墨墨导读：本文详述硬件掉电后，Oracle集群无法启动的诡异故障处理过程。

一、 问题描述

现象：硬件掉电后，Oracle集群无法启动。

[root@rac2 ~]# crsctl stat res -tCRS-4535: Cannot communicate with Cluster Ready ServicesCRS-4000: Command Status failed, or completed with errors.[root@rac2 ~]# crsctl start crsCRS-4640: Oracle High Availability Services is already activeCRS-4000: Command Start failed, or completed with errors.

二、 故障处理

查看集群组件发现ora.asm状态为offline

root@rac2 ~]# crsctl stat res -t -init--------------------------------------------------------------------------------NAME           TARGET  STATE        SERVER                   STATE_DETAILS--------------------------------------------------------------------------------Cluster Resources--------------------------------------------------------------------------------ora.asm      1        ONLINE  OFFLINE                               Instance Shutdownora.cluster_interconnect.haip      1        ONLINE  ONLINE       rac2ora.crf      1        ONLINE  ONLINE       rac2ora.crsd      1        ONLINE  OFFLINEora.cssd      1        ONLINE  ONLINE       rac2ora.cssdmonitor      1        ONLINE  ONLINE       rac2ora.ctssd      1        ONLINE  ONLINE       rac2                     OBSERVERora.diskmon      1        OFFLINE OFFLINEora.drivers.acfs      1        ONLINE  ONLINE       rac2ora.evmd      1        ONLINE  INTERMEDIATE rac2ora.gipcd      1        ONLINE  ONLINE       rac2ora.gpnpd      1        ONLINE  ONLINE       rac2ora.mdnsd      1        ONLINE  ONLINE       rac2

查看grid alert日志发现磁盘组没有mount

[ohasd(4329)]CRS-2769:Unable to failover resource 'ora.diskmon'.2018-05-08 04:12:24.940:[cssd(4576)]CRS-1707:Lease acquisition for node rac2 number 2 completed2018-05-08 04:12:26.188:[cssd(4576)]CRS-1605:CSSD voting file is online: /dev/asmdisk/oraasm-OCR_0000; details in /u01/app/11.2.0/grid/log/rac2/cssd/ocssd.log.2018-05-08 04:12:28.723:[cssd(4576)]CRS-1601:CSSD Reconfiguration complete. Active nodes are rac1 rac2 .2018-05-08 04:12:30.617:[ctssd(4660)]CRS-2401:The Cluster Time Synchronization Service started on host rac2.2018-05-08 04:12:30.617:[ctssd(4660)]CRS-2407:The new Cluster Time Synchronization Service reference node is host rac1.2018-05-08 04:12:32.348:[ohasd(4329)]CRS-2767:Resource state recovery not attempted for 'ora.diskmon' as its target state is OFFLINE2018-05-08 04:12:32.348:[ohasd(4329)]CRS-2769:Unable to failover resource 'ora.diskmon'.

查看asm_alert，出现ORA-00600 [kfrValAcd30]的报错

NOTE: GMON heartbeating for grp 2GMON querying group 2 at 6 for pid 23, osid 5727NOTE: cache opening disk 0 of grp 2: DATA_0000 path:/dev/asmdisk/oraasm-ASM_0000NOTE: F1X0 found on disk 0 au 2 fcn 0.0NOTE: cache opening disk 1 of grp 2: DATA_0001 path:/dev/asmdisk/oraasm-ASM_0001NOTE: F1X0 found on disk 1 au 2 fcn 0.0NOTE: cache opening disk 2 of grp 2: DATA_0002 path:/dev/asmdisk/oraasm-ASM_0002NOTE: F1X0 found on disk 2 au 2 fcn 0.0NOTE: cache opening disk 3 of grp 2: DATA_0003 path:/dev/asmdisk/oraasm-ASM_0003NOTE: cache mounting (first) normal redundancy group 2/0x877A96CD (DATA)* allocate domain 2, invalid = TRUENOTE: attached to recovery domain 2NOTE: starting recovery of thread=1 ckpt=8.390 group=2 (DATA)Errors in file /u01/app/grid/diag/asm/+asm/+ASM2/trace/+ASM2_ora_5727.trc  (incident=50111):ORA-00600: internal error code, arguments: [kfrValAcd30], [DATA], [1], [8], [390], [9], [390], [], [], [], [], []ORA-15017: diskgroup "ASM" cannot be mountedORA-15063: ASM discovered an insufficient number of disks for diskgroup "ASM"Incident details in: /u01/app/grid/diag/asm/+asm/+ASM2/incident/incdir_50111/+ASM2_ora_5727_i50111.trcUse ADRCI or Support Workbench to package the incident.See Note 411.1 at My Oracle Support for error and packaging details.Errors in file /u01/app/grid/diag/asm/+asm/+ASM2/trace/+ASM2_ora_5727.trc:ORA-00600: internal error code, arguments: [kfrValAcd30], [DATA], [1], [8], [390], [9], [390], [], [], [], [], []ORA-15017: diskgroup "ASM" cannot be mountedORA-15063: ASM discovered an insufficient number of disks for diskgroup "ASM"Abort recovery for domain 2NOTE: crash recovery signalled OER-600ERROR: ORA-600 signalled during mount of diskgroup DATANOTE: cache dismounting (clean) group 2/0x877A96CD (DATA)NOTE: messaging CKPT to quiesce pins Unix process pid: 5727, image: oracle@rac2 (TNS V1-V3)

查看ASM2_ora_5727_i50111.trc

kfdp_query(DATA): 6----- Abridged Call Stack Trace -----ksedsts()+465<-kfdp_query()+530<-kfdPstSyncPriv()+585<-kfgFinalizeMount()+1630<-kfgscFinalize()+1433<-kfgForEachKfgsc()+285<-kfgsoFinalize()+135<-kfgFinalize()+398<-kfxdrvMount()+5558<-kfxdrvEntry()+2207<-opiexe()+20624<-opiosq0()+3932<-kpooprx()+274<-kpoal8()+842<-opiodr()+917<-ttcpip()+2183<-opitsk()+1710<-opiino()+969<-opiodr()+917<-opidrv()+570<-sou2o()+103<-opimai_real()+133<-ssthrdmain()+265<-main()+201<-__libc_start_main()+253----- End of Abridged Call Stack Trace -----2018-05-17 21:54:44.321623 : Start recovery for domain=2, valid=0, flags=0x4NOTE: starting recovery of thread=1 ckpt=8.390 group=2 (DATA)Incident 50111 created, dump file: /u01/app/grid/diag/asm/+asm/+ASM2/incident/incdir_50111/+ASM2_ora_5727_i50111.trcORA-00600: internal error code, arguments: [kfrValAcd30], [DATA], [1], [8], [390], [9], [390], [], [], [], [], []ORA-15017: diskgroup "ASM" cannot be mountedORA-15063: ASM discovered an insufficient number of disks for diskgroup "ASM"
recovery of group DATA failed due to the following error(s):ORA-00600: internal error code, arguments: [kfrValAcd30], [DATA], [1], [8], [390], [9], [390], [], [], [], [], []ORA-15017: diskgroup "ASM" cannot be mountedORA-15063: ASM discovered an insufficient number of disks for diskgroup "ASM"2018-05-17 21:54:45.016907 : Abort recovery for domain 2, flags = 0x42018-05-17 21:54:45.017799 : kjb_abort_recovery: domain flags=0x0, valid=0NOTE: messaging CKPT to quiesce pins Unix process pid: 5727, image: oracle@rac2 (TNS V1-V3)kfdp_dismount(): 7----- Abridged Call Stack Trace -----ksedsts()+465<-kfdp_dismountInt()+389<-kfdp_dismount()+11<-kfgTermCache()+347<-kfgRecoverDismount()+373<-kfgRecoverMount()+277<-kfgscDelete()+2742<-kssdel()+155<-kfgscFinalize()+1211<-kfgForEachKfgsc()+285<-kfgsoFinalize()+135<-kfgFinalize()+398<-kfxdrvMount()+5558<-kfxdrvEntry()+2207<-opiexe()+20624<-opiosq0()+3932<-kpooprx()+274<-kpoal8()+842<-opiodr()+917<-ttcpip()+2183<-opitsk()+1710<-opiino()+969<-opiodr()+917<-opidrv()+570<-sou2o()+103<-opimai_real()+133<-ssthrdmain()+265<-main()+201<-__libc_start_main()+253----- End of Abridged Call Stack Trace -----ASM name of disk 0x9f1d9488 (2:0:DATA_0000:/dev/asmdisk/oraasm-ASM_0000) is being clearedASM name of disk 0x9f1d9808 (2:1:DATA_0001:/dev/asmdisk/oraasm-ASM_0001) is being clearedASM name of disk 0x9f1d9108 (2:2:DATA_0002:/dev/asmdisk/oraasm-ASM_0002) is being clearedASM name of disk 0x9f1d8d88 (2:3:DATA_0003:/dev/asmdisk/oraasm-ASM_0003) is being clearedNOTE: messaging CKPT to quiesce pins Unix process pid: 5727, image: oracle@rac2 (TNS V1-V3)kfdp_dismount(): 8----- Abridged Call Stack Trace -----ksedsts()+465<-kfdp_dismountInt()+389<-kfdp_dismount()+11<-kfgTermCache()+347<-kfgRecoverDismount()+373<-kfgRecoverMount()+277<-kfgscDelete()+2742<-kssdct()+258<-kfgsoDelete()+721<-kssdel()+155<-kssdch()+6849<-ksures()+52<-opiosq0()+4721<-kpooprx()+274<-kpoal8()+842<-opiodr()+917<-ttcpip()+2183<-opitsk()+1710<-opiino()+969<-opiodr()+917<-opidrv()+570<-sou2o()+103<-opimai_real()+133<-ssthrdmain()+265<-main()+201<-__libc_start_main()+253----- End of Abridged Call Stack Trace -----

搜索mos也发现了几个Bug：

Bug 27288230 - ORA-600: [kfrvalacd30] in ASM SMON (Doc ID 27288230.8)Bug 19064132 : ORA-600 [KFRVALACD30] MOUNTING DISKGROUP AFTER RESTARTING SERVERSORA-600 [KFRVALACD30] in ASM (Doc ID 2123013.1)bug在11.2.0.1/11.2.0.3/112.0.4/12.1/18.1 都存在。

Oracle认为这是存储或者OS问题导致asm acd block的元数据不一致了，可能导致ASM元数据的主辅扩展区都发生损坏。这个损坏会导致rebalance挂起或不断尝试失败，或者阻止磁盘组被挂载。

strace出现ORA 600 [kfrValAcd30]报错，但没发现其他可用的信息

write(25, "string_value", 12)           = 12write(25, ">", 1)                       = 1write(25, "ORA 600 [kfrValAcd30]", 21)  = 21write(25, "</", 2)                      = 2write(25, "string_value", 12)           = 12write(25, ">", 1)                       = 1write(25, "</", 2)                      = 2write(25, "pkey_dbgristih", 14)         = 14write(25, ">", 1)                       = 1write(25, "\n", 1)                      = 1write(25, "<", 1)                       = 1write(25, "creatm_dbgristih", 16)       = 16write(25, ">", 1)                       = 1write(25, "<", 1)                       = 1write(25, "time", 4)                    = 4write(25, ">", 1)                       = 1

ASM的Active Change Directory(ACD)简单来说就相当于asm元数据重做记录，需要说明的是，每一个asm实例，都用有其自己的ACD目录，也就说，如果你是双节点的rac，那么就有84m的ACD 目录信息。

ACD主要是记录了这样一些数据：线程号，block号，opcode，sequences，记录长度等信息。

当处于运行的asm实例，突然crash后，那么重启asm实例以后，asm可以根据ACD信息去进行”instance recover”，从而保证能够正常启动asm实例。

如果acd的信息出现损坏，磁盘组将不能够被mount。

使用kfed来读取元数据，首先定位到active change directory 所在AU

[grid@rac2 ~]$  kfed read  /dev/dm-0  aun=2 blkn=3|grep .xptr.au |head -1kfffde[0].xptr.au:                    4 ; 0x4a0: 0x00000004

从上面信息可以看出，ACD元数据和数据应该包含在其中一个AU 4。

查看au 4的信息

grid@rac2 ~]$  kfed read /dev/dm-0 aun=4 blkn=1|morekfbh.endian:                          1 ; 0x000: 0x01kfbh.hard:                          130 ; 0x001: 0x82kfbh.type:                            8 ; 0x002: KFBTYP_CHNGDIRkfbh.datfmt:                          1 ; 0x003: 0x01kfbh.block.blk:                       1 ; 0x004: blk=1kfbh.block.obj:                       3 ; 0x008: file=3kfbh.check:                    17400326 ; 0x00c: 0x01098206kfbh.fcn.base:                        0 ; 0x010: 0x00000000kfbh.fcn.wrap:                        0 ; 0x014: 0x00000000kfbh.spare1:                          0 ; 0x018: 0x00000000kfbh.spare2:                          0 ; 0x01c: 0x00000000kfracdb.aba.seq:                      2 ; 0x000: 0x00000002kfracdb.aba.blk:                      0 ; 0x004: 0x00000000kfracdb.ents:                         2 ; 0x008: 0x0002kfracdb.ub2spare:                     0 ; 0x00a: 0x0000kfracdb.lge[0].valid:                 1 ; 0x00c: V=1 B=0 M=0kfracdb.lge[0].chgCount:              1 ; 0x00d: 0x01kfracdb.lge[0].len:                  52 ; 0x00e: 0x0034kfracdb.lge[0].kfcn.base:             1 ; 0x010: 0x00000001kfracdb.lge[0].kfcn.wrap:             0 ; 0x014: 0x00000000kfracdb.lge[0].bcd[0].kfbl.blk:       0 ; 0x018: blk=0kfracdb.lge[0].bcd[0].kfbl.obj:       4 ; 0x01c: file=4kfracdb.lge[0].bcd[0].kfcn.base:      0 ; 0x020: 0x00000000kfracdb.lge[0].bcd[0].kfcn.wrap:      0 ; 0x024: 0x00000000kfracdb.lge[0].bcd[0].oplen:          4 ; 0x028: 0x0004  ---表示长度，类似logfile dump的LENkfracdb.lge[0].bcd[0].blkIndex:       0 ; 0x02a: 0x0000kfracdb.lge[0].bcd[0].flags:         28 ; 0x02c: F=0 N=0 F=1 L=1 V=1 A=0 C=0kfracdb.lge[0].bcd[0].opcode:       212 ; 0x02e: 0x00d4  --opcode，类似数据库实例中的update/delete/insert操作的opcode编号kfracdb.lge[0].bcd[0].kfbtyp:         9 ; 0x030: KFBTYP_COD_BGO  --操作类型，类似数据库实例中的update/delete/insert等类型kfracdb.lge[0].bcd[0].redund:        17 ; 0x031: SCHE=0x1 NUMB=0x1  --这里表示冗余级别,17是unport,18是mirror,19表示highkfracdb.lge[0].bcd[0].pad:        63903 ; 0x032: 0xf99fkfracdb.lge[0].bcd[0].KFRCOD_CRASH:   1 ; 0x034: 0x00000001kfracdb.lge[0].bcd[0].au[0]:         46 ; 0x038: 0x0000002ekfracdb.lge[0].bcd[0].disks[0]:       0 ; 0x03c: 0x0000kfracdb.lge[1].valid:                 1 ; 0x040: V=1 B=0 M=0kfracdb.lge[1].chgCount:              1 ; 0x041: 0x01

check等信息属于hash值，每隔3s都会更新一次，可能是由于突然掉电，cache里的信息没有更新到磁盘中导致。

[grid@rac2 ~]$  kfed read /dev/dm-2 aun=4 blkn=0|grep ckptkfracdc.ckpt.seq:                     8 ; 0x018: 0x00000008kfracdc.ckpt.blk:                   390 ; 0x01c: 0x00000186[grid@rac2 ~]$  kfed read /dev/dm-0 aun=4 blkn=0|grep ckptkfracdc.ckpt.seq:                    70 ; 0x018: 0x00000046kfracdc.ckpt.blk:                   255 ; 0x01c: 0x000000f

三、 问题解决

通过kfed merge手工修改的方法恢复

kfed read /dev/dm-2 aun=4 blkn=0 >acd.txtkfed merge /dev/dm-2 aun=4 blkn=0  text=acd.txt

手工mount磁盘组

SQL>  alter diskgroup data mount; alter diskgroup data mount*ERROR at line 1:ORA-00600: internal error code, arguments: [kfrValAcd30], [DATA], [1], [8],[390], [9], [390], [], [], [], [], []
SQL> alter diskgroup data mount;Diskgroup altered.

如果kfed merge无法修复，请及时联系恩墨紧急救援（400-660-8755），以通过odu对asm底层进行恢复数据文件，从而最大限度恢复数据。

墨天轮原文链接：https://www.modb.pro/db/26884（复制到浏览器中打开或者点击“阅读原文”）

推荐阅读：144页！分享珍藏已久的数据库技术年刊

数据和云

ID：OraNews

如有收获，请划至底部，点击“在看”，谢谢！

点击下图查看更多 ↓

云和恩墨大讲堂 | 一个分享交流的地方

长按，识别二维码，加入万人交流社群

请备注：云和恩墨大讲堂

  点个“在看”

你的喜欢会被看到❤