每个推荐RedGate工具的人都要加1分，还有额外的建议和警告。

SqlCompare also has a decently documented API: so you can, for instance, write a console app which syncs your source controlled scripts folder with a CI integration testing database on checkin, so that when someone checks in a change to the schema from their scripts folder it's automatically deployed along with the matching application code change. This helps close the gap with developers who are forgetful about propagating changes in their local db up to a shared development DB (about half of us, I think :) ).

A caveat is that with a scripted solution or otherwise, the RedGate tools are sufficiently smooth that it's easy to forget about SQL realities underlying the abstraction. If you rename all the columns in a table, SqlCompare has no way to map the old columns to the new columns and will drop all the data in the table. It will generate warnings but I've seen people click past that. There's a general point here worth making, I think, that you can only automate DB versioning and upgrade so far - the abstractions are very leaky.

我们不存储数据库模式，而是存储对数据库的更改。我们所做的是存储模式更改，以便为数据库的任何版本构建更改脚本，并将其应用于客户的数据库。我写了一个数据库实用程序，它与我们的主应用程序一起分发，它可以读取脚本，并知道需要应用哪些更新。它还具有足够的智能，可以根据需要刷新视图和存储过程。

关于这个主题，Martin Fowler写了一篇我最喜欢的文章http://martinfowler.com/articles/evodb.html。我选择不像alumb和其他人建议的那样将模式转储放在版本控制下，因为我想要一种简单的方法来升级我的生产数据库。

对于一个web应用程序，我将有一个生产数据库实例，我使用两种技术:

数据库升级脚本

一个序列数据库升级脚本，其中包含将模式从版本N移动到版本N+1所需的DDL。(这些会进入你的版本控制系统。)一个_version_history_表，类似于

create table VersionHistory (
    Version int primary key,
    UpgradeStart datetime not null,
    UpgradeEnd datetime
    );

每次运行与新版本对应的升级脚本时获取一个新条目。

这可以确保很容易查看数据库模式的版本，并且数据库升级脚本只运行一次。同样，这些不是数据库转储。相反，每个脚本表示从一个版本迁移到下一个版本所需的更改。它们是应用于生产数据库以“升级”它的脚本。

开发人员沙箱同步

备份、清理和收缩生产数据库的脚本。在每次升级到生产DB后运行此命令。 在开发人员的工作站上恢复(并在必要时调整)备份的脚本。每个开发人员在每次升级到生产DB后都会运行此脚本。

警告:我的自动化测试运行在模式正确但空的数据库上，因此这个建议不能完全满足您的需求。

典型的解决方案是根据需要转储数据库并备份这些文件。

根据您的开发平台，可能会有可用的开源插件。编写自己的代码来完成这个任务通常是相当简单的。

注意:您可能希望备份数据库转储，而不是将其放入版本控制中。在版本控制中，文件可能会迅速变大，并导致整个源代码控制系统变慢(我正在回忆CVS的恐怖故事)。

为了更快地将数据转储到源代码控制系统，您可以使用sysobjects中的版本信息来查看自上次以来哪些对象发生了更改。

设置:在每个数据库中增量地创建一个表，以保存上次检查时的版本信息(第一次运行时为空)。如果要重新扫描整个数据结构，请清除此表。

IF ISNULL(OBJECT_ID('last_run_sysversions'), 0) <> 0 DROP TABLE last_run_sysversions
CREATE TABLE last_run_sysversions (
    name varchar(128), 
    id int, base_schema_ver int,
    schema_ver int,
    type char(2)
)

正常运行模式:您可以从此sql中获取结果，并为您感兴趣的sql脚本生成sql脚本，并将它们放入您选择的源代码控制中。

IF ISNULL(OBJECT_ID('tempdb.dbo.#tmp'), 0) <> 0 DROP TABLE #tmp
CREATE TABLE #tmp (
    name varchar(128), 
    id int, base_schema_ver int,
    schema_ver int,
    type char(2)
)

SET NOCOUNT ON

-- Insert the values from the end of the last run into #tmp
INSERT #tmp (name, id, base_schema_ver, schema_ver, type) 
SELECT name, id, base_schema_ver, schema_ver, type FROM last_run_sysversions

DELETE last_run_sysversions
INSERT last_run_sysversions (name, id, base_schema_ver, schema_ver, type)
SELECT name, id, base_schema_ver, schema_ver, type FROM sysobjects

-- This next bit lists all differences to scripts.
SET NOCOUNT OFF

--Renamed.
SELECT 'renamed' AS ChangeType, t.name, o.name AS extra_info, 1 AS Priority
FROM sysobjects o INNER JOIN #tmp t ON o.id = t.id
WHERE o.name <> t.name /*COLLATE*/
AND o.type IN ('TR', 'P' ,'U' ,'V')
UNION 

--Changed (using alter)
SELECT 'changed' AS ChangeType, o.name /*COLLATE*/, 
       'altered' AS extra_info, 2 AS Priority
FROM sysobjects o INNER JOIN #tmp t ON o.id = t.id 
WHERE (
   o.base_schema_ver <> t.base_schema_ver
OR o.schema_ver      <> t.schema_ver
)
AND  o.type IN ('TR', 'P' ,'U' ,'V')
AND  o.name NOT IN ( SELECT oi.name 
         FROM sysobjects oi INNER JOIN #tmp ti ON oi.id = ti.id
         WHERE oi.name <> ti.name /*COLLATE*/
         AND oi.type IN ('TR', 'P' ,'U' ,'V')) 
UNION

--Changed (actually dropped and recreated [but not renamed])
SELECT 'changed' AS ChangeType, t.name, 'dropped' AS extra_info, 2 AS Priority
FROM #tmp t
WHERE    t.name IN ( SELECT ti.name /*COLLATE*/ FROM #tmp ti
         WHERE NOT EXISTS (SELECT * FROM sysobjects oi
                           WHERE oi.id = ti.id))
AND  t.name IN ( SELECT oi.name /*COLLATE*/ FROM sysobjects oi
         WHERE NOT EXISTS (SELECT * FROM #tmp ti
                           WHERE oi.id = ti.id)
         AND   oi.type  IN ('TR', 'P' ,'U' ,'V'))
UNION

--Deleted
SELECT 'deleted' AS ChangeType, t.name, '' AS extra_info, 0 AS Priority
FROM #tmp t
WHERE NOT EXISTS (SELECT * FROM sysobjects o
                  WHERE o.id = t.id)
AND t.name NOT IN (  SELECT oi.name /*COLLATE*/ FROM sysobjects oi
         WHERE NOT EXISTS (SELECT * FROM #tmp ti
                           WHERE oi.id = ti.id)
         AND   oi.type  IN ('TR', 'P' ,'U' ,'V'))
UNION

--Added
SELECT 'added' AS ChangeType, o.name /*COLLATE*/, '' AS extra_info, 4 AS Priority
FROM sysobjects o
WHERE NOT EXISTS (SELECT * FROM #tmp t
                  WHERE o.id = t.id)
AND      o.type  IN ('TR', 'P' ,'U' ,'V')
AND  o.name NOT IN ( SELECT ti.name /*COLLATE*/ FROM #tmp ti
         WHERE NOT EXISTS (SELECT * FROM sysobjects oi
                           WHERE oi.id = ti.id))
ORDER BY Priority ASC

注意:如果在任何数据库中使用非标准排序规则，则需要将/* COLLATE */替换为数据库排序规则。即COLLATE Latin1_General_CI_AI

每个推荐RedGate工具的人都要加1分，还有额外的建议和警告。

SqlCompare also has a decently documented API: so you can, for instance, write a console app which syncs your source controlled scripts folder with a CI integration testing database on checkin, so that when someone checks in a change to the schema from their scripts folder it's automatically deployed along with the matching application code change. This helps close the gap with developers who are forgetful about propagating changes in their local db up to a shared development DB (about half of us, I think :) ).

A caveat is that with a scripted solution or otherwise, the RedGate tools are sufficiently smooth that it's easy to forget about SQL realities underlying the abstraction. If you rename all the columns in a table, SqlCompare has no way to map the old columns to the new columns and will drop all the data in the table. It will generate warnings but I've seen people click past that. There's a general point here worth making, I think, that you can only automate DB versioning and upgrade so far - the abstractions are very leaky.