Oracle 11G rac VOTE盘冗余问题

yueyangflash

11.2.0.3for aix 6.1的RAC

两个阵列，通过ASM做冗余，但是过程中遇到一个问题。

通过以下语句创建磁盘组，但是关闭一个存储后整个集群崩溃

create  DISKGROUP dg1
  FAILGROUP controller1 DISK '/dev/ocr12','/dev/ocr13'
      FAILGROUP controller2 DISK '/dev/ocr21','/dev/ocr22'；

[grid@db1 ~]$ crsctl query css votedisk
##  STATE    File Universal Id                File Name Disk group
--  -----    -----------------                --------- ---------
1. ONLINE   6b45c62f4a044f09bf5c2cdea834224c (/dev/ocr12) [DG1]
2. ONLINE   a5787225a3d34f22bf7df72042f172f4 (/dev/ocr21) [DG1]
3. ONLINE   4b9e619a58954f49bf2ff3dc69df7e16 (/dev/ocr22) [DG1]
Located 3 voting disk(s).

原因是投票磁盘没有超过半数，不知道有没有方案规避这个问题。谢谢！

yueyangflash

原帖由 whiterain 于 2012-7-11 22:36 发表
nfs毕竟不是非常好的方案，实在不行用powerha 做 concurrentvg 然后做lv镜像，把votingdisk 放到lv上吧。

应该会采用这种，谢谢！

yueyangflash

NFS的方案测试了，新建一个NFSserver，可以正常的关闭任意一个存储，但是客户认为不稳定性的因素比较大，目前放弃用ASM镜像VOTE的方式了。

whiterain

nfs毕竟不是非常好的方案，实在不行用powerha 做 concurrentvg 然后做lv镜像，把votingdisk 放到lv上吧。

Maclean Liu(刘相兵

2个 FAILGROUP中的disk是否属于不同的2个存储？

在11.2 中你可以考虑使用 Quorum failure group

ODM DATA:


OCR and voting files stored in ASM disk groups - ASM determines redundancy for OCR and voting files based on disk group  redundancy - quorum only failure groups will store only PST and voting files, not  database files

Voting Disk
Stored in selected ASM disks
‘crsctl’ used to specify a disk group for Voting disks
ASM auto creates1/3/5 Voting Disks based on Ext/Normal/High redundancy on Failure Groups
Quorum failure group

We can always use ASMCA to add a 3rd Voting File on NFS to a Extended Cluster using Oracle ASM. However, in cases, where ASMCA has a launch problem or X Windows is not working or you have lost existing NFS voting disk then we can use following sql command to manually add NFS voting disk to an Extended Cluster.

We need to use "alter diskgroup add quorum" syntax to add quorum disk,


SQL>alter diskgroup <DISKGROUP_NAME> add quorum failgroup <FAILGROUP_NAME> DISK 'NFS_path_to_disk';

e.g.-
Connect to any ASM Instance and execute following command -

SQL>alter diskgroup CRS add quorum failgroup FGQ DISK '/votedisk/vote_stnsp0506';


Here are excerpts from ASM Alert log upon running the above SQL command:-


+ASM_Alert.log

-----------
ALTER SYSTEM SET asm_diskstring='ORCL:*','/votedisk/*' SCOPE=BOTH SID='*';
..
SQL> ALTER DISKGROUP CRS ADD QUORUM DISK '/votedisk/vote_stnsp0506' SIZE 500M
NOTE: Assigning number (1,3) to disk (/votedisk/vote_stnsp0506)
..
SUCCESS: ALTER DISKGROUP CRS ADD QUORUM DISK '/votedisk/vote_stnsp0506' SIZE 500M
NOTE: Attempting voting file refresh on diskgroup CRS
NOTE: Voting file relocation is required in diskgroup CRS
NOTE: Attempting voting file relocation on diskgroup CRS
------------

After adding quorum disk either manually or via ASMCA, the Oracle Clusterware will automatically notice the disk group change and perform the necessary voting disk relocation.

Before running the above "Alter diskgroup" SQL command:-


$crsctl query css votedisk
## STATE File Universal Id File Name Disk group
-- ----- ----------------- --------- ---------
1. ONLINE 3e1836343f534f51bf2a19dff275da59 (/dev/sdg10) [CRS]
2. ONLINE 138cbee15b394f3ebf57dbfee7cec633 (/dev/sdf11) [CRS]
3. ONLINE 63514dd105e44fb0bfa741f20126e61c (/dev/sdf10) [CRS]


After running above "Alter diskgroup" SQL command:-


$ crsctl query css votedisk
## STATE File Universal Id File Name Disk group
-- ----- ----------------- --------- ---------
1. ONLINE 3e1836343f534f51bf2a19dff275da59 (/dev/sdg10) [CRS]
2. ONLINE 138cbee15b394f3ebf57dbfee7cec633 (/dev/sdf11) [CRS]
3. ONLINE 462722bd24c94f70bf4d90539c42ad4c (/votedisk/vote_stnsp0506) [CRS]



quorum failgroup  可以基于NFS ， 添加后 任意一个normal failgroup 失败，不造成votedisk 半数以上无法访问

aks5201314

表决磁盘超过半数主要是保护OCR盘数据的安全，对整个集群稳定运行很重要。不建议采取其他方案规避。