|< Day Day Up >|| |
At Horatio's Woodscrews, the system administrators finally installed the new network card. With the system down, they took the opportunity to install new drivers for their storage area network (SAN) device. Then they contacted the DBA to let him know he could start the database.
The DBA starts the database, and begins to run through the system checks to make sure all the applications will start up correctly. Then he initiates all the reports that failed because of the shutdown. Everything seems to be going smoothly, until he notices that one of the reports has an error:
ORA-1578: ORACLE data block corrupted (file # 121, block # 68)
As the DBA began to investigate the problem, the system administrators called him back: the new SAN driver is causing problems, and he should shut down the database immediately. But the damage has already been done-database corruption.
But there is only one corrupt block, in the WS_APP_DATA01.DBF file. Just one block out of thousands of perfectly good blocks. With datablock corruption, the solution to the problem is to restore the entire datafile from backup, and then apply archivelogs to recover the file. Because the datafile contains parts of the Woodscrew, woodscrew_orders, and woodscrew_inventory tables, these objects will be offline until the file can be restored and recovered. The brief outage for the hardware fix now has been extended to include the restore of a file from backup, and then the application of the archivelogs to that file. The tablespace will not be available until recovery completes.
When RMAN is utilized for backups, you can use those backups to restore a single block from the last good backup, and then perform media recovery on that block (apply archive log changes, if there are any). You can do it for a single block or for a list of corrupt blocks, and the tablespace stays online while you do recovery. In fact, the corrupt table stays online, too. We discuss this in Chapter 8.
|< Day Day Up >|| |