It's time to re-arrange your thinking. Consider your SQL Servers not as regular database servers but as data centers,...
said Kent Erickson. "Think of them as critical business applications," with the goal of achieving the highest availability possible. Better overall business processes will result, he said.
But, how exactly do you transform your thinking, and what steps are best to obtain high availability? Erickson, director of product management for SQL at San Jose, Calif.-based NetIQ, Corp., outlined several dos and don'ts for SQL Server management. NetIQ announced its new SQL Server Management Suite on Feb. 4. The suite contains three new products: Configuration Manager, Recover Manager and Diagnostic Manager. It will be available in March.
Do employ key management practices, such as automated monitoring. That way, relying on specific manual processes to fix problems is not necessary. Further, by automating, you can predict overall performance metrics, Erickson said.
Do select tools that fix problems completely. Rather than constantly putting out fires, stop them at the core. This is faster than writing a unique solution that will be indecipherable to the next DBA who administers the server, he said.
Do use a change management product. This will ensure that all servers are configured the same way. For example, an NT administrator might change some settings on a server in an effort to increase availability. By accident, he could trigger a faulty recovery mode, which could reboot the server and change its configuration. The server's database administrator, however, may never know this happened. The end result is that the DBA never fixes the configuration, and the server is out of synch with the others, Erickson said.
Do implement good rollout processes. Once testing is complete, make sure the production center matches the test environment. Change aspects that are not the same before rollout is complete.
Don't implement something that has never been tested. The majority of problems occur, he said, because administrators don't fully know what they're doing, in Erickson's opinion.
Do get in the right frame of mind. You don't need to focus your efforts on writing code to solve every problem. If you do, "there are hundreds of things you they will need to write," he said. Instead, focus on implementing and administering management products.
So, do focus on top-level business needs, rather than furthering your own technical aptitude. "Brilliance is only a prerequisite for the job," Erickson said.
Don't create an irreversible change on the server. Erickson recalled one DBA who accidentally deleted a row from a spreadsheet that contained a list of people living in Tennessee. He really meant to delete the list of people living in Texas. This error caused several hours of downtime as the DBA worked to get the deleted information restored from back up.
Do track usage by end users. You can't test end user usage in the test environment, so keep close tabs on them after deployment, Erickson said. Do collect your SQL server's performance information today, he said. That way, you'll know when changes in performance are drastic. That information can also help with predictive analysis.
Do watch servers closely and be alerted when processes are slowing down. When you see a problem, do react as quickly as possible, Erickson concluded.
FOR MORE INFORMATION
SearchWindowsManageability has a whole Best Web Links section devoted to dos and don'ts.
Please let us know what subject you'd like to see some new dos and don'ts on. E-mail Editor@searchWindowsManageability.com