ASM在线迁移LUN遇到的问题

ASM在线迁移LUN遇到的问题 在2个节点上都配置好了RAW之后，确保在两边ll /dev/raw都能看到新配置好的raw， 在节点1用命令行的方式添加新LUN并删除旧的RAW，执行了一会报错： SQL Alter diskgroup ZXDG add disk /dev/raw/raw4,/dev/raw/raw5 drop disk /dev/r

ASM在线迁移LUN遇到的问题

在2个节点上都配置好了RAW之后，确保在两边ll /dev/raw都能看到新配置好的raw，
在节点1用命令行的方式添加新LUN并删除旧的RAW，执行了一会报错：
SQL> Alter diskgroup ZXDG add disk &＃39;/dev/raw/raw4&＃39;,&＃39;/dev/raw/raw5&＃39; drop disk &＃39;/dev/raw/raw1&＃39; rebalance power 4 nowait;
Alter diskgroup ZXDG add disk &＃39;/dev/raw/raw4&＃39;,&＃39;/dev/raw/raw5&＃39; drop disk &＃39;/dev/raw/raw1&＃39; rebalance power 4 nowait
*
ERROR at line 1:
ORA-15032: not all alterations performed
ORA-15054: disk "/DEV/RAW/RAW1" does not exist in diskgroup "ZXDG"

SQL>
在两边的v$asm_disk中查看，发现节点1上已经能看到新LUN了，但节点2上却看不到新LUN。
SQL> select disk_number from v$asm_disk;

DISK_NUMBER
-----------
0
1
0

于是在节点2上重新添加了一次：
SQL>
SQL>
SQL> Alter diskgroup ZXDG add disk &＃39;/dev/raw/raw4&＃39;,&＃39;/dev/raw/raw5&＃39; ;
Alter diskgroup ZXDG add disk &＃39;/dev/raw/raw4&＃39;,&＃39;/dev/raw/raw5&＃39;
*
ERROR at line 1:
ORA-15032: not all alterations performed
ORA-15031: disk specification &＃39;/dev/raw/raw5&＃39; matches no disks
ORA-15025: could not open disk &＃39;/dev/raw/raw5&＃39;
ORA-27041: unable to open file
Linux-x86_64 Error: 6: No such device or address
Additional information: 42
Additional information: 255
Additional information: -750856672
ORA-15031: disk specification &＃39;/dev/raw/raw4&＃39; matches no disks
ORA-15025: could not open disk &＃39;/dev/raw/raw4&＃39;
ORA-27041: unable to open file
Linux-x86_64 Error: 6: No such device or address
Additional information: 42
Additional information: 255
Additional information: -750856672

上面的报错一般是raw的属主、权限等问题，也可能是节点2识别raw有问题（ll /dev/raw的结果有很大的欺骗性），
检查发现raw的属主、权限都没问题，那么是不是节点2未能正确识别lun呢，在节点2上执行partprobe之后再次添加disk：
SQL>
SQL>
SQL> Alter diskgroup ZXDG add disk &＃39;/dev/raw/raw4&＃39;,&＃39;/dev/raw/raw5&＃39;
2 ;
Alter diskgroup ZXDG add disk &＃39;/dev/raw/raw4&＃39;,&＃39;/dev/raw/raw5&＃39;
*
ERROR at line 1:
ORA-15032: not all alterations performed
ORA-15033: disk &＃39;/dev/raw/raw4&＃39; belongs to diskgroup "ZXDG"
ORA-15033: disk &＃39;/dev/raw/raw5&＃39; belongs to diskgroup "ZXDG"

这回报ORA-15032与ORA-15033了，显示raw4/raw5已经属于DG的一部分了，
但再次查看v$asm_disk，发现raw4,raw5的状态有异常，group_number为0，mount_status为closed，没有分配name：
group_number disk_number mount_status header_status mode_status state path name
1 0 cached member online dropping /dev/raw/raw1 ZXDG_0000
0 1 closed member online normal /dev/raw/raw4
0 2 closed member online normal /dev/raw/raw5
而此时raw1已经处于dropping状态了，因为raw4/raw5没有被正常add到dg，所以无法对数据做rebalance，此时不允许将raw1从dg中drop掉，所以状态一直为dropping。
以防万一，先undrop：
SQL> Alter diskgroup ZXDG undrop disks ;

Diskgroup altered.

由于add raw的时候asm并有为raw4/raw5分配name,也无法正常将之从dg中删除：
SQL> Alter diskgroup ZXDG drop disk &＃39;/dev/raw/raw4&＃39;;
Alter diskgroup ZXDG drop disk &＃39;/dev/raw/raw4&＃39;
*
ERROR at line 1:
ORA-15032: not all alterations performed
ORA-15054: disk "/DEV/RAW/RAW4" does not exist in diskgroup "ZXDG"
读取3个raw的metadata，庆幸的是发现metadata并没有出现混乱的情况，如下：
[oracle@zxdb01 ~]$ kfed read /dev/raw/raw1 > /tmp/raw1
[oracle@zxdb01 ~]$ kfed read /dev/raw/raw4 > /tmp/raw4
[oracle@zxdb01 ~]$ kfed read /dev/raw/raw5 > /tmp/raw5

[oracle@zxdb01 ~]$ diff /tmp/raw4 /tmp/raw5
6c6
---
> kfbh.block.obj: 2147483650 ; 0x008: TYPE=0x8 NUMB=0x2
20c20
---
> kfdhdb.dsknum: 2 ; 0x024: 0x0002
23c23
---
> kfdhdb.dskname: ZXDG_0002 ; 0x028: length=10
25c25
---
> kfdhdb.fgname: ZXDG_0002 ; 0x068: length=10
[oracle@zxdb01 ~]$ diff /tmp/raw1 /tmp/raw4
6,7c6,7
---
> kfbh.block.obj: 2147483649 ; 0x008: TYPE=0x8 NUMB=0x1
> kfbh.check: 3389207210 ; 0x00c: 0xca0332aa
20c20
---
> kfdhdb.dsknum: 1 ; 0x024: 0x0001
23c23
---
> kfdhdb.dskname: ZXDG_0001 ; 0x028: length=10
25c25
---
> kfdhdb.fgname: ZXDG_0001 ; 0x068: length=10
27,30c27,30
---
> kfdhdb.crestmp.hi: 33000305 ; 0x0a8: HOUR=0x11 DAYS=0x1b MNTH=0x2 YEAR=0x7de
> kfdhdb.crestmp.lo: 2735401984 ; 0x0ac: USEC=0x0 MSEC=0x2bb SECS=0x30 MINS=0x28
> kfdhdb.mntstmp.hi: 33000305 ; 0x0b0: HOUR=0x11 DAYS=0x1b MNTH=0x2 YEAR=0x7de
> kfdhdb.mntstmp.lo: 2735433728 ; 0x0b4: USEC=0x0 MSEC=0x2da SECS=0x30 MINS=0x28
35,36c35,36
---
> kfdhdb.dsksize: 102398 ; 0x0c4: 0x00018ffe
> kfdhdb.pmcnt: 2 ; 0x0c8: 0x00000002
39c39
---
> kfdhdb.f1b1locn: 0 ; 0x0d4: 0x00000000
[oracle@zxdb01 ~]$
[oracle@zxdb01 ~]$ diff /tmp/raw1 /tmp/raw5
6,7c6,7
---
> kfbh.block.obj: 2147483650 ; 0x008: TYPE=0x8 NUMB=0x2
> kfbh.check: 3389207210 ; 0x00c: 0xca0332aa
20c20
---
> kfdhdb.dsknum: 2 ; 0x024: 0x0002
23c23
---
> kfdhdb.dskname: ZXDG_0002 ; 0x028: length=10
25c25
---
> kfdhdb.fgname: ZXDG_0002 ; 0x068: length=10
27,30c27,30
---
> kfdhdb.crestmp.hi: 33000305 ; 0x0a8: HOUR=0x11 DAYS=0x1b MNTH=0x2 YEAR=0x7de
> kfdhdb.crestmp.lo: 2735401984 ; 0x0ac: USEC=0x0 MSEC=0x2bb SECS=0x30 MINS=0x28
> kfdhdb.mntstmp.hi: 33000305 ; 0x0b0: HOUR=0x11 DAYS=0x1b MNTH=0x2 YEAR=0x7de
> kfdhdb.mntstmp.lo: 2735433728 ; 0x0b4: USEC=0x0 MSEC=0x2da SECS=0x30 MINS=0x28
35,36c35,36
---
> kfdhdb.dsksize: 102398 ; 0x0c4: 0x00018ffe
> kfdhdb.pmcnt: 2 ; 0x0c8: 0x00000002
39c39
---
> kfdhdb.f1b1locn: 0 ; 0x0d4: 0x00000000
[oracle@zxdb01 ~]$

这种情况下完全可以用add disk force的方式强制加入dg:
SQL> Alter diskgroup ZXDG add disk &＃39;/dev/raw/raw4&＃39; name ZXDG_01 force ;

Diskgroup altered.
SQL> Alter diskgroup ZXDG add disk &＃39;/dev/raw/raw5&＃39; name ZXDG_01 force ;

Diskgroup altered.

再看状态，已经成功加入：
name group_number disk_number mount_status header_status mode_status state path
ZXDG_0000 1 0 CACHED MEMBER ONLINE NORMAL /dev/raw/raw1
ZXDG_02 1 2 CACHED MEMBER ONLINE NORMAL /dev/raw/raw5
ZXDG_01 1 1 CACHED MEMBER ONLINE NORMAL /dev/raw/raw4

可以安全的drop raw1了,设置并发为10的rebalance：
SQL>
SQL> Alter diskgroup ZXDG drop disk ZXDG_0000 rebalance power 10 nowait;

Diskgroup altered.

SQL>
此时通过v$asm_disk可以查看rebalance的进度：
name free_mb
ZXDG_0000 155446
ZXDG_02 55532
ZXDG_01 54982

速度还是挺快的。

值得庆幸的是在故障的过程中，metadata是没有损坏的，如果metadata发生损坏，则需要对metadata进行修复，这是个很繁琐的步骤。
当然如果有大量的停机时间的话，也可以对dg进行重建以修复这个问题。
估计经常运营WINDOWS服务器的童鞋在碰到这种情况的时候很有想重启一把asm的冲动，但如果是metedata发生损坏的情况，千万不能去尝试重启asm(尤其不能将所有节点的asm同时重启)，
因为很可能会导致dg无法被挂载，这会对解决问题带来不必要的麻烦。