RAC Renewal OCR / Voting disk

My test environment: 

Master System: winXP T400 
Virtual Software: GXS vmware3.2.1 
guest system: as5u1 

oracle db: oracle10.2.1 

As the cause of the accident two days ago, my colleagues, under the new plug power supply, so that my test environment rac unexpected power outages, resulting in a vote (votingdisk) virtual disk fails, causing systems to get up. How do? I was thinking is that until the last moment, determined not to reload. Recovery process is also familiar with the process. 

As the votingdisk disk failure, the system can not start, reporting votingdisk disk is not recognized, in order to allow the system to start, I votingdisk disk in vmware in re-add it again, before the system can restart. System up, the node rac1 gnome encountered the bug, better not affect the normal work rac. 

Because my votedisk damage, and there is no backup, we can only rebuild votedisk 

votedisk damage can only be reconstructed in two ways votedisk 

1. If votedisk back up, down with the backup and recovery 

eg: 
Backup votedisk: dd if = / dev/raw/raw2 f = / tmp / votedisk.bak 
Recovery votedisk: dd if = / tmp / votedisk.bak f = / dev/raw/raw2 

2. No votedisk backup, then it can only reinstall clusterware the (votedisk the information is deleted or added in the installation of clusterware node node to add the information to go in and have ocssd process maintenance) 

According to my description of the failure phenomenon, now can only reinstall clusterware to re-initialize votedisk, has achieved the purpose of restoration 

This is the fastest way 

Steps: 

1. Stop all node clusterware stack 

[Root @ rac2 ~] # su - oracle 
[Oracle @ rac2 ~] $ su 
Password: 
[Root @ rac2 oracle] # crsctl stop crs 

2. Backup directory for each node clusterware 

[Root @ rac1 oracle] # mv / u01/crs / u01/crsbak 

3. In order to avoid the new configuration information to configure and old conflicts, the best clear the configuration information must clusterware 

Clean up the old configuration information (in each node is running) 

3.1 crs from start to delete information 

Modify / etc / inittab, remove the following three lines. 

h1: 2: respawn: / etc / init.evmd run> / dev / null 2> & 1 dev / null 
h2: 2: respawn: / etc / init.cssd fatal> / dev / null 2> & 1 dev / null 
h3: 2: respawn: / etc / init.crsd run> / dev / null 2> & 1 dev / null 

linux: 

rm / etc / oracle / * 
rm-f / etc / init.d / init.cssd 
rm-f / etc / init.d / init.crs 
rm-f / etc / init.d / init.crsd 
rm-f / etc / init.d / init.evmd 
rm-f / etc/rc2.d/K96init.crs 
rm-f / etc/rc2.d/S96init.crs 
rm-f / etc/rc3.d/K96init.crs 
rm-f / etc/rc3.d/S96init.crs 
rm-f / etc/rc5.d/K96init.crs 
rm-f / etc/rc5.d/S96init.crs 
rm-Rf / etc / oracle / scls_scr 
rm-f / etc / inittab.crs 
cp / etc / inittab.orig / etc / inittab 

ps-ef | grep init.d 

/ Etc / init.d / init.crsd 
/ Etc / init.d / init.evmd 
/ Etc / init.d / init.cssd 

3.2 clean up the old configuration information 

rm-rf / etc / oracle / * (delete ocr.loc) 
rm-rf / var / tmp / .oracle or / tmp / .oracle 

3.3 Use dd clear vote disk and ocr (raw device) 

dd if = / dev / zero f = / dev / votedisk_device bs = 8192 count = 2560 
dd if = / dev / zero f = / dev / ocr_device bs = 8192 count = 12800 

Reference: 

Clear 10g RAC CRS method 
http://6month.itpub.net/post/37672/470422 

4. Were performed at each node $ CRS_HOME / install / rootdelete.sh 

5. In any node on the implementation of the script. $ CRS_HOME / install / rootdeinstall.sh, only one node can run 

Crs above configuration is deleted, the following is re-install crs, execute the following command to check if there is no return value, it can continue to install 

ps-e | grep-i 'ocs [s] d' 
ps-e | grep-i 'cr [s] d.bin' 
ps-e | grep-i 'ev [m] d.bin' 

6. And in step 5, a node with the script. $ CRS_HOME / root.sh 

7. In other node implementation of the $ CRS_HOME / root.sh, and then note the last node of the output and, finally, run ". / VIPCA" 

8. With netca reconfigure listener, to confirm the registration to ocr 

This time with crs_stat-t-v can see the listener, ONS, GSD, VIP registration to ocr, the also need to asm, the database also registered in ocr 

9. ASM registered to ocr 

srvctl add asm-n rac1-i + ASM1-o $ ORACLE_HOME 
srvctl add asm-n rac2-i + ASM2-o $ ORACLE_HOME 

10. Start ASM 

srvctl start asm-n rac1 
srvctl start asm-n rac2 

Usually starts when the last instance will asm error, generally because of rac can not confirm which network to use as a private interconnect 
So can two ASM instance pfile to add the following two parameters 

Add two parameters in the file / u01/app/oracle/product/10.2.0/db_1/dbs/init + ASM2.ora 

+ ASM1.cluster_interconnects = '192 .168.0.31 ' 
+ ASM2.cluster_interconnects = '192 .168.0.22 ' 

ip must write right, or have problems, and then restart the asm instance, the problem can be solved 

11. Manual registration database objects (database name and db_name to the same case) 

srvctl add database-d RAC-o $ ORACLE_HOME 

12. Manual registration of two instances of objects (instances of the same name to and isntance_name case) 

srvctl add instance-d RAC-i rac1-n rac1 
srvctl add instance-d RAC-i rac2-n rac2 

13. Modify instances and examples of correspondence between asm 

srvctl modify instance-d RAC-i RAC1-s + ASM1 
srvctl modify instance-d RAC-i RAC2-s + ASM2 

14. Start the database (in this process will usually error, so as to hand-asm instance of the specified private interconnect) 

sql> alter system set cluster_interconnects = '192 .168.0.31 'scope = spfile sid =' * ' 
sql> alter system set cluster_interconnects = '192 .168.0.22 'scope = spfile sid =' * ' 

Then start the database 

srvctl start database-d RAC 

15. Check ons is normal 

[Root @ rac2 oracle] # ps-ef | grep ons 
root 4008 10066 0 08:39 pts / 0 00:00:00 grep ons 
root 5660 1 0 06:42? 00:00:00 sendmail: accepting connections 
oracle 16107 1 0 07:21? 00:00:00 / u01/crs/oracle/product/10.2.0/crs/opmn/bin/ons-d 
oracle 16108 16107 0 07:21? 00:00:00 / u01/crs/oracle/product/10.2.0/crs/opmn/bin/ons-d 

[Root @ rac2 oracle] # onsctl ping 
Number of onsconfiguration retrieved, numcfg = 0 
ons is not running ... 
[Root @ rac2 oracle] # onsctl start 
Number of onsconfiguration retrieved, numcfg = 0 
Number of onsconfiguration retrieved, numcfg = 0 
onsctl: ons started 
[Root @ rac2 oracle] # onsctl ping 
Number of onsconfiguration retrieved, numcfg = 0 
ons is running ... 
[Root @ rac2 oracle] # 

If ons can not start, you can use the following command, you can also directly edit the file $ CRS_HOME / opmn / conf / ons.config 

racgons add_config hostname1: port hostname2: port 
racgons remove_config hostname1: port hostname2: port 

16. Check whether the network command oifcfg normal 

eg: add interface configuration 

oifcfg setif-global eth0/192.168.2.0: public 
oifcfg setif-global eth1/192.168.0.0: cluster_interconnect 

17. With the following two commands verify the correctness of configuration 

1. [Oracle @ rac1 ~] $ / tmp/10201_clusterware_linux32/clusterware/cluvfy/runcluvfy.sh stage-post crsinst-n rac1, rac2 

2. Crs_stat-t-v 

This votingdisk already successfully re-configured, but also encountered in the process are many problems, the following is the summary of issues, including the third issue left a lot of detours, with a small half-time, other problems are are resolved quickly. 

1. As votingdisk disk failure, so re-add the removed disk in vmware 
2. Crs start-up problems (time synchronization) 
3. Asm2 not start instance of the problem (the point written as a comma) 
4. Can not start a database problem 
5. Onsctl startup problems 
6. Change / dev / raw is the main problem 
7. Oifcfg home network problems 

Specific problems and solutions, please refer to another article: http://blog.csdn.net/wyzxg/archive/2010/05/09/5572418.aspx 

Reference document: 

Reconstruction votedisk / ocr 
http://www.laoxiong.net/10g_rebuild_crs_rac.html 
http://www.dbasoul.com/2010/700.html 

------ End ------

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/230160/viewspace-675870/，如需转载，请注明出处，否则将追究法律责任。

转载于:http://blog.itpub.net/230160/viewspace-675870/