Recovering from Database Problems

I l @ ve RuBoard

Tools such as BMC PATROL can detect failures of key database processes and restart them automatically. MC/ServiceGuard provides a similar capability, and is also able to restart an application on another system.

Assuming successful backups exist, databases such as Oracle can recover from a disk failure without losing data. This requires a trained database administrator, because the process typically is not automated. The Oracle Recovery Manager product can be used to manage the backup, restore, and recovery processes.

After a system failure, Oracle can automatically recover data to a consistent state by performing full recovery. This is done by ensuring that committed changes are applied to the database and that uncommitted changes are undone. This can be a time-consuming activity. If the system problem is persistent, the workload may need to be moved to another system. For HP-UX systems, MC/ServiceGuard provides the capability of automatically moving a workload. To avoid the downtime associated with a single node failure, Oracle Parallel Server can be used. HP's MC/LockManager provides high availability for HP-UX Oracle Parallel Server environments.

In addition to system failures, human error, such as deleting the wrong rows in a database table, can lead to downtime. Successful database backups are critical, to ensure that the system can be returned to a good, consistent state.

To determine how to recover from Oracle database problems, refer to the following documents from Oracle: Oracle7 Server Administrator's Guide , Oracle7 Server Tuning, Oracle7 Server Reference , and Oracle7 Server SQL Reference . Similar manuals exist for later versions of Oracle as well.

I l @ ve RuBoard