|
Re: Re: 请小心风险
最初由 834508 发布
[B]
当初在3个阵列设定逻辑卷时没有考虑array spaning, 即在
阵列间实现I/O balance. 目前用sar发现三个阵列的I/O负荷
很不平衡,因此决定重新配置阵列,即在配置逻辑卷时
在阵列间实现条块化,同时保证对上层软件的透明(卷组,
逻辑卷raw device名称保持不变)
目前系统已经在实际运行中,数据库估计有将近400G的数据
,我已经用exp分别将几个大表和truncate大表后的整库备份出
来。我的问题是如何用imp实现全库恢复。因为此时数据库已
经关闭,而好像imp必须在database open 情况下运行,那么此时
我怎样才能把数据库完全重建呢?如果在mount后运行create
database 以同样名字创建一个数据库后在imp,我又担心不能
原样恢复。
实际上每个阵列已经RAID-5, 目前所作的是在RAID-5的基础上再做
RAID-0, 所以应该没有你所说的问题。 [/B]
Reorganizing a large production database require much down time. So I suggest you check your application first instead of reorganizing the disk arrays. If you still want to do it, you don’t need to use the exp and imp method. According to the raw device, as you mentioned above you can use dd to do a cold or hot backup, and then reorganize the disk array, after that use dd to restore the whole database. You still have another choice, use RMAN to backup and restore the database. You
said that you use RAID 5 and then RAID 0, this way I never heard before, it's very interesting, why? I suggest you use RAID 0+1 to get a better load balance and still gain high availability.
You can write a PL/SQL script to generate the backup scripts. For example:
host dd if=<raw_file> of=<fs_file> bs=65536 count=<v$datafile.bytes/65536) + 16>
And if you use RMAN, you can gain a better performance than use dd. Please reference RMAN manuals to backup & restore the database. |
|