« 《Oracle DBA诊断案例》与读者的建议 | Blog首页 | 纪念EYGLE.com BLOG写作四周年 »

关于ASM NORMAL REDUNDANCY的故障处理

今天有朋友问到关于ASM的一个问题，当使用NORMAL REDUNDANCY方式创建磁盘组后，如果一个磁盘组掉线，或者使用不同盘阵，一个盘阵掉线，如何进行之后的处理，数据库运行是否会受到影响。

关于这个内容，理论上是没有问题的，至于实际操作则应极其慎重，在NORMAL REDUNDANCY模式下使用ASM的用户并不多见。

而且在Metalink上还刚刚看到一个BUG：5999619，在HP-UX PA-RISC (64-bit) 11.11平台
使用Database Version 10.2.0.3.0 ，有用户报告：ASM HANGS WHENEVER THE DISKS IN A FAILGROUP ARE NOT AVAILABLE

Metalink Note:395712.1对一些ASM常见问题做了总结，其中包括如下解答：

SQL> CREATE DISKGROUP dgroup1 NORMAL REDUNDANCY
FAILGROUP controller1 DISK '/devices/diska1', '/devices/diska2', '/devices/diska3', '/devices/diska4'
FAILGROUP controller2 DISK '/devices/diskb1', '/devices/diskb2', '/devices/diskb3', '/devices/diskb4';

in this example we have 2 failure groups called controller1,controller2. There are 2 possible situations of failure can happen.

Situation #1:
-----------------
A disk in controller2 has failed or damaged.
In this case, if we supposed that disk '/devices/diskb3' has been damaged, ASM will mirror all the data in that disk to be available in the other 3 disks '/devices/diskb1', '/devices/diskb2', '/devices/diskb4'. ASM collected the file extents or data from the other failure group controller1 not from controller2 which means every file extent in controller1 failure group will be mirrored in controller2 failure group. The situation here may come to a critical one if the free space in '/devices/diskb1', '/devices/diskb2', '/devices/diskb4' can not hold the amount of data from the damaged disk '/devices/diskb3'. So some of the file extents located in controller1 failure group will be mirrored in controller2 failure group an others will not according to the lack of space in controller2 failure group. Database will still be working.

Situation #2:
-----------------
A whole failure group has been damaged.
In this case, if we supposed that controller1 failure group has been damaged, ASM will not mirror any thing because the mirroring happens among failure groups not among disks - No MIRRORING in a single failure group - . Database will still be working. This will lead that controller2 failure group will contain the primary extents. If any thing wrong happened to this data - primary extents - , the ASM will stop immediately.

By eygle on 2008-12-18 14:14 | Comments (0) | Advanced | 2118 |

ASM