关机重启导致 ASM 磁盘丢失数据库无法启动

2023年 10月 6日 51.2k 0

问题现象

事情是这样的,各位看官且看之,中秋国庆期间测试机房由于线路变更需要断电,故需要关掉所以测试机器,等待变更完成后,测试环境的 RAC 有一个节点数据库启动不了,检查发现节点 2 的 ARCH 磁盘没有正常挂载。

登录到 ASM 实例,查看 ASM 磁盘及路径发现归档盘 ARCH 没挂载,磁盘路径为 ‘/dev/sde’。

su - grid
sqlplus / as sysasm

Connected to:
Oracle Database 19c Enterprise Edition Release 19.0.0.0.0 - Production
Version 19.15.0.0.0

SQL> set lin 1000 pagesize 999
col PATH for a30
col NAME for a15
col FAILGROUP for a15
select GROUP_NUMBER,DISK_NUMBER,OS_MB/1024,TOTAL_MB/1024,FREE_MB/1024,NAME,FAILGROUP,PATH,FAILGROUP_TYPE,header_status,state from v$asm_disk order by 1;
select GROUP_NUMBER,NAME,STATE,TYPE,TOTAL_MB/1024,FREE_MB/1024,USABLE_FILE_MB/1024,REQUIRED_MIRROR_FREE_MB,HOT_USED_MB,COLD_USED_MB/1024 from v$asm_diskgroup;
SQL> SQL> SQL>
GROUP_NUMBER DISK_NUMBER OS_MB/1024 TOTAL_MB/1024 FREE_MB/1024 NAME FAILGROUP PATH FAILGRO HEADER_STATU STATE
------------ ----------- ---------- ------------- ------------ --------------- --------------- ------------------------------ ------- ------------ --------
0 0 100 0 0 /dev/sde REGULAR MEMBER NORMAL
2 2 200 200 5.91796875 DATA_0002 DATA_0002 /dev/sdi REGULAR MEMBER NORMAL
2 0 200 200 5.8828125 DATA_0000 DATA_0000 /dev/sdc REGULAR MEMBER NORMAL
2 1 200 200 5.81640625 DATA_0001 DATA_0001 /dev/sdd REGULAR MEMBER NORMAL
3 0 3 3 2.6640625 OCR_0000 OCR_0000 /dev/sdf REGULAR MEMBER NORMAL
3 1 3 3 2.66015625 OCR_0001 OCR_0001 /dev/sdg REGULAR MEMBER NORMAL
3 2 3 3 2.66015625 OCR_0002 OCR_0002 /dev/sdh REGULAR MEMBER NORMAL

7 rows selected.

SQL>
GROUP_NUMBER NAME STATE TYPE TOTAL_MB/1024 FREE_MB/1024 USABLE_FILE_MB/1024 REQUIRED_MIRROR_FREE_MB HOT_USED_MB COLD_USED_MB/1024
------------ --------------- ----------- ------ ------------- ------------ ------------------- ----------------------- ----------- -----------------
0 ARCH DISMOUNTED 0 0 0 0 0 0
2 DATA MOUNTED EXTERN 600 17.6171875 17.6171875 0 0 582.382813
3 OCR MOUNTED NORMAL 9 7.984375 2.4921875 3072 0 1.015625

然后通过操作系统层 lsblk 命令查看 /dev/sde 大小 100G,然后通过 ASM 命令挂载 ARCH 磁盘却不成功,报错丢失磁盘。

[root@jiekexu2 dev]# lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 200G 0 disk
|-sda1 8:1 0 1G 0 part /boot
`-sda2 8:2 0 199G 0 part
|-rootvg-lvroot 253:0 0 191.1G 0 lvm /
`-rootvg-lvswap 253:1 0 15.9G 0 lvm [SWAP]
sdb 8:16 0 8G 0 disk
`-rootvg-lvswap 253:1 0 15.9G 0 lvm [SWAP]
sdc 8:32 0 200G 0 disk
sdd 8:48 0 200G 0 disk
sde 8:64 0 100G 0 disk
sdf 8:80 0 3G 0 disk
sdg 8:96 0 3G 0 disk
sdh 8:112 0 3G 0 disk
sdi 8:128 0 200G 0 disk
sdj 8:144 0 100G 0 disk
sdk 8:160 0 200G 0 disk
sr0 11:0 1 1024M 0 rom

jiekexu2:/home/grid(+ASM2)$ asmcmd
ASMCMD> mount ARCH
ORA-15032: not all alterations performed
ORA-15040: diskgroup is incomplete
ORA-15042: ASM disk "1" is missing from group number "1" (DBD ERROR: OCIStmtExecute)
ASMCMD> exit

下面我们通过 ASM 实例的 alter 日志,切实看到 ARCH 盘有两块磁盘,一块是 (/dev/sde),另一块则为空()看不到。那么我们来看看节点 1 的 ARCH 磁盘及相关权限吧。

节点 1/2 权限对比
通过节点 1 发现 ARCH 磁盘组有两块磁盘(/dev/sde)和(/dev/sdj),属组则为 grid:asmadmin,但是节点 2 (/dev/sde)权限属组正常,(/dev/sdj)属组为 root:disk,这显然是有问题的,正因为这里的属组权限不对,上面最开始查看磁盘 ARCH 盘时仅显示一块(/dev/sde)。

解决问题

下面检查一下 udev 配置文件是否出现问题,然后我们修改(/dev/sdj)属组为 grid:asmadmin,并重新触发 udev 规则,然后再尝试去挂载磁盘组。

jiekexu2:/home/grid(+ASM2)$ more /etc/udev/rules.d/99-oracle-asmdisks.rules
KERNEL=="sd*", ACTION=="add|change", SUBSYSTEM=="block", PROGRAM=="/lib/udev/scsi_id -g -u -d /dev/$name", RESULT=="36000c29bd4635c80a3011779bc1a7f99", SYMLINK+="asmdisks/asmdiskb", OWNER=
"grid", GROUP="asmadmin", MODE="0660"
KERNEL=="sd*", ACTION=="add|change", SUBSYSTEM=="block", PROGRAM=="/lib/udev/scsi_id -g -u -d /dev/$name", RESULT=="36000c2947f3b25a26a2bd31496cd2f59", SYMLINK+="asmdisks/asmdiskc", OWNER=
"grid", GROUP="asmadmin", MODE="0660"
KERNEL=="sd*", ACTION=="add|change", SUBSYSTEM=="block", PROGRAM=="/lib/udev/scsi_id -g -u -d /dev/$name", RESULT=="36000c29ad3307c8874aaeb6b5c83f2f8", SYMLINK+="asmdisks/asmdiskd", OWNER=
"grid", GROUP="asmadmin", MODE="0660"
KERNEL=="sd*", ACTION=="add|change", SUBSYSTEM=="block", PROGRAM=="/lib/udev/scsi_id -g -u -d /dev/$name", RESULT=="36000c29756eabbaea24471e77a46d1bf", SYMLINK+="asmdisks/asmdiske", OWNER=
"grid", GROUP="asmadmin", MODE="0660"
KERNEL=="sd*", ACTION=="add|change", SUBSYSTEM=="block", PROGRAM=="/lib/udev/scsi_id -g -u -d /dev/$name", RESULT=="36000c296f7ef4f660d24ecdc32fd5216", SYMLINK+="asmdisks/asmdiskf", OWNER=
"grid", GROUP="asmadmin", MODE="0660"
KERNEL=="sd*", ACTION=="add|change", SUBSYSTEM=="block", PROGRAM=="/lib/udev/scsi_id -g -u -d /dev/$name", RESULT=="36000c29ef6de29e246eec5d21ae13f74", SYMLINK+="asmdisks/asmdiskg", OWNER=
"grid", GROUP="asmadmin", MODE="0660"
KERNEL=="sd*", ACTION=="add|change", SUBSYSTEM=="block", PROGRAM=="/lib/udev/scsi_id -g -u -d /dev/$name", RESULT=="36000c2995925bb888b965036ad9d9807", SYMLINK+="asmdisks/asmdiski", OWNER=
"grid", GROUP="asmadmin", MODE="0660"
KERNEL=="sd*", ACTION=="add|change", SUBSYSTEM=="block", PROGRAM=="/lib/udev/scsi_id -g -u -d /dev/$name", RESULT=="36000c294342f0d7dd282c4a632d82c47", SYMLINK+="asmdisks/asmdiskj", OWNER=
"grid", GROUP="asmadmin", MODE="0660"
jiekexu2:/home/grid(+ASM2)$ ll /dev/sde
brw-rw---- 1 grid asmadmin 8, 64 Oct 1 09:46 /dev/sde
jiekexu2:/home/grid(+ASM2)$ ll /dev/sdj
brw-rw---- 1 root disk 8, 144 Oct 1 09:19 /dev/sdj
jiekexu2:/home/grid(+ASM2)$ exit
logout
[root@jiekexu2 ~]# udevadm control --reload-rules
[root@jiekexu2 ~]# ll /dev/sdj
brw-rw---- 1 root disk 8, 144 Oct 1 09:19 /dev/sdj
[root@jiekexu2 ~]# ll /dev/sde
brw-rw---- 1 grid asmadmin 8, 64 Oct 1 09:46 /dev/sde
[root@jiekexu2 ~]# udevadm trigger
[root@jiekexu2 ~]# ll /dev/sdj
brw-rw---- 1 root disk 8, 144 Oct 1 09:50 /dev/sdj
[root@jiekexu2 ~]# ll /dev/sde
brw-rw---- 1 grid asmadmin 8, 64 Oct 1 09:50 /dev/sde
[root@jiekexu2 ~]# chown grid:asmadmin /dev/sdj
[root@jiekexu2 ~]# ll /dev/sdj
brw-rw---- 1 grid asmadmin 8, 144 Oct 1 09:50 /dev/sdj

接下来我们则登录 ASM 实例,查看磁盘路径可以正常显示 sdj 和 sde,也可以正常挂载 ARCH。

su - grid
sqlplus / as sysasm
alter diskgroup ARCH mount;

登录数据库发现数据库实例已经自启动了。

另外一套 RAC 也是节点 2 数据库无法启动,DATA 磁盘组无法正常挂载。经查看 ASM 实例日志也是 DATA 磁盘组缺失一块磁盘,缺失的磁盘权限也变成 root:disk,如法炮制,修改属组重新挂载即可恢复正常。

2023-10-01T09:18:17.259212+08:00
ERROR: /* ASMCMD cguid:cae29900570ecf7ebfd17fbf776d2840 cname:jieke-rac-scan nodename:jieke-rac-87 */ALTER DISKGROUP data MOUNT
2023-10-01T09:18:17.650518+08:00
ASM Health Checker found 1 new failures
2023-10-01T10:21:30.690039+08:00
SQL> /* ASMCMD cguid:cae29900570ecf7ebfd17fbf776d2840 cname:jieke-rac-scan nodename:jieke-rac-87 */ALTER DISKGROUP DATA MOUNT
2023-10-01T10:21:30.795687+08:00
NOTE: cache registered group DATA 1/0x36302669
NOTE: cache began mount (not first) of group DATA 1/0x36302669
NOTE: Assigning number (1,0) to disk (/dev/asmdisks/asm-data)
2023-10-01T10:21:31.238030+08:00
GMON querying group 1 at 38 for pid 37, osid 59371
2023-10-01T10:21:31.269748+08:00
NOTE: Assigning number (1,1) to disk ()
2023-10-01T10:21:31.285726+08:00
GMON querying group 1 at 39 for pid 37, osid 59371
2023-10-01T10:21:31.286854+08:00
NOTE: cache dismounting (clean) group 1/0x36302669 (DATA)
NOTE: messaging CKPT to quiesce pins Unix process pid: 59371, image: oracle@jieke-rac-87 (TNS V1-V3)
NOTE: dbwr not being msg'd to dismount
NOTE: LGWR not being messaged to dismount
NOTE: cache dismounted group 1/0x36302669 (DATA)
NOTE: cache ending mount (fail) of group DATA number=1 incarn=0x36302669
NOTE: cache deleting context for group DATA 1/0x36302669
2023-10-01T10:21:31.386823+08:00
GMON dismounting group 1 at 40 for pid 37, osid 59371
2023-10-01T10:21:31.387863+08:00
NOTE: Disk DATA_0000 in mode 0x7f marked for de-assignment
ERROR: diskgroup DATA was not mounted
ORA-15032: not all alterations performed
ORA-15040: diskgroup is incomplete
ORA-15042: ASM disk "1" is missing from group number "1"

以上两套 RAC 关机重启为啥会导致共享磁盘属组发生变化,暂时没法排查了,也不知道该如何排查了,就先这样吧。下面在记录一下这套 RAC 以前出现的另一个问题。

DROPPED 状态的磁盘

几个月前的一天,突然发现此 RAC 有一块 OCR 盘出现故障,状态变为 FORCING ,磁盘 name 则以下划线命名 _DROPPED_0000_OCR ,查看 ASM 和集群日志均没有发现异常的地方。查看 CRS 集群和数据库均是正常状态,也可以正常启动关闭集群,但就是这块盘的状态是异常的。查询信息如下图所示:

解决办法

使用 lsblk 查看 sdg sdh sdi 三块 3G 的磁盘及对应的权限没有问题。

# lsblk
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda 8:0 0 100G 0 disk
|-sda1 8:1 0 1G 0 part /boot
`-sda2 8:2 0 99G 0 part
|-rootvg-lvroot 253:0 0 91.1G 0 lvm /
`-rootvg-lvswap 253:1 0 7.9G 0 lvm [SWAP]
sdb 8:16 0 100G 0 disk
`-u01vg-lvu01 253:2 0 100G 0 lvm /u01
sdc 8:32 0 50G 0 disk
sdd 8:48 0 300G 0 disk
sde 8:64 0 100G 0 disk
sdf 8:80 0 50G 0 disk
sdg 8:96 0 3G 0 disk
sdh 8:112 0 3G 0 disk
sdi 8:128 0 3G 0 disk
sdj 8:144 0 300G 0 disk
sr0 11:0 1 1024M 0 rom

[root@jiekexu-rac1 /]# ll /dev/asmdisks/asm-ocr3
lrwxrwxrwx 1 root root 6 Jul 1 16:51 /dev/asmdisks/asm-ocr3 -> ../sdi
[root@jiekexu-rac1 /]# ll /dev/sdi
brw-rw---- 1 grid asmadmin 8, 128 Jul 1 17:12 /dev/sdi

直接强制将第三块盘添加进 OCR 磁盘组即可。

SQL> alter diskgroup OCR add disk '/dev/asmdisks/asm-ocr3' force;

Diskgroup altered.

再次查看磁盘组状态则正常,只不过是作为一块新盘添加进去,从添加进去的磁盘名 OCR-0003 就可以看到,因为原来的名字为 OCR-0000。

全文完,希望可以帮到正在阅读的你,如果觉得此文对你有帮助,可以分享给你身边的朋友,同事,你关心谁就分享给谁,一起学习共同进步~~~

欢迎关注我的公众号【JiekeXu DBA之路】,第一时间一起学习新知识!
————————————————————————————
公众号:JiekeXu DBA之路
CSDN :https://blog.csdn.net/JiekeXu
墨天轮:https://www.modb.pro/u/4347
腾讯云:https://cloud.tencent.com/developer/user/5645107
————————————————————————————

相关文章

Oracle如何使用授予和撤销权限的语法和示例
Awesome Project: 探索 MatrixOrigin 云原生分布式数据库
下载丨66页PDF,云和恩墨技术通讯(2024年7月刊)
社区版oceanbase安装
Oracle 导出CSV工具-sqluldr2
ETL数据集成丨快速将MySQL数据迁移至Doris数据库

发布评论