Every DBA needs to know how to do their work under pressure, while facing crisis after crisis. Our most important contribution is to keep the database running. It′s an on-call function; you never know what might happen. Half the branch was up all night last night restoring a database because of a failure. Data recovery is very important, and so is performance tuning and problem solving. IT tends to start at the back end and work out to see where the problem lies, so generally, we′re one of the first areas that will be contacted when a problem occurs.
There are hundreds of possible sources of downtime hiding in the crooks and crannies of your company. Many outage situations cascade into each other, where an attempt to resolve a problem creates a new outage and quickly spirals further and further out of control. DBAs should always try to obtain critical diagnostics even when under pressure to bring the database up or press before system reboot.
数据挖掘研究院
This article describes some of the steps for a thorough post-mortem that a DBA in crisis mode should follow, especially in crises that happened unknowingly, caused database outage or company dollars. It is very important that some of these tasks listed below be performed before you resort to such drastic measures as restarting the database or server it self. I classify these tasks into three categories, tested on 10g and Linux environments. Please ensure you can run the tasks on your specific environment before you use them.
数据挖掘工具
- hanganalyze;
- system state dump;
- Errorstack if a specific process is at fault.
After the system has been stabilized and when there is time, open a Service Request with Oracle. Don′t forget to collect necessary trace and log files listed in the end of this article and upload them for the Oracle to work on your TAR
A. Hanganalyze:
HANGANALYZE helps you to determine if a performance analysis will be required to diagnose the real problem when a HANG is taking place
