.ToString("X2")); } } catch (Exception ex) { } //如果修改后的照片等于默认照片则还要标记为删除 if (sb.ToString().Equals(DefaultImage.EmpPhoto)) { dataSyncSqls.Append( string.Format(SynchronizeSqlStr.DeleteSQLs, tempParams[0], tempParams[3], dt.Rows[i]["PrimaryKey"].ToString()) + "\r\n" + string.Format(SynchronizeSqlStr.InsertSQLs, tempParams[0], tempParams[1], tempParams[2], dt.Rows[i]["TableName"].ToString(), tempParams[4], dt.Rows[i]["PrimaryKey"].ToString()) + "\r\n" + string.Format("UPDATE {0} SET [MODIFY_TIME] = '" + dtNow + "', [RECORD_COUNTER] = " + maxCount + ", DELETE_FLAG = 1 WHERE {1} = '{2}'", tempParams[0], tempParams[3], dt.Rows[i]["PrimaryKey"].ToString()) + ";" ); } else { dataSyncSqls.Append( string.Format(SynchronizeSqlStr.DeleteSQLs, tempParams[0], tempParams[3], dt.Rows[i]["PrimaryKey"].ToString()) + "\r\n" + string.Format(SynchronizeSqlStr.InsertSQLs, tempParams[0], tempParams[1], tempParams[2], dt.Rows[i]["TableName"].ToString(), tempParams[4], dt.Rows[i]["PrimaryKey"].ToString()) + "\r\n" + string.Format("UPDATE {0} SET [MODIFY_TIME] = '" + dtNow + "', [RECORD_COUNTER] = " + maxCount + " WHERE {1} = '{2}'", tempParams[0], tempParams[3], dt.Rows[i]["PrimaryKey"].ToString()) + ";" ); } } else { // 注意:这里更新 RECORD_COUNTER 时不能用 MAX([RECORD_COUNTER]) +1,因为这里是先删除原记录(如果删除的记录的 RECORD_COUNTER 的值刚好最大,那么把 MAX([RECORD_COUNTER]) +1 也还是原来的值),然后复制新记录。 // 而利用 ID 自增长的特点,让 [RECORD_COUNTER] = [ID] 则刚好满足 [RECORD_COUNTER] 的要求 dataSyncSqls.Append( string.Format(SynchronizeSqlStr.DeleteSQLs, tempParams[0], tempParams[3], dt.Rows[i]["PrimaryKey"].ToString()) + "\r\n" + string.Format(SynchronizeSqlStr.InsertSQLs, tempParams[0], tempParams[1], tempParams[2], dt.Rows[i]["TableName"].ToString(), tempParams[4], dt.Rows[i]["PrimaryKey"].ToString()) + "\r\n" + string.Format("UPDATE {0} SET [MODIFY_TIME] = '" + dtNow + "', [RECORD_COUNTER] = " + maxCount + " WHERE {1} = '{2}'", tempParams[0], tempParams[3], dt.Rows[i]["PrimaryKey"].ToString()) + ";" ); } } SQLHelper.ExecuteNonQuery(_connStr, CommandType.Text, dataSyncSqls.ToString()); // 清理中间表里的数据 clearSqls.Append("DELETE FROM [SKEP_DAS].[dbo].[SKEP_SYNC] WHERE [PrimaryKey] = '" + dt.Rows[i]["PrimaryKey"] + "';"); } catch (Exception e) { logger.LogError("拼接脚本时出错:\r\n" + e.ToString() + " " + DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss:fff") + "\r\n" + dataSyncSqls.ToString()); break; } // 拼接完脚本后即可删除[SKEP_DAS].[dbo].[SKEP_SYNC]表中的临时数据 if(clearSqls.Length > 0) { // 执行拼接好的脚本 SQLHelper.ExecuteNonQuery(_connStr, CommandType.Text, clearSqls.ToString()); } } #endregion return Task.FromResult(true); } catch (Exception ex) { logger.LogError("执行同步程序时出错:\r\n" + ex.ToString() + " " + DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss:fff")); return Task.FromResult(false); } } private bool IsNullDataSet(DataSet ds) { if (ds != null && ds.Tables != null && ds.Tables.Count > 0 && ds.Tables[0] != null && ds.Tables[0].Rows.Count > 0) return false; return true; } } }
注:visual studio中提供了SSIS工具进行数据同步,但笔者不推荐该方式,因为配置太多且出错后无法调试,推荐通过程序进行同步
相关资源:
https://www.dataintegration.info/data-synchronization
SQL Server中的Merge Into - 西瓜豆酱 - 博客园 (cnblogs.com)
posted on 2019-07-02 16:34 wangzhiliang 阅读(2224) 评论(0) 编辑 收藏 举报