The most successful scheme I've ever used on a project has combined backups and differential SQL files. Basically we would take a backup of our db after every release and do an SQL dump so that we could create a blank schema from scratch if we needed to as well. Then anytime you needed to make a change to the DB you would add an alter scrip to the sql directory under version control. We would always prefix a sequence number or date to the file name so the first change would be something like 01_add_created_on_column.sql, and the next script would be 02_added_customers_index. Our CI machine would check for these and run them sequentially on a fresh copy of the db that had been restored from the backup.

我们还准备了一些脚本，开发人员可以用一个命令将本地db重新初始化到当前版本。

我已经在http://dbdeploy.com/上使用了ThoughtWorks的dbdeploy工具。它鼓励使用迁移脚本。在每个版本中，我们将更改脚本合并到一个文件中，以简化理解，并允许dba“支持”更改。

我们维护由ER工具(PowerAMC)生成的DDL(有时是DML)脚本。

我们有一个shell脚本工作台，它重命名以主干分支上的数字开头的脚本。 提交每个脚本并标记bugzilla编号。

这些脚本在需要时与应用程序代码合并到发布分支中。

我们有一个记录脚本及其状态的表。 在部署工具的每次安装中，每个脚本都按顺序执行，并记录在此表中。

Your project team can have a DBA to whom every developer would forward their create alter, delete, insert/update (for master data) sql statements. DBAs would run those queries and on successfully making the required update would add those statements to a text file or a spreadsheet. Each addition can be labeled as a savepoint. Incase you revert back to a particular savepoint, just do a drop all and run the queries uptil the labelled savepoint. This approach is just a thought... a bit of fine tuning here would work for your development environment.

我通过保存创建/更新脚本和生成sampledata的脚本来完成。

是的，我们通过保留SQL作为构建的一部分来做到这一点——我们保留DROP。sql,创造。sql,用户。sql,值。SQL和版本控制，所以我们可以恢复到任何带标签的版本。

我们还有ant任务，可以在需要时重新创建db。

此外，SQL还被标记为与之配套的源代码。