Tek-Tips is the largest IT community on the Internet today!

Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!

  • Congratulations gkittelson on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Can't start databases...

Status
Not open for further replies.

THoey

IS-IT--Management
Jun 21, 2000
2,142
US
Okay, previously I had a thread that mentioned that our Oracle services seemed to have disappeared when we did a full restore from tape backup. What led up to this is that we had gotten new HD's for our NT server. I shutdown Oracle completely, we did a complete backup of the smaller HD's, swapped out and configured the new drives, and then did a restore from tape. Somehow, the services for the four instances did not get written into the registry. What we ended up doing is re-restoring from the back up and then everything appeared as normal. Well, appearances can be deceiving...

Of the four instances, only one will come up to a Open status. The other three will go so far as to be mounted, but then return the following error:

ORA-00600: internal error code, arguments: [2662], [0], [1028815], [0], [1029230], [8388610], [], []

I have done some research into this error in the past, and it usually means a corrupt system table. Anyone know of any tool or way to restore the system tables, with out blowing everything else away? Luckily, I wasn't too far into these new database instances. I can delete all of the corrupt instances, rebuild and start reloading, but it is a pain.

Any ideas that don't have anything to do with arsenic will be greatly appreciated. "Alcohol is our friend, and it's about time we had more friends over."

Terry M. Hoey
 
I just lost an Oracle 7 database on an NT server and had to have the LAN admin restore the files from tape. I restored the following files from backup - control, datafiles, pwdctl. I did not restore the redo logs - I just started writing on them after the restore. I used the init.ora file from before the db failure. Doing it this way the db came back up.

On another NT server the harddrive had to be replaced as part of an upgrade. I shut down the Oracle 8i db, and copied all the following to another server: datafiles, redo logs, control, init.ora, pwdctl. Then I had to reinstall oracle to get all the registry entries restored. Once oracle was restored, I deleted all the inital pwdctl, init.ora, redo logs, and datafiles set up in the vanila database. I copied the saved files to the appropriate drive, and restarted the database. I kept the SID the same for both the old installation and the new oracle installation. Then I restarted the database.

In your case you may have to do the same reinstall process modiufied slightly for each of the 3 db's that did not get restored. I think you have to use the oradim program and that will create the db name in the registry and set up the db services. Check the ora doc on that.

For whatever reason neither db had archiving turned on. Now I am not relying on just the image backups anymore for several reasons - and am moving to setup hotbackups using export and RMAN.

THere are several problems with the image backups as a restore mechanism:
1. The datafiles may get large on your db and restoring large datafiles from image backup restore tapes over the LAN can take a lo-o-o-o-ng time depending on your environment and hardware. It took more than 15 hours to restore the ora7 db whose sum total of datafiles occupied about 5gb.

The LAN Admin has complained about some datafiles that get as big as 5gb slowing down his image full backup runs - because these large files have to be moved across the LAN.
2. The exported oracle backups are considerably smaller than image backups speeding up up both backup and recovery.

3. In my case the ora7 database was in use for a few years before I arrived 10 months ago - datafiles were scattered over all three hd's, and buried in a mishmash of directories and levels of directories - so finding the files and restoring them from the backup was more prone to error in datafile path identification, and the datafiles were distributed across the entire backup of the harddrive, rather than having all datafiles concentrated together. Altogether there were more than 30 datafiles to restore.

Hope this helps - good luck

Perhaps
 
Thanks for the info. I ended up getting into the one database that I "needed" to with some help from Oracle. Once in, I did a complete export, blew away the instance, recreated it and then imported the information. Didn't trust it in the state it was in. The rest of the instances, I just recreated as they were too new to worry about.

I have also set up some scripts to perform exports of the two instances I currently have on line, one daily and one weekly. The daily one, I am keeping a weeks worth of backups, and the weekly one, I am keeping three weeks worth of.

Thanks again for the info. I will keep it all in mind... Terry M. Hoey
 
Whenever I set up a schema I set the datafiles to max out at a number such as 1 to 2 GB so the backup services do not get overwhelmed with exteremely large files. If this is exceeded I go to a second datafile for the tablespace. Either I do it this way or I set max out the db user at somewhere between 200mb and 2gb for each ts they use. The downside of this is that more than one user may use the same tablespace.

Glad to help

Ray
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top