以前都是手工右击索引重新生成,但是索引太多,操作起来费时费力,索引在网上找了个存储过程,自己整理了一下,执行的时候只需要选择相应的数据库,运行exec sp_rebuild_index即可,如下。
USE [master]
GO
SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
CREATE proc [dbo].[sp_rebuild_index]
(
@Rebuild_Fragmentation_Percent smallint = 5 -- 当逻辑碎片百分比 > 5% 重新生成索引
)
as
begin
/* 调用方法:
1.针对当前实例所有数据库: exec sys.sp_MSforeachdb 'use ?;exec sp_rebuild_index'
2.针对当前数据库: exec sp_rebuild_index
*/
--对系统数据库不作重新组织索引和重新生成索引
if (db_name() in ('master','model','msdb','tempdb')) return;
--如果逻辑碎片(索引中的无序页)的百分比 = 5% ,就不作重新组织索引和重新生成索引
if not exists(select 1 from sys.dm_db_index_physical_stats(db_id(),null,null,null,null) a where a.index_id>0 and a.avg_fragmentation_in_percent > @Rebuild_Fragmentation_Percent) return
print replicate('-',60)+char(13)+char(10)+replicate(' ',14)+N'对数据库 '+quotename(db_name())+N' 进行索引优化'+replicate(' ',20)+char(13)+char(10)
declare @sql nvarchar(2000),@str nvarchar(2000)
declare cur_x cursor for
select 'alter index '+quotename(a.name)+' on '+quotename(object_schema_name(a.object_id))+'.'+quotename(object_name(a.object_id))+' rebuild;' as [sql]
,N'重新生成索引:' +quotename(object_schema_name(a.object_id))+'.'+quotename(object_name(a.object_id))+'.'+quotename(a.name) as [str]
from sys.indexes a
inner join sys.dm_db_index_physical_stats(db_id(),null,null,null,null) b on b.object_id=a.object_id
and b.index_id=a.index_id
where a.index_id>0
and b.avg_fragmentation_in_percent > @Rebuild_Fragmentation_Percent
order by object_name(a.object_id),a.index_id
open cur_x
fetch next from cur_x into @sql,@str
while (@@fetch_status = 0)
begin
print @sql
exec(@sql)
print @str
fetch next from cur_x into @sql,@str
end
close cur_x
deallocate cur_x
end