Avoiding Database Problems

I l @ ve RuBoard

You can take several proactive steps to avoid database problems. System and database server failures can be difficult to predict, but some of the root causes can be monitored . You can monitor how many rows a table gains over time, to monitor database growth. Gradual performance degradation or an available tablespace falling below critical thresholds could be an indication that a failure is imminent. Loss of data can be avoided through the use of data redundancy. Several techniques are available, such as disk mirroring. This is not sufficient, however, unless the failure of a disk mirror, for example, is quickly detected and repaired. The EMS HA Monitors product is one tool for detecting mirror failures.

The database operator should also periodically check system and database log files (such as Oracle's alert file) to look for failure information or unusual events.

Tools for monitoring system performance can be found in Chapter 4. A tool such as PerfView can be used to discover any resource utilization and performance trends. This information can be used by a database administrator to project database growth and plan for increased capacity. Various monitoring tools, such as BMC PATROL or SMART Plug-Ins, can be used to gather resource threshold information. If a database is out of available space, you need to archive or delete unneeded data, or allocate more space.

Some tools are available to help you properly tune your system. Oracle provides an add-on product, Oracle Expert, which is an expert system that uses configuration and tuning rules specific for Oracle databases. BMC PATROL has built-in self-tuning and load-balancing capabilities, such as compressing log files and resizing extents.

Monitoring backups and security are two additional proactive steps that you can take. Database backups should be verified periodically to ensure that data can be restored when needed. By maintaining audit logs of system access, you can more easily detect security intrusions and respond before experiencing database outages.

I l @ ve RuBoard